Page 1
Historical influences dominate the compositionof regenerating plant communities in abandoned citrusgroves in north-central Florida
Graeme S Cumming AElig Ann George
Received 28 May 2008 Accepted 20 May 2009 Published online 11 June 2009
Springer Science+Business Media BV 2009
Abstract The question of what determines plant
community composition is fundamental to the study
of plant community ecology We examined the
relative roles of historical land use landscape con-
text and the biophysical environment as determinants
of plant community composition in regenerating
citrus groves in north-central Florida Results were
interpreted in light of plant functional traits Herba-
ceous and woody plants responded differently to
broad-scale variables herbs correlated most strongly
with surrounding land cover at a scale of 8 km while
the only significant determinant of woody species
distributions was local land use history There were
significant correlations between herbaceous species
and spatial context habitat isolation environmental
variables and historical variables Partial Mantel
tests indicated that each variable provided a unique
contribution in explaining some of the variation in the
herbaceous dataset The correlation between woody
plants and local historical variables remained signif-
icant even with other effects corrected for In the
herbaceous community species composition was
linked to functional traits much as expected from
classical theory While spatial influences in our study
system are important for both woody and herbaceous
plants the primary determinant of plant community
composition in regenerating citrus groves is historical
land use Our results suggest that the fine-scale
mechanisms of local competition tolerance and
facilitation invoked by many classical studies may
ultimately be less important than land use history in
understanding current plant community composition
in regenerating agricultural areas
Keywords Succession Restoration Biodiversity Path dependency Community composition Functional trait
Introduction
The abandonment of agricultural land is a common
phenomenon in many parts of the world It is often
assumed that abandoned land will follow a traditional
old-field successional trajectory gradually returning
to its pre-cultivation state (Egler 1954) The likeli-
hood that this occurs is however contingent on a
number of variables that influence successional
outcomes across a range of different scales Local
regional and historical processes matter to differing
degrees in any one instance and their relative
G S Cumming A George
Department of Wildlife Ecology and Conservation
University of Florida Gainesville FL 32611 USA
G S Cumming (amp)
Percy FitzPatrick Institute DSTNRF Center
of Excellence University of Cape Town Rondebosch
Cape Town 7701 South Africa
e-mail graemecumminguctacza
123
Landscape Ecol (2009) 24957ndash970
DOI 101007s10980-009-9368-2
importance may be difficult to predict a problem that
is exacerbated by differences in the traditions and
theoretical approaches of fine- and broad-scale ana-
lysts (eg Connell and Slayter 1977 MacArthur and
Wilson 1967)
Important local influences on plant communities in
abandoned agricultural lands include biological inter-
actions such as tolerance facilitation or competition
(eg (Connell and Slayter 1977 Grime 1973 Grime
1979 Tilman 1994) and within-site resource avail-
ability (Mou et al 2005 Verheyen and Hermy 2001)
Local-scale influences are in many ways the best
understood aspect of plant community ecology in
part because they are amenable to controlled exper-
imentation over short time periods
At broader spatial scales regional variation in
habitats becomes more important and species occur-
rences may be more obviously dispersal-limited
Post-agricultural sites vary in their internal quality
and species composition and are surrounded by
different land cover types with different permeabil-
ities to dispersing propagules (Gustafson and Gardner
1996 Ricketts 2001) Studies of regenerating old-
fields have typically found that propagules with short
mean dispersal distances were strongly affected by
habitat isolation while wind dispersed species with
small light propagules were able to spread indepen-
dently of landscape context (eg Dzwonko and
Loster 1992)
Historical influences on community composition
include both past land use practices and historical
events such as perturbations or the introduction of
invasive species Large infrequent perturbations
(eg hurricanes or hot fires) can have lasting impacts
that may override local or regional environmental
control of vegetation patterns (de Blois et al 2001
Flinn and Vellend 2005) and many land-uses leave
legacies that persist for many years even once they
have been discontinued For instance bulldozing
remnant vegetation after land abandonment not only
reduces or eliminates established populations (Motz-
kin et al 1996) but also creates harsh microsite
conditions which favour colonization by opportunis-
tic ruderal species (De Steven 1991) and herbicide
or fertiliser use may favour species adapted to
nutrient additions (Aviron et al 2005)
Understanding the roles of local regional and
historical factors and their interplay with the differ-
ent life history strategies of plants is one of the
central goals of plant community ecology In this
paper we explore the scale dependencies of succes-
sional dynamics in abandoned citrus groves in north-
central Florida We contrast three hypotheses that
explain successional dynamics in post-agricultural
landscapes They include (1) that the signature of
previous land use is the primary determinant of the
subsequent composition of the plant community
within a given range of biophysical conditions (2)
that plant succession is determined primarily by the
existing biophysical template including such vari-
ables as soil type and water availability and (3) that
plant succession is driven primarily by broad-scale
variation in the surrounding environment through its
influence on such processes as the availability of
propagules from potential colonising species These
hypotheses are considered for herbaceous plants and
woody plants respectively and in relation to plant life
history traits
Methods
Study system
Citrus has traditionally been one of the most
economically valuable agricultural commodities in
North-central Florida providing around 25 of the
worldrsquos citrus at its peak (Weischet and Caviedes
1987) Two-thirds of the statersquos citrus crop was
grown on the Central Ridge in peninsular Florida
where the slightly elevated topography and well-
drained sandy soils provided an ideal disease-free
environment (Ewel 1990) In the 1980s a series of
freezes resulted in the abandonment of many citrus
groves in the north-central region (Miller 1991)
Recent abandonment has been driven by canker
(Gottwald et al 2001) and socioeconomic pressures
(Ewel 1990) The majority of groves were converted
to pine plantations or pasture after the freezes but a
small proportion (our study system) was left to
regenerate naturally
Field methods
We studied 66 abandoned citrus groves across six
counties (Hillsborough Lake Marion Pasco Polk
and Putnam) in north-central Florida (Fig 1) Study
groves varied in time since abandonment and size
958 Landscape Ecol (2009) 24957ndash970
123
The lsquolsquooldestrsquorsquo groves were abandoned prior to the
freeze decade in the mid 1960s while the lsquolsquoyoungestrsquorsquo
groves were abandoned in 2003 The year of aban-
donment was determined through informal interviews
with landowners or long-term residents living around
the sites Grove area ranged from 05ndash60 ha with a
single sampling unit placed in each grove at least
1 km from other sampling sites We did not consider
groves smaller than half a hectare because of
concerns about edge effects (Gehlhausen et al
2000 Hansen and Clevenger 2005) An abandoned
grove was considered a discrete unit if it was
surrounded on all sides by different land-use classes
We sampled only in sites where no subsequent
management (eg periodic mowing or grazing) had
been undertaken and we selected areas that had
similar fertilizer and herbicide regimes while in
production Land-use history information was
obtained from interviews with local caretakers and
UFrsquos Institute for Food and Agricultural Sciences
(IFAS) extension agents
We used a variant of the Gentry-Boyle transect
methodology (Boyle 1996) to quantify community
composition for woody and herbaceous species
Different sampling techniques were selected for
woody and herbaceous plants respectively (Bonham
1989) The basic sampling unit consisted of six
parallel 2 9 50 m belt transects running in a North-
South direction with the mid-point of each separated
by 10 m We located grove boundaries and then
chose a central starting point away from any edge
that became the SW corner of the first belt transect
Citrus trees were planted 25 feet (approx 8 m) apart
prior to the freezes in the 80 s and 125 feet (approx
4 m) apart after the freeze decade (R Wigul personal
communication) Transects were thus out of phase
Fig 1 Map of north-central Florida showing the distribution of sampling sites within Hillsborough Lake Marion Pasco Polk and
Putnam counties
Landscape Ecol (2009) 24957ndash970 959
123
with historical lines of citrus trees UTM grid
coordinates for the outer corners of the first and last
belt transect (A-D) were obtained using a hand-held
Garmin GPS 76S Average soil organic matter
content was obtained by taking soil samples with a
hand-held auger to a depth of 8 inches at four
randomly chosen locations at each sampling site
Samples were analysed by Waters Agricultural Labs
Woody species
Woody species were censused by counting all
individuals rooted within 1 m on either side of a
50 m tape We assumed that each rooted stem
belonged to a separate individual unless stems joined
at the base of the trunk We identified all woody
plants (trees shrubs and lianas) that were at least
25 cm diameter at breast height (dbh) defined as
13 m from the base of the stem Trees and shrubs
above 25 cm dbh were included if the center of the
base of the trunk lay within 1 m of either side of the
tape Lianas (here defined as ground-rooted epiphytes
or creepers) were included if at least one of their roots
entered the soil within the transect and if they had a
diameter of at least 25 cm at any point along the
stem Total dbh for stems that divided above ground
on a single trunk was calculated using the sum of
individual stem diameters
Herbaceous species
All herbaceous plants were identified in 30 (1 9
1 m) quadrats within the six belt transects Five
quadrants were located along the length of each belt
transect at intervals of 10 m (5 15 25 35 and 45 m
respectively) Because of the difficulty in recognizing
individuals of stoloniferous or rhizomatous species
we instead recorded the presence of species in each
quadrat and calculated a relative frequency matrix
(number of occurrences out of 30) for each sampling
