Conservation of remnant populations of Colchicum autumnale – The relative importance of local habitat quality and habitat fragmentation

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2

ava i lab le at www.sc ienced i rec t . com

journa l homepage : www. e lsev ier . com/ loca te /ac toec

Original article

Conservation of remnant populations of Colchicum autumnale –The relative importance of local habitat quality and habitatfragmentation

Dries Adriaensa,*, Hans Jacquemynb, Olivier Honnayb, Martin Hermya

aUniversity of Leuven, Division of Forest, Nature and Landscape, Celestijnenlaan 200 E, B-3001 Leuven, BelgiumbUniversity of Leuven, Biology Department, Plant Ecology, Kasteelpark Arenberg 31, B-3001 Leuven, Belgium

a r t i c l e i n f o

Article history:

Received 13 February 2008

Accepted 20 August 2008

Published online 21 October 2008

Keywords:

Population size

Plant reproductive capacity

Habitat isolation

Clonal growth

Habitat restoration

Microsite availability

* Corresponding author. Tel.: þ32 16 32 97 36E-mail addresses: [email protected]

kuleuven.be (O. Honnay), martin.hermy@ee1146-609X/$ – see front matter ª 2008 Elsevdoi:10.1016/j.actao.2008.08.003

a b s t r a c t

Semi-natural habitat is extremely vulnerable to habitat fragmentation and degradation

since its socio-economic value has decreased substantially during the last century in most

parts of Europe. We evaluated the relative effects of habitat fragmentation and local

environmental conditions on population structure and reproductive performance of the

long-lived corm geophyte Colchicum autumnale in 17 highly fragmented populations. Habitat

isolation did not affect patch occupancy, population structure or plant performance. In

contrast, population size and local environment strongly affected population structure and

reproductive performance. Densities of all life stages increased with increasing population

size. Large populations also showed a higher reproductive performance and a larger

proportion of new recruits. Relationships with local growth conditions pointed towards the

importance of an open grassland sward for flower and fruit set and the presence of

microsites for successful sexual recruitment. These results suggest that the distribution of

C. autumnale consists of an assemblage of basically unconnected populations that are

remnants of formerly larger populations. This is in accordance with the species’ ability to

grow clonally, allowing long-term persistence under deteriorating conditions that occurred

during a long period of habitat fragmentation. In conclusion, our results indicate that local

habitat and population size are more important than habitat fragmentation (i.e. calcareous

grassland isolation and surface area) and argue in favour of a management that is

primarily focused on local habitat restoration. This is preferentially accomplished by

reintroducing grazing practices, complemented by regular setback of spontaneous

succession towards forest.

ª 2008 Elsevier Masson SAS. All rights reserved.

1. Introduction for biodiversity. Due to socio-economic changes during the 19th

In most western-European countries, habitat fragmentation is

a widespread phenomenon that has profound consequences

; fax: þ32 16 32 97 60.om (D. Adriaens), hans.j

s.kuleuven.be (M. Hermy)ier Masson SAS. All rights

and 20th century, large and species-rich grassland areas, once

extensively managed by man, were abandoned and converted

tomore profitable landuses with mainlyagricultural, forestry or

[email protected] (H. Jacquemyn), olivier.honnay@bio..

reserved.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 270

urbanization purposes. As a consequence, many semi-natural

grasslands were destroyed and the size of the remaining areas

further diminished, while the distances between remnants

increased. In many cases, their habitat quality also seriously

deteriorated due to the cessation of traditional management

practices and the onset of natural succession, driving many

species to local extinction (e.g. Fischer and Stocklin, 1997; Maes

and Van Dyck, 2001; Wenzel et al., 2006).

The consequences of the different components of habitat

fragmentation (pure area loss, decline in fragment area and

increasing isolation) may be far-reaching. While the result of

pure area loss is obvious, small habitat remnants can only

harbour small populations that suffer from increased envi-

ronmental, demographic and genetic stochasticity (Matthies

et al., 2004; Munzbergova, 2006). In general, the impact of

severe environmental disturbance is most pronounced,

whereas demographic stochasticity only affects very small

populations (Menges and Dolan, 1998; Holsinger, 2000;

Matthies et al., 2004). Inbreeding and loss of genetic variation

can further lead to reduced fitness and evolutionary potential,

eventually causing further population decline (Ellstrand and

Elam, 1993; Newman and Pilson, 1997; Fischer and Matthies,

1998). Moreover, the inflow of genes and seeds by pollinators

or seed dispersers is hampered by increasing isolation, pre-

venting small populations from being ‘rescued’ and extinct

habitat patches from being recolonized (Brown and Kodric-

Brown, 1977; Hanski, 1998). Not only rare species but also

common species have been found to suffer from habitat

fragmentation (Honnay and Jacquemyn, 2007).

However, the impact of habitat fragmentation may take

years to become manifest (Ellstrand and Elam, 1993). Plant

species may respond slowly to changing landscape configu-

ration. They can still be present as remnant populations

despite current conditions that do no longer fulfill their

ecological requirements, contributing to a so-called extinction

debt (Tilman et al., 1994; Helm et al., 2006; Vellend et al., 2006).

This is especially true for long-lived and slow-growing species

that can reproduce vegetatively. Because habitat fragmenta-

tion often implies a decline in habitat quality, it is also

important to discriminate between the effects of habitat

fragmentation per se and local environmental conditions on

long-term population viability, to ensure that management

actions are deployed as efficiently as possible to sustain and

reinforce fragmented populations. These effects can be eval-

uated by measuring changes in vegetative and reproductive

capacity, population structure or long-term demography.

Although results from long-term demographic monitoring are

likely to be most informative, gathering these data is labour-

intensive and requires many years. A single-time census of

population structure, i.e. the relative proportion of different

ontogenetic stages in a species’ life cycle, combined with

measuring various life-cycle characteristics, has been pro-

posed as a more feasible way to quickly assess population

viability (Oostermeijer et al., 1994), in particular when it was

shown that population demographic structures are strongly

correlated with population growth rates obtained from demo-

graphic monitoring (e.g. Oostermeijer et al., 1996; Colling and

Matthies, 2006). As a result, assessing the population structure

has repeatedly been used to evaluate the success of mana-

gement interventions by changing habitat characteristics

(Oostermeijer et al., 1994; Hegland et al., 2001), to evaluate the

performance of species in different habitat types (Endels et al.,

2004; Mroz, 2006) and to measure the impact of habitat isola-

tion and population size on species persistence (Hooftman

and Diemer, 2002; Lienert et al., 2002).

