Effect of spatial scale on factors limiting species distributions in dry grassland fragments

Authors

  • ZUZANA MÜNZBERGOVÁ

    Corresponding author
    1. Institute of Botany, Academy of Sciences of the Czech Republic, CZ-252 43 Prùhonice, Czech Republic, and Department of Botany, Faculty of Science, Charles University, Benátská 2, CZ-12801 Praha 2, Czech Republic
    Search for more papers by this author

Z. Münzbergová (fax +42 0267 750031, e-mail zuzmun@natur.cuni.cz).

Summary

  • 1Distribution of plant species in fragmented landscapes is the result of both seed and site availability. Little is known about how their relative importance differs at different spatial scales.
  • 2I sowed seeds of eight dry grassland species into 22 localities that differed in occupancy by these species and followed seedling establishment over 3 years. I compared the number of emerging seedlings at three scales: between previously occupied and previously unoccupied localities, between occupied and unoccupied blocks within occupied localities, and between plots with and without seed addition within occupied blocks.
  • 3At the two larger scales, I also studied the relationship of the number of seedlings and of the distributions of adult plants to environmental factors.
  • 4Both seed and site availability are important in structuring the distribution of these species, but site availability becomes less important with increasing spatial scale. The intensity of this effect is, however, species specific.
  • 5The relationship between environmental factors and pattern of species distribution is also clearly scale dependent, and differs between seedlings and adults. Whilst abiotic factors are the main determinants of seedling distribution, the occurrence of adult individuals is best predicted by the occurrence of other species. This suggests that the present distribution of species in the landscape is determined mainly by historical factors.
  • 6Conclusions based on the importance of seed and site availability for species distribution in natural communities at one scale cannot be extrapolated to other scales. Only comparisons on multiple scales can provide a full understanding of factors affecting species distribution at landscape level.

Introduction

Patterns of species distributions are determined by the balance between colonization and extinction (MacArthur & Wilson 1967; Grubb 1977; Hanski 1997; Tilman 1997). The ability to colonize a site depends on both a species’ ability to reach the site and its ability to establish there, leading to alternate hypotheses. One of them assumes that communities are not saturated with species and that diversity is limited mainly by availability of seeds, and thus by dispersal traits (Eriksson & Ehrlén 1992; van der Meijden et al. 1992; Ackerman et al. 1996; Eriksson 1996; Hanski 1997; Ehrlén & Eriksson 2000; Foster & Tilman 2003). The other assumes that communities are potentially saturated so that local biotic interactions and site availability, reflected in traits related to establishment, are predominant (Grubb 1977; Brown 1984; van der Maarel & Sykes 1993; Foster 2001).

Species distributions in a landscape are the result of processes operating at both local and regional scales, with availability of seeds and sites for germination playing a role at both levels (e.g. local-scale seed production, seed predation and density dependence will interact with seed dispersal ability and habitat availability at the regional scale). Because there is no clear link between the different scales, conclusions about the relative importance of seed and site availability are expected to be scale dependent. Recently an increasing number of studies have used sowing experiments to determine the relative importance of seed and site availability for single species distributions (e.g. Houle & Phillips 1989; Eriksson & Ehrlén 1992; van der Meijden et al. 1992; Ackerman et al. 1996; Ehrlén & Eriksson 2000; Turnbull et al. 2000; Foster & Tilman 2003). Most studies have, however, been carried out at a single scale, and the relative importance of different types of limitations at different spatial scales is still largely unknown (Ehrlén & Eriksson 2000).

The importance of site availability implies that species distributions should depend on how environmental conditions vary with locality. The ultimate aim of many sowing experiments is therefore to identify environmental factors that affect the germination success of a species (e.g. Burke & Grime 1996; Ehrlén & Eriksson 2000; Franzén 2002; Stevens et al. 2004). As well as explaining distribution patterns, such experiments allow identification of suitable, but unoccupied, habitats for use in metapopulation studies. Identification of such key factors is, however, not straightforward. First, it has been repeatedly shown that habitat requirements of seedlings can differ from those of adult plants, and that critical phases may appear later in development (Losos 1995; Ehrlén & Eriksson 2000; Gustafsson et al. 2002). Only combined information on seedlings and adults will capture the full effect of the factors concerned. Secondly, many plant species are able to modify their environments and differences in abiotic and biotic factors between occupied and unoccupied habitats may therefore be a consequence, rather than the cause, of any difference (e.g. Charley & West 1975; Franco-Pinaza et al. 1996; Münzbergová & Ward 2002). Even though there have been many attempts to relate either seedling numbers or adult distribution to environmental variables (e.g. Ehrlén & Eriksson 2000; Verheyen & Hermy 2001), the relationship between the two stages is often unknown (but see Howard & Goldberg 2001; Jutila 2003) and may depend on spatial scale.