unit For upright species presence was determined if
the plant was rooted within the quadrat for species
with a creeping habit an individual was counted if a
shoot fell within the quadrat (Greig-Smith 1983) A
few morphologically similar species were difficult to
identify without flowering or fruiting parts and were
classified as single operational taxonomic units
Specimens of both woody and herbaceous species
were brought to the University of Florida Herbarium
in the Florida Museum of Natural History for
identification and keyed to species with the help of
expert consultation (R Abbott K Perkins and W
Judd) using Wunderlin and Hansen (2003) Thirty-
four tree species and eighty-eight herbaceous species
were identified in total
Remotely sensed data
We used a land cover map produced by the Florida
Fish and Wildlife Conservation Commission (FWC)
(Stys et al 2004) The FWC map contains 43
categories comprising 27 natural classes and 16
disturbed classes We merged classes of interest into
five broad categories xeric uplands (xeric oaksand
pine scrubsandhill) mesic forest (mixed hardwood
pinehardwood hammock and forestpinelands) shrub-
landgrassland (shrub and brushlandgrasslanddry
prairieimproved and unimproved pasture) citrus
and low-impact urban These classes were chosen as
potential propagule sources that might shape succes-
sional communities in regenerating groves All other
land cover classes were removed from the analysis
The accuracy of the classification was checked by the
collection of training points in the field on land cover
patches with a minimum of 100 m2 area Point
locations were recorded with a Garmin 76S GPS
receiver The overall accuracy for the classification
was 824 with a Kappa statistic of 079 (Table 1)
We used ArcView 33 (ESRI 2002) to create
multiple buffers around sampled groves at distances
of 05 1 15 2 4 6 8 and 10 km with buffer radii
being used as a proxy for scale (Holland et al 2004)
This range was selected to reflect the potential
Table 1 Results of the accuracy assessment carried out on the
merged Florida fish and wildlife conservation commission land
cover map
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 100 097 096
Urban 076 067 061
Xeric uplands 076 086 084
Grassland 070 082 080
Mesic forest 091 077 073
Citrus 077 096 095
The overall classification accuracy is 824 with a Kappa
statistic of 079
960 Landscape Ecol (2009) 24957ndash970
123
dispersal distances of both woody and herbaceous
species We extracted the area of each land cover
type within each buffer and created a site by habitat
diversity matrix for each of the eight scales of
analysis
We obtained Landsat 7 Enhanced Thematic Map-
per (ETM) images for path 16row 40 path 17row
40 and path 16row 41 in March 2003 The historic
path 16row 40 and path 17row 40 images were
obtained from February and April 1988 and the path
16row 41 scene from April 1989 Radiometric
calibration was carried out on each image to correct
for sensor related sources of variance variation in
atmospheric conditions and differences in solar angle
at the time of image acquisition (Jensen 1996) The
2003 image was classified by creating training
signatures on the image using homogeneous ground
cover types The training sample data collected to
verify the accuracy of the FWC land cover map were
used to conduct an accuracy assessment for the
supervised classification The overall accuracy for the
classification was 703 (Table 2) The same signa-
ture file was used on the 198889 images The area of
each land cover category within an 8 km radius
around each grove was tabulated to create a site by
land cover area matrix at each time step For each
site the change in area was obtained by subtracting
the area of that class in the 198889 image from its
corresponding area in the 2003 image
Functional trait assessment
To link our analysis more directly to the ecology of
the plants in our study sites we undertook a simple
analysis of plant species composition by functional
trait Data were assembled for each species on
dispersal mode life-span presence of woody stems
vegetative propagation and seed bank persistence
(Tables 3 4)
Data analysis
We used Mantel tests at different scales to establish
whether species composition correlated with the
amount of suitable habitat in the landscape The
herbaceous dataset (HERB) was tabulated as a
frequency matrix of each herbaceous species occur-
rence out of a possible 30 quadrants at a site Data for
the woody species matrix (TREE) were collected as
abundance of individuals per site but to make the
results of the Mantel tests comparable we transformed
these data into a relative abundance matrix
We used Spearmanrsquos ranked correlation coefficient
in the Mantel analysis as recommended by Dietz
(1983) Many species were represented by very few
individuals since these data do not usefully reflect
ecological preference or site suitability (Legendre and
Legendre 1998) we removed species that occurred in
fewer than 5 (B 3) of the 66 sampling sites (see also
McCune and Grace 2002) We calculated dissimilar-
ities for woody and herbaceous species composition
using Bray-Curtis distance (Legendre and Legendre
1998) The analysis included 37 herbaceous species
and 14 woody species
At larger scales of analysis buffer overlap can
influence the independence of samples We used the
program Focus (Holland et al 2004) with 1000
iterations at each scale to select spatially indepen-
dent sites Nonparametric Mantel tests were carried
out on 25 randomly chosen site combinations at each
scale The mean Mantelrsquos r statistic and standard
deviation were plotted against the scale of analysis
and statistical significance was determined using
randomization procedures (Fortin and Dale 2005)
Since we used 8 nested scales (05ndash10 km) a0 was
obtained by dividing 005 by 8 outcomes to give a
Bonferroni-corrected significance criterion for the
Mantel r- statistic of P 000625 (a0 = 0058) The
P-values were plotted against the correlation coeffi-
cients and the proportion of P-values less than
000625 tabulated We defined significant r-statistics
as those for which at least 50 of the P-values were
less than 000625
Table 2 Results of the accuracy assessment carried out on the
supervised classification of the mosaiced 2003 Landsat ETM
images
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 097 1000 1
Urban 076 043 034
Xeric uplands 056 088 085
Grassland 055 050 042
Mesic forest 091 077 073
Citrus 047 078 072
The overall classification accuracy is 7027 and the
corresponding Kappa statistic is 06443
Landscape Ecol (2009) 24957ndash970 961
123
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 2
importance may be difficult to predict a problem that
is exacerbated by differences in the traditions and
theoretical approaches of fine- and broad-scale ana-
lysts (eg Connell and Slayter 1977 MacArthur and
Wilson 1967)
Important local influences on plant communities in
abandoned agricultural lands include biological inter-
actions such as tolerance facilitation or competition
(eg (Connell and Slayter 1977 Grime 1973 Grime
1979 Tilman 1994) and within-site resource avail-
ability (Mou et al 2005 Verheyen and Hermy 2001)
Local-scale influences are in many ways the best
understood aspect of plant community ecology in
part because they are amenable to controlled exper-
imentation over short time periods
At broader spatial scales regional variation in
habitats becomes more important and species occur-
rences may be more obviously dispersal-limited
Post-agricultural sites vary in their internal quality
and species composition and are surrounded by
different land cover types with different permeabil-
ities to dispersing propagules (Gustafson and Gardner
1996 Ricketts 2001) Studies of regenerating old-
fields have typically found that propagules with short
mean dispersal distances were strongly affected by
habitat isolation while wind dispersed species with
small light propagules were able to spread indepen-
dently of landscape context (eg Dzwonko and
Loster 1992)
Historical influences on community composition
include both past land use practices and historical
events such as perturbations or the introduction of
invasive species Large infrequent perturbations
(eg hurricanes or hot fires) can have lasting impacts
that may override local or regional environmental
control of vegetation patterns (de Blois et al 2001
Flinn and Vellend 2005) and many land-uses leave
legacies that persist for many years even once they
have been discontinued For instance bulldozing
remnant vegetation after land abandonment not only
reduces or eliminates established populations (Motz-
kin et al 1996) but also creates harsh microsite
conditions which favour colonization by opportunis-
tic ruderal species (De Steven 1991) and herbicide
or fertiliser use may favour species adapted to
nutrient additions (Aviron et al 2005)
Understanding the roles of local regional and
historical factors and their interplay with the differ-
ent life history strategies of plants is one of the
central goals of plant community ecology In this
paper we explore the scale dependencies of succes-
sional dynamics in abandoned citrus groves in north-
central Florida We contrast three hypotheses that
explain successional dynamics in post-agricultural
landscapes They include (1) that the signature of
previous land use is the primary determinant of the
subsequent composition of the plant community
within a given range of biophysical conditions (2)
that plant succession is determined primarily by the
existing biophysical template including such vari-
ables as soil type and water availability and (3) that
plant succession is driven primarily by broad-scale
variation in the surrounding environment through its
influence on such processes as the availability of
propagules from potential colonising species These
hypotheses are considered for herbaceous plants and
woody plants respectively and in relation to plant life
history traits
Methods
Study system
Citrus has traditionally been one of the most
economically valuable agricultural commodities in
North-central Florida providing around 25 of the
worldrsquos citrus at its peak (Weischet and Caviedes
1987) Two-thirds of the statersquos citrus crop was
grown on the Central Ridge in peninsular Florida
where the slightly elevated topography and well-
drained sandy soils provided an ideal disease-free
environment (Ewel 1990) In the 1980s a series of
freezes resulted in the abandonment of many citrus
groves in the north-central region (Miller 1991)
Recent abandonment has been driven by canker
(Gottwald et al 2001) and socioeconomic pressures
(Ewel 1990) The majority of groves were converted
to pine plantations or pasture after the freezes but a
small proportion (our study system) was left to
regenerate naturally
Field methods
We studied 66 abandoned citrus groves across six
counties (Hillsborough Lake Marion Pasco Polk
and Putnam) in north-central Florida (Fig 1) Study
groves varied in time since abandonment and size
958 Landscape Ecol (2009) 24957ndash970