We sampled the population structure and plant perfor-

mance of the long-lived Colchicum autumnale in calcareous

grasslands that have had a long history of habitat fragmen-

tation and cover a range of habitat characteristics. The impact

of population isolation and size, and local environmental

conditions was studied in order to quantify the relative

importance of habitat fragmentation and local environment,

respectively. Since many of its primary habitat has been

destroyed, the remaining populations in calcareous grass-

lands are important for regional species survival.

More specifically, we addressed the following questions:

– How are populations of C. autumnale structured and how

do they perform in a fragmented and heterogeneous

landscape?

– What is the relative importance of habitat fragmentation

vs. local variables in determining Colchicum population

structure?

– If local environment is important, which variables most

influence plant performance and population structure?

– What management can be the most beneficial for the

long-term survival of Colchicum in calcareous grasslands?

2. Materials and methods

2.1. Study area

The study area (80 km2) is located in the south-western part of

an elongated belt of calcareous outcrops in south-western Bel-

gium and consists of hilly areas surrounded by small valleys

(Fig. 1). After a period of socio-economic change in the second

half of the 20th century, traditional management was aban-

doned. Of the 4000 ha of calcareous grasslands that existed in

1775, only 70 ha remain today, as 70 small and isolated frag-

ments in the mainly agricultural and forested landscape matrix.

Most fragments are actively managed as nature reserves by

rotational sheep and goat grazing. We refer to Adriaens et al.

(2006) for a more detailed description of the study area.

2.2. Study species

C. autumnale or meadow saffron is an iteroparous and peren-

nial geophyte with a corm. Preferring moist and fertile valley

meadows, it also occurs in dry calcareous semi-natural

grasslands. Nowadays, the latter have become crucial for

regional survival of Colchicum because its optimal habitat has

become very scarce due to agricultural intensification and

conversion, and because C. autumnale has been regarded as

a pest and toxic plant species in meadows (Butcher, 1954).

Colchicum rapidly disappears after intensive mowing, grazing,

treading, draining or after ploughing and heavy fertilization

(Mroz, 2006). Colchicum also rarely occurs in road verges or

forest margins within the study area. However, these sites are

characterised by very low numbers of reproducing

Fig. 1 – Overview of the calcareous grasslands in the study area. Fragments that are occupied by Colchicum autumnale are

indicated in black. Fragments that were sampled for this study bear their respective label (see Table 1). The other occupied

fragments are marked with an asterisk.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 71

individuals. In terms of ecological connectivity, their impor-

tance is thus considered minimal (Mroz, 2006).

Meadow saffron has a hysteranthous life cycle: flowers

(1–5) appear from August to October and are followed by 2–5

leaves only in early spring (March), together with the fruits

that ripen until June. In mid-July, the leaves wither and only

the dry fruits remain visible aboveground (Butcher, 1954).

Flowers produce nectar and are pollinated by flies and

butterflies. Autogamy is possible if cross-pollination has not

taken place (Butcher, 1954). Seed dispersal is myrmecocho-

rous due to the presence of an elaiosome (Persson, 1993;

personal observation). Vegetative reproduction occurs by

daughter corms that occasionally split off the mother corm

(Jaehn and Roux, 1986; Frankova et al., 2004). The dispersal of

both vegetative and regenerative offspring is thus very limited

and results mainly in locally clumped distributions of indi-

viduals. Plants grown from seed rarely flower before the fifth

year, while vegetative offspring can be reproductive already in

the first year (Poutaraud and Champay, 1995). Seed banks are

transient (Thompson et al., 1997).

2.3. Data collection

Patch occupancy (i.e. the number of occupied calcareous

grassland fragments) of C. autumnale was determined based on

complete species lists of 64 grassland fragments and further

field observations (see Adriaens et al., 2006). In total, 58 plots,

varying in size from 1 to 4 m2, were randomly established in

populations of 17 grassland fragments (Table 1). All individuals

within a grassland fragment were considered to belong to

a single population, given the mainly clumped occurrence of C.

autumnale within the grassland fragments. The number of plots

was determined by the size and spread of the population. To

determine stage structure, i.e. plant density (number of indi-

viduals divided by plot area) and proportion of each different

life stage, we used a classification that discriminated between

seedlings and juveniles (S&J; single leaf), and vegetative (VA)

and reproductive adults (RA; two or more leaves, the latter

flowering in autumn and possibly fruiting in spring) (cf Mroz,

2006). Proportions among life stages were calculated.

In autumn 2004, the exact location of all reproductive

adults within each plot was measured with high-precision

differential GPS (up to the nearest cm) and the number of

flowers per plant was counted. During June 2005, all plots were

revisited to count and locate seedlings, juveniles and vegeta-

tive adults. An individual (S, J, VA or RA) could be easily

discriminated by the flowers or leaf/leaves originating from

a single sheath at ground level. Fruits (i.e. seed pods) of

a maximum of five adults were collected in the neighbour-

hood of the plots to estimate seed production within each plot.

Seed mass was determined as average seed mass after drying

to constant mass at 30 �C. From these measurements,

7 quantitative plant performance variables that represent

reproductive capacity were calculated for each plot: flower,

Table 1 – Code, name, coordinates, plot number, area, isolation (NN5) and population size of the 17 calcareous grasslandfragments in which Colchicum was sampled

Code Name Longitude Latitude Plot number Area (ha) NN5 (km) Population size

Bo Bonnerieu 4� 430 0200E 50� 060 2200N 5 1.36 2.92 44

CA Champ d’Al Veau 4� 400 0500E 50� 060 2700N 6 1.62 1.18 300

CoT Coupu Tienne 4� 410 5200E 50� 070 2500N 4 1.47 1.99 59

CTL Aux Couturelles 4� 400 0100E 50� 070 4800N 3 0.18 1.70 60

H Les Hurees 4� 370 2600E 50� 060 4300N 5 1.07 2.36 32

L2 Lernoee 4� 410 1600E 50� 080 1200N 4 0.56 2.22 10

MB1 Montagne aux Buis 4� 340 2000E 50� 050 3300N 4 0.47 1.32 230

MS Mallonsart 4� 380 2200E 50� 060 4900N 2 0.67 1.84 50

PL Petites Lorines 4� 400 2000E 50� 060 4800N 3 0.50 1.42 180

RM2 Roche Madoux 4� 370 5200E 50� 050 0600N 4 0.59 1.10 500

RO La Rosiere 4� 310 4500E 50� 040 5100N 4 0.53 2.23 10

T1 Transoi 4� 380 0900E 50� 050 0800N 2 1.24 0.94 320

TAP1 Tienne au Porey 4� 340 4300E 50� 050 1200N 3 0.57 1.01 99

TDM Tienne du Moulin 4� 360 0500E 50� 040 3100N 3 0.27 2.06 15

TL Tienne du Lion 4� 300 4700E 50� 040 1700N 2 4.07 3.33 133

TP1 Tienne Pele 4� 340 1100E 50� 050 0200N 2 0.22 1.22 30

TP2 Tienne Pele 4� 340 2500E 50� 050 0500N 2 0.09 1.11 21

Population size was counted as the number of flowering individuals during autumn 2004.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 272

fruit and seed density, average number of flowers and fruits

per plant, fruit set and seed mass per square meter. Seed

density was calculated as the number of fruits in the plot,

multiplied by the average of the seed counts from the

collected fruits and divided by plot area. Multiplying seed

density with average mass per seed from the collected fruits

yielded seed mass per square meter. Fruit set was obtained as

the ratio of fruits (2005) and flowers (2004) per plot.

