Quick viewing(Text Mode)

Distribution Patterns of Land Snails in Ugandan Rain Forests Support The

Distribution Patterns of Land Snails in Ugandan Rain Forests Support The

Journal of Biogeography (J. Biogeogr.) (2008) 35, 1759–1768

ORIGINAL Distribution patterns of land snails in ARTICLE Ugandan rain forests support the existence of Pleistocene forest refugia Torsten Wronski and Bernhard Hausdorf*

Zoological Museum, University Hamburg, ABSTRACT Hamburg, Germany Aim We investigated whether the biogeographical patterns expected if the East African fauna was affected by cycles of contraction to refugia and expansion of ranges, as has been previously hypothesized, can be found in the land snail fauna of rain forests in . Location The and the Lake Victoria forest belt in Uganda. Methods Snails and slugs were sampled in 60 plots in 13 rain forests, and small species were extracted from 5-L leaf litter samples. Relative species richness was calculated by rarefaction. The influence of putative determinants of species richness was examined by bivariate correlation and multiple regression. Clustering and nestedness were tested by Monte Carlo simulations with a null model that considers the range size distribution, the species richness distribution of the forests, and the spatial autocorrelation of the occurrences of each species. Biotic elements were determined by model-based Gaussian clustering. Results A total of 169 land snail species were recorded from 13 Ugandan rain forests. Relative species richness increases with rainfall and altitude, and decreases with evaporation and distance from the putative East Congolian refugia. Mean annual rainfall and distance from the putative East Congolian refugia were included in the best multiple regression model. The distribution areas of the Ugandan land snails are significantly clustered. Two montane, two lowland and a northern biotic element were found. The mean range extension increases with increasing distance from the putative East Congolian refugia. Moreover, the ranges of the Ugandan land snails are significantly nested. The centre of the sets of nested subsets is in the Virunga Mountains, close to the putative East Congolian refugia. Main conclusions The decrease of diversity with increasing distance from the putative East Congolian refugia, the clustering and nestedness of ranges, and the range size increase with increasing distance from the refugia indicate that the East African land snail fauna was affected by cycles of contraction to refugia and expansion of ranges. The significant clustering and nestedness cannot be explained by current environmental conditions. Given the environmental history, it can be supposed that the lowland elements expanded post-glacially, whereas the ranges of the montane species are

*Correspondence: Bernhard Hausdorf, probably currently contracting. Zoological Museum, University Hamburg, Keywords Martin-Luther-King-Platz 3, 20146 Hamburg, Germany. Biogeography, biotic elements, land snails, nestedness, Rapoport effect, refuge E-mail: [email protected] theory, species diversity, Uganda, vicariance.

ª 2008 The Authors www.blackwellpublishing.com/jbi 1759 Journal compilation ª 2008 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2008.01933.x T. Wronski and B. Hausdorf