The objective of this study is to answer three questions. (i) How important is availability of (a) seeds and (b) sites for species distribution at the three spatial scales (within a patch, between patches within occupied localities and between localities)? (ii) How does the relationship between seedling numbers and environmental factors vary with spatial scale? (iii) Does the relationship observed for established plants at each scale differ from that for seedlings?

I performed a sowing experiment and followed seedling recruitment of eight dry grassland species at 22 localities over 3 years. Recruitment success was compared at three spatial scales (between localities occupied and unoccupied by focal species, between blocks occupied and unoccupied by focal species within occupied localities, and between plots with and without seed addition within occupied blocks). Seedling recruitment and adult distribution were also related to environmental parameters and actual species occurrence at the block and locality scales.

Dry chalk grasslands have been the subject of many recent studies on the importance of site and seed availability for species diversity on the local scale (e.g. Hutchings & Stewart 2002; Kalamees & Zobel 2002; Tofts & Silvertown 2002). Assessing their relative importance also at the regional scale can thus contribute not only to our knowledge of scale dependency in general, but also to specific conservation issues for these habitats.

Methods

site description

This study was carried out in northern Bohemia, Czech Republic, where the geology and geomorphology (usually steep slopes on foothills) results in low water permeability and continuous erosion. Chalk grasslands have developed on Tertiary marine sediments and occur as distinct localities surrounded mainly by agricultural fields. All the localities studied have been abandoned for at least the last 30 years and occur within a region of approximately 870 km2 (the distance between the two most distant localities was 41 km, Table 1).

Table 1.  List of localities used in the sowing experiment. Table provides geographical position of the localities, and estimates of number of flowering individuals at occupied localities. Unoccupied localities are marked with zero. Management regime in 1952 was determined from old maps. Division into communities follows unpublished data of M. Chytrý available at http://www.sci.muni.cz/botany/chytry/mapy-veg-cr
LatitudeLongitudeAnemoneAsterCirsiumCoronillaGlubulariaL. flavumL. tenuifoliumScorzoneraManagement in 1952Community
50°32′22.8″14°18′48.1″  0   0  0   01850   0330   0Field, pastureScabioso ochroleuceaBrachypodietum pinnati
50°32′25.8″14°19′9.3″  0   0  3   0   5  20  0   0Field 
50°31′46.6″14°14′13.9″ 43   0520   0   0   0  01630Gardens 
50°33′26.6″14°8′40.6″  0 145 80   5   0   0  0  54Pasture 
50°33′10.4″14°10′16.4″  0   0  0   0   0 280  0 387Pasture 
50°33′54.7″14°20′55.8″  3   0  0   0   0   0  0   0Pasture 
50°31′42.7″14°20′40.8″  0 180  0   0  40  40  0   0Field, pasture 
50°24′39.4″13°56′41.4″  0   0  0   0   0   0  0   0PastureAssociation with Bromus erectus
50°30′25.4″14°19′20″  0 290  0   0   0   0670   0Pasture 
50°23′24.5″14°5′9.8″650   0  0   0   0   0  0   0Orchad 
50°23′55.8″14°10′42.8″  0   0  0   5   0   0  0   0Field, pasture 
50°25′47.6″13°52′58.6″  0   0  0  30   0   0 80   0Pasture 
50°31′56.8″13°48′3″ 20 230 80   0   0 350  0  12Pasture 
50°23′28.4″14°9′8.1″ 30   0  0 560   0   0  0   0Field, pasture 
50°23′28.3″14°3′36.5″ 12   0  0  55  80   0  0   2PastureCirsio pannoniciSeslerietum albicantis
50°33′0.4″14°13′59.1″ 14   0  0  49   0   0  0   0Pasture 
50°31′39.1″14°19′40.3″420   1  0   0 750   0540   0Field, pasture 
50°31′50.8″14°15′36.4″  0 550  01100   0   0730   0Pasture 
50°29′45.9″13°58′39.6″  0  66  0   0   0   0340  48Pasture 
50°32′34.8″14°5′19.7″1203450960140014006000  01980Pasture 
50°28′19.1″13°52′42.8″ 80   0  0  40   0   0  0  76Pasture 
50°28′1.7″14°18′37.7″  0 220  0  95   0   0420   0Vineyard 
No. of occupied localities 10   9  5  10   5   5  7   8  

study species

Eight species were selected to meet the following criteria: (i) they are restricted to dry chalk grasslands, i.e. they do not occur in any other habitat type in the study region; (ii) they are neither very common (so suitable unoccupied localities are likely), nor rare (so that it was possible to collect enough seeds for the experiment without affecting the source populations). Anemone sylvestris, Aster amellus, Coronilla vaginalis, Cirsium pannonicum, Globularia punctata, Linum flavum, Linum tenuifolium and Scorzonera hispanica, all perennial forbs, were selected (Table 2). Nomenclature follows Tutin et al. (1964–83).