123
The lsquolsquooldestrsquorsquo groves were abandoned prior to the
freeze decade in the mid 1960s while the lsquolsquoyoungestrsquorsquo
groves were abandoned in 2003 The year of aban-
donment was determined through informal interviews
with landowners or long-term residents living around
the sites Grove area ranged from 05ndash60 ha with a
single sampling unit placed in each grove at least
1 km from other sampling sites We did not consider
groves smaller than half a hectare because of
concerns about edge effects (Gehlhausen et al
2000 Hansen and Clevenger 2005) An abandoned
grove was considered a discrete unit if it was
surrounded on all sides by different land-use classes
We sampled only in sites where no subsequent
management (eg periodic mowing or grazing) had
been undertaken and we selected areas that had
similar fertilizer and herbicide regimes while in
production Land-use history information was
obtained from interviews with local caretakers and
UFrsquos Institute for Food and Agricultural Sciences
(IFAS) extension agents
We used a variant of the Gentry-Boyle transect
methodology (Boyle 1996) to quantify community
composition for woody and herbaceous species
Different sampling techniques were selected for
woody and herbaceous plants respectively (Bonham
1989) The basic sampling unit consisted of six
parallel 2 9 50 m belt transects running in a North-
South direction with the mid-point of each separated
by 10 m We located grove boundaries and then
chose a central starting point away from any edge
that became the SW corner of the first belt transect
Citrus trees were planted 25 feet (approx 8 m) apart
prior to the freezes in the 80 s and 125 feet (approx
4 m) apart after the freeze decade (R Wigul personal
communication) Transects were thus out of phase
Fig 1 Map of north-central Florida showing the distribution of sampling sites within Hillsborough Lake Marion Pasco Polk and
Putnam counties
Landscape Ecol (2009) 24957ndash970 959
123
with historical lines of citrus trees UTM grid
coordinates for the outer corners of the first and last
belt transect (A-D) were obtained using a hand-held
Garmin GPS 76S Average soil organic matter
content was obtained by taking soil samples with a
hand-held auger to a depth of 8 inches at four
randomly chosen locations at each sampling site
Samples were analysed by Waters Agricultural Labs
Woody species
Woody species were censused by counting all
individuals rooted within 1 m on either side of a
50 m tape We assumed that each rooted stem
belonged to a separate individual unless stems joined
at the base of the trunk We identified all woody
plants (trees shrubs and lianas) that were at least
25 cm diameter at breast height (dbh) defined as
13 m from the base of the stem Trees and shrubs
above 25 cm dbh were included if the center of the
base of the trunk lay within 1 m of either side of the
tape Lianas (here defined as ground-rooted epiphytes
or creepers) were included if at least one of their roots
entered the soil within the transect and if they had a
diameter of at least 25 cm at any point along the
stem Total dbh for stems that divided above ground
on a single trunk was calculated using the sum of
individual stem diameters
Herbaceous species
All herbaceous plants were identified in 30 (1 9
1 m) quadrats within the six belt transects Five
quadrants were located along the length of each belt
transect at intervals of 10 m (5 15 25 35 and 45 m
respectively) Because of the difficulty in recognizing
individuals of stoloniferous or rhizomatous species
we instead recorded the presence of species in each
quadrat and calculated a relative frequency matrix
(number of occurrences out of 30) for each sampling
unit For upright species presence was determined if
the plant was rooted within the quadrat for species
with a creeping habit an individual was counted if a
shoot fell within the quadrat (Greig-Smith 1983) A
few morphologically similar species were difficult to
identify without flowering or fruiting parts and were
classified as single operational taxonomic units
Specimens of both woody and herbaceous species
were brought to the University of Florida Herbarium
in the Florida Museum of Natural History for
identification and keyed to species with the help of
expert consultation (R Abbott K Perkins and W
Judd) using Wunderlin and Hansen (2003) Thirty-
four tree species and eighty-eight herbaceous species
were identified in total
Remotely sensed data
We used a land cover map produced by the Florida
Fish and Wildlife Conservation Commission (FWC)
(Stys et al 2004) The FWC map contains 43
categories comprising 27 natural classes and 16
disturbed classes We merged classes of interest into
five broad categories xeric uplands (xeric oaksand
pine scrubsandhill) mesic forest (mixed hardwood
pinehardwood hammock and forestpinelands) shrub-
landgrassland (shrub and brushlandgrasslanddry
prairieimproved and unimproved pasture) citrus
and low-impact urban These classes were chosen as
potential propagule sources that might shape succes-
sional communities in regenerating groves All other
land cover classes were removed from the analysis
The accuracy of the classification was checked by the
collection of training points in the field on land cover
patches with a minimum of 100 m2 area Point
locations were recorded with a Garmin 76S GPS
receiver The overall accuracy for the classification
was 824 with a Kappa statistic of 079 (Table 1)
We used ArcView 33 (ESRI 2002) to create
multiple buffers around sampled groves at distances
of 05 1 15 2 4 6 8 and 10 km with buffer radii
being used as a proxy for scale (Holland et al 2004)
This range was selected to reflect the potential
Table 1 Results of the accuracy assessment carried out on the
merged Florida fish and wildlife conservation commission land
cover map
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 100 097 096
Urban 076 067 061
Xeric uplands 076 086 084
Grassland 070 082 080
Mesic forest 091 077 073
Citrus 077 096 095
The overall classification accuracy is 824 with a Kappa
statistic of 079
960 Landscape Ecol (2009) 24957ndash970
123
dispersal distances of both woody and herbaceous
species We extracted the area of each land cover
type within each buffer and created a site by habitat
diversity matrix for each of the eight scales of
analysis
We obtained Landsat 7 Enhanced Thematic Map-
per (ETM) images for path 16row 40 path 17row
40 and path 16row 41 in March 2003 The historic
path 16row 40 and path 17row 40 images were
obtained from February and April 1988 and the path
16row 41 scene from April 1989 Radiometric
calibration was carried out on each image to correct
for sensor related sources of variance variation in
atmospheric conditions and differences in solar angle
at the time of image acquisition (Jensen 1996) The
2003 image was classified by creating training
signatures on the image using homogeneous ground
cover types The training sample data collected to
verify the accuracy of the FWC land cover map were
used to conduct an accuracy assessment for the
supervised classification The overall accuracy for the
classification was 703 (Table 2) The same signa-
ture file was used on the 198889 images The area of
each land cover category within an 8 km radius
around each grove was tabulated to create a site by
land cover area matrix at each time step For each
site the change in area was obtained by subtracting
the area of that class in the 198889 image from its
corresponding area in the 2003 image
Functional trait assessment
To link our analysis more directly to the ecology of
the plants in our study sites we undertook a simple
analysis of plant species composition by functional
trait Data were assembled for each species on
dispersal mode life-span presence of woody stems
vegetative propagation and seed bank persistence
(Tables 3 4)
Data analysis
We used Mantel tests at different scales to establish
whether species composition correlated with the
amount of suitable habitat in the landscape The
herbaceous dataset (HERB) was tabulated as a
frequency matrix of each herbaceous species occur-
rence out of a possible 30 quadrants at a site Data for
the woody species matrix (TREE) were collected as
abundance of individuals per site but to make the
results of the Mantel tests comparable we transformed
these data into a relative abundance matrix
We used Spearmanrsquos ranked correlation coefficient
in the Mantel analysis as recommended by Dietz
(1983) Many species were represented by very few
individuals since these data do not usefully reflect
ecological preference or site suitability (Legendre and
Legendre 1998) we removed species that occurred in
fewer than 5 (B 3) of the 66 sampling sites (see also
McCune and Grace 2002) We calculated dissimilar-
ities for woody and herbaceous species composition
using Bray-Curtis distance (Legendre and Legendre
1998) The analysis included 37 herbaceous species
and 14 woody species
At larger scales of analysis buffer overlap can
influence the independence of samples We used the
program Focus (Holland et al 2004) with 1000
iterations at each scale to select spatially indepen-
dent sites Nonparametric Mantel tests were carried
out on 25 randomly chosen site combinations at each
scale The mean Mantelrsquos r statistic and standard
deviation were plotted against the scale of analysis
and statistical significance was determined using
randomization procedures (Fortin and Dale 2005)
Since we used 8 nested scales (05ndash10 km) a0 was
obtained by dividing 005 by 8 outcomes to give a
Bonferroni-corrected significance criterion for the
Mantel r- statistic of P 000625 (a0 = 0058) The
P-values were plotted against the correlation coeffi-
cients and the proportion of P-values less than
000625 tabulated We defined significant r-statistics
as those for which at least 50 of the P-values were
less than 000625
Table 2 Results of the accuracy assessment carried out on the
supervised classification of the mosaiced 2003 Landsat ETM
images
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 097 1000 1
Urban 076 043 034
Xeric uplands 056 088 085
Grassland 055 050 042
Mesic forest 091 077 073
Citrus 047 078 072
The overall classification accuracy is 7027 and the
corresponding Kappa statistic is 06443
Landscape Ecol (2009) 24957ndash970 961
123
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 3
The lsquolsquooldestrsquorsquo groves were abandoned prior to the
freeze decade in the mid 1960s while the lsquolsquoyoungestrsquorsquo
groves were abandoned in 2003 The year of aban-
donment was determined through informal interviews
with landowners or long-term residents living around
the sites Grove area ranged from 05ndash60 ha with a
single sampling unit placed in each grove at least
1 km from other sampling sites We did not consider
groves smaller than half a hectare because of
concerns about edge effects (Gehlhausen et al
2000 Hansen and Clevenger 2005) An abandoned
grove was considered a discrete unit if it was
surrounded on all sides by different land-use classes
