Evaluating management interventions in small populations of a perennial herb Primula vulgaris using spatio-temporal analyses of point patterns Hans Jacquemyn 1 *, Patrick Endels 2 , Olivier Honnay 1 and Thorsten Wiegand 3 1 Division of Plant Ecology and Systematics, University of Leuven, Arenbergpark 31, B-3001 Heverlee, Belgium; 2 Division of Forest, Nature and Landscape Research, University of Leuven, Celestijnenlaan 200E, 3001 Leuven, Belgium; and 3 UFZ Helmholtz Centre for Environmental Research – UFZ, Department of Ecological Modeling, PF 500136, D-04301 Leipzig, Germany Summary 1. In high-intensity agricultural landscapes, small landscape elements such as hedgerows, ditch banks, and rows of pollard trees may represent the last refuge of many plant and animal species, some of them being rare or even threatened with extinction. However, due to their small size and low habitat quality, long-term population survival cannot be ascertained and often active manage- ment is needed to maintain viable populations of species forced to survive in these small landscape elements. 2. Population models are needed to assess the threats to species at risk and to evaluate alternative management actions. Here, we present a methodology to evaluate management interventions using spatio-temporal analyses of point patterns. We apply this method to several populations of prim- rose Primula vulgaris in Flanders, where it is rare and predominantly occurs along ditch banks. 3. The effects of ditch bank clearing on the establishment success of seedlings was investigated by comparing spatial patterns of seedling recruitment, survival and mortality between populations that were grazed and populations that were severely disturbed by mechanical clearing of ditch banks followed by annual mowing. A total of 884 seedlings were mapped and monitored during 4 consecutive years (1999–2002). 4. In all populations, plants showed significant clustering, but in cleared sites only seedlings were significantly clustered around adults. Spatial patterns of mortality varied according to the manage- ment intervention. In grazed sites, mortality was almost random, whereas in cleared sites we found clear evidence for strong negative density-dependent mortality. There was no evidence that the pres- ence of adults affected survival of recruits in any of the sites studied. 5. Synthesis and applications. This study shows that the analysis of spatial point patterns contributes to our understanding of the population dynamics of plant species occurring in different environments. The approach can be broadly applied to other plant species to elucidate the processes that determine the number of individuals that establish and persist into later life stages and will help conservation managers to refine management strategies intended to conserve or restore plant populations. In the case of P. vulgaris, increasing the availability of microsites is most likely to result in increased growth rates, as it results in increased recruitment and establishment of recruits. Key-words: O-ring statistics, pair correlation function, point pattern analysis, Primula, random mortality hypothesis, recruitment Introduction In most of north-western Europe, the intensification of agricul- ture practices has forced many formerly widespread plant and animal species to survive in semi-natural habitat remnants, such as hedgerows, ditch banks, and rows of pollard trees (Sto- ate et al. 2001; Robinson & Sutherland 2002). However, the long-term survival of plant populations along small-landscape elements cannot be guaranteed because of their small area and low habitat quality (Kleijn & Verbeek 2000; Blomqvist et al. 2003). Specific management interventions are often needed to *Correspondence author. E-mail: [email protected]Journal of Applied Ecology 2010, 47, 431–440 doi: 10.1111/j.1365-2664.2010.01778.x Ó 2010 The Authors. Journal compilation Ó 2010 British Ecological Society
10
Embed
Evaluating management interventions in small populations of a perennial herb Primula vulgaris using spatio-temporal analyses of point patterns
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Evaluating management interventions in small
populations of a perennial herb Primula vulgaris using
spatio-temporal analyses of point patterns
Hans Jacquemyn1*, Patrick Endels2, Olivier Honnay1 and Thorsten Wiegand3
1Division of Plant Ecology and Systematics, University of Leuven, Arenbergpark 31, B-3001 Heverlee, Belgium;2Division of Forest, Nature and Landscape Research, University of Leuven, Celestijnenlaan 200E, 3001 Leuven,
Belgium; and 3UFZ Helmholtz Centre for Environmental Research – UFZ, Department of Ecological Modeling,
PF 500136, D-04301 Leipzig, Germany
Summary
1. In high-intensity agricultural landscapes, small landscape elements such as hedgerows, ditch
banks, and rows of pollard trees may represent the last refuge of many plant and animal species,
some of them being rare or even threatened with extinction. However, due to their small size and
low habitat quality, long-term population survival cannot be ascertained and often active manage-
ment is needed to maintain viable populations of species forced to survive in these small landscape
elements.