Table 2.  Main characteristics of the studied species and the importance of different types of limitations for species distribution at three spatial scales based on results of this study. D indicates dispersal limitation, S indicates seed limitation, H indicates habitat limitation, M indicates microsite limitation and Other indicates limitation by other factors. Last row contains data on proportion of occupied localities from all suitable localities. Suitable locality is defined as locality with at least one seedling surviving in the third year of the experiment
SpeciesAnemone sylvestrisAster amellusCirsium pannonicumCoronilla vaginalisGlobularia punctataLinum flavumLinum tenuifoliumScorzonera hispanica
Descriptive parameters
 FamilyRanunculaceaeAsteraceaeAsteraceaeFabaceaeGlobulariaceaeLinaceaeLinaceaeAsteraceae
 Main dispersal agentWindWindWindNoNoNoNoWind
 Clonal propagationYesYesYesNoYesNoNoRarely
 Seed germination ability (%)21241936458065100
 Decrease in germination after 3 years9058100192210098100
 Scarification requiredNoNoNoYesNoNoNoNo
 Terminal velocity (m/s)0.670.800.422.401.351.911.501.78
 Attachment ability (%)977976558302130
 Seed weight (mg)0.50.61.63.80.20.80.68.0
 Seed production per 1 m217361866141425386810 510692189
Results
 Locality scaleNo germinationDDD, H?D, H?D, otherDD, H?
 Block scalein the fieldDDH, DD, H?D, otherDD
 W/out seed addition SSMMMSS
 Proportional habitat occupancy 0.410.230.480.230.240.350.32

sowing experiments

The sowing experiment was established at 22 different localities in the study area (Table 1). The localities were selected subjectively to cover the whole range of dry chalk grassland types that could be found in the region (Table 1).

At each locality I subjectively selected five distinct patches of homogenous vegetation and established one block in each of these. Each block consisted of four 1-m2 plots, separated by 1 m, that were divided into nine subplots 0.33 m × 0.33 m for a total of 36 subplots in each block. These subplots were randomly assigned to nine different treatments (sowing of one of the eight species or control) to yield four subplots per species treatment per block, half of which were planted in 2000, and the remainder in 2001.

Seeds were collected from two source populations for each species (one for Coronilla and Linum flavum) and each was used for one of the two replicates within each block in each year. I sowed 100 seeds (50 seeds for Coronilla and Scorzonera) in undisturbed vegetation in each subplot in late September of 2000 or 2001. Germination and survival of seedlings were followed yearly in late May/early June from 2001 to 2003. At each census, I counted the number of seedlings of the target species in each subplot, as well as the number of seedlings of the other study species (these values served as controls). The number of seedlings of all studied species was also counted in the control subplots. Controls at the unoccupied localities enabled estimation of possible secondary dispersal of seeds between the subplots, but none was detected for any of the species.

environmental parameters

I recorded a set of parameters to capture the most important differences between localities and blocks. They were also expected to serve as correlates of past management regimes of the localities, because they can be both its cause (e.g. vineyards are often on south or south-west facing slopes) and its consequence (altered productivity and physical soil properties).

Parameters at the block scale were above-ground biomass collected in 15 × 15 cm plots placed just outside the sowing plots at peak standing crop, inclination, aspect, soil water holding capacity, bulk density, species composition and percentage of bare ground. Each parameter was measured four times per block, once for each of the four 1-m2 plots, and block means were used in analyses. Data on inclination and aspect were used to calculate potential direct solar radiation. This was done by summation of cosines of angles between the sun and the locality surface over the whole day at 15-minute intervals. The calculation was done on the twenty-first day of each month between December and June. As some of these values are highly correlated, only data on direct radiation in January and June, and the average over the whole period, were used. These data provide summary information on irradiation and consequently temperature and moisture throughout the year (Zimmermann & Kienast 1999). Soil samples of 100 cm3 were collected and bulk density was estimated from the weight of a given volume of soil dried at 105 °C. Soil water holding capacity was measured as the amount of water bound in the soil monolith after standing on a constantly wet filter paper for 24 hours (amount of water retained per gram of dry soil).

Parameters at the locality scale included presence of the target species and total species composition of the locality. I also used mean of block level parameters to describe differences between localities.

Species composition of naturally occurring adult individuals was estimated as cover of single species visually assigned to the Braun-Blaquet cover scale. Data on species composition of both the localities and the blocks were used to derive Beals’ index of sociological favourability (Beals 1984; further called Beals index value, see Münzbergová & Herben 2004 for its use). This index calculates the probability of encountering a species at a locality using data on the presence of other species at that locality and information on patterns of co-occurrence of the target species with other species. To estimate the species co-occurrence patterns, I used 2984 relevés on species composition of dry grasslands in the Czech Republic from the Czech national phytosociological data base (Chytrý & Rafajová 2003). Data on species composition at both spatial scales were also summarized with detrended correspondence analysis (DCA) using the program canoco (ter Braak & Šmilauer 1998). DCA is a multivariate technique assuming bell-shaped species distributions along an underlying environmental gradient that enables one to extract a few ordination axes that summarize maximum variation in species composition. Species composition summarized as the first and second ordination axes (subsequently referred to as AX1 and AX2) from this analysis were then used as additional environmental variables. AX1 and AX2 scores were calculated for each focal species separately, with that species excluded from the analysed matrix.

