Journal of Biogeography (J. Biogeogr.) (2012) 39, 1101- ORIGINAL ARTICLE Dispersal limitation is stronger in communities of microorganisms than macroorganisms across Central European cities Milan Chytrý1, Zdeňka Lososová1'2 , Michal Horsák1, Bohuslav Uher1, Tomáš Čejka3, Jiří Danihelka1'4, Karel Fajmon1, Ondřej Hájek1, Lucie Juřičková5, Kateřina Kintrová1, Deana Láníková1'4, Zdenka Otýpková1, Vladimír Řehořek1 and Lubomír Tichý1 'Department of Botany and Zoology, Masaryk University, Kotlářská 2, CZ-611 37 Brno, Czech Republic, 2Department of Biology, Masaryk University, Poříčí 7, CZ-603 00 Brno, Czech Republic, 3Institute of Zoology, Slovak Academy of Sciences, Dúbravská cesta 9, SK-841 04 Bratislava, Slovakia, ^Department of Vegetation Ecology, Institute of Botany, Academy of Sciences of the Czech Republic, Lidická 25/27, CZ-657 20 Brno, Czech Republic, 5Department of Zoology, Charles University in Prague, Viničná 7, CZ-128 44 Praha, Czech Republic "Correspondence: Zdeňka Lososová, Department of Botany and Zoology, Masaryk University, Kotlářská 2, CZ-611 37 Brno, Czech Republic. E-mail: lososova@ped.muni.cz ABSTRACT Aim It is generally believed that communities of small organisms, or those with small propagules, are structured mainly by local niche-based processes, and less by dispersal limitation. Conversely, weaker environmental and stronger spatial structure, indicating dispersal limitation, are expected to occur more frequently in communities of large organisms. However, this hypothesis has rarely been tested by comparing spatial and environmental effects across groups of organisms of different size (or with different size of propagules) sampled at the same set of sites. Here, we test it in urban environments. Location Thirty-two cities in 10 countries of Central Europe and Benelux. Methods We compared effects of spatial location and climate on species composition of different groups of organisms sampled in corresponding types of urban habitats. The studied groups were: (1) subaerial cyanobacteria and algae, (2) vascular plants, (3) land snails; and subgroups of vascular plants with different life form and dispersal mode, namely: (4) herbs, (5) animal-dispersed trees and shrubs, and (6) wind-dispersed trees and shrubs. Data were analysed by variation partitioning based on redundancy analysis (RDA) with principal coordinates of neighbour matrices (PCNM). Eighteen PCNM eigenvectors (expressing spatial effects) and mean annual temperature, July-January temperature difference and annual precipitation sum (expressing environmental effects) were used as explanatory variables. Results Pure effects of climate on species composition, indicating niche-based processes, were not significant for any group or subgroup of the studied organisms. In contrast, pure effects of space, indicating dispersal limitation, were significant for all groups and subgroups except herbs. Surprisingly, the community of cyanobacteria/algae possessed much stronger spatial structure independent of climate than communities of larger organisms, although cyanobacteria/algae had the lowest beta diversity among the studied cities. Main conclusions We hypothesize that the community of subaerial cyanobacteria/algae is structured by natural processes which involve dispersal limitation, whereas communities of urban plants and snails are influenced by human-assisted dispersal of their propagules between cities, which results in weaker dispersal limitation. Our study indicates that dispersal vectors can be more important for community structure than size of organisms or of their propagules. Keywords Algae, body size, community structure, cyanobacteria, environmental filtering, land snails, spatial structure, urban ecology, variation partitioning, vascular plants. © 2011 Blackwell Publishing Ltd http://wileyonlinelibrary.com/journal/jbi doi:10.1111/J.1365-2699.2011.02664.x 1101 M. Chytrý ef al. INTRODUCTION The current theory of community ecology understands the species composition of local communities to be an outcome of both local niche-based processes (Hutchinson, 1957; Chase & Leibold, 2003) and regional processes such as speciation, extinction, migration and dispersal (Ricklefs & Schluter, 1993; Zobel, 1997), but their relative importance in different contexts is still poorly known. The key concept for understanding these processes is that of the metacommunity, denned as a set of local communities connected by the dispersal of multiple species (Hubbell, 2001). In the metacommunity framework, mechanisms underlying niche-based versus dispersal-limited community structures are described respectively by the species-sorting model and the neutral model (Leibold et al, 2004). The species-sorting model is derived from the theory of species composition change along environmental gradients (Whittaker, 1967). It assumes that environment is heterogeneous across