2. Population models are needed to assess the threats to species at risk and to evaluate alternative
management actions. Here, we present a methodology to evaluate management interventions using
spatio-temporal analyses of point patterns. We apply this method to several populations of prim-
rosePrimula vulgaris in Flanders, where it is rare and predominantly occurs along ditch banks.
3. The effects of ditch bank clearing on the establishment success of seedlings was investigated by
comparing spatial patterns of seedling recruitment, survival andmortality between populations that
were grazed and populations that were severely disturbed by mechanical clearing of ditch banks
followed by annual mowing. A total of 884 seedlings were mapped and monitored during 4
consecutive years (1999–2002).
4. In all populations, plants showed significant clustering, but in cleared sites only seedlings were
significantly clustered around adults. Spatial patterns of mortality varied according to the manage-
ment intervention. In grazed sites, mortality was almost random, whereas in cleared sites we found
clear evidence for strong negative density-dependent mortality. There was no evidence that the pres-
ence of adults affected survival of recruits in any of the sites studied.
5. Synthesis and applications. This study shows that the analysis of spatial point patterns
contributes to our understanding of the population dynamics of plant species occurring in different
environments. The approach can be broadly applied to other plant species to elucidate the processes
that determine the number of individuals that establish and persist into later life stages and will help
conservation managers to refine management strategies intended to conserve or restore plant
populations. In the case ofP. vulgaris, increasing the availability of microsites is most likely to result
in increased growth rates, as it results in increased recruitment and establishment of recruits.
Key-words: O-ring statistics, pair correlation function, point pattern analysis, Primula,
random mortality hypothesis, recruitment
Introduction
Inmost of north-western Europe, the intensification of agricul-
ture practices has forced many formerly widespread plant and
animal species to survive in semi-natural habitat remnants,
such as hedgerows, ditch banks, and rows of pollard trees (Sto-
ate et al. 2001; Robinson & Sutherland 2002). However, the
long-term survival of plant populations along small-landscape
elements cannot be guaranteed because of their small area and
low habitat quality (Kleijn & Verbeek 2000; Blomqvist et al.
2003). Specific management interventions are often needed to*Correspondence author. E-mail: [email protected]
Journal of Applied Ecology 2010, 47, 431–440 doi: 10.1111/j.1365-2664.2010.01778.x
� 2010 The Authors. Journal compilation � 2010 British Ecological Society
maintain viable plant populations in agricultural landscapes.
One way of evaluating changes in population viability after
intervention is to usematrix populationmodels (e.g.Brys et al.
2004; Pfeifer et al. 2006). These approaches, however, provide
little insight in the processes that determine why particular life
stages establish and survive into the next life stage and others
do not.
There are two contrasting explanations for the distribution
and abundance of individuals within a population (reviewed in
Turnbull, Crawley & Rees 2000; Clark et al. 2007; Poulsen
et al. 2007). The first states that the distribution of juveniles
and adults is primarily determined by the number of seeds that
are dispersed into a population. The alternative viewpoint is
that population size and the distribution of juveniles and adults
are primarily determined by the quality of suitable sites for
recruitment and the effect of density-dependent survival within
these sites. Recent meta-analyses of seed addition experiments
(Clark et al. 2007; Poulsen et al. 2007) have shown that supple-
mental addition of seeds to populations almost always results
in increased seedling recruitment, indicating that most plant
populations are seed limited. However, the effects of seed addi-
tion were relatively small and most seeds failed to recruit into
the seedling stage, suggesting that establishment limitation,
rather than seed limitation, is the most important factor
limiting the distribution and abundance of individuals (Clark
et al. 2007).