species parameters

To be able to interpret the observed differences between species, I collected data on life-history traits related to seed production, seed germination ability, seed bank and seed dispersal. To estimate a species ability to survive in the seed bank, 10 nylon bags, each containing 100 seeds, were buried at five places in three localities in late September 2000. Two bags from each place were excavated in late September 2002 and 2003 and the seeds were tested for viability. The seeds were regularly watered with distilled water on Petri dishes and kept in a growth chamber under a fluctuating regime (12 hours light at 20 °C, 12 hours dark at 10 °C). Germinated seeds were regularly removed. The dishes were kept until all the seeds germinated or decayed (approximately 6 months). The same procedure was used to estimate the viability of fresh seeds to provide a baseline from which to estimate the decline in germination over time. Seed weight was estimated by weighing 10 groups of 10 seeds from three source populations. Seed production was estimated as number of flowering plants m−2 counted in five quadrats in each of three populations of the species, and multiplying it by seed production per plant estimated in these three populations.

To estimate species dispersal ability, I estimated each species’ terminal velocity and attachment ability. Data on terminal velocity for Aster, Coronilla, Globularia and L. tenuifolium were taken from the DIASPORUS data base (Bonn et al. 2000) and measured for other species using the method described in Tackenberg et al. (2003). Attachment ability, used as an estimate of the ability to disperse via exozoochory, was assessed by gently placing a piece of sheep fur over a tray containing 100 seeds, removing it and counting the number of attached seeds (see Fischer et al. 1996).

data analysis

To take into account seedlings occurring naturally at the previously occupied blocks, the average number of seedlings of target species in all subplots within a block where the target species was not sown was subtracted from the number of seedlings found in the subplot with that species added. All the analyses were done both with and without this correction, but there was little difference in the overall patterns and only results without the subtraction are shown. There was also little difference between results of the experiments started in the two subsequent years, so only results for the first year are presented.

Tests including seedling numbers were done using a generalized linear model with Poisson distribution; tests evaluating adult occurrence used a binomial distribution model. All tests were performed using S-PLUS (2000).

factors limiting species distribution

Results of the sowing experiment in relationship to habitat occupancy were evaluated by comparing the number of seedlings at three spatial scales: (i) at occupied and unoccupied localities (using sum of all seedlings found in all plots at each locality as dependent variable); (ii) at occupied and unoccupied blocks within occupied localities; and (iii) at subplots with and without seed addition (controls) within occupied blocks. In both comparisons (i) and (ii), an equal number of seedlings in the two groups was considered an indication of dispersal limitation at that spatial scale. A significantly lower but non-zero number of seedlings in the ‘unoccupied’ variant in each pair was considered an indication that both dispersal and habitat limitation occurred. Absence of recruitment only in the unoccupied variant was considered an indication of habitat limitation. An equal number of seedlings in the two groups in the third comparison was considered an indication of microsite limitation, while a higher number of seedlings in plots with seeds added was considered an indication of seed limitation.

terminology

There is little consistency in use of the terms seed, dispersal, microsite and habitat limitation in the literature (Münzbergová & Herben, unpublished observations). Here I use seed limitation to describe an increase in number of seedlings after extra seeds were added to places where the species occurs (sensuEriksson & Ehrlén 1992) and microsite limitation to describe an absence of any increase in the number of seedlings at occupied places after seed addition (sensuGrieshop & Nowierski 2002). I use dispersal and habitat limitation as alternative terms to seed and microsite limitation defined at the regional scale. A species is dispersal limited if it can establish at sites following seed addition where it does not occur (sensuEriksson 1998) and habitat limited if it cannot (sensuMulligan & Gignac 2001).

The terminology used here strictly relies on a distinction between local and regional scales. Local scale in this definition refers to a homogeneous area occupied by a continuous population of the species of interest and assumes that the dispersal curve is flat over that area. Regional scale refers to distinct patches at least partly isolated from other patches (dispersal curve is not flat). It can refer to both distinct localities and distinct patches within one locality.

effect of environmental variables

Tests of the effect of the environmental variables were carried out separately for seedlings and naturally occurring adult plants (further referred to as adults) and at the block and locality scales. To adjust for reduced number of degrees of freedom in tests at the locality scale, the sum of all seedlings or frequency of adults was calculated for each locality and used as the dependent variable. To test the effect of environmental variables I performed a stepwise regression combining back and forward approaches, determined the optimal model and recorded the percentage of total variance explained by that model. This was done for all the environmental parameters together, as well as separately for biotic (Beals index value, AX1, AX2) and abiotic (all the others) parameters. While the percentage of variance explained by the model was not sensitive to definition of the original model, the exact formulation of the optimal model was (cf. Graham 2003). Comparison of the single factors included in the optimal model between the two scales and between seedlings and adults could thus be misleading, and I therefore also tested the significance of single factors. This was done using log-likelihood ratio estimates by comparing a baseline model and a model including the term of interest. The baseline model included seed source and, for the test of differences in environmental variables between blocks, it also included the locality code. To correct for the high number of tests on the same data the significance levels of these tests were adjusted using the Bonferroni correction.