Local environmental variables were collected during June

2005 within each plot. Both the biotic and abiotic environment

was sampled. For each plot, the potential annual direct radi-

ation (MJ cm�2 yr�1) was calculated based on its latitude, slope

and aspect using the formula of McCune and Keon (2002).

Canopy closure (%), i.e. the proportion of the sky hemisphere

obscured by vegetation, was evaluated as the mean of 4

measurements with a spherical densiometer in each wind

direction at 1 m height (Jennings et al., 1999). Soil depth above

rock material (cm) was determined as the mean of 10

measurements with a metal pin. For each plot, cover (%) of

bare ground, rock, litter, moss, small woody debris, dominant

grasses, sedges and shrubs was also estimated. Small woody

debris was often present after management interventions, e.g.

shrub chopping. Shrub height (cm) was calculated as the

mean height of all different species.

Habitat fragmentation was determined within a GIS

(ArcView 3.2a; ESRI, 2000). Grassland fragment area was

calculated based on digitized topographical maps and aerial

photographs, complemented with field surveys. Population

isolation was computed as the edge-to-edge distance to the

nearest grassland fragment with a Colchicum population

using the Nearest Features extension (Jenness, 2007).

Connectivity measures, derived from Hanski’s Incidence

Function Model (IFM; Hanski, 1994), proved to be highly

correlated with population size. To prevent inter-correla-

tions among independent variables in our analyses, we

therefore preferred to use the average of five simple edge-to-

edge distances to the nearest Colchicum populations (isola-

tion variable NN5).

Total population size was counted as the number of flow-

ering individuals in 2004 within each grassland fragment.

2.4. Data analysis

In a first step, the impact of grassland fragment area and

isolation on patch occupancy of all C. autumnale populations in

the study area was assessed using multiple logistic regression.

To reduce the number of environmental variables we

conducted a preliminary correlation analysis, consisting of

both a principal component analysis (PCA with VARIMAX

rotation) and a simple correlation analysis (Pearson’s corre-

lation coefficients) among individual environmental,

population-structural and plant performance variables. Due

to the nested structure of the plots within each grassland

fragment, we applied linear mixed models with grassland

fragments as random factor. Population structure and plant

performance were treated as independent variables, while

population size and isolation and the selected local environ-

mental variables were considered as covariates. Preliminary

analyses ensured us that random slopes could be left out.

Hence, only the random intercept was retained to evaluate the

among-fragment variance component.

All variables were transformed prior to analysis (Table 2

and Appendices 1 and 2) and all analyses were performed in

SPSS 13.0 for Windows (SPSS, 2005).

3. Results

3.1. Habitat occupancy and configuration

Within the study area, C. autumnale occurred in 26 of the 64

investigated grassland fragments (40.6%, Fig. 1), of which 17

were sampled for this study (Table 1). Average area of the

occupied grasslands was 0.911 ha, ranging from 0.087 to

4.07 ha (Table 1). Occupied patches were on average 720 m

Table 2 – Factor loadings of the four principal components (eigenvalue > 1) for 13 local environmental variables of 58 plotsof Colchicum after VARIMAX rotation

Transformation PC1 PC2 PC3 PC4

Eigenvalue 2.896 2.223 1.591 1.234

Variance explained 22.275 17.098 12.236 9.494

PotDirRad (MJ cm�2 yr�1) – �0.382 �0.123 0.113 �0.492

Soil depth (cm) x2 �0.146 �0.107 0.837 �0.089

Bare ground (%) x�1/2 0.235 �0.708 �0.200 �0.163

Rock (%) x�1/2 �0.198 �0.134 �0.779 �0.151

Moss (%) sinh(x1/2 ) 0.124 0.637 �0.137 0.034

Litter (%) – 0.102 0.572 0.320 �0.193

Small woody debris (%) x�1/2 0.120 �0.625 0.274 0.040

Vegetation height (cm) – 0.064 0.584 0.049 0.601

Grasses (%) – �0.441 0.576 0.285 0.039

Sedges (%) x�1/2 �0.172 �0.154 0.091 0.804

Canopy closure (%) sin�1(x1/2 ) 0.826 0.005 0.053 �0.032

Shrub height (cm) log(x) 0.754 0.027 0.123 �0.013

Shrub cover (%) log(x) 0.850 �0.212 �0.098 0.058

Loadings in italic indicate significant ( p� 0.05) correlations between variables and principal components. Each variable is attributed to a single

principal component based on the most significant correlation, indicated in bold.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 73

apart from each other. Average distance to the nearest 5

occupied patches (NN5 avg) was 1572 m.

Patch occupancy was not determined by the area of the

grassland fragment (c2¼ 0.379, p¼ 0.538), neither by its isola-

tion (c2¼ 0.481; p¼ 0.488) nor by their interaction (c2¼ 0.679;

p¼ 0.410).

0

100

200

300

400

500

600

P2 TL P1 DM B1 PL RO L2 P1 M2

MS H T1 Bo oT CA TL

Po

pu

latio

n size

0

0.5

1

1.5

2

2.5

3

3.5

Iso

latio

n (km

)

Population sizeNN5 (km)

3.2. Environmental gradients

The first environmental gradient was characterised by

decreasing local light availability due to increasing shrub

cover (Table 2), and to a lesser extent by decreasing global light

interception (PotDirRad, intrinsic to local physiography).

Increasing amount of soil cover by woody debris, litter,

mosses and dominant grasses defined the second principal

component. The first two principal components explained

39.3% of the variance among the 13-variable dataset. A third

component was highly correlated with soil depth and super-

ficial rock cover and added 12.2% to the explained variance.

Although highly informative to understand the major

environmental gradients that were present in the sampled

population of C. autumnale, the PCs did not explain any vari-

ation in the mixed model analyses and were therefore

replaced by the individual environmental variables that

proved to explain most of the variance in the dataset: shrub

cover, cover of small woody debris and soil depth. These

variables showed both highest factor loadings on each of the

first three orthogonal PCA axes with eigenvalues exceeding 1

and exhibited the highest correlations with population

structure and plant performance. These 3 variables were

therefore used in further analyses.