We sampled only in sites where no subsequent
management (eg periodic mowing or grazing) had
been undertaken and we selected areas that had
similar fertilizer and herbicide regimes while in
production Land-use history information was
obtained from interviews with local caretakers and
UFrsquos Institute for Food and Agricultural Sciences
(IFAS) extension agents
We used a variant of the Gentry-Boyle transect
methodology (Boyle 1996) to quantify community
composition for woody and herbaceous species
Different sampling techniques were selected for
woody and herbaceous plants respectively (Bonham
1989) The basic sampling unit consisted of six
parallel 2 9 50 m belt transects running in a North-
South direction with the mid-point of each separated
by 10 m We located grove boundaries and then
chose a central starting point away from any edge
that became the SW corner of the first belt transect
Citrus trees were planted 25 feet (approx 8 m) apart
prior to the freezes in the 80 s and 125 feet (approx
4 m) apart after the freeze decade (R Wigul personal
communication) Transects were thus out of phase
Fig 1 Map of north-central Florida showing the distribution of sampling sites within Hillsborough Lake Marion Pasco Polk and
Putnam counties
Landscape Ecol (2009) 24957ndash970 959
123
with historical lines of citrus trees UTM grid
coordinates for the outer corners of the first and last
belt transect (A-D) were obtained using a hand-held
Garmin GPS 76S Average soil organic matter
content was obtained by taking soil samples with a
hand-held auger to a depth of 8 inches at four
randomly chosen locations at each sampling site
Samples were analysed by Waters Agricultural Labs
Woody species
Woody species were censused by counting all
individuals rooted within 1 m on either side of a
50 m tape We assumed that each rooted stem
belonged to a separate individual unless stems joined
at the base of the trunk We identified all woody
plants (trees shrubs and lianas) that were at least
25 cm diameter at breast height (dbh) defined as
13 m from the base of the stem Trees and shrubs
above 25 cm dbh were included if the center of the
base of the trunk lay within 1 m of either side of the
tape Lianas (here defined as ground-rooted epiphytes
or creepers) were included if at least one of their roots
entered the soil within the transect and if they had a
diameter of at least 25 cm at any point along the
stem Total dbh for stems that divided above ground
on a single trunk was calculated using the sum of
individual stem diameters
Herbaceous species
All herbaceous plants were identified in 30 (1 9
1 m) quadrats within the six belt transects Five
quadrants were located along the length of each belt
transect at intervals of 10 m (5 15 25 35 and 45 m
respectively) Because of the difficulty in recognizing
individuals of stoloniferous or rhizomatous species
we instead recorded the presence of species in each
quadrat and calculated a relative frequency matrix
(number of occurrences out of 30) for each sampling
unit For upright species presence was determined if
the plant was rooted within the quadrat for species
with a creeping habit an individual was counted if a
shoot fell within the quadrat (Greig-Smith 1983) A
few morphologically similar species were difficult to
identify without flowering or fruiting parts and were
classified as single operational taxonomic units
Specimens of both woody and herbaceous species
were brought to the University of Florida Herbarium
in the Florida Museum of Natural History for
identification and keyed to species with the help of
expert consultation (R Abbott K Perkins and W
Judd) using Wunderlin and Hansen (2003) Thirty-
four tree species and eighty-eight herbaceous species
were identified in total
Remotely sensed data
We used a land cover map produced by the Florida
Fish and Wildlife Conservation Commission (FWC)
(Stys et al 2004) The FWC map contains 43
categories comprising 27 natural classes and 16
disturbed classes We merged classes of interest into
five broad categories xeric uplands (xeric oaksand
pine scrubsandhill) mesic forest (mixed hardwood
pinehardwood hammock and forestpinelands) shrub-
landgrassland (shrub and brushlandgrasslanddry
prairieimproved and unimproved pasture) citrus
and low-impact urban These classes were chosen as
potential propagule sources that might shape succes-
sional communities in regenerating groves All other
land cover classes were removed from the analysis
The accuracy of the classification was checked by the
collection of training points in the field on land cover
patches with a minimum of 100 m2 area Point
locations were recorded with a Garmin 76S GPS
receiver The overall accuracy for the classification
was 824 with a Kappa statistic of 079 (Table 1)
We used ArcView 33 (ESRI 2002) to create
multiple buffers around sampled groves at distances
of 05 1 15 2 4 6 8 and 10 km with buffer radii
being used as a proxy for scale (Holland et al 2004)
This range was selected to reflect the potential
Table 1 Results of the accuracy assessment carried out on the
merged Florida fish and wildlife conservation commission land
cover map
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 100 097 096
Urban 076 067 061
Xeric uplands 076 086 084
Grassland 070 082 080
Mesic forest 091 077 073
Citrus 077 096 095
The overall classification accuracy is 824 with a Kappa
statistic of 079
960 Landscape Ecol (2009) 24957ndash970
123
dispersal distances of both woody and herbaceous
species We extracted the area of each land cover
type within each buffer and created a site by habitat
diversity matrix for each of the eight scales of
analysis
We obtained Landsat 7 Enhanced Thematic Map-
per (ETM) images for path 16row 40 path 17row
40 and path 16row 41 in March 2003 The historic
path 16row 40 and path 17row 40 images were
obtained from February and April 1988 and the path
16row 41 scene from April 1989 Radiometric
calibration was carried out on each image to correct
for sensor related sources of variance variation in
atmospheric conditions and differences in solar angle
at the time of image acquisition (Jensen 1996) The
2003 image was classified by creating training
signatures on the image using homogeneous ground
cover types The training sample data collected to
verify the accuracy of the FWC land cover map were
used to conduct an accuracy assessment for the
supervised classification The overall accuracy for the
classification was 703 (Table 2) The same signa-
ture file was used on the 198889 images The area of
each land cover category within an 8 km radius
around each grove was tabulated to create a site by
land cover area matrix at each time step For each
site the change in area was obtained by subtracting
the area of that class in the 198889 image from its
corresponding area in the 2003 image
Functional trait assessment
To link our analysis more directly to the ecology of
the plants in our study sites we undertook a simple
analysis of plant species composition by functional
trait Data were assembled for each species on
dispersal mode life-span presence of woody stems
vegetative propagation and seed bank persistence
(Tables 3 4)
Data analysis
We used Mantel tests at different scales to establish
whether species composition correlated with the
amount of suitable habitat in the landscape The
herbaceous dataset (HERB) was tabulated as a
frequency matrix of each herbaceous species occur-
rence out of a possible 30 quadrants at a site Data for
the woody species matrix (TREE) were collected as
abundance of individuals per site but to make the
results of the Mantel tests comparable we transformed
these data into a relative abundance matrix
We used Spearmanrsquos ranked correlation coefficient
in the Mantel analysis as recommended by Dietz
(1983) Many species were represented by very few
individuals since these data do not usefully reflect
ecological preference or site suitability (Legendre and
Legendre 1998) we removed species that occurred in
fewer than 5 (B 3) of the 66 sampling sites (see also
McCune and Grace 2002) We calculated dissimilar-
ities for woody and herbaceous species composition
using Bray-Curtis distance (Legendre and Legendre
1998) The analysis included 37 herbaceous species
and 14 woody species
At larger scales of analysis buffer overlap can
influence the independence of samples We used the
program Focus (Holland et al 2004) with 1000
iterations at each scale to select spatially indepen-
dent sites Nonparametric Mantel tests were carried
out on 25 randomly chosen site combinations at each
scale The mean Mantelrsquos r statistic and standard
deviation were plotted against the scale of analysis
and statistical significance was determined using
randomization procedures (Fortin and Dale 2005)
Since we used 8 nested scales (05ndash10 km) a0 was
obtained by dividing 005 by 8 outcomes to give a
Bonferroni-corrected significance criterion for the
Mantel r- statistic of P 000625 (a0 = 0058) The
P-values were plotted against the correlation coeffi-
cients and the proportion of P-values less than
000625 tabulated We defined significant r-statistics
as those for which at least 50 of the P-values were
less than 000625
Table 2 Results of the accuracy assessment carried out on the
supervised classification of the mosaiced 2003 Landsat ETM
images
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 097 1000 1
Urban 076 043 034
Xeric uplands 056 088 085
Grassland 055 050 042
Mesic forest 091 077 073
Citrus 047 078 072
The overall classification accuracy is 7027 and the
corresponding Kappa statistic is 06443
Landscape Ecol (2009) 24957ndash970 961
123
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 4
with historical lines of citrus trees UTM grid
coordinates for the outer corners of the first and last
belt transect (A-D) were obtained using a hand-held
Garmin GPS 76S Average soil organic matter
content was obtained by taking soil samples with a
hand-held auger to a depth of 8 inches at four
randomly chosen locations at each sampling site
Samples were analysed by Waters Agricultural Labs
Woody species
Woody species were censused by counting all
individuals rooted within 1 m on either side of a
50 m tape We assumed that each rooted stem
belonged to a separate individual unless stems joined
at the base of the trunk We identified all woody
plants (trees shrubs and lianas) that were at least
25 cm diameter at breast height (dbh) defined as
13 m from the base of the stem Trees and shrubs
above 25 cm dbh were included if the center of the
base of the trunk lay within 1 m of either side of the
tape Lianas (here defined as ground-rooted epiphytes
or creepers) were included if at least one of their roots
entered the soil within the transect and if they had a
diameter of at least 25 cm at any point along the
stem Total dbh for stems that divided above ground
on a single trunk was calculated using the sum of
individual stem diameters
Herbaceous species
All herbaceous plants were identified in 30 (1 9
1 m) quadrats within the six belt transects Five