To estimate the consistency of the effect of single environmental variables I defined two types of contrasts, viz. contrast of significance of environmental variables between the two scales separately for seedlings and adults and between seedlings and adults separately for the two scales. Then I calculated the Jaccard similarity coefficient for each contrast by comparing the number of environmental variables having a significant effect in both tests of that contrast and the total number of significant tests in that contrast.

The relationships between proportional habitat occupancy and plant traits were tested using linear regression, with single species as data points. Because there were only eight species in this study the test is rather weak and the results should be taken only as an indication of possible patterns that need further exploration.

Results

All species except Anemone germinated and established successfully in at least some localities. A total of 2437 seedlings of Aster, 1893 seedlings of Cirsium, 200 seedlings of Coronilla, 3546 seedlings of Globularia, 6262 seedlings of L. flavum, 1755 seedlings of L. tenuifolium and 953 seedlings of Scorzonera germinated in the first year of the experiment. Seedlings of several species formed flowers, with 31 flowering individuals of L. tenuifolium in the second and 29 in the third year, 2 and 20 for Coronilla, and 1 Cirsium flowering in the third year.

factors limiting species distribution

Three different types of pattern of seedling numbers were distinguished at the locality scale. (i) In Aster and Cirsium there was no difference in number of seedlings between occupied and unoccupied localities in any of the 3 years (Fig. 1), indicating dispersal limitation. (ii) In Coronilla, Globularia, L. tenuifolium and Scorzonera there was a higher number of seedlings at the occupied localities in the first years of the experiment, but the difference disappeared in all of them in the last year (Fig. 1), indicating that, overall, dispersal was the most important limitation. (iii) In L. flavum there was a higher number of seedlings at the unoccupied than at the occupied localities in the second and third years (Fig. 1), suggesting that some other factor determined species distributions.

Figure 1.

Seedling numbers at occupied and unoccupied localities over the 3 years of the sowing experiment. Graph shows mean ± 1 SD. Asterisks above points indicate significant differences between occupied and unoccupied localities in that species and year. Significance was estimated using a generalized linear model with Poisson distribution separately for each species, and using sum of seedlings in all plots within one locality as the dependent variable. Each test was based on 22 observations.

Four different types of pattern of seedling numbers were distinguished in the comparison between blocks within occupied localities. (i) Occupied blocks had higher number of seedlings than unoccupied blocks in all 3 years for Coronilla (Fig. 2), indicating a combination of dispersal and habitat limitation. (ii) Occupied blocks had higher number of seedlings than unoccupied blocks but only in the first year for Globularia (Fig. 2), suggesting that dispersal was probably the more important of the two types of limitation. (iii) Occupied blocks had lower number of seedlings than unoccupied blocks for L. flavum in the second and third year (Fig. 2), probably a result of some other type of limitation. (iv) Finally, there was no difference between occupied and unoccupied blocks in the other species (Fig. 2), indicating dispersal limitation.

Figure 2.

Seedling numbers at occupied and unoccupied blocks within occupied localities over the 3 years of the sowing experiment. Graph shows mean ± 1 SD. Asterisks above points indicate significant differences between occupied and unoccupied blocks in that species and year. Significance was estimated using a generalized linear model with Poisson distribution separately for each species and using number of seedlings per plot as the dependent variable. There were 90 observations for Aster, 50 for Cirsium, 100 for Coronilla, 50 for Globularia, 50 for L. flavum, 70 for L. tenuifolium and 70 for Scorzonera.

Two different types of patterns could be distinguished in the comparison between subplots with seeds added and without seed addition within occupied blocks. (i) Higher number of seedlings could be found at subplots with seeds added than at subplots without seed addition in Aster, Cirsium, L. tenuifolium and Scorzonera (Fig. 3), indicating seed limitation. (ii) No differences occurred for the numbers of seedlings between subplots in Coronilla, Globularia and L. flavum (Fig. 3), indicating microsite limitation.

Figure 3.