T C T T M TA R C

Grassland fragment

Fig. 2 – Size (number of flowering adults) and isolation

(average of the 5 nearest distances to neighbour

populations, NN5) of the 17 sampled Colchicum autumnale

populations (r [ L0.099; p [ 0.705).

3.3. Population size and structure of C. autumnale

Population size, measured as the number of flowering individ-

uals, variedfrom10to500 (mean:123,SD:139) (Fig.2).Population

size was not affected by grassland area (r¼ 0.395, p¼ 0.117,

n¼ 17) or landscape isolation (r¼�0.448, p¼ 0.071, n¼ 17), but

was significantly correlated with average soil depth (r¼ 0.326,

p¼ 0.012, n¼ 58), grass cover (r¼ 0.347, p¼ 0.008, n¼ 58) and

shrub height (r¼�0.438, p¼ 0.001, n¼ 58) (Fig. 3). No relation-

ships between isolation and environment were observed.

Average plant density varied from 3.31 reproductive adults

(RA) per square meter to densities of 21.87 and 72.11 m�2 for

vegetative adults (VA) and subadults (seedlings and juveniles,

S&J), respectively (Fig. 4a). However, population structure was

highly variable. Especially the number of subadults showed

a high coefficient of variation of 309%, and values of 118% and

128% for vegetative and reproductive adults, respectively. The

skewed distribution of each of the life stages is revealed by the

medians and quartiles. In 75% of the plots subadult densities

did not exceed 16.72 m�2. For RA and VA these densities were

4.25 and 26.87 m�2, respectively (Fig. 5). Thus, proportionally

and on average, vegetative adults were the most abundant

(64.7%), followed by the subadult (35.1%) and reproductive

stage (15.4%) (Fig. 4b).

Soil depth (cm2)

Po

pu

latio

n size

1

10

100

1000

Grass cover (%)

0 50 100 150 200 250 0 20 40 60 80 100

Po

pu

latio

n size

1

10

100

1000

Shrub height (cm)

10 100 1000

Po

pu

latio

n size

1

10

100

1000

ba

c

Fig. 3 – Population size in relation to (a) soil depth (r [ 0.326, p [ 0.012, square transformation), (b) grass cover (r [ 0.347,

p [ 0.008) and (c) shrub height (r [ L0.438, p [ 0.001) for 58 plots of Colchicum autumnale.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 274

3.4. Determinants of population structureand plant performance

None of the population structure variables was related to land-

scape isolation in the mixed model analysis (Appendix 1). On the

contrary, population structure was in all instances significantly

determined by population size, even if controlled for environ-

mental relations, with the exception of the proportion of repro-

ductive adults (Fig. 6a–c, Appendix 1). Large populations showed

high densities of all life stages. Proportionally, however,

increasing population size implied a larger proportion of the

seedling and subadult stage, while the proportion of vegetative

adults significantly diminished, suggesting low seedling recruit-

ment in small populations. The proportion of flowering adults

0

20

40

60

80

100

120

140

Life stage

Pla

nt d

en

sity

(m

-2),

av

era

ge

± s

t.e

rro

r

TotalS&JVARA

a

Fig. 4 – Average densities (±st. error) (a) and proportions (±st. err

in 58 plots in 17 sampled populations of Colchicum autumnale. S

reproductive adult.

was not related to population size. The amount of small woody

debris was positively related to all structural variables except the

density of reproductive adults and the proportion of vegetative

adults, with subadult density (Fig. 6d) and number of subadults

per reproductive adult most positively affected. The ratio

subadults/adults was negatively affected by the amount of

debris, although the number of offspring per reproductive adult

significantly increased with increasing amount of debris. The

cover of shrubs turned out to be negatively related to the number

of offspring per adult, although the share of subadults increased

with shrub cover. Soil depth and bare ground were nowhere of

importance.

Similar trends were observed for the plant performance

variables. Flower, fruit and seed density were all positively

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Life stage

Pla

nt p

ro

po

rtio

n,

av

era

ge

± s

t.e

rro

r

S&JVARA

b

or) (b) for each life stage, based on the population structure

&J: seedlings and juveniles; VA: vegetative adult; RA:

RAVAS&J

1.000

100

10

1

-0,1

S&J/RAS&J/ARAVAS&J

100

10

1

-0,1

a b

Fig. 5 – Box-plots of (a) plant densities (mL2) and (b) proportions of the three classified life stages of Colchicum autumnale. S&J:

seedlings and juveniles, VA: vegetative adults, RA: reproductive adults, A: VA D RA.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 75

linked with population size (Fig. 7a–c). The number of

flowers per plant and seed mass per square meter were also

positively affected by population size (Appendix 2). Fruit set,

however, decreased with increasing population size (Fig. 7b).

Landscape isolation only affected seed density, seed mass

and flower number per plant, whereas local environment

was important for the number of flowers per plant and fruit

Population size

10 100 1000Seed

lin

g &

ju

ven

ile d

en

sity (m

-2)

0.1

1

10

100

1000

10000

Population size

10 100 1000Rep

ro

du

ctive ad

ult d

en

sity (m

-2)

0.01

0.1

1

10

100

a

c

Fig. 6 – Relationship between population size and density of (a)

(t [ 3.547, p [ 0.001) and (c) reproductive adults (t [ 3.584, p [ 0

most significant (d, t [ 4.138, p [ 0.002). For a complete overvie

Appendix 1.

set (Appendix 2). Fruit set was promoted by higher amounts

of debris while plants in more shaded conditions exhibited

more flowers.

The intercept of the random factor (fragment) was never

significant, stressing that the among-fragment variance

component was negligible, while the among-plot variance

within each fragment was always significant.

Population size

10 100 1000

Veg

etative ad

ult d

en

sity (m

-2)

0.1

1

10

100

1000

Small woody debris

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6Seed

lin

g &

ju

ven

ile d

en

sity (m

-2)

0.1

1

10

100

1000

10000

b

d

subadults (t [ 3.964, p [ 0.003), (b) vegetative adults

.003). Impact of small woody debris on subadult density is

w of all statistics regarding population structure, see

Population size

10 100 1000

Flo

wer d

en

sity (m

-2)

0.1

1

10

100

Population size

10 100 1000

Fru

it d

en

sity (m

-2)

0.1

1

10

100

Population size

10 100 1000

Seed

d

en

sity (m

-2)

1

10

100

1000

10000

Population size

10 100 1000

Fru

it set +

1

1.5

2

1

a b

dc

Fig. 7 – Relationship of population size with density of (a) flowers (t [ 3.632, p [ 0.003), (b) fruits (t [ 3.057, p [ 0.004) and (c)

seeds (t [ 3.651, p [ 0.005), and with (d) fruit set (t [ L2.301, p [ 0.039). For a complete overview of all statistics regarding

plant performance, see Appendix 2.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 276

4. Discussion

4.1. No effects of population isolation

Landscape isolation did not affect patch occupancy nor

population structure or plant performance of C. autumnale,

with the notable exception of seed density and mass, which

responded positively to increasing isolation. Because seed

dispersal is very likely limited to very short distances from the

mother plant, these results suggest that seed exchange

between populations is very limited. The studied populations

may be seen as remnant populations (Eriksson, 2000), capable

of surviving unfavourable environmental conditions for many

years. Indeed, the species shows several features that enable

persistence during long periods of unfavourable conditions.