INTRODUCTION

Refuge theory assumes that the recent distribution of organ- isms is influenced by past, usually Pleistocene, environmental changes that resulted in the contraction of ranges into refugia or the expansion of ranges from refugia (Haffer, 1969, 1982, 1997; Mayr & O’Hara, 1986). Refugia are areas that are less affected by environmental changes than the surrounding , so that organisms that become extinct elsewhere can survive there. Moreover, it has often been assumed that the contraction of ranges into refugia resulted in speciation by vicariance (Haffer, 1969, 1982, 1997; Mayr & O’Hara, 1986). The existence and importance of Pleistocene refugia in the northern , which were heavily affected by Pleisto- cene glaciations, is universally accepted. However, the exis- tence and role of refugia in the tropics remains controversial (Nichol, 1999; Smith et al., 2005; Noonan & Wray, 2006). Usually the existence of refugia has been inferred only from Figure 1 Map of Uganda showing the locations of the 13 forests the recent distributions of species richness and endemic species. studied: 1, Mgahinga; 2, Echuya, 3, Bwindi, 4, Kalinzu; 5, Kasyoha- However, several other patterns in the distribution of organisms Kitomi; 6, Maramagambo; 7, Kibale; 8, Semliki; 9, Matiri; 10, are expected to emerge if retraction to refugia and expansion Bugoma; 11, Budongo, 12, Mpanga; 13, Mabira. Dashed line: from refugia are processes that affect recent biogeography. Such approximate eastern border of the putative East Congolian refugia patterns include the clustering of ranges (Hausdorf & Hennig, (modified from Hamilton, 1976: fig. 1; Maley, 1996: fig. 5, 2001: 2004), the nestedness of ranges (Hulte´n, 1937; Daubenmire, fig. 5; Jolly et al., 1997). 1975; Hausdorf & Hennig, 2003a), and an increase of average range extension with increasing distance from the refugia cutting. However, the plots in the Mabira and Bugoma forests (a Rapoport effect: Pfenninger, 2004; Hausdorf, 2006). were situated in recreation forest and in heavily encroached We investigated whether the patterns expected if the fauna forest patches, respectively, and sampling in Maramagambo was affected by cycles of contraction to refugia and expansion of Forest was carried out in colonizing forest. Sampling was ranges can be found in the land snail fauna of rain forests in the carried out in March and April 2006, during the rainy season, Albertine rift region in Uganda. We did not consider the when snails and slugs are most active. A combination of visual potential role of refugia in speciation. Land snails are good searching and sorting a standardized volume of litter is the model organisms for historical biogeography because they are most efficient method for land snail inventories, if repeated poor active dispersers that might reflect historical patterns visits to sites are not possible (Emberton et al., 1996; Cameron better than mammals or flying insects. The Albertine rift region & Pokryszko, 2005). Thus, we collected all living slugs and is especially suitable for such a study because Pleistocene refugia snails as well as their empty shells for 2 h at each plot, and in and post-glacial recolonization of adjacent areas have been addition collected c. 5 L of surface leaf litter and soil into proposed for this region (Hamilton, 1976; Diamond & Ham- plastic bags for later sampling. The litter samples were dried, ilton, 1980; Struhsaker, 1981; Crowe & Crowe, 1982; Rodgers fractioned by sieving, and sorted. More details of the sampling et al., 1982; Kingdon, 1990; Maley, 1996, 2001; Tattersfield, plots and lists of the sampled species will be published 1996; Grubb, 2001; Linder, 2001) and because it is one of the elsewhere (Wronski & Hausdorf, in preparation). centres of high biodiversity in (Pomeroy, 1993). The material will be kept in the Zoological Museum at the University Hamburg in Germany. A reference collection has been deposited at the Zoological Museum of Makerere MATERIALS AND METHODS University, Kampala, in Uganda. Study area and sampling Measuring species richness We sampled 60 plots, each 20 m · 20 m, in 13 forests within protected areas along the Albertine Rift Valley and in the forest In order to survey the land snail fauna of a forest exhaustively it belt around Lake Victoria in Uganda (Fig. 1, Table 1). Plots is certainly insufficient to sample a few plots (Cameron & were selected so that diverse aspects within a forest were Pokryszko, 2005). Therefore, we standardized species richness to covered and according to accessibility. The number of plots account for the number of sampled specimens in a forest. We varied among forests, as some sites considered in the sampling used the approach proposed by Prendergast et al. (1993) to plan could not be visited because of heavy rainfall during the calculate relative species richness. First, the number of speci- study period or because of political instability. Most plots were mens was rarefied for each pair of forests to the lower number in primary forest that had no more than limited selective collected in one of the forests using EcoSim ver. 7 (Gotelli &

1760 Journal of Biogeography 35, 1759–1768 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd Distribution patterns of land snails in Uganda

Entsminger, 2006). A pairwise measure of relative species richness, corrected for sampling effort, of forest A compared with forest B is calculated as the number of species expected in A divided by that expected in B for the same number of specimens. Distance to East Congolian refugia (km) The overall index of relative species richness for a forest is then calculated as the geometric mean of all such pairwise values. Mean altitude (m) Putative determinants of species richness

We tested seven abiotic variables as potential determinants of

Mean soil pH species richness (Table 2), namely mean annual rainfall, mean monthly evaporation (averaged over the year; calculated ) 2 estimates from meteorological data and corrected evapora-

Forest area (km tion-pan data), mean annual maximum temperature (the three climatic parameters were taken from the Atlas of Uganda, Uganda Department of Land and Surveys, 1967), forest area (mainly from Howard et al., 2000), soil pH (from Langdale- Brown et al., 1964), mean altitude (measured with GPS), and

Mean monthly evaporation (mm) shortest distance from the putative East Congolian refugia (Fig. 1) as shown in the reconstructions of Hamilton (1976): C)

fig. 1 and Maley, 1996: fig. 5, 2001: fig. 5) (see also Jolly et al., 1997), using bivariate correlation and multiple regression analyses. We chose the best multiple regression model among all subsets of variables based on the Akaike information

Mean annual maximum temperature ( criterion, as well as on leave-one-out cross-validation (squared loss) (Hastie et al., 2001).