Seedling numbers at plots with seeds added and without seed addition (controls) within occupied blocks over the 3 years of the sowing experiment. Graph shows mean ± 1 SD. Asterisks above points indicate significant differences between plots with seeds added and without seed addition within occupied blocks. Significance was estimated using a generalized linear model with Poisson distribution separately for each species and using number of seedlings per plot as the dependent variable. There were 22 observations for Aster, 14 for Cirsium, 36 for Coronilla, 26 for Globularia, 20 for L. flavum, 24 for L. tenuifolium and 34 for Scorzonera.

effect of environmental variables

The percentage of total variation that was explained by environmental factors was higher for the naturally occurring adult distribution than for seedlings (Table 3). This difference was especially strong at the block scale. While abiotic parameters generally contributed more to variance in the number of seedlings, adult species distributions were much more closely related to biotic parameters (Table 3). There was no consistent trend in the correlations of seedlings with environmental variables over time, with patterns decreasing, increasing or peaking in the second year and showing scale dependence (Table 3).

Table 3.  Percentage variance in seedling numbers in the first, second and third year of the experiment and in adult occurrence explained by two groups of environmental variables, abiotic parameters and parameters describing species composition. Total denotes total variation explained by both of these groups of parameters together. The increase of this value compared with the sum of the two previous columns is due to variance that could possibly be explained by both of the two factors. Variance decomposition at the block scale is based on 220 observations; variance decomposition at the locality scale is based on 22 observations. Bold values mark the higher value of the two neighbouring columns showing the contribution of species composition and abiotic parameters
ScaleSpeciesYear 1Year 2Year 3Adults
Species compositionAbiotic parametersTotalSpecies compositionAbiotic parametersTotalSpecies compositionAbiotic parametersTotalSpecies compositionAbiotic parametersTotal
BlockAster0 1 4 0 0 1 0 2 435 045
Cirsium0 1 2 0 1 1 0 0 035 060
Coronilla0 3 3 0 7 7 0 3 4 02468
Globularia1 810 01617 01214 1 033
L. flavum0 0 2 01215 01214252364
L. tenuifolium01619 41829 02626 0 2 7
Scorzonera0 1 1 0 0 0 0 0 043 070
LocalityAster22533 21517 0202210 212
Cirsium21820 02829 03132 9 242
Coronilla01214121027 63149 2 735
Globularia33541 31825 02828 3 074
L. flavum02929 14649 0434360 060
L. tenuifolium22934 05761 04879 01017
Scorzonera5141910112113 821 7 232

There was little consistency in the significance of single environmental factors between seedlings and adults, and between the two scales (Table 4). The consistency was lowest for the seedling vs. adult comparison at the locality scale, and it was highest between the scales for adults (Table 4).

Table 4.  Correspondence between importance of single environmental factors for explaining the number of seedlings in the third year of the experiment and adult occurrence at block and locality scale. Jaccard is a value of the Jaccard coefficient comparing significances of single factors between seedlings and adults at the block and locality scales, and between the scales for seedlings and adults. n gives number of significant tests on which the comparison is based. The comparison is based on all 11 environmental variables
 Block scaleLocality scaleSeedlingsAdults
NJaccardNJaccardNJaccardNJaccard
Aster 70.57 40.00 70.5740.50
Cirsium 40.00 50.00 40.0050.80
Coronilla 30.00100.00120.3310.00
Globularia100.40 60.00120.6740.50
L. flavum 40.50 80.00100.4021.00
L. tenuifolium 40.00 60.00100.800
Scorzonera 20.00 30.00 30.0020.00

Overall, the Beals index value was the most successful factor explaining number of seedlings and adult occurrence at the two scales; it was significant in 46% of all comparisons. It was followed by percentage of bare ground (significant in 39% of all comparisons), bulk density (36%), biomass (32%), soil water holding capacity (29%), vegetation composition on AX1 (21%), vegetation composition on AX2 (18%), potential solar direct radiation in June (18%), mean potential solar radiation (14%), inclination (11%) and potential solar radiation in January (7%) (Appendices 1 and 2).

locality occupancy and species traits

Actual locality occupancy, based on the results of the sowing experiment, varied between 23 and 48%. There was no relationship between this proportion and any of the plant traits measured (for terminal velocity F1,5 = 0.97, P = 0.37, for attachment ability F1,5 = 0.95, P = 0.38, for seed weight F1,5 = 0.32, P = 0.59, for seed production F1,5 = 1.28, P = 0.31, see Table 1 for absolute values).

Discussion

factors limiting species distribution

Seven out of eight species used in this study recruited well both at occupied and previously unoccupied patches within occupied localities as well as at the unoccupied localities. The patterns of recruitment, however, differed markedly between the three spatial scales as well as between species. Hence any conclusion about the relative importance of different types of limitations for species distribution is scale as well as species dependent.

At the locality scale, only Aster and Cirsium showed unequivocal evidence for dispersal limitation (no difference between occupied and unoccupied localities) but, surprisingly, these are the two species with the best wind dispersal ability. Coronilla, Globularia, L. tenuifolium and Scorzonera show this pattern only after the first year. An alternative explanation for dispersal limitation for these species is that occupied localities are more favourable and allow faster development, so that seedlings reach a stage where they suffer high post-establishment mortality, whereas the less developed seedlings at unoccupied localities never reach this phase. However, individuals often differ markedly in size between localities, and size does not seem to be a good indicator of plant maturity in this system. The higher recruitment at unoccupied than at some occupied localities in L. flavum was similar to the pattern found by Ehrlén & Eriksson (2000) for Lathyrus vernus. It may be due either to the presence of specific pathogens or predators only at the occupied localities, or a correlated change of environmental conditions following establishment of the species at the occupied localities. In the latter case adult distribution would reflect past, but not present, habitat suitability (also see below).