quadrants were located along the length of each belt
transect at intervals of 10 m (5 15 25 35 and 45 m
respectively) Because of the difficulty in recognizing
individuals of stoloniferous or rhizomatous species
we instead recorded the presence of species in each
quadrat and calculated a relative frequency matrix
(number of occurrences out of 30) for each sampling
unit For upright species presence was determined if
the plant was rooted within the quadrat for species
with a creeping habit an individual was counted if a
shoot fell within the quadrat (Greig-Smith 1983) A
few morphologically similar species were difficult to
identify without flowering or fruiting parts and were
classified as single operational taxonomic units
Specimens of both woody and herbaceous species
were brought to the University of Florida Herbarium
in the Florida Museum of Natural History for
identification and keyed to species with the help of
expert consultation (R Abbott K Perkins and W
Judd) using Wunderlin and Hansen (2003) Thirty-
four tree species and eighty-eight herbaceous species
were identified in total
Remotely sensed data
We used a land cover map produced by the Florida
Fish and Wildlife Conservation Commission (FWC)
(Stys et al 2004) The FWC map contains 43
categories comprising 27 natural classes and 16
disturbed classes We merged classes of interest into
five broad categories xeric uplands (xeric oaksand
pine scrubsandhill) mesic forest (mixed hardwood
pinehardwood hammock and forestpinelands) shrub-
landgrassland (shrub and brushlandgrasslanddry
prairieimproved and unimproved pasture) citrus
and low-impact urban These classes were chosen as
potential propagule sources that might shape succes-
sional communities in regenerating groves All other
land cover classes were removed from the analysis
The accuracy of the classification was checked by the
collection of training points in the field on land cover
patches with a minimum of 100 m2 area Point
locations were recorded with a Garmin 76S GPS
receiver The overall accuracy for the classification
was 824 with a Kappa statistic of 079 (Table 1)
We used ArcView 33 (ESRI 2002) to create
multiple buffers around sampled groves at distances
of 05 1 15 2 4 6 8 and 10 km with buffer radii
being used as a proxy for scale (Holland et al 2004)
This range was selected to reflect the potential
Table 1 Results of the accuracy assessment carried out on the
merged Florida fish and wildlife conservation commission land
cover map
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 100 097 096
Urban 076 067 061
Xeric uplands 076 086 084
Grassland 070 082 080
Mesic forest 091 077 073
Citrus 077 096 095
The overall classification accuracy is 824 with a Kappa
statistic of 079
960 Landscape Ecol (2009) 24957ndash970
123
dispersal distances of both woody and herbaceous
species We extracted the area of each land cover
type within each buffer and created a site by habitat
diversity matrix for each of the eight scales of
analysis
We obtained Landsat 7 Enhanced Thematic Map-
per (ETM) images for path 16row 40 path 17row
40 and path 16row 41 in March 2003 The historic
path 16row 40 and path 17row 40 images were
obtained from February and April 1988 and the path
16row 41 scene from April 1989 Radiometric
calibration was carried out on each image to correct
for sensor related sources of variance variation in
atmospheric conditions and differences in solar angle
at the time of image acquisition (Jensen 1996) The
2003 image was classified by creating training
signatures on the image using homogeneous ground
cover types The training sample data collected to
verify the accuracy of the FWC land cover map were
used to conduct an accuracy assessment for the
supervised classification The overall accuracy for the
classification was 703 (Table 2) The same signa-
ture file was used on the 198889 images The area of
each land cover category within an 8 km radius
around each grove was tabulated to create a site by
land cover area matrix at each time step For each
site the change in area was obtained by subtracting
the area of that class in the 198889 image from its
corresponding area in the 2003 image
Functional trait assessment
To link our analysis more directly to the ecology of
the plants in our study sites we undertook a simple
analysis of plant species composition by functional
trait Data were assembled for each species on
dispersal mode life-span presence of woody stems
vegetative propagation and seed bank persistence
(Tables 3 4)
Data analysis
We used Mantel tests at different scales to establish
whether species composition correlated with the
amount of suitable habitat in the landscape The
herbaceous dataset (HERB) was tabulated as a
frequency matrix of each herbaceous species occur-
rence out of a possible 30 quadrants at a site Data for
the woody species matrix (TREE) were collected as
abundance of individuals per site but to make the
results of the Mantel tests comparable we transformed
these data into a relative abundance matrix
We used Spearmanrsquos ranked correlation coefficient
in the Mantel analysis as recommended by Dietz
(1983) Many species were represented by very few
individuals since these data do not usefully reflect
ecological preference or site suitability (Legendre and
Legendre 1998) we removed species that occurred in
fewer than 5 (B 3) of the 66 sampling sites (see also
McCune and Grace 2002) We calculated dissimilar-
ities for woody and herbaceous species composition
using Bray-Curtis distance (Legendre and Legendre
1998) The analysis included 37 herbaceous species
and 14 woody species
At larger scales of analysis buffer overlap can
influence the independence of samples We used the
program Focus (Holland et al 2004) with 1000
iterations at each scale to select spatially indepen-
dent sites Nonparametric Mantel tests were carried
out on 25 randomly chosen site combinations at each
scale The mean Mantelrsquos r statistic and standard
deviation were plotted against the scale of analysis
and statistical significance was determined using
randomization procedures (Fortin and Dale 2005)
Since we used 8 nested scales (05ndash10 km) a0 was
obtained by dividing 005 by 8 outcomes to give a
Bonferroni-corrected significance criterion for the
Mantel r- statistic of P 000625 (a0 = 0058) The
P-values were plotted against the correlation coeffi-
cients and the proportion of P-values less than
000625 tabulated We defined significant r-statistics
as those for which at least 50 of the P-values were
less than 000625
Table 2 Results of the accuracy assessment carried out on the
supervised classification of the mosaiced 2003 Landsat ETM
images
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 097 1000 1
Urban 076 043 034
Xeric uplands 056 088 085
Grassland 055 050 042
Mesic forest 091 077 073
Citrus 047 078 072
The overall classification accuracy is 7027 and the
corresponding Kappa statistic is 06443
Landscape Ecol (2009) 24957ndash970 961
123
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 5
dispersal distances of both woody and herbaceous
species We extracted the area of each land cover
type within each buffer and created a site by habitat
diversity matrix for each of the eight scales of
analysis
We obtained Landsat 7 Enhanced Thematic Map-
per (ETM) images for path 16row 40 path 17row
40 and path 16row 41 in March 2003 The historic
path 16row 40 and path 17row 40 images were
obtained from February and April 1988 and the path
16row 41 scene from April 1989 Radiometric
calibration was carried out on each image to correct
for sensor related sources of variance variation in
atmospheric conditions and differences in solar angle
at the time of image acquisition (Jensen 1996) The
2003 image was classified by creating training
signatures on the image using homogeneous ground
cover types The training sample data collected to
verify the accuracy of the FWC land cover map were
used to conduct an accuracy assessment for the
supervised classification The overall accuracy for the
classification was 703 (Table 2) The same signa-
ture file was used on the 198889 images The area of
each land cover category within an 8 km radius
around each grove was tabulated to create a site by
land cover area matrix at each time step For each
site the change in area was obtained by subtracting
the area of that class in the 198889 image from its
corresponding area in the 2003 image
Functional trait assessment
To link our analysis more directly to the ecology of
the plants in our study sites we undertook a simple
analysis of plant species composition by functional
trait Data were assembled for each species on
dispersal mode life-span presence of woody stems
vegetative propagation and seed bank persistence
(Tables 3 4)
Data analysis
We used Mantel tests at different scales to establish
whether species composition correlated with the
amount of suitable habitat in the landscape The
herbaceous dataset (HERB) was tabulated as a
frequency matrix of each herbaceous species occur-
rence out of a possible 30 quadrants at a site Data for
the woody species matrix (TREE) were collected as
abundance of individuals per site but to make the
results of the Mantel tests comparable we transformed
these data into a relative abundance matrix
We used Spearmanrsquos ranked correlation coefficient
in the Mantel analysis as recommended by Dietz
(1983) Many species were represented by very few
individuals since these data do not usefully reflect
ecological preference or site suitability (Legendre and
Legendre 1998) we removed species that occurred in
fewer than 5 (B 3) of the 66 sampling sites (see also
McCune and Grace 2002) We calculated dissimilar-
ities for woody and herbaceous species composition
using Bray-Curtis distance (Legendre and Legendre
1998) The analysis included 37 herbaceous species
and 14 woody species
At larger scales of analysis buffer overlap can
influence the independence of samples We used the
program Focus (Holland et al 2004) with 1000
iterations at each scale to select spatially indepen-
dent sites Nonparametric Mantel tests were carried
out on 25 randomly chosen site combinations at each
scale The mean Mantelrsquos r statistic and standard
deviation were plotted against the scale of analysis
and statistical significance was determined using
randomization procedures (Fortin and Dale 2005)
Since we used 8 nested scales (05ndash10 km) a0 was
obtained by dividing 005 by 8 outcomes to give a
Bonferroni-corrected significance criterion for the
Mantel r- statistic of P 000625 (a0 = 0058) The
P-values were plotted against the correlation coeffi-
cients and the proportion of P-values less than
000625 tabulated We defined significant r-statistics
as those for which at least 50 of the P-values were
less than 000625
Table 2 Results of the accuracy assessment carried out on the
supervised classification of the mosaiced 2003 Landsat ETM
images
Producerrsquos
accuracy
Userrsquos
accuracy
Kappa
statistic
Water 097 1000 1
Urban 076 043 034
Xeric uplands 056 088 085
Grassland 055 050 042