One way of investigating the relative importance of dispersal
and establishment processes in determining the distribution
and abundance in plant populations is the study of spatial
patterns in recruitment, survival andmortality, especially when
temporal changes in these patterns are analysed (e.g. Turnbull
et al. 2004).Whereas spatial patterns of seed arrival depend on
the mechanisms of seed dispersal, seed recruitment depends on
the interaction of multiple biotic and abiotic factors. Biotic
factors relate to seed competition, predation or pathogens,
whereas abiotic factors refer to gaps, resources or microsites
(Clark et al. 1999). Competitive interactions can be further
partitioned according whether they cause density-dependent
or density-independent mortality between seed arrival and
recruitment into a later life stage (Clark, Macklin & Wood
1998). Intra-specific competition can be intense, particularly
when seeds are strongly aggregated at small spatial scales
(Rees, Grubb & Kelly 1996; Turnbull et al. 2004). While
several mechanisms could generate this pattern, local seed
dispersal resulting in seedlings being clustered around adults or
seedlings clumped together at very high local densities, leading
to intra-specific aggregation is most likely for species with
limited dispersal capacities. This implies that patterns of seed
dispersal and the subsequent spatial arrangement of seedlings
within a population may largely determine patterns of growth
and mortality and thus may have important implications for
populationmanagement.
Few studies have investigated processes that determine the
number of individuals that establish and persist into later life
stages (but see Kenkel 1988; De la Cruz et al. 2008). In this
paper, patterns of seedling establishment were investigated in
several populations of the perennial Primula vulgaris by
comparing seedling recruitment, mortality and survival into
later life stages. In Flanders, the species is rare and occurs pre-
dominantly in small landscape elements that are subjected to
different disturbance regimes. Previous demographic research
using matrix models already showed that seed germination,
seedling establishment and survival were the most important
factors determining variation in growth rates (Endels et al.
2007). By examining changes in spatial patterns in natural pop-
ulations along ditch banks we aimed to investigate to what
extent density-dependent or density-independent mortality
between seed arrival and recruitment into a later life stage
determined seedling establishment success. The results of
this study may therefore not only contribute to a better
understanding of the driving forces (dispersal, competition and
management) that shape spatial patterns in plant populations,
but also allow conservation strategies to be refined. The
methodology should be widely applicable as a valuable
addition tomatrix populationmodelling.
Materials and methods
SPECIES
Primula vulgaris Huds. is a small, long-lived, herbaceous, diploid
perennial with a North Atlantic and Mediterranean distribution
(Jacquemyn et al. 2009). In Flanders, Belgium, it is considered a very
rare and declining species: only 89 populations, ranging in size from 1
to 1219 individuals, were found in 1999.Most populations (58 in total
or 68%) contained fewer than 50 individuals (Endels et al. 2002).
AlthoughP. vulgaris is regarded as a typical woodland species (Whale
1984) associated with newly opened gaps (Valverde & Silvertown
1995), most populations (85%) in Flanders occur on ditch banks
along hedgerows and forest edges (Endels et al. 2002). Plants flower
in early spring (March–May) and are mainly pollinated byHymenop-
tera (mostly bumblebees) andDiptera (Jacquemyn et al. 2009). Fruits
ripen around the middle of June and have an elaiosome, which is
attractive to ants and rodents who may actively harvest seeds and
disperse them across the landscape (Valverde & Silvertown 1995).
However, seeds usually fall directly to the ground in the immediate
vicinity of the mother plant (barochory; Cahalan & Gliddon 1985).
Vegetative spread is restricted and only occurs within very short
distances from the mother plant through production of lateral
rosettes. Although these individual rosettes can die off, individual
plants are relatively long-lived (10–30 years; Boyd, Silvertown &
Tucker 1990).