At the block scale, recruitment was predominantly dispersal limited, with clear evidence for Aster, Cirsium, L. tenuifolium and Scorzonera and evidence for both habitat and dispersal limitation in Coronilla. In Globularia, a higher number of seedlings was found at occupied blocks only in the first year, which suggests that dispersal limitation replaces a combination of dispersal and habitat limitation over time. In L. flavum there was, as at the locality scale, a higher number of seedlings in the unoccupied than in the occupied blocks.

Within occupied blocks, seed addition enhanced recruitment in over half (four out of seven) the species, indicating seed limitation. This is congruent with findings in many other studies (e.g. Cavers & Harper 1967; Eriksson & Ehrlén 1992; Clark et al. 1998; Ehrlén & Eriksson 2000; Jakobsson & Eriksson 2000), although the remaining three showed the much rarer lack of an effect. In Coronilla 3 years may not have been long enough to break seed dormancy, so that most of the extra seeds may germinate later (Thompson et al. 1997), but no dormancy was observed in Globularia and L. flavum, and therefore the pattern can be regarded as true microsite limitation.

Comparison between different scales demonstrates that conclusions about importance of limitation by seed and site availability for species distribution depend on the spatial scale used, with limitation by site availability becoming more important with decreasing spatial scale. There was, however, also a strong interaction between scale and species. Although the seven different species studied have a similar ecology and patterns of distribution there was high variation in the underlying mechanisms. This weakens the predictive power of patterns of distribution for the underlying processes (e.g. Quintana-Ascencio & Menges 1996; Bastin & Thomas 1999; Dupré & Ehrlén 2002).

effect of environmental variables

Different environmental factors were important for distribution of seedlings and adult plants at both locality and block scales. This discrepancy may be due to the fact that: (i) small and large plants are affected by different factors, or selective forces appear later in species development (Losos 1995; Ehrlén & Eriksson 2000; Gustafsson et al. 2002); (ii) adult plants may modify their environment (Charley & West 1975; Franco-Pinaza et al. 1996); or (iii) adult plant distribution may reflect random dispersal events in the past, or past landscape structure.

The higher variation explained by environmental variables for adults than for seedlings supports explanations (i) and (ii) over explanation (iii). It is generally thought that success of all seedlings depends on simple abiotic parameters, such as water balance, while survival of adults is determined by more complex factors that can be captured in the species composition. This would predict an increase in the importance of biotic parameters over time for seedlings. The reverse is, however, true for several species, suggesting that the species composition of the sites may be the result of past events or of environmental conditions that no longer prevail, but which affected establishment of all species (Eriksson 1996) (i.e. explanation (iii)).

Studies that relate germination success at the locality scale to environmental variables usually conclude that no such factors can be identified (e.g. Eriksson & Ehrlén 1992; Ehrlén & Eriksson 2000), whereas studies testing the effect of environmental factors within localities usually succeed (e.g. Kelly 1989; Peart 1989; Robinson et al. 1995; Tilman 1997; Franzén 2002; Jutila & Grace 2002; Xiong et al. 2003). The higher success in this study in explaining patterns between, rather than within, localities may be due to a much wider range of distances between localities than in many such comparisons. All the localities are, however, relatively small and homogeneous so that within-locality differences are rather small. This discrepancy demonstrates that the relative importance of different types of limitations and of environmental variables is dependent on the selection of localities for the sowing experiment. Larger variation among localities would increase importance of habitat limitation as well as the importance of environmental variables. Similar patterns would have held if the localities used in the study had been more heterogeneous. Scale dependency is usually neglected, but this unavoidable variation must be considered when interpreting the results. Most published studies of this type, however, do not explicitly state what spatial scale was used and consider its relevance for the question being studied (e.g. Ehrlén & Eriksson 2000; Gustafsson et al. 2002).

Acknowledgements

I would like to thank Tomáš Herben, Johan Ehrlén, Ove Eriksson, Deborah Goldberg, Paul Foster, Vigdis Vandvik, members of the Plant Ecology Discussion Group at the Department of Ecology and Evolutionary Biology, University of Michigan, three anonymous referees and David Gibson for critical comments on the manuscript, TomášČerný, Michal Ducháèek and Katka Scharffová for help in the field, Tomáš Herben for providing the program to calculate potential direct solar radiation, Daniela Münzbergová, Paul Foster and especially Lindsay Haddon for language revision and Jana Kubcová for help in the laboratory. This study was supported by grants GAČR no. 206/02/05, GAUK no. 156/201 and by doctoral grant GACČR no. 206/03/H137.