Mesic forest 091 077 073
Citrus 047 078 072
The overall classification accuracy is 7027 and the
corresponding Kappa statistic is 06443
Landscape Ecol (2009) 24957ndash970 961
123
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 6
The relative influences of environmental contex-
tual and historical factors on community composition
were explored using Mantel and partial Mantel tests
We assembled matrices describing local environmen-
tal variables (ENV) site history variables (HIST)
habitat isolation (ISO) and landscape change from
the late 1980s to present day (CHANGE) Soils data
on pH and average water holding capacity (ENV)
were obtained from a USDA-NRCS state-wide soil
geographic database (NRCS 1991) and soil organic
matter was measured directly Site elevation was
recorded at the SW corner point using a GPS unit
Climate data (temperature precipitation) were
obtained from DAYMET an interpolated fine-reso-
lution climatological model of the contiguous US
(Thornton 1999) We obtained the variation in
minimum daily air temperature from 1980ndash1997 at
each sampling location with the assumption that
locations with high fluctuations in daily minimum air
temperature had a greater risk of freezing Time since
abandonment and whether a grove was bulldozed
following abandonment were used as local historical
descriptors (HIST) Of the 66 sites 20 were bull-
dozed and 31 were not Information on 15 sites could
not be rigorously verified but available evidence
suggested that they had not been bulldozed and so
they were assigned a value of lsquo0rsquo (ie not bulldozed)
Based on the scaling analysis we used FRAG-
STATS (McGarigal and Marks 1995) to quantify
habitat isolation (ISO) within 8 km buffers Using the
5-category reclassified FWC map we created a
matrix of habitat isolation using the proximity index
Table 3 Functional traits assessed for plants in our study sites
Characteristic Description References
Dispersal mode
Endozoochory A dispersal mode for diaspores with fleshy fruits and
high lipid content
Flora of North America series (2005)
(van der Pijl 1982)
(Holm et al 1977)
R Abbott (2006) personal communication
httpedisifasufleduHS175
wwwefloraorg
httpplantsusdagovindexhtml
httpwwwplantatlasusfedu
httpwwwfloridatacom
Epizoochory Dispersal via hooks barbs or viscid coatings that
attach externally to animal dispersers Wind-
dispersal
Anemochory Diaspores were defined as anemochorous only if
they had specific morphological adaptations for
dispersal such as balloons plumes or wings
Barochory Gravity dispersal
Autochory
(active)
Dispersal by explosive dehiscence
Life span
Annual Species that complete their life-cycle within a year
Perennial Species that that maintain persistent above-ground
structures or store persistent buds atbelow the
soil surface (Raunkier 1934) in (McIntyre et al
1995)
Woody stems Species that maintained a perennating structure and
had high levels of stem lignification
Vegetative
propagation
Stoloniferous or rhizomatous species
Seed bank
persistence
The ability to store viable seeds in the soil through
either physical dormancy (delayed germination
due to an inherent trait of the seed itself) or
enforced dormancy the lsquolsquoarrested growth of non-
dormant seeds due to environmental conditionsrsquorsquo
(Baskin and Baskin 1998)
(Cohen et al 2004 Kalmbacher et al 2004 Looney
and Gibson 1995 Scheiner and Sharitz 1986
Thompson et al 2003) N Bissett (2006) personal
communication)
Duration of fertility Number of months out of the year that a species is
fertile over its entire range
(Wunderlin and Hansen 2003)
Nativeexotic The determination of whether or not a species is
native to Florida
httpwwwplantatlasusfedu
962 Landscape Ecol (2009) 24957ndash970
123
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 7
Ta
ble
4B
reak
do
wn
of
fun
ctio
nal
trai
tsb
yh
erb
aceo
us
pla
nt
spec
ies
for
ou
rst
ud
ysi
tes
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ab
rus
pre
cato
riu
sL
0
00
01
01
01
10
4
Am
ara
nth
us
viri
du
s1
00
00
10
00
10
12
Am
bro
sia
art
emis
iifo
lia
00
01
01
01
01
16
Am
ple
op
sis
arb
ore
a0
00
01
01
01
11
9
Bid
ens
alb
ava
rra
dia
ta0
00
10
10
00
11
12
Ch
am
aes
yce
hir
ta0
10
00
10
00
11
12
Co
mm
elin
ad
iffu
sa0
00
00
11
10
10
9
Co
nyz
aca
na
den
sis
00
10
01
00
01
11
2
Cyn
od
on
da
ctyl
on
10
00
00
10
00
09
Cyp
eru
sg
lob
ulo
sus
10
00
00
11
01
16
Dry
ma
ria
cho
rda
ta0
00
10
11
10
01
12
Em
ilia
son
chif
oli
a0
01
00
10
00
10
12
Eu
pa
tori
um
com
po
siti
foli
um
00
10
00
11
10
16
Fro
elic
hia
flo
rid
an
a0
01
00
10
10
11
6
Het
ero
thec
asu
ba
xill
ari
s0
01
00
10
00
11
12
Hyp
tis
mu
tab
ilis
00
00
00
10
00
01
2
Ind
igo
fera
hir
suta
10
01
01
00
01
01
2
La
nta
na
cam
ara
00
00
10
10
10
01
2
Lep
idiu
mvi
rgin
icu
m1
00
00
10
00
11
3
Mo
mo
rdic
ach
ara
nti
a0
00
01
11
00
00
9
Mo
na
rda
pu
nct
ata
00
00
01
10
10
06
Op
un
tia
hu
mif
usa
00
00
10
10
01
16
Pa
nic
um
ma
xim
um
10
00
01
11
01
09
Pa
rth
eno
ciss
us
qu
inq
uef
oli
a0
00
01
01
01
01
3
Pa
spa
lum
no
tatu
m1
00
00
01
10
10
9
Pa
ssifl
ora
inca
rna
ta0
00
01
01
01
01
12
Ph
ylla
nth
us
ten
ellu
s0
10
00
10
00
10
12
Rh
ynch
elyt
rum
rep
ens
00
10
01
00
00
19
Ric
ha
rdia
sca
bra
10
00
01
01
01
01
2
Ru
bu
sa
rgu
tus
00
00
10
10
10
13
Landscape Ecol (2009) 24957ndash970 963
123
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 8
metric which has been shown to generally outper-
form isolation measures that rely only on dis-
tance (Schumaker 1996) Area-based and distance-
weighted area metrics such as proximity index ignore
small patches with large perimeters that would
otherwise bias predictors of dispersal success
Changes in land-use around groves (CHANGE)
were quantified by comparing the area of change in
land cover types from the late 80 s to the present day
The spatial context matrix (SPCO) used geographic
coordinates from each sampling site to test the effects
of spatial autocorrelation on species distributions in
abandoned groves We used the x and y coordinates
from the SW corner point as representative of the
50 9 50 m plot and trend surface analysis an
extension of multiple regression to fit a cubic
regression model (Borcard et al 1992 Legendre
1990) in CANOCO (ter Braak and Smilauer 2002)
using the species data (HERBTREE) as the response
variable matrix Using a forward selection method
the model picked significant terms and these were
retained for the partial Mantel tests The final spatial
context matrix included the following terms
z frac14 b1xthorn b2ythorn b3x2 thorn b5y2
where x and y are geographic coordinates b1b5 are
coefficients and z is the response variable matrix As
recommended by Legendre and Legendre (1998) the
landscape and environmental matrices (ISO ENV and
CHANGE) was log transformed to approximate nor-
mal distributions and converted into similarity matri-
ces using Gowerrsquos coefficient The Jaccard index a
metric coefficient appropriate for multistate variables
was used for the historical matrix (HIST) while the
spatial context matrix (SPCO) was converted to a
distance matrix using a Euclidean distance measure
Mantel tests were first carried out on the land-
scape environmental historical and spatial context
matrices to determine correlations among predictor
variables The individual effect of each matrix on
woody and herbaceous species composition was then
determined using partial Mantel tests The partial
Mantel test allows for comparison between two
distance matrices while controlling for the effects of
all others (Smouse et al 1986) We corrected levels
of significance using Holmrsquos procedure (Holm 1979)
rather than the Bonferroni correction as the latter is
considered overly conservative in this context
(Legendre and Legendre 1998) The scaling analysisTa
ble
4co
nti
nu
ed
Sp
ecie
sB
aro
cho
rou
sA
uto
cho
rou
sA
nem
och
oro
us
Ep
izo
och
oro
us
En
do
zoo
cho
rou
sA
nn
ual
Per
enn
ial
Clo
nal
Wo
od
yS
eed
ban
k
Nat
ive
Du
rati
on
fert
ilit
y
Ru
mex
ha
sta
tulu
s0
00
10
10
00
11
3
Sid
arh
om
bif
oli
a1
00
00
11
00
11
9
Sm
ila
xa
uri
cula
ta0
00
01
01
11
01
12
Sm
ila
xb
on
a-n
ox
00
00
10
11
10
11
2
So
lid
ag
oto
rtif
oli
a0
01
00
01
10
01
6
Ure
na
lob
ata
00
01
00
10
00
01
2
Vit
isro
tun
dif
oli
a0
00
01
01
01
01
4
lsquolsquo1
rsquorsquoin
dic
ates
mem
ber
ship
ina
cate
go
ry
lsquolsquo0
rsquorsquoin
dic
ates
no
n-m
emb
ersh
ip
Th
ed
esig
nat
ion
of
cate
go
ries
and
the
sou
rces
fro
mw
hic
hth
ese
dat
aw
ere
der
ived
are
giv
enin
Tab
le3
964 Landscape Ecol (2009) 24957ndash970
123
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 9
and computation of Mantel and partial Mantel tests
were carried out using the R package (Development
Core Team 2005)
Lastly variations in plant functional groups based
on time since abandonment were examined by
creating a weighted measure of traits based on
species abundance (Peco et al 2005) We used the
conventional method of multiplying the site by
species matrix by the species by life-history traits
matrix to obtain a matrix of sampling sites by
vegetation traits Since each trait was coded as a
dummy variable weights were directly representative
of species abundances at each site As many life-
history characteristics are known to be correlated
(Weiher et al 1999) we clustered species into two
groups based on the seven life-history traits identi-
fied Sites were grouped in three age classes based on
time since abandonment B10 years 11ndash25 years
[25 years The abundance of representative traits
was summed across sites and separately plotted as a
function of these age classes Differences in numbers
of sites in each age class were corrected for by
plotting the relative proportions of each functional
group
Results
The Mantel tests indicate that there was a general
increase in the correlation between landcover vari-
ables and herbaceous vegetation composition with
increasing scale The highest correlation for this
stratum (r = 045 n = 14 proportion of significant
P-values = 072) occurred at 8 km (Table 5 Fig 2)
Mantel r-statistics were significant only at 8 and
10 km (Fig 3) Since there were a greater proportion
of significant P-values with an 8 km buffer radius
this was the scale at which landscape variables were
extracted for use in further analyses The community
composition of woody species showed no significant
correlations with land cover variables at any scale
and a general decreasing trend in correlation coeffi-
cients as scale increased
As predicted the correlations between predictor
variables revealed a significant relationship between
spatial context and all other matrices (Table 6) There
was also an association between local management
history and habitat isolation The results of the
Mantel tests show that influences of local landscape
and spatial variables differ according to vegetation
strata There were significant correlations between
the herbaceous species matrix and spatial context
(r = 010 P = 0007 n = 66) habitat isolation (r =
013 P = 0016 n = 66) environmental (r = 013
P = 0026 n = 66) and historical variables (r =
039 P = 0001 n = 66) Partial Mantel tests
revealed that each matrix provided a unique contri-