STUDY SITES AND DATA COLLECTION
During 1999–2002, population surveys of P. vulgariswere carried out
at four locations (Appendix S1). Two populations (G1 and G2) were
grazed, whereas the other two (MCL1 and MCL2) occurred at sites
that were cleared mechanically (with scraping of ditch banks and
partial removal of the vegetation) once every three years, followed by
annual mowing and removal of all litter in subsequent years.
Distances between the studied populations were small (0Æ5–2 km),
implying that they experienced similar climatic conditions. All sites
were visited once a year during spring at peak flowering (mid March
to early April). During each survey, all plants were mapped to the
nearest cm, enabling to locate each individual throughout the
sampling period and to locate mortality events within each site. In all
432 H. Jacquemyn et al.
� 2010 The Authors. Journal compilation � 2010 British Ecological Society, Journal of Applied Ecology, 47, 431–440
relationships of both treatments are similar at years 2001 and
2002.
ANALYSIS 2 : RECRUIT SURVIVAL
Mortality of recruits under the grazing treatment was almost
random, with the exception of a small scale aggregation of
dead recruits at year 2001 (Fig. 3a), an additional small-scale
pattern in dead recruits (Figs. 3d vs. e) and negative small-scale
density dependent effect (Fig. 3f). All three significant depar-
tures from the randommortality hypothesis are consistent with
small scale density dependent mortality under the scramble
type (although the tendency of segregation of dead and surviv-
ing recruits was not significant; Fig. 3c). However, in year 2002
recruit mortality was entirely random (Fig. 3g–l).
The 2001 mortality pattern under the mowing treatment
contrasted sharply with that under the grazing treatment. Due
to the extreme clustering of recruits (Supporting Information
Appendix S1, Fig. 1b–d), we observe strong departures from
random labelling in all test statistics (Figs. 4a–f; Table 1).
Dead recruits are clustered (Fig. 4a), surviving recruits are
more regular than the pre-mortality pattern (Fig. 4b), surviv-
ing and dead recruits are strongly segregated at scales up to
20 cm (Fig. 4c), there is a strong additional pattern in dead
recruits (Figs. 4d vs. e), and strong negative small-scale density
dependence in mortality (Fig. 4f). These departures from ran-
dommortality are consistent with density dependent mortality
of the scramble type. Curiously, the 2002mortality in themow-
ing treatment shows exactly the same characteristics as the
2001 mortality in the grazing treatment: a significant small
scale aggregation of dead recruits (Fig. 4g), an additional
small-scale pattern in dead recruits (Figs. 4j vs. k) and a
(weakly) significant negative small-scale density dependent
effect (Fig. 4l). Again, all three departures from random
labelling in year 2002 are consistent with small scale density
dependentmortality.
ANALYSIS 3 : SURVIVAL OF RECRUITS VS. ADULTS
There was no clear indication that adults influenced the sur-
vival of recruits although there was a tendency towards posi-
tive interactions: the probability of survival tended to be higher
in the neighbourhood of adults (Fig. 5b–d; Table 1). However,
at a scale of 40 cm, which was also the scale of maximal attrac-
tion of recruits around adults, the test statistic indicated a
weakly significant positive effect at approximately 40 cm,
slightly outside the simulation envelopes under the mowing
treatment for year 2001. This additional (weak) pattern in sur-
vival was also depicted by the test statistic g21(r) - g22(r) which
indicated additional aggregation of surviving recruits condi-
tionally on the pre-mortality pattern (Fig. 4e).