First, the corm structure provides extensive starch reserves

that can be used for survival. Second, these reserves also allow

the production of vegetative offspring (Frankova et al., 2003,

2004), even when environmental conditions gradually deteri-

orate (Mroz, 2006). Both features increase the probability of

survival as a remnant population (Eriksson, 2000).

4.2. Population size effects and relationshipswith local environment

Both population structure and plant performance were

strongly influenced by population size. For population

structure, the effects of local environment were comparable

with those of population size regarding statistical strength,

while plant performance tended to be much less influenced by

environmental conditions.

Larger populations (mostly on deeper soils with a dense

sward and little shrub) were characterised by increasing

densities of all life stages, confirming the often observed,

clumped distribution of plants within a habitat fragment.

Moreover, increasing population size implied a higher share

of seedlings and juveniles, at the expense of vegetative

adults. This is in contrast with other studies that detected

a very low germination and recruitment rate (Poutaraud and

Champay, 1995), while vegetative reproduction can be

considerable. Reserve buds on the mother corm have a 0–

30% probability to grow into new individuals under field

conditions (Godet, 1987). Mroz (2006) also found that 26% of

the mother corms produced two daughter corms in

a competitive environment (road verges), compared with

only 9% in frequently mown hay meadows. The observations

of Mroz (2006) strongly suggest an increased clonal growth in

suboptimal habitat. Although this may ensure long-term

persistence in remnant habitat, the area occupied by the

population will barely grow, also because sexual reproduc-

tion is almost non-existent in these conditions and seed

dispersal distance is very limited. Moreover, environmental

stochasticity is generally much higher in small, remnant

habitat patches (Menges, 1998), increasing the risk of

a sudden extinction.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 77

Little is known about the flower and seed ecology of

C. autumnale. Until now, most research concentrated on

seed production and seed alkaloid content for medicinal

purposes (Poutaraud and Girardin, 2005, 2006). Results

point towards increasing seed yield under competitive-free

growth conditions, caused by a yearly vegetative pop-

ulation increase of about 50%, despite a slight decrease in

seed yield per plant due to increasing intraspecific

competition among ramets (Poutaraud and Girardin, 2003).

It is thus suggested that generative reproduction has

a tendency to decline after long periods of vegetative

growth. Despite the fact that population growth is lower in

grassland (about 15% yr�1), although with considerably

more fluctuations during population build-up, an analo-

gous reduction of sexual reproduction can be expected

here as well, as confirmed by our results. First, our data

showed that competition for light significantly reduced the

number of flowers per plant, which is confirmed by other

studies (Godet, 1987; Mroz, 2006). Secondly, fruit set was

higher in open vegetation, as quantified by high amounts

of woody debris associated with recent shrub removal

management interventions. Thus, our results suggest

a significant decrease in sexual reproduction when grass-

land habitat quality deteriorates. Surprisingly, fruit set was

higher in small populations. Normally, smaller populations

receive less pollinator attention, resulting in lower fruit

and seed set (e.g. Jennersten, 1988). Although higher visi-

tation rates by pollinators can be observed in smaller

populations (e.g. Bosch and Waser, 1999; Mustajarvi et al.,

2001; Goverde et al., 2002), this generally promotes

inbreeding (Richards, 2000) that negatively affects offspring

fitness (Lennartsson, 2002). However, no data are available

on the influence of selfing and outcrossing rates on fruit

set, seed production and offspring performance of C.

autumnale. It must be noted, however, that pollinator

activity during anthesis of C. autumnale is generally low

due to the often unfavourable weather conditions during

September and October, probably promoting autonomous

self-pollination. Despite this uncertainty, the lower fruit set

in large populations was compensated for by a higher

density of reproductive adults (Appendix 1) each having

more flowers. Consequently, seed production per unit area

significantly increased (Appendix 2).

We observed high variation in recruitment (Fig. 5), even

between plots within fragments. Subadult density was

highly promoted by the cover of small woody debris.

Probably, small woody debris provides shelter and assures

sufficient soil surface humidity to enable germination.

Therefore, we assume that the availability of very specific

microsites might be limiting successful germination and

establishment, possibly in combination with deteriorating

seed quality in small populations. However, a more

detailed study of the seed and recruitment ecology would

be of interest.

4.3. Management guidelines

Although small fragments did not necessarily harbour small

populations, and no effects of landscape isolation were found,

it is likely that small and isolated fragments are especially

prone to deterioration of habitat quality because of rapid

shrub encroachment during natural succession towards

forest. Moreover, these fragments often lack conservation

protection status, making them even more vulnerable to this

threat. C. autumnale can survive beneath a dense canopy, but

the number of flowering and vegetative adults steadily

decreases and fruit set is significantly diminished, leading to

a shrinking population size and a lack of sexual reproduction.

Long-term survival may thus strongly depend on recurrent

phases with sexual reproduction, given that (1) population

genetic variability may be low due to long periods of stasis or

gradual population regression during which genotypes

disappear, (2) dispersal capacity of daughter ramets is

extremely low, (3) whole genotypes can be destroyed at once

due to the clumped nature of mortality in response to

disturbance and (4) C. autumnale does not form a persistent

seed bank. This was also observed in a detailed long-term

study of the bulb species Allium tricoccum in the forests of

Quebec (Nault and Gagnon, 1993).

Since highest densities of subadults were observed after

recent shrub removal, with the resulting small woody debris

still present, cyclic management with shrub removal followed

by recurrent regrowth might favour C. autumnale populations.

This treatment indeed allows for phases of both vegetative

(during encroachement) and sexual (directly after clearance)

population growth. Management has to look for a balance

between both modes of reproduction in order to ensure long-

term survival. However, to ensure open grassland conditions

during which seedlings can establish and reach the adult

state, periodic shrub removal should be accompanied by

extensive grazing or late-summer mowing, i.e. when plants

are dormant, to reduce excessive competition. Grazing can

occur almost year-round because plants are avoided by cattle

due to their poisonous nature. Excessive trampling has to be

avoided though, especially during anthesis. Additionally,

small disturbances by animal treading can also increase

microsite availability, stimulating seed germination and

seedling establishment.