landscapes, and local species composition is an outcome of the adaptation of different species to particular environmental qualities and local interspecific interactions. Thus, beta diversity is generated by environmental heterogeneity and environmental control of species distributions (Legendre et al, 2005). In contrast, the neutral model assumes spatially homogeneous environments and equivalence of species with respect to their environmental requirements (Bell, 2001; Hubbell, 2001); under this model, dispersal limitation is the only factor responsible for variation in species composition among local communities. The role of dispersal limitation has traditionally been inferred from the observation of decreasing similarity in species composition with geographical distance, known as the distance decay of similarity (Qian et al, 1998; Nekola & White, 1999; McKinney, 2004, 2006; Soininen et al, 2007; La Sorte et al, 2008; Leng et al, 2010). However, not only species composition, but also environment, can become less similar with increasing distance, and in such cases, distance decay can result from niche-based processes rather than from dispersal limitation. It is therefore necessary to measure environmental variables relevant to the studied community, and partition variation in its species composition into fractions explained by the effect of spatial distance, the effect of environment, and the shared effect of both, the latter of which is sometimes called spatially structured environmental variation (Borcard et al, 1992). Spatial effects can operate on various scales, including migration and range extension on large scales and mass effect or sink-source dynamics on smaller scales. Any of these processes can be important at some scales but insignificant at others (Verleyen et al, 2009). Recently developed methods such as principal coordinates of neighbour matrices (PCNM; Borcard & Legendre, 2002) or Moran's eigenvector maps (Dray et al, 2006) enable spatial structure to be described across multiple scales through a series of eigenvector-based variables, which are used as predictors in variation partitioning models, e.g. using redundancy analysis (RDA). Cottenie (2005) suggested that a combination of significant pure effects of environmental variables and non-significant pure effects of spatial variables in variation partitioning indicates a community structure based on the species-sorting model, while the opposite pattern indicates the neutral or patch-dynamics model. However, the shared effect of environmental and spatial variables can also be very important in these considerations (Smith & Lundholm, 2010). It is generally believed that small organisms (especially microorganisms) or those with small propagules are widespread over large areas, and their occurrence at particular sites is explained mainly or entirely by environmental qualities (Finlay et al, 1996; Fenchel & Finlay, 2004; Green & Bohan-nan, 2006; Van der Gucht et al, 2007). Consequently, their community structure would be determined by species sorting. The reasoning behind this assumption is that the small size of individuals or propagules can facilitate long-distance passive dispersal, and that the large population sizes and short generation times typical of small organisms result in high dispersal rates. In contrast, larger organisms, or those with large propagules, are supposed to be dispersal-limited, and therefore, besides environmental qualities, their current distribution is also strongly dependent on the history of their evolution and migration. Contrary to the common belief that for microorganisms 'everything is everywhere, but, the environment selects' (de Wit & Bouvier, 2006), some recent studies demonstrate that communities of unicellular organisms can possess a notable spatial structure (Martiny et al, 2006; Telford et al, 2006; Pommier et al, 2007; Vyverman et al, 2007; Verleyen et al, 2009; Heino et al, 2010). To date, there have been few comparative studies of communities of organisms of different size (e.g. Hillebrand et al, 2001; Cottenie, 2005), and even fewer studies based on a set of samples of differently-sized organisms from the same set of sites, which would enable a comparison of the relative effects of spatial and environmental variables on community structure. Of these, Beisner et al (2006) have demonstrated that communities of bacteria, plankton and fish in Canadian lakes tend to be less environmentally and more spatially structured with increasing size of organism. Hájek et al (2011) found a stronger spatial structure for organisms with large propagules (vascular plants and molluscs) than for those with small propagules (diatoms and bryophytes) in Carpathian fens. Astorga et al (2011) revealed much stronger environmental than spatial structure in diatoms, bryophytes and macroinvertebrates in Finnish streams. Sattler et al (2010) also observed a weak spatial structure in a study of spiders, bees and birds within Swiss cities, but this could be due to the restricted spatial extent of their study. Generally, these studies support the notion that communities of larger organisms or those with larger propagules tend to have a stronger spatial structure, but distinct spatial structures are often also present in communities of organisms with small propagules. Here, we compare the relative effects of spatial and environmental factors on the community structure of three 1102 Journal of Biogeography 39, 1101-1111 © 2011 Blackwell Publishing Ltd Dispersal limitation in Central European cities groups of organisms possessing different sized individuals and propagules: (1) subaerial cyanobacteria and algae, (2) vascular plants, and (3) land snails, each of which were consistently sampled in 32 cities of Central Europe and Benelux. Cities are very suitable for a study such as this, because urban habitats are created and strongly influenced by humans, and are therefore very similar across large geographical distances (McKinney, 2006). Given that we sampled corresponding habitats in each city and merged the samples from individual habitats for each city, the main environmental differences between samples from different cities were climatic. We hypothesized that, due to their smaller propagule size, the community of cyanobacteria and algae would be mainly structured by climate, without any purely spatial effects, thus conforming to the species-sorting model of the metacommu-nity with no dispersal limitation. In contrast, the communities of plants and snails would be partly structured by purely spatial effects not associated with climatic differences, indicating dispersal limitation. MATERIALS AND METHODS Species data We sampled 32 cities, each with more than 100,000 inhabitants, in Belgium, the Netherlands, Germany, Poland, the Czech Republic, Slovakia, Switzerland, Austria, Slovenia and Hungary (for simplicity, this region is referred to as Central Europe throughout this paper; Fig. 1). The ranges of climatic characteristics across these cities were as follows: mean annual temperature 7.9-11.2 °C, January-July temperature difference 14.2-23.0 °C, and annual precipitation sum 544-1312 mm. Sampling was performed from June to August in 2007-2009. Samples of subaerial cyanobacteria and algae were collected on building stone (preferably sandstone) and tree bark, each of these two substrates being both sampled in the city centre (usually on a historical square) and in a city park. Sampled surfaces were 1 cm2 large, located at a height of 120-150 cm on a building wall or a tree trunk. Sampling sites on buildings in the city centre were selected from both north-facing and south-facing walls, while sites on stones in parks and trees were selected from the north-facing side. Five samples from each city were sampled (centre-stone-north-facing, centre-stone-south-facing, centre-tree, park-stone, park-tree). Samples were cultivated in Petri dishes and tubes containing the culture medium BBM (Smith & Bold, 1966) under controlled conditions (temperature 21 °C, humidity 60%, irradiance 10.3 irmol s"1 m"2, light 660 lx, PhAR 2.14 W m"2 and a photoperiod of 12:12 h light : dark). The initial sample incubation in Petri dishes (7-21 days) was followed by isolation of clones. The morphometric data were measured from both natural populations and isolates maintained in Journal of Biogeography 39, 1101-1111 © 2011 Blackwell Publishing Ltd 1103 M. Chytry ef al. cultures. Samples were observed and documented using the Olympus BX 50 light microscope equipped with Lucia Image Analysis and the Olympus SZH stereomicroscope. Species composition (presence/absence) of vascular plants and land snails was recorded in seven 1-ha plots in each city, each of them representing one urban habitat type: (1) square in the historical city centre, (2) 19th-century boulevard, (3) residential area with closed building pattern of family houses and private gardens, (4) residential area with open building pattern with blocks of flats, (5) city park, (6) unmanaged mid-successional site with perennial herbaceous vegetation and scattered shrubs or trees, and (7) recently disturbed site (see Lososova et al, 2011a for details). At each site, all land snails and all spontaneously established vascular plant species were recorded, including garden escapes and seedlings of spontaneously regenerating planted trees and shrubs. Plant species represented only by deliberately planted individuals were not recorded. Plants were classified according to their life form and dispersal mode as (1) herbs, (2) animal-dispersed trees and shrubs, and (3) wind-dispersed trees and shrubs. Snails were searched for by eye in all appropriate microhabitats within each site. All live snail individuals, as well as empty shells with