Discussion
SPATIAL CLUSTERING OF SEEDLINGS AND ADULTS
In plant populations, significant spatial structure (a departure
from complete spatial randomness) is the norm and plants in
mesic environments are commonly clustered together in
groups of conspecifics (Law et al. 2000). The results from our
study are similar to these general observations, as in all four
populations both adults and recruits showed significant spatial
clustering. Three major causes of spatial structure (aggrega-
tion or segregation) have been identified (Coomes et al. 1999):
MCL adults
0
5
10
15
20MCL recruits 00
1
10
100
1000
G adults
Spatial scale r [m] Spatial scale r [m] Spatial scale r [m] Spatial scale r [m] 0·0 0·2 0·4 0·6 0·8 1·0 0·0 0·2 0·4 0·6 0·8 1·0 0·0 0·2 0·4 0·6 0·8 1·0 0·0 0·2 0·4 0·6 0·8 1·0
O-r
ing
stat
istic
O(r
) [n
o pl
ants
m–2
]O
-rin
g st
atis
tic O
(r)
[no
plan
ts m
–2]
O-r
ing
stat
istic
O(r
) [n
o pl
ants
m–2
]O
-rin
g st
atis
tic O
(r)
[no
plan
ts m
–2]
0
10
20
30
40G recruits 00
1
10
100
1000
MCL surviving recruits 01
G surviving recruits 01
MCL surviving recruits 02
G surviving recruits 02
(a) (b) (c) (d)
(e) (f) (g) (h)
Fig. 1. Analysis of the univariate spatial patterns of adults and recruits (00: recruits present in 2000, 01: surviving recruits in 2001; 02: surviving
recruits in 2002) under the two contrasting management treatments, grazing (G) and mowing plus clearing (MCL). The data were contrasted to
the null model of complete spatial randomness. The O-ring statisticO(r) giving the density of plants at distance r away from a typical plant of the
pattern (dots), simulation envelopes (black solid line) being the 5th lowest and highest values of theO-ring statistic taken from the 199 simulations
of null model, and the average O-ring statistic under the null model (grey solid line). Note the logarithmic scale of the y-axis for the recruit pat-
terns (b–d and f–h). The patterns are aggregated if the O-ring statistic is above the simulation envelopes.
Spatial population dynamics in Primula vulgaris 435
� 2010 The Authors. Journal compilation � 2010 British Ecological Society, Journal of Applied Ecology, 47, 431–440
Fig. 2. Analysis of the recruit-adults relationship. Test of independence between the pattern of recruits in years 2000, 2001, and 2002 and the adult
pattern under the two contrasting management treatments grazing (G) and mowing plus clearing (MCL). The data were compared to toroidal
shift as null model. The bivariate O-ring statisticO12(r) giving the density of recruits at distance r away from a typical adults plant (dots), simula-
tion envelopes (black solid line) being the 5th lowest and highest values of theO12(r) taken from the 199 simulations of null model, and the aver-
age O12(r) under the null model (grey solid line). Recruits are significantly aggregated around adults if the O12(r) is above the simulation
envelopes.
Table 1. Results of goodness-of-fit tests. We used 199 Monte Carlo
simulations of a null model to assess significant departures of the
observed data from the null model. The table shows the rank of the
test statistic ui for the observed pattern (i = 0) within the simulated
patterns (i = 1, …, 199). If rank u0 > 190 the null model is rejected
at the distance interval of interest on a 5% level (*), and if
rank > 198 at a 1% level (**). The distance intervals were 0–1 m
(Figs 1, 2 and 5) and 0–30 cm (Figs 3 and 4)
Panel Fig 1 Fig 2 Fig 3 Fig 4 Fig 5
A 200** 90 199** 200** 164
B 200** 110 112 200** 168
C 200** 159 144 200** 118
D 200** 199** 199** 200** 120
E 200** 168 6 200**
F 200** 172 195* 200**
G 200** 137 200**
H 200** 109 119
I 97 124
J 89 200**
K 106 47
L 121 191*
436 H. Jacquemyn et al.
� 2010 The Authors. Journal compilation � 2010 British Ecological Society, Journal of Applied Ecology, 47, 431–440
Fig. 3. Analysis of survival for recruits under the grazing treatment. We compared the spatial pattern of surviving (index 2) and dead recruits
(index 1) at years 2001 (00–01) and 2002 (01–02) with the null model of randommortality. We used different test statistics based on pair correla-
tion functions to test the random mortality null model; each test statistic depicts a different biological effect. The observed test statistics (dots),
simulation envelopes (black solid line) being the 5th lowest and highest values of the test statistic taken from the 199 simulations of null model,
and the average test statistic under the null model (grey solid line).