bution in explaining some of the variation in the
herbaceous dataset even when controlling for all
other variables The only significant determinant of
woody species distribution was local historical vari-
ables (r = 037 P = 0001 n = 66) and this corre-
lation remained robust even when removing the
effects of all other predictor matrices
Our cluster analysis of functional traits produced
two emergent groups that were distinguished based
Table 5 Proportion of significant P-values (corrected P 0006) for each run of the Mantel test
Scale (km) Number of
buffers
Mean
r-statistic
Proportion
significant
P-values
05 20 018 016
1 20 021 024
15 20 025 036
2 20 028 044
4 20 025 036
6 16 030 032
8 14 045 072
10 12 045 052
0
005
01
015
02
025
03
035
04
045
05
0 2 4 6 8 10
scale
r-st
atis
tic
herbaceous
woody
Fig 2 Results of Mantel tests at different scales showing the
differential response of woody vs herbaceous vegetation to
landscape variables Filled squares woody vegetation cleardiamonds herbaceous vegetation error bars standard devia-
tion The mechanisms underlying the mirror-image appearance
of these curves are discussed in the text and in Fig 4
Landscape Ecol (2009) 24957ndash970 965
123
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 10
on dispersal mode persistence and stem woodiness
Species in Cluster 1 were woody perennials mainly
dispersed via endozoochory Species in Cluster 2 were
epizoochorous autochorous barochorous or anem-
ochorous Plants in this group stored seeds in the seed
bank and were largely non-woody annuals There were
approximately equal proportions of native and exotic
species in each group The effects of time since
abandonment on functional aspects of community
composition are obvious from the relative abundance
of life-history traits at different sites with woody
plants increasing and annuals decreasing (Fig 4)
Comparisons of dispersal modes with age
showed that the oldest sites were dominated by
endozoochorous species and that this proportion was
significantly lower than in sites abandoned10 years
(Z = -0803 P = 0005) or between 11ndash25 years
(Z = -219 P = 0027) Conversely less than 10
of species in sites [25 years old were comprised of
epizoochorous and anemochorous propagules There
were approximately equal proportions of epizooch-
orous propagules in the two younger age classes and
anemochorous species did not differ significantly
between these two age classes Contrary to expecta-
tions the highest proportion of autochorous species
occurred in sites abandoned for less than 10 years
Based on the distribution of life-span with age there
was a significantly higher proportion of perennials in
00 01 02 03 04
00
02
04
06 05km
Correlation coefficient
p v
alu
e
03 04 05 06 07
000
002
004
006
10km
Correlation coefficient
p v
alu
e
03 04 05 06
000
002
004
8km
Correlation coefficient
p v
alu
e
015 025 035 045
000
005
010
015
6km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
4km
Correlation coefficient
p v
alu
e
005 015 025 035
000
010
020
2km
Correlation coefficient
p v
alu
e
01 02 03 04
00
01
02
03
04
1km
Correlation coefficient
p v
alu
e
00 01 02 03 04
00
02
04
06
15km
Correlation coefficient
p v
alu
e
Fig 3 Scatterplots of Mantel correlation coefficients against P-values at 8 scales (05ndash10 km) for each of the 25 randomly chosen
subset combinations Points below the horizontal dashed line are statistically significant at P = 000625
Table 6 Results of Mantel and partial Mantelrsquos tests showing
Spearmanrsquos correlation coefficients between matrices for
spatial context (SPCO) landscape isolation (ISO) land-used
history (HIST) local environment (ENV) landscape change
(CHANGE) and relative abundances of woody (TREE) and
herbaceous (HERB) species (ALL) indicates that coefficients
were calculated using partial correlations to remove the effects
of all other matrices
SPCO ENV HIST CHANGE ISO SPCOALL ENVALL HISTALL CHANGEALL ISOALL
SPCO 0293 0275 0281 0356 0256 0199 0298 0306
ENV 0077 0057 0136 -0011 -0023 0034
HIST 0036 0252 -0025 0168
CHANGE -0008 -0114
HERB 0301 0199 0456 0099 0269 0101 0133 0393 0056 0127
TREE 0064 0004 0354 0030 -0048 0008 -0009 0370 0013 -0145
P 005
966 Landscape Ecol (2009) 24957ndash970
123
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 11
the oldest compared to the youngest sites (Z =
-2469 P = 0011) while the youngest sites were
dominated by annuals
Discussion
We considered the relative roles of local regional and
historical variation as drivers of community compo-
sition in abandoned citrus groves in northern Florida
We found that over the time scales of the analysis the
effects of past land-use and the time since disturbance
were the primary correlates of vegetation composi-
tion and structure Species distributions in abandoned
groves of different ages could at least in part be
attributed to their life-history traits Younger sites
were dominated by weedy annuals which are long-
distance wind dispersed and have a large seed bank
(eg Holt et al 1995 Jacquemyn et al 2001) As
time since abandonment increases herbaceous peren-
nials and less mobile species are able to colonize
Previous studies have found that birds are the
predominant dispersers of pioneer woody species
(Duncan and Chapman 1999) especially for succes-
sional sites in which remnant trees remain after
cultivation Citrus trees left standing in abandoned
groves provide structural complexity which attracts
birds and consequently increases the input of verte-
brate-dispersed species to a site Consequently the
decision about whether to abandon a grove with trees
left standing or trees removed has important conse-
quences for the later successional trajectory of the
site
A number of local mechanisms could potentially
contribute to the patterns that we observed Land
clearing following abandonment alters a range of
environmental and biotic conditions that can signif-
icantly affect understory plants (Aide et al 1995
Motzkin et al 1996) Likely processes influencing
community dynamics following land abandonment
include exposure of weedy seeds from the seed bank
following bulldozing exclusion of woody seedlings
through competition (Berkowitz et al 1995 Davis
et al 1998) allelopathic effects of leaf litter (Mol-
ofsky and Augspurger 1992) competition for space
(Sydes and Grime 1981) and the impacts of
decreases in rainfall and light in the understory (Ko
and Reich 1993)
It is important to note that plant communities in
abandoned citrus groves are not necessarily returning
to their pre-agricultural states Citrus groves were
historically planted on sand pine scrub and fire-
adapted sandhill habitat Sand pine scrub is domi-
nated by an overstory of sand pine (Pinus clausa)
with an understory of scrub oaks and little herbaceous
cover (Myers 1985) Sandhill communities are char-
acterized by a ground cover of wire-grass (Aristida
stricta) with scattered longleaf (P palustris) or
Florida slash pine (P elliotti var densa) as well as
turkey oak (Quercus laevis) and scrub hickory (Carya
floridana) in the overstory (Menges et al 1993
Myers and White 1987) The alteration of natural fire
regimes through long-term agricultural use and
subsequent abandonment allows the invasion of
propagules that would otherwise be unable to persist
with periodic burning Evergreen oaks such as
Quercus virginiana Q geminata and Q laurifolia
were commonly found in abandoned groves but
these species do not typically make up the flora of
sandhill habitat Hardwoods can alter community-
level fire dynamics by making the community less
prone to burning (Myers 1985) Our data do not on
their own make a rigorous case for the idea that
different sites may be locked in to alternative stable
states but the potential for alternative states exists
and would be interesting to explore further
0
01
02
03
04
05
06
07
08
09
oldest (1964-1975) mid (1980-1990) youngest (1991-2003)
Age class
prop
ortio
n
Fig 4 Breakdown by time since abandonment of the propor-
tions of species in different functional trait clusters Cluster 1
(grey) consists primarily of woody perennials that are mainly
dispersed via endozoochory Cluster 2 (stippled) consists
primarily of non-woody annuals that store seeds in the seed
bank Bars indicate standard deviation across sites in the
analysis This figure shows a clear successional shift from
weedy to woody species through time
Landscape Ecol (2009) 24957ndash970 967
123
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 12
Habitat isolation was correlated with species
occurrence in the herbaceous stratum If the matrix
between patches were homogeneous and formed a
binary landscape of suitable vs unsuitable habitat
(Ricketts 2001) patch accessibility would become a
function of habitat configuration The movement of
dispersers across the matrix between habitat patches
is however influenced by the dispersal ability of
species as well as the spatial arrangement of land
cover types The tree species that were analyzed are
dispersed via potentially long-distance bird mammal
and wind-dispersal primary vectors highly vagile
species are less likely to be influenced by immediate
landscape context (eg Aviron et al 2005) Herba-
ceous vegetation by contrast had a higher variation
in known dispersal modes and distribution patterns
appeared to be a function of underpitch distance as
well as dispersal capability
Ecological processes are affected by their location
in space and their interactions with neighbouring
units (Levin 1992) Patterns in human land-use have
also been shown to occur non-randomly in space in
response to socioeconomic and environmental drivers
(Brooks et al 2002 Overmars et al 2003) Spatial
dependence in our measured local and landscape
variables is indicated by significant correlations
between the spatial context matrix (SPCO) and each
predictor matrix once potentially confounding fac-
tors are controlled for (Table 3)
Our results indicate that path dependency (ie the
influence of historical land use and perturbations) is
the dominant factor driving plant community com-
position in regenerating citrus groves Herbaceous
communities in our sites responded more strongly to
both local and landscape variables than woody
species and species composition related loosely to
dispersal modes While land abandonment on its own
does not assure recovery to pre-agricultural condi-
tions structural complexity and species richness in
abandoned agricultural areas increased with age and
the identity of species in post-agricultural sites often
differs from unimpacted sites Ultimately an under-
standing of the long-term effects of both past land-use
and spatial influences on current community compo-
sition will be critical for predicting and managing the
recovery of post-agricultural landscapes
Acknowledgments We are grateful to the many people who
assisted in different ways with this research particularly Jane
South worth and Doria Gordon for advice on various technical
issues and Richard Abbott Kent Perkins and Walter Judd for
assistance with identifying plants Valuable help with finding