5. Concluding remarks

We found evidence that meadow saffron did not suffer

from a long history of habitat fragmentation. In combina-

tion with its high clonal growth capacity, these results

suggest that C. autumnale currently occurs as a collection of

basically unconnected, remnant populations within the

landscape. Therefore, survival of these population

remnants is especially conditional upon local growth

conditions, as suggested by the strong effects of population

size and environment on population structure and repro-

ductive plant performance. Since reproductive success is

especially dependent on open vegetation with sufficient

microsite availability for seed germination and establish-

ment, extensive grazing is suggested as the most beneficial

management for population expansion. However, to

unravel the ultimate reasons behind the impact of pop-

ulation size on reproductive success and to assess future

response to changing environment, further insight into the

reproductive mechanisms of C. autumnale is needed.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 278

Acknowledgements

This study was funded by a grant from the Flemish Fund for

Scientific Research (FWO) and the BIOCORE project EV/01/26A,

Appendix 1

Estimates of the fixed effects of population size (popsize), isolvariables on population structure of Colchicum autumnaleParameter Estimate Std. error Df

S&J

Intercept �1.636 1.126 15

Popsize 0.822 0.207 9.2

NN5 �0.001 0.015 9

Soil depth �0.001 0.002 22

Debris 1.025 0.248 9.1

Shrub cover 0.343 0.225 50

Rand. int. 0.001 0.085

Res 0.404 0.104

A

Intercept 0.019 0.682 52

Popsize 0.462 0.125 52

NN5 �0.004 0.009 52

Soil depth 0.000 0.001 52

Debris 0.456 0.150 52

Shrub cover 0.008 0.137 52

Rand. int. – –

Res 0.149 0.029

VA

Intercept 0.101 0.690 52

Popsize 0.451 0.127 52

NN5 �0.006 0.009 52

Soil depth 0.000 0.001 52

Debris 0.473 0.152 52

Shrub cover 0.007 0.138 52

Rand. int. – –

Res 0.153 0.030

RA

Intercept �0.840 0.534 24

Popsize 0.362 0.101 14

NN5 0.010 0.007 15

Soil depth 0.000 0.001 41

Debris 0.176 0.119 21

Shrub cover 0.038 0.101 52

Rand. int. 0.008 0.015

Res 0.075 0.018

propS&J

Intercept �0.671 0.603 17

Popsize 0.269 0.117 8.6

NN5 0.000 0.008 9.7

Soil depth 0.000 0.001 44

Debris 0.418 0.135 19

Shrub cover 0.219 0.109 50

Rand. int. 0.016 0.027

Res 0.085 0.022

propVA

Intercept 1.539 0.688 28

Popsize �0.279 0.131 17

financed by the Federal Belgian Science Policy. Thanks to

L.Woue, L.-M. Delescaille and forester ir. J-P. Scohy for their kind

cooperation with BIOCORE. Thanks also to Jan Butaye, Eric Van

Beek, Tom Neels and Katrien Piessens for field assistance.

ation (NN5) and the three most influential environmental

t/Wald Z p n Transf.

58 log(x)

�1.453 0.167

3.964 0.003**

�0.081 0.938

�0.811 0.426

4.138 0.002**

1.524 0.134

0.015 0.988

3.897 0.000***

58 log(x)

0.028 0.977

3.682 0.001***

�0.404 0.688

0.033 0.974

3.043 0.004**

0.061 0.952

– –

5.099 0.000***

58 log(x)

0.146 0.885

3.547 0.001***

�0.656 0.514

�0.246 0.807

3.116 0.003**

0.049 0.961

– –

5.099 0.000***

58 log(x)

�1.574 0.128

3.584 0.003**

1.451 0.167

0.377 0.708

1.478 0.154

0.374 0.710

0.519 0.604

4.292 0.000***

58 sin�1(x�1/2 )

�1.113 0.281

2.307 0.048*

�0.054 0.958

�0.079 0.937

3.103 0.006**

2.009 0.050*

0.610 0.542

3.852 0.000***

58 sin�1(x�1/2 )

2.237 0.034*

�2.140 0.047*

Appendix 1 (continued)

Parameter Estimate Std. error Df t/Wald Z p n Transf.

NN5 �0.001 0.009 18 �0.137 0.893

Soil depth 0.000 0.001 44 0.137 0.892

Debris �0.229 0.154 25 �1.491 0.148

Shrub cover �0.016 0.130 52 �0.122 0.903

Rand. int. 0.014 0.022 0.622 0.534

Res 0.124 0.028 4.447 0.000***

propRA 58 x�1/2

Intercept 1.125 0.230 52 4.891 0.000***

Popsize 0.042 0.042 52 1.000 0.322

NN5 �0.002 0.003 52 �0.682 0.498

Soil depth 0.000 0.000 52 �0.473 0.719

Debris 0.111 0.051 52 2.203 0.032*

Shrub cover 0.069 0.046 52 1.494 0.141

Rand. int. – – – –

Res 0.017 0.003 5.099 0.000***

S&J/A 58 x�1/2

Intercept 2.066 0.507 18 4.075 0.001***

Popsize �0.272 0.098 8.8 �2.765 0.022*

NN5 0.000 0.007 9.9 �0.050 0.961

Soil depth 0.000 0.001 45 0.219 0.827

Debris �0.364 0.113 20 �3.213 0.004**

Shrub cover �0.203 0.092 50 �2.208 0.032*

Rand. int. 0.012 0.019 0.621 0.535

Res 0.060 0.015 3.865 0.000***

S&J/RA 53 log(x)

Intercept �0.579 0.893 47 �0.649 0.520

Popsize 0.479 0.166 47 2.896 0.006**

NN5 �0.012 0.011 47 �1.070 0.290

Soil depth �0.001 0.001 47 �0.732 0.468

Debris 0.816 0.196 47 4.155 0.000***

Shrub cover 0.252 0.192 47 1.312 0.196

Rand. int. – – – –

Res 0.239 0.049 4.848 0.000***

The intercept of the random factor (rand. int.; i.e. grassland fragment) and the residual model variance are given (res). S&J: seedlings and

juveniles, A: adult, VA: vegetative adult, RA: reproductive adult, prop: proportion (i.e. proportion of total number of individuals S&JþA), res:

residual. Wald Z-statistic is valid for variance of random intercept and residual variance. Parameters were transformed prior to analysis (see

Table 2, population size: log(x), isolation NN5: square root). * 0.01< p� 0.05, ** 0.001< p� 0.01, *** p� 0.001.

Appendix 2

Estimates of the fixed effects of population size (popsize), isolation (NN5) and the three most influential environmentalvariables on plant traits of Colchicum autumnale

Parameter Estimate Std. error Df t/Wald Z p n Transf.