Appendices

Appendix 1

Table 5.  Results of analysis of the effect of selected variables on number of seedlings at block and locality scales in the third year of the experiment. Table shows significance values from log-likelihood tests comparing the baseline model (model including source of the seeds at both scales and locality code at the block scale) with the model including also the term of interest. Significant values are in bold (significance level is adjusted using Bonferoni correction). If the direction of the response makes sense, the sign of the relationship, if significant, is given in brackets. Results at the block scale are based on 220 observations; results at the locality scale are based on 22 observations
VariableAsterCirsiumCoronillaGlobulariaL. flavumL. tenuifoliumScorzonera
BlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocality
Beals index value0.0400.0140.1700.7280.025< 0.001< 0.0010.275< 0.0010.2120.2580.0080.452< 0.001
     (+)(+) (–)    (+)
AX10.025< 0.0010.8300.2720.334< 0.0010.0080.0860.8800.3020.1800.0190.206< 0.001
AX20.3900.6480.0640.0140.098< 0.0010.7600.0130.520< 0.0010.058< 0.0010.5830.016
Biomass0.0020.0170.9500.0030.009< 0.001< 0.001< 0.0010.003< 0.001< 0.001< 0.0010.6980.245
(+)    (–)(–)(–) (–)(–)(–)  
Inclination0.3800.7450.2500.1210.303< 0.0010.004< 0.0010.3820.2500.498< 0.0010.3360.900
     (+) (–)   (–)  
Potential solar radiation in January0.3600.7720.7400.1000.946< 0.001< 0.0010.0110.1460.0540.0120.0240.4830.074
     (+)(+)       
Potential solar radiation in June< 0.001< 0.0010.037< 0.0010.6620.5390.9100.0410.238< 0.0010.3600.0190.3120.001
(+)(+) (+)     (+)    
Bulk density< 0.001< 0.0010.990< 0.001< 0.001< 0.001< 0.001< 0.0010.010< 0.001< 0.001< 0.0010.9000.054
(+)(+) (+)(+)(+)(+)(+) (+)(+)(+)  
Soil water holding capacity0.7300.0140.089< 0.0010.341< 0.001< 0.001< 0.0010.111< 0.001< 0.001< 0.0010.5700.002
   (+) (+)(–)(+) (–)(–)(–) (–)
Mean potential solar radiation0.4100.2340.2000.0080.452< 0.001< 0.0010.008< 0.0010.0020.7980.3240.6650.015
     (–)(–) (+)(+)    
Percentage of bare ground0.0010.075< 0.0010.0170.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.0010.1130.313
(+) (+) (+)(+)(+)(+)(+)(+)(+)(+)  

Appendix 2

Table 6.  Results of analysis of the effect of selected variables on occurrence of adult plants at the block and locality scales. Table shows significance values of log-likelihood tests in logistic regression comparing the baseline model (model including source of the seeds at both scales and locality code at the block scale) with the model including also the term of interest. Significant values are in bold (significance level is adjusted using Bonferoni correction). If the direction of the response makes sense, the sign of the relationship, if significant, is given in brackets. Results at the block scale are based on 220 observations; results at the locality scale are based on 22 observations
FactorAsterCirsiumCoronillaGlobulariaL. flavumL. tenuifoliumScorzonera
BlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocalityBlockLocality
Beals index value0.003< 0.001< 0.001< 0.001< 0.0010.2930.002< 0.001< 0.001< 0.0010.2210.1840.0060.212
(+)(+)(+)(+)(+) (+)(+)(+)(+)    
AX10.1550.699< 0.001< 0.0010.0080.6020.0490.0950.1340.0400.7200.063< 0.0010.103
AX20.5310.2040.0070.3740.8250.013< 0.0010.0660.6320.7300.5650.0380.0020.016
Biomass0.5410.0080.0040.5890.9250.2200.0040.0580.7080.0820.7940.0060.1550.355
  (–)   (–)       
Inclination0.0240.2970.5740.3770.1620.1850.3110.5240.9930.1570.1390.3880.6230.899
Potential solar radiation in January0.0980.1450.0130.7320.5300.2610.5550.0760.1430.1000.5150.0340.2760.930
Potential solar radiation in June< 0.0010.6410.5140.3890.3810.1580.2920.0780.5600.0330.6400.6850.2980.053
(+)             
Bulk density0.7360.3540.9420.9170.9080.0400.8580.6680.5880.3000.9650.6050.0830.180
Soil water holding capacity0.2420.0690.5660.9080.0580.0620.6840.5580.5860.1680.1790.6490.8800.228
Mean potential solar radiation0.5700.1670.1760.6050.7870.7200.8270.2410.4450.4100.1930.8970.6080.642
Percentage of bare ground0.0040.9560.1510.9910.0100.9420.3490.4200.0940.6680.8780.2440.2610.028
(–)             

Ancillary