intact periostracum, were considered. For each city and each group of organisms, a pooled list of all species recorded in all the studied habitats was prepared and analyses were based on these pooled species lists. For all species groups, presence/ absence data were analysed. Explanatory variables Variables analysed for their effects on species composition were: (1) geographical distance between cities, and (2) climatic data (from Hijmans et al, 2005), including mean annual temperature (°C), annual temperature range expressed as the difference between July and January mean temperature (°C) and annual precipitation sum (mm). To quantify the effects of the spatial component on species composition, we produced a matrix of between-city geographical distances (truncated at a distance determined by the minimum spanning tree technique), which we analysed using principal coordinates of neighbour matrices (PCNM; Borcard & Legendre, 2002). The calculations were done in the R program (http://www. r-project.org), using the 'PCNM' package (https://r-forge. r-project.org/R/?group_id=195). The 18 eigenvectors corresponding to positive eigenvalues obtained in this analysis were used as independent variables to explain the effect of spatial distance between cities on changes in species composition. Statistical analysis We calculated alpha (a), beta (P) and gamma (y) diversity of all the studied groups of organisms in Central European cities, where cx-diversity was denned as the mean number of species recorded in single cities, P-diversity as the mean dissimilarity in species composition between cities, and y-diversity as the cumulative number of all species found in the 32 studied cities. We used the mean Jaccard dissimilarity index as a measure of |3-diversity (Koleff et al, 2003). Sample-based rarefaction curves (Gotelli & Colwell, 2001), calculated according to the analytical formula published by Colwell et al (2004), were used to compare species richness of the studied species groups. This calculation was performed using the juice program, version 7 (Tichy, 2002). To visualize the main patterns in species composition among the studied cities, we used principal components analysis (PCA) with Hellinger-transformed presence/absence species data (Legendre & Gallagher, 2001) in the canoco program, version 4.5 (ter Braak & Smilauer, 2002). We then determined relative effects of climate and space on species composition of each group of organisms using variation partitioning based on the above-mentioned climatic and spatial variables. Three climatic variables and 18 spatial variables with positive eigenvalues obtained from the PCNM analysis were included in redundancy analyses (RDA) with the Hellinger-transformed presence/absence species data. The variation partitioning algorithm proposed by Peres-Neto et al (2006) was used to balance the possible bias in estimation of the variation explained by climate and space, which might occur because of different number of explanatory variables for these two groups (Kromrey & Hines, 1995). The significance of the pure effects of climatic variables and spatial distance was tested by Monte Carlo tests with 999 permutations. These calculations were done in the R program, using the 'vegan' package, version 1.17-2 (https://r-forge.r-project.Org/R/ ?group_id=68). There is an ongoing debate on the relative effectiveness of various methods that have been proposed to disentangle the environmental and spatial effects on community composition, namely the raw-data approach, such as canonical ordination with spatial explanatory variables (e.g. PCNM eigenvectors) versus the distance-based approach, such as correlation between distance matrices (Legendre et al, 2005, 2008; Tuomisto & Ruokolainen, 2006, 2008). Here we use the raw-data approach, because it is more appropriate for the question that we ask, i.e. whether the variation in community composition among sites is due to variation in environmental conditions or to neutral community dynamics (Legendre et al, 2008). However, recent comparisons of different methods based on simulated data (Gilbert & Bennett, 2010; Smith & Lundholm, 2010) have demonstrated that raw-data approaches tend to underestimate the environmental component of community variation, whereas distance-based approaches consistently underestimate the spatial component. Therefore we additionally analysed our data following a distance-based approach, namely Mantel statistic of matrix correlation and partial matrix correlation. We used matrices of Jaccard dissimilarities in species composition between cities, a matrix of geographical distances between cities and a matrix of Euclidean distances calculated based on the three climatic variables. Significances of matrix correlations were tested using Monte Carlo permutation test with 999 permutations (using the 'vegan' package in the R program). 