Flower density 57 log(x)

Intercept �1.363 0.666 22 �2.048 0.053

Popsize 0.472 0.130 12 3.632 0.003**

NN5 0.017 0.009 13 1.845 0.088

Soil depth 0.001 0.001 47 0.757 0.453

Debris 0.202 0.149 26 1.355 0.187

Shrub cover 0.007 0.120 50 0.062 0.951

Rand. int. 0.021 0.027 0.780 0.435

Res 0.101 0.025 4.108 0.000***

Flower/plant 45 –

Intercept �0.866 0.270 39 �3.202 0.003**

Popsize 0.167 0.048 39 3.462 0.001**

(continued on next page)

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 79

Appendix 2 (continued)

Parameter Estimate Std. error Df t/Wald Z p n Transf.

NN5 0.010 0.003 39 3.004 0.005**

Soil depth 0.001 0.000 39 1.609 0.116

Debris 0.057 0.062 39 0.909 0.369

Shrub cover 0.168 0.060 39 2.805 0.008**

Rand. int. – – – –

Res 0.019 0.004 4.416 0.000***

Fruit density 53 log(x)

Intercept �1.944 0.845 47 �2.300 0.026*

Popsize 0.479 0.157 47 3.057 0.004**

NN5 0.021 0.011 47 1.912 0.062

Soil depth 0.002 0.001 47 1.480 0.146

Debris 0.140 0.186 47 0.752 0.456

Shrub cover 0.239 0.182 47 1.312 0.196

Rand. int. – – – –

Res 0.214 0.044 4.848 0.000***

Fruit/plant 53 log(x)

Intercept �0.083 0.267 27 �0.312 0.757

Popsize �0.011 0.052 16 �0.211 0.836

NN5 0.003 0.004 17 0.792 0.439

Soil depth 0.001 0.000 37 1.427 0.162

Debris �0.036 0.060 28 �0.593 0.558

Shrub cover 0.015 0.052 46 0.284 0.778

Rand. int. 0.003 0.004 0.961 0.337

Res 0.015 0.004 4.167 0.000***

Fruit set 47 log(x)

Intercept 0.385 0.183 22 2.106 0.047*

Popsize �0.091 0.040 13 �2.301 0.039*

NN5 �0.001 0.003 13 �0.445 0.664

Soil depth 0.000 0.000 �0.413

Debris 0.092 0.042 31 2.212 0.035*

Shrub cover �0.024 0.033 35 �0.710 0.482

Rand. int. 0.003 0.002 1.349 0.177

Res 0.005 0.001 3.743 0.000***

Seed density 53 log(x)

Intercept �1.977 1.207 24 �1.638 0.115

Popsize 0.853 0.255 13 3.351 0.005**

NN5 0.051 0.017 15 2.950 0.010*

Soil depth 0.000 0.002 45 0.125 0.901

Debris �0.002 0.263 38 �0.009 0.993

Shrub cover 0.201 0.201 41 0.996 0.325

Rand. int. 0.141 0.092 1.530 0.126

Res 0.210 0.052 4.029 0.000***

Seed mass m�2 53 log(x)

Intercept �1.216 1.190 24 �1.022 0.317

Popsize 0.850 0.250 13 3.401 0.005**

NN5 0.050 0.017 15 2.928 0.011*

Soil depth 0.000 0.002 45 0.197 0.845

Debris 0.047 0.260 38 0.182 0.857

Shrub cover 0.202 0.200 41 1.008 0.319

Rand. int. 0.133 0.089 1.489 0.137

Res 0.209 0.052 4.022 0.000***

The intercept of the random factor (rand. int.; i.e. grassland fragment) and the residual model variance are given (res). Leaf length, breadth and

area are cumulative variables, i.e. measurement� leaf number. Wald Z-statistic is valid for variance of random intercept and residual variance.

Parameters were transformed prior to analysis (see Table 2, population size: log(x), isolation NN5: square root). * 0.01< p� 0.05, ** 0.001< p� 0.01,

*** p� 0.001.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 280

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 2 81

r e f e r e n c e s

Adriaens, D., Honnay, O., Hermy, M., 2006. No evidence of a plantextinction debt in highly fragmented calcareous grasslands inBelgium. Biological Conservation 133, 212–224.

Bosch, M., Waser, N.M., 1999. Effects of local density onpollination and reproduction in Delphinium nuttallianum andAconitum columbianum (Ranunculaceae). American Journal ofBotany 86, 871–879.

Brown, J.H., Kodric-Brown, A., 1977. Turnover rates in insularbiogeography – effect of immigration on extinction. Ecology58, 445–449.

Butcher, R.W., 1954. Colchicum autumnale L. Journal of Ecology 42,249–257.

Colling, G., Matthies, D., 2006. Effects of habitat deterioration onpopulation dynamics and extinction risk of an endangered,long-lived perennial herb (Scorzonera humilis). Journal ofEcology 94, 959–972.

Ellstrand, N.C., Elam, D.R., 1993. Population geneticconsequences of small population size – implications forplant conservation. Annual Review of Ecology andSystematics 24, 217–242.

Endels, P., Adriaens, D., Verheyen, K., Hermy, M., 2004. Populationstructure and adult plant performance of forest herbs in threecontrasting habitats. Ecography 27, 225–241.

Eriksson, O., 2000. Functional roles of remnant plant populationsin communities and ecosystems. Global Ecology andBiogeography 9, 443–449.

ESRI, 2000. ArcView 3.2a. Enviroinmental Systems ResearchInstitute Inc., Redlands.

Fischer, M., Matthies, D., 1998. Effects of population size onperformance in the rare plant Gentianella germanica. Journal ofEcology 86, 195–204.

Fischer, M., Stocklin, J., 1997. Local extinctions of plants inremnants of extensively used calcareous grasslands 1950–1985. Conservation Biology 11, 727–737.

Frankova, L., Komjathyova, H., Boka, K., Gasparıkova, O.,Psenak, M., 2003. Biochemical and physiological aspects ofdevelopmental cycle of Colchicum autumnale L. BiologiaPlantarum 47, 509–516.

Frankova, L., Cibırova, K., Boka, K., Gasparıkova, O., Psenak, M.,2004. The role of the roots in the life strategy of Colchicumautumnale. Biologia, Bratislava 59, 87–93.

Godet X., 1987. Biologie du colchique (Colchicum autumnale).Multiplication vegetative par voie traditionnelle et in vitro. Ph.D.thesis, University of Blaise Pascal, Clermont-Ferrand.

Goverde, M., Schweizer, K., Baur, B., Erhardt, A., 2002. Small-scalehabitat fragmentation effects on pollinator behaviour:experimental evidence from the bumblebee Bombus veteranuson calcareous grasslands. Biological Conservation 104, 293–299.