and accessing abandoned groves was provided by the
University of Floridarsquos Institute of Food and Agricultural
Sciences (IFAS) extension services We also thank the many
farmers who discussed prior land use with us This research
was supported by the University of Florida and a T-STAR
research grant to GSC
References
Aide TM Zimmerman JK Herrera L Rosario M (1995) Forest
recovery in abandoned tropical pastures along in Puerto
Rico For Ecol Manage 7777ndash86 doi1010160378-1127
(95)03576-V
Aviron S Burel F Baudry J Schermann N (2005) Carabid
assemblages in agricultural landscapes impacts of habitat
features landscape context at different spatial scales and
farming intensity Agric Ecosyst Environ 108205ndash217
doi101016jagee200502004
Baskin CC Baskin JM (1998) Seeds ecology biogeography
and evolution of dormancy and germination Academic
Press San Deigo
Berkowitz AR Canham CD Kelly VR (1995) Competition vs
facilitation of tree seedling growth and survival in early suc-
cessional communities Ecology 761156ndash1168 doi102307
1940923
Bonham CD (1989) Measurements for terrestrial vegetation
Wiley New York
Borcard D Legendre L Drapeau P (1992) Partialling out the
spatial component of ecological variation Ecology
731045ndash1055 doi1023071940179
Boyle BL (1996) Changes on altitudinal and latitudinal gra-
dients in neotropical montane forests Washington Uni-
versity St Louis p 275
Brooks JS Bliss JC Spies TA (2002) Land ownership and
landscape structure a spatial analysis of sixty-six Oregon
(USA) Coast Range watersheds Landscape Ecol 17685ndash
697 doi101023A1022977614403
Cohen S Braham R Sanchez F (2004) Seed bank viability in
disturbed longleaf pine sites Restor Ecol 12503ndash515
doi101111j1061-2971200400382x
Connell JH Slayter RO (1977) Mechanisms of succession in
natural communities and their role in community stability
and organization Am Nat 1111119ndash1144 doi101086
283241
Davis MA Wrage KJ Reich PB (1998) Competition between
tree seedlings and herbaceous vegetation support for a
theory of resource supply and demand J Ecol 86652ndash
661 doi101046j1365-2745199800087x
de Blois S Domon G Bouchard A (2001) Environmental
historical and contextual determinants of vegetation
cover a landscape perspective Landscape Ecol 16421ndash
436 doi101023A1017548003345
De Steven D (1991) Experiments in mechanisms of tree
establishment in old-field succession seedling emergence
Ecology 721066ndash1075 doi1023071940606
968 Landscape Ecol (2009) 24957ndash970
123
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 13
Development Core Team R (2005) R A language and envi-
ronment for statistical computing 221 edn Foundation
for Statistical Computing Vienna Austria
Dietz EJ (1983) Permutation tests for association between two
distance matrices Syst Zool 3221ndash26 doi102307
2413216
Duncan RS Chapman CA (1999) Seed dispersal and potential
forest succession in abandoned agriculture in tropical
Africa Ecol Appl 9998ndash1008 doi1018901051-
0761(1999)009[0998SDAPFS]20CO2
Dzwonko Z Loster S (1992) Species richness and seed dis-
persal to secondary woods in Southern Poland J Biogeogr
19195ndash204 doi1023072845505
Egler FE (1954) Vegetation science concepts 1 initial floristic
composition a factor in old-field vegetation development
Plant Ecol 4412ndash417 doi101007BF00275587
Ewel JJ (1990) Introduction In Myers RL Ewel JJ (eds)
Ecosystems of Florida University Press of Florida
Gainesville
Flinn KM Vellend M (2005) Recovery of forest plant com-
munities in post-agricultural landscapes Front Ecol
Environ 3243ndash250
Fortin MJ Dale M (2005) Spatial analysis a guide for ecolo-
gists Cambridge University Press Cambridge
Gehlhausen SM Schwartz MW Augspurger CK (2000) Veg-
etation and microclimate edge effects in two mixed-
mesophytic forest fragments Plant Ecol 14721ndash35
doi101023A1009846507652
Gottwald TR Hughes G Graham JH Sun X Riley T (2001) The
citrus canker epidemic in Florida the scientific basis of
regulatory eradication policy for an invasive species Phy-
topathology 9130ndash34 doi101094PHYTO200191130
Greig-Smith P (1983) Quantitative plant ecology 3rd edn
University of California Press Berkeley
Grime JP (1973) Competitive exclusion in herbaceous vege-
tation Nature 242344ndash347 doi101038242344a0
Grime JP (1979) Plant strategies and vegetation processes
Wiley New York
Gustafson EJ Gardner RH (1996) The effect of landscape
heterogeneity on the probability of patch colonization
Ecology 7794ndash107 doi1023072265659
Hansen MJ Clevenger AP (2005) The influence of disturbance
and habitat on the presence of non-native plant species
along transport corridors Biol Conserv 125249ndash259
doi101016jbiocon200503024
Holland JD Bert DG Fahrig L (2004) Determining the spatial
scale of speciesrsquo response to habitat Bioscience 54
227ndash233 doi1016410006-3568(2004)054[0227DTSSOS]
20CO2
Holm S (1979) A simple sequentially rejective multiple test
procedure Scand J Stat 665ndash70
Holm LG Plunkett DL Panche JV Herberger JP (1977) The
worldrsquos worst weeds distribution and biology The Uni-
versity Press of Hawaii Honolulu
Holt RD Robinson GR Gaines MS (1995) Vegetation
dynamics in an experimentally fragmented landscape
Ecology 761610ndash1624 doi1023071938162
Jacquemyn H Butaye J Dumortier M Hermy M Lust N
(2001) Effects of age and distance on the composition of
mixed deciduous forest fragments in an agricultural
landscape J Veg Sci 12635ndash642 doi1023073236903
Jensen JR (1996) Introductory digital image processing A
remote sensing perspective 2nd edn Prentice-Hall Upper
Saddle River NJ
Kalmbacher R Cellinese N Martin F (2004) Seeds obtained by
vacuuming the soil surface after fire Native Plants J
6233ndash241
Ko LJ Reich PB (1993) Oak tree effects on soil and herba-
ceous vegetation in savannas and pastures in Wisconsin
Am Midl Nat 13031ndash42 doi1023072426272
Legendre L (1990) Quantitative methods and biogeographic
analysis In Garbary DJ South RR (eds) Evolutionary
biogeography of the marine algae of the North Atlantic
vol G 22 Springer Berlin Germany pp 9ndash34
Legendre P Legendre L (1998) Numerical ecology 2nd edn
Elsevier Amsterdam
Levin S (1992) The problem of pattern and scale in ecology
the Robert H MacArthur Award lecture Ecology
731943ndash1967 doi1023071941447
Looney PB Gibson DJ (1995) The relationship between the
soil seed bank and above-ground vegetation of a coastal
barrier island J Veg Sci 6825ndash836 doi1023073236396
MacArthur RH Wilson EO (1967) The theory of island bio-
geography Princeton University Press Princeton NJ
McCune B Grace JB (2002) Analysis of ecological commu-
nities MjM Software Design Oregon
McGarigal K Marks BJ (1995) FRAGSTATS spatial pattern
analysis program for quantifying landscape structure
USDA Forest Service General
McIntyre S Lavorel S Tremont RM (1995) Plant life-history
attributes their relationship to disturbance response in
herbaceous vegetation J Ecol 8331ndash44 doi102307
2261148
Menges ES Abrahamson WG Givens KT Gallo NP Layne
JN (1993) Twenty years of vegetation change in five long-
unburned Florida plant communities J Veg Sci 4375ndash
386 doi1023073235596
Miller KA (1991) Response if Florida citrus growers to the
freezes of the 1980s Clim Res 1133ndash144 doi103354
cr001133
Molofsky J Augspurger CK (1992) The effect of leaf litter on
early seedling establishment in a tropical forest Ecology
7368ndash77 doi1023071938721
Motzkin G Foster D Allen A (1996) Controlling site to
evaluate history Vegetation patterns of a New England
sand plain Ecol Monogr 66345ndash365
Mou P Jones RH Guo D Lister A (2005) Regeneration
strategies disturbance and plant interactions as organizers
of vegetation spatial patterns in a pine forest Landscape
Ecol 20971ndash987 doi101007s10980-005-7007-0
Myers RL (1985) Fire and the dynamic relationship between
Florida sandhill and sand pine scrub vegetation Bull
Torrey Bot Club 112241ndash252 doi1023072996539
Myers RL White DL (1987) Landscape history and changes in
sandhill vegetation in North-Central and South-Central
Florida Bull Torrey Bot Club 11421ndash32
NRCS (1991) State Soil Geographic (STATSGO) Database for
the state of Florida In Natural Resources Conservation
Service United States Department of Agriculture
Overmars KP de Koning GHJ Veldkamp A (2003) Spatial
autocorrelation in multi-scale land-use models Ecol
Modell 164257ndash270 doi101016S0304-3800(03)00070-X
Landscape Ecol (2009) 24957ndash970 969
123
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123
Page 14
Peco B de Pablos I Traba J Levassor C (2005) The effect of
grazing abandonment on species composition and func-
tional traits the case of dehesa grasslands Basic Appl
Ecol 6175ndash183 doi101016jbaae200501002
Raunkier C (1934) The life form of plants and statistical plant
geography Oxford University Press Oxford United
Kingdom
Ricketts TS (2001) The matrix matters effective isolation in
fragmented landscapes Am Nat 15887ndash99 doi101086
320863
Scheiner RL Sharitz RR (1986) Seed bank dynamics in a
southeastern riverine swamp Am J Bot 731022ndash1030
doi1023072444121
Schumaker NH (1996) Using landscape indices to predict
habitat connectivity Ecology 771210ndash1225 doi102307
2265590
Smouse PE Long JC Sokal RR (1986) Multiple-regression
and correlation extensions of the Mantel test of Matrix
correspondence Syst Zool 35627ndash632 doi102307
2413122
Stys B et al (2004) Florida vegetation and land cover data
derived from 2003 Landsat ETM imagery Florida Fish
and Wildlife Conservation Commission Tallahasee FL
Sydes C Grime P (1981) Effects of tree leaf litter on herba-
ceous vegetation in deciduous wodland II An experi-
mental investigation J Ecol 69249ndash262 doi102307
2259829
ter Braak CJF Smilauer P (2002) CANOCO reference manual
and Cano Draw for Windows userrsquos guide software for
canonical community ordination 45th edn Microcom-
puter Power Ithaca NY
Thompson K Ceriani RM Bakker J Bekker R (2003) Are seed
dormancy and persistence in the soil related Seed Sci Res
1397ndash100 doi101079SSR2003128
Thornton P (1999) DAYMET US Data Center for Daily
Surface Weather Data and Climatological Summaries In
University of Montana Numerical Terradynamic Simu-
lation Group
Tilman D (1994) Competition and biodiversity in spatially
structured habitats Ecology 752ndash16 doi1023071939377
van der Pijl L (1982) Principles of dispersal in higher plants
3rd edn Springer Berlin
Verheyen K Hermy M (2001) The relative importance of
dispersal limitation of vascular plants in secondary forest
succession in Muizen Forest Belgium J Ecol 89829ndash
840 doi101046j0022-0477200100596x
Weiher E van der Werf A Thompson K Roderick M Garnier
E Eriksson O (1999) Challenging Theophrastus a com-
mon core list of plant traits for functional ecology J Veg
Sci 10609ndash620 doi1023073237076
Weischet W Caviedes CN (1987) Citrus in Florida Erdkunde
41210ndash226 doi103112erdkunde19870304
Wunderlin RP Hansen BF (2003) Guide to the vascular plants
of Florida 2nd edn University Press of Florida Gainesville
970 Landscape Ecol (2009) 24957ndash970
123