1104 Journal of Biogeography 39, 1101-1111 © 2011 Blackwell Publishing Ltd Dispersal limitation in Central European cities In addition to the RDA-based variation partitioning, we visualized the relationships between Jaccard dissimilarities in species composition and geographical distances between the cities using scatter plots (an opposite pattern to the plots of distance decay of similarity; Nekola & White, 1999). Because the observations in dissimilarity or distance matrices are not independent, the significance of this relationship could not be tested using traditional regression techniques. Therefore, we computed Mantel permutation tests using the 'vegan' package of the R program. Following Baselga (2010), we compared the slopes of this relationship among different organism groups by bootstrapping, using a case resampling approach with 10,000 iterations. The P-value of the null hypothesis of no difference between the slopes was estimated from the proportion of iterations for which the slope of one group was smaller than the slope of the other group. RESULTS Species lists from the 32 studied cities contained on average 59 species of cyanobacteria/algae, 303 vascular plants (among them 250 herbs, 31 animal-dispersed trees and shrubs, and 21 wind-dispersed trees and shrubs), and 22 land snails (Table 1). Numbers of species found in each city were not correlated between the studied groups of organisms; for example, cities with many snail species were not particularly rich in species of cyanobacteria/algae or plants (correlation coefficients: cyano-bacteria/algae-plants -0.14 n.s.; cyanobacteria/algae-snails 0.20 n.s.; plants-snails -0.07 n.s.). P-diversity, calculated as the mean dissimilarity in species composition between cities, was largest for land snails, followed by plants and cyanobacteria/algae (Table 1). This means that Central European urban communities of cyanobacteria/algae shared a larger proportion of widespread species than plant communities, and snail communities had the lowest proportion of species shared between different cities. For subgroups of vascular plants, the largest P-diversity was in animal-dispersed trees and shrubs, followed by herbs and wind-dispersed trees and shrubs. Like the mean species richness of the cities, the cumulative species richness across all cities (y-diversity) was also highest for plants and lowest for snails (Table 1). The shape of the rarefaction curves (Fig. 2) differed among the studied groups. For cyanobacteria/algae, a steep increase followed by a levelling of the rarefaction curve indicated the lowest P-diversity of their communities. However, PCA ordination revealed a distinct geographical gradient associated with changes in species composition of cyanobacteria/algae, beginning with Hungarian and Slovak cities near the top of the scatter plot and ending with Dutch and Belgian cities on the right side (Fig. 3a). Very similar but less pronounced geographical gradients were found for the other species groups (Fig. 3b-f). Dissimilarity in species composition within all the studied groups of organisms increased with geographical distance between the cities (Fig. 4, Mantel test, P < 0.001). This increase was significantly higher for cyanobacteria/algae and snails than for plants. There were no differences in the slopes for the subgroups of plants. For all the studied groups of organisms, spatial distribution of the cities explained a larger proportion of variation in 0 5 10 15 20 25 30 0 I...............................I 0 5 10 15 20 25 30 Number of cities Figure 2 Sample-based rarefaction curves showing an increase in the total number of species of (a) cyanobacteria/algae, plants and snails; (b) herbs, animal-dispersed trees and shrubs, and wind-dispersed trees and shrubs recorded in Central European cities with increasing number of cities sampled. Note the square-root transformation of the vertical axis. Table 1 Mean numbers of species (± standard deviation) found in single Central European cities (a-diversity), mean Jaccard dissimilarities between species lists from individual cities (ß-diversity), and total numbers of species found in all the studied cities (y-diversity). Cyanobacteria/algae Plants Land snails Herbs Animal-dispersed trees and shrubs Wind-dispersed trees and shrubs No. of species per city 59 ± 11 303 ± 25 22 + 5 250 ± 21 31+6 21 + 0.2 Mean Jaccard dissimilarity 0.55 0.59 0.64 0.59 0.63 0.56 Total no. of species 129 1196 87 979 120 87 Journal of Biogeography 39, 1101-1111 © 2011 Blackwell Publishing Ltd 1105 M. Chytrý ef al. F Bratisl. 4 Debrecei Oldenburg .......'S........ Q Salzburg Regensburg # l «Hamburg O "^rlaNe^o^ss^F^bur? «»" Würzburg1"' Bern 1 _ Innsbruck1-* u Geneve Stuttgart1 (a) cyanobacteria/algae Antwerpen 0 28% Budapest Debrecen Bratislava Szczecin PrahaO Ha Amsterdam a Jntwemen Hamburg w , „ ^ Q a Utrecht • •Groningen ow Oldenburg e Köln Ostrava* Geneve Wiirzburg \ Regensburg fCNr Ljubljana\V Stuttgart Ma (d) herbs ■ ---0---------------- nnsbruck Chemnitz :reiburg^Kassel Augsburg k #^ # Salzburg 17% Budapest Debrecen Ko4ic^ Krakow^'" ] Bratislav? °Bm0 4í"í Prahám _ " I Amsterdam utrecnl O o Antwerpen O _Gronm9erl O ° Oldenburg Man Wiirzburg OOChemnitz ■sbruck Köln oVtmz °* OKassel ^Augsburg Ljubljana^ C» u Osalzburg Stuttgarte Geneve' Freiburg (b) plants 17 % (c) snails Amsterdam Utrecht .ktL Brati slava BrntP Krakbw Praha * $ Q Halle Szczecin • "aphemmtz WoartQOMa,'r "O*™ Regensburg Q Salzburg ln°oo o ^ München ^ Fret*""6"" ! (e) animal-dispersed trees and shrubs □Oldenburg 17% axis 1 nbnick C„ sterdam Os Maribor ° in^t) Würzburg ; g^U^I --------Q,q---------- Bratislava© 1 Praha ' Stuttgart _ O <£Í Halle Linz* Freiburg L Au gsburg Salzburg Q ^ C| ^Maml q Ostrava Würzburg Regensburg Ol 26% Ljubljana* Kassel ■ Salzburg Halle - •#Freib*rg • Groninger ueneveuQ l .Chemnitz q Regensbwrg Oq (f) wind-dispersed trees 21 % species composition than did climate. Redundancy analysis (RDA) detected significant gross effects of climate on species composition of each of the studied groups, but these effects were shared with spatial effects, whereas pure effects of climate were non-significant (Fig. 5). The total variation in species composition jointly explained by both climate and space, as well as the variation explained by the pure effects of space, were highest for cyanobacteria/algae, followed by snails and plants. For subgroups of vascular plants, the pure effects of space were only significant for trees and shrubs, but not for herbs. The distance-based approach (Mantel tests, not shown) gave the same results as the raw-data approach (RDA): pure effects of climate were non-significant, while pure effects of distance were significant for all the studied organism groups and subgroups. DISCUSSION We demonstrated that the community of subaerial cyanobac-teria and algae in Central European cities has a strong spatial structure, with a considerable spatial component of variation in species composition that is independent of the variation in climate. This is consistent with recent studies showing that not Figure 3 Ordination diagrams of the 32 Central European cities studied based on species composition of individual groups of organisms (principal components analysis; PCA). Variations explained cumulatively by the first and second ordination axes are given in percentages of total inertia. only communities of macroorganisms, but also those of microorganisms, such as bacteria (Martiny et al, 2006; Pom-mier et al, 2007) or diatoms (Telford et al, 2006; Vyverman et al, 2007; Verleyen et al, 2009; Heino et al, 2010; Hájek et al, 2011), are spatially structured, at least at larger spatial scales (Verleyen et al, 2009). This pattern suggests that although microscopic cyanobacteria and algae are believed to have generally better dispersal abilities, as may be deduced from their large distribution ranges and small propagules (Rindi & Guiry, 2004; Neustupa & Škaloud, 2008; but see Jenkins et al, 2007), their community composition and |3-diversity is still structured by dispersal limitation. Moreover, spatial structure of the community of cyanobacteria and algae was found to be much stronger than that of the communities of vascular plants and land snails across the studied cities. Redundancy analyses with PCNM-based spatial variables identified a weaker pure effect of space on plants and snails than on cyanobacteria/algae. This is surprising given that many vascular plant species possess heavy seeds with poor dispersal abilities (Klotz et al, 2002), and land snails are traditionally believed to have low dispersal rates (Baur & Baur, 1995), especially those with body size larger than c. 10 mm (McClain & Nekola, 2008; Cameron et al, 2010). The entire 1106 Journal of Biogeography 39, 1101-1111 © 2011 Blackwell Publishing Ltd Dispersal limitation in Central European cities Figure 4 Relationships between the Jaccard indices of dissimilarity in species composition and geographical distance. Dissimilarity values resulting from 496 pairwise comparisons among the 32 Central European cities studied are plotted against between-city geographical distances. All relationships are significantly positive (Mantel test, P < 0.001); they are illustrated by the lines fitted by the least-square method. Differences in the slopes among the organism groups were tested by bootstrapping; the same letters indicate no differences between groups (P < 0.05). (a) cyanobacteria/algae B slope 0.00018 0 200 400 600 800 1000 1200 1400 (b) plants slope 0.00014 0 200 400 600 800 1000 1200 1400 (c) snails „oo 4 slope 0.00021 (d) herbs slope 0.00014 200 400 600 800 1000 1200 1400 (e) animal-dispersed A trees and shrubs slope 0.00013 200 400 600 800 1000 1200 1400 (f) wind-dispersed trees and shrubs -50» Wim fOOOO