Hanski, I., 1998. Metapopulation dynamics. Nature 396, 41–49.Hanski, I., 1994. A practical model of metapopulation dynamics.

Journal of Animal Ecology 63, 151–162.Hegland, S.J., Van Leeuwen, M., Oostermeijer, J.G.B., 2001.

Population structure of Salvia pratensis in relation tovegetation and management of Dutch dry floodplaingrasslands. Journal of Applied Ecology 38, 1277–1289.

Helm, A., Hanski, I., Partel, M., 2006. Slow response of plantspecies richness to habitat loss and fragmentation. EcologyLetters 9, 72–77.

Holsinger, K.E., 2000. Demography and extinction in smallpopulations. In: Young, A.G., Clarke, G.M. (Eds.), Genetics,Demography and Viability of Fragmented Populations.Cambridge University Press, Cambridge, pp. 55–74.

Honnay, O., Jacquemyn, H., 2007. Susceptibility of common andrare plant species to the genetic consequences of habitatfragmentation. Conservation Biology 21, 823–831.

Hooftman, D.A.P., Diemer, M., 2002. Effects of small habitat sizeand isolation on the population structure of common wetlandspecies. Plant Biology 4, 720–728.

Jaehn, F., Roux, J., 1986. Architecture and annual cycle of theAutumn crocus (Colchicum autumnale L.). Bulletin de laSociete Botanique de France-Lettres Botaniques 133,225–233.

Jenness, J., 2007. Nearest Features. Jenness Enterprises.Jennersten, O., 1988. Pollination in Dianthus deltoides

(Caryophyllaceae) – effects of habitat fragmentation onvisitation and seed set. Conservation Biology 2, 359–366.

Jennings, S.B., Brown, N.D., Sheil, D., 1999. Assessing forestcanopies and understorey illumination: canopy closure,canopy cover and other measures. Forestry 72, 59–73.

Lienert, J., Diemer, M., Schmid, B., 2002. Effects of habitatfragmentation on population structure and fitnesscomponents of the wetland specialist Swertia perennisL. (Gentianaceae). Basic and Applied Ecology 3, 101–114.

Lennartsson, T., 2002. Extinction thresholds and disrupted plant–pollinator interactions in fragmented plant populations.Ecology 83, 3060–3072.

Maes, D., Van Dyck, H., 2001. Butterfly diversity loss in Flanders(North Belgium): Europe’s worst case scenario? BiologicalConservation 99, 263–276.

Matthies, D., Brauer, I., Maibom, W., Tscharntke, T., 2004.Population size and the risk of local extinction: empiricalevidence from rare plants. Oikos 105, 481–488.

McCune, B., Keon, D., 2002. Equations for potential annual directincident radiation and heat load. Journal of Vegetation Science13, 603–606.

Menges, E.S., 1998. Evaluating extinction risks in plantpopulations. In: Fiedler, P.L., Kareiva, P.M. (Eds.), ConservationBiology for the Coming Decade. Chapman and Hall, New York,pp. 49–65.

Menges, E.S., Dolan, R.W., 1998. Demographic viability ofpopulations of Silene regia in midwestern prairies:relationships with fire management, genetic variation,geographic location, population size and isolation. Journal ofEcology 86, 63–78.

Mroz, L., 2006. Variation in stage structure and fitness traitsbetween road verge and meadow populations of Colchicumautumnale (Liliaceae): effects of habitat quality. Acta SocietatisBotanicorum 75, 69–78.

Munzbergova, Z., 2006. Effect of population size on the prospectof species survival. Folia Geobotanica 41, 137–150.

Mustajarvi, K., Siikamaki, P., Rytkonen, S., Lammi, A., 2001.Consequences of plant population size and density for plant–pollinator interactions and plant performance. Journal ofEcology 89, 80–87.

Nault, A., Gagnon, D., 1993. Ramet demography of Alliumtricoccum, a spring ephemeral, perennial forest herb. Journal ofEcology 81, 101–119.

Newman, D., Pilson, D., 1997. Increased probability of extinction dueto decreased genetic effective population size: experimentalpopulations of Clarkia pulchella. Evolution 51, 354–362.

Oostermeijer, J.G.B., Vaneijck, M.W., Dennijs, J.C.M., 1994.Offspring fitness in relation to population size and geneticvariation in the rare perennial plant species Gentianapneumonanthe (Gentianaceae). Oecologia 97, 289–296.

Oostermeijer, J.G.B., Brugman, M.L., De Boer, E.R., den Nijs, J.C.M.,1996. Temporal and spatial variation in the demography ofGentiana pneumonanthe, a rare perennial herb. Journal ofEcology 84, 153–166.

Persson, K., 1993. Reproductive strategies and evolution inColchicum. In: Proceedings of the Fifth OPTIMA Meeting,Istanbul, 8–15 September 1986, pp. 397–414.

Poutaraud, A., Champay, N., 1995. Le Colchique: une plantemedicinale a domestiquer. Revue Suisse d’Agriculture 27, 93–100.

a c t a o e c o l o g i c a 3 5 ( 2 0 0 9 ) 6 9 – 8 282

Poutaraud, A., Girardin, P., 2006. Agronomical and chemicalvariability of Colchicum autumnale accessions. CanadianJournal of Plant Science 86, 547–555.

Poutaraud, A., Girardin, P., 2005. Influence of chemicalcharacteristics of soil on mineral and alkaloid seed contents ofColchicum autumnale. Environmental and Experimental Botany54, 101–108.

Poutaraud, A., Girardin, P., 2003. Seed yield and components ofalkaloid of Meadow saffron (Colchicum autumnale) in naturalgrassland and under cultivation. Canadian Journal of PlantScience 83, 23–29.

Richards, C.M., 2000. Inbreeding depression and genetic rescue ina plant metapopulation. American Naturalist 155, 383–394.

SPSS, 2005. SPSS 13.0 for Windows. SPSS Inc., Chicago.

Thompson, K., Bakker, J.P., Bekker, R.M., 1997. The Soil SeedBanks of North-West Europe: Methodology, Density andLongevity. Cambridge University Press, Cambridge.

Tilman, D., May, R.M., Lehman, C.L., Nowak, M.A., 1994.Habitat destruction and the extinction debt. Nature 371,65–66.

Vellend, M., Verheyen, K., Jacquemyn, H., Kolb, A., Van Calster, H.,Peterken, G., Hermy, M., 2006. Extinction debt of forest plantspersists for more than a century following habitatfragmentation. Ecology 87, 542–548.

Wenzel, M., Schmitt, T., Weitzel, M., Seitz, A., 2006. The severedecline of butterflies on western German calcareousgrasslands during the last 30 years: a conservation problem.Biological Conservation 128, 542–552.