Tests for clustering of distribution areas

Mean annual rainfall (mm) We applied the test for clustering proposed by Hausdorf & Hennig (2003b) and Hennig & Hausdorf (2004) to the species · forests matrix. As dissimilarity measure between the ranges of the examined taxa we used the geco coefficient (Hennig & Hausdorf, 2006), which takes the geographic Mean number of individuals per plot ± SE distances between the occurrences of the taxa into account and is robust against incomplete sampling:

0P P 1 min uðdRða; bÞÞ min uðdRða; bÞÞ B C 1 Ba2A b 2 B b2B a 2 A C dGðA; BÞ¼ @ þ A: Mean number of species per plot ± SE 2 jjA jjB

A, B ˝ R denote the distribution areas of two taxa and are

Relative species richness* subsets of the set of geographic units of the study region R.|A|

Table 2 Bivariate relationships between relative land snail species richness and abiotic variables of forests. n = 13 for all analyses.

Number of species rP (1993). Mean annual rainfall (mm) 0.694 0.008 et al. Mean annual maximum temperature (C) )0.539 0.057 Number of plots Mean monthly evaporation (mm) )0.571 0.041 Mean altitude (m) 0.598 0.031 Acidity (pH) )0.491 0.088

Number of recorded species and specimens, calculated relative species richness, and characteristics of the investigated Ugandan forests. Area (km2)* 0.141 0.645 Distance from refugia (km)* )0.563 0.045 Table 1 Forest MgahingaEchuyaBwindiKalinzu 6Kasyoha-Kitumi 4Maramagambo 5 5Kibale 57 4 5SemlikiMatiri 57 46Bugoma 66 1.16 44Budongo 56 5 3Mpanga 1.31Mabira 0.90 4 22.0 4 1.59 ± 9.7 0.83 5 1.12 68*Following Prendergast 45 28.0 ± 5 28.4 3.6 ± 3.1 17.8 169.5 33 ± ± 11.9 5 54 108.0 23.8 30.4 ±*log ± 7.0 60 4.4 1.16 151.5 1663 ±10 0.89 240.4 50.8 ± 40 104.8 120.0-transformed. ± 98.9 0.64 348.8 222.2 ± 44 1.04 ± 172.7 1375 1313 38.0 116.4 ± 1.09 5.1 1875 26.3 ± 1.5 1313 1375 21.0 0.83 17.3 ± 31.5 4.6 ± 8.2 571.6 0.79 32.6 ± ± 412.9 356.0 351.8 23.3 21.0 ± 137.2 20.0 23.3 ± 1313 3.0 258.3 28.8 ± 23.3 340.8 351.8 165.8 1313 ± ± 23.6 218.1 255.1 ± 2.5 118.8 1188 1250 1375 165.8 ± 68.3 282.4 26.3 ± 122.9 118.8 111.3 31.3 1313 118.8 146.8 122.9 1438 26.3 28.8 28.8 34 390 4.5 35 26.3 321 142.2 443 28.8 137.5 137 4.5 6.0 3.7 2570 142.2 146.8 146.8 6.3 6.0 1256 558 2295 40 1870 220 135.4 1428 988 137.8 100 6.0 54 401 55 825 7.3 40 80 85 5.8 6.0 5.8 1330 5 686 300 1285 1073 85 1108 5.8 5.8 35 110 150 95 1214 1260 245 325

Journal of Biogeography 35, 1759–1768 1761 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd T. Wronski and B. Hausdorf

is the number of geographic units of A. dR can be defined as the cases are defined by variables of metric scale. We therefore the geographic distance between geographic units. u is a performed a non-metric multidimensional scaling (NMDS; monotone increasing transformation with u(0) = 0. For geo- Kruskal, 1964) on the matrix of geco distances. Furthermore, graphical distances we used the following transformation u mclust requires an initial estimation of noise, that is, points that weights down the differences between large distances: that do not fit in any cluster, which was carried out using the software nnclean (Byers & Raftery, 1998) as suggested by d Fraley & Raftery (1998). As tuning constant (the number of : d f max dR uðdÞ¼uf ðdÞ¼ f max dR ; 0 f 1: nearest neighbours taken into account) for nnclean we chose 1 : d > f max dR ) k = number of species 40 1, rounded up to the next integer. That is, uf is linear for distances smaller than f times (we set Four NMDS dimensions were used. f = 0.1) the diameter (maximum geographical distance) of the considered region R, and larger geographical distances are Tests for nestedness of distribution areas treated as ‘very far away’, encoded by uf = 1. The geco coefficient is a generalization of the Kulczynski dissimilarity, The cases in which the occurrences of a species in the 13 forests which is obtained with f = 0 and uf (0) = 0 (see Hennig & form a subset of the occurrences of another species, that is, the Hausdorf, 2006). number of supersets, is used as a test statistic for nestedness Clustering of ranges means that dissimilarities between ranges following Hausdorf & Hennig (2003a). Several other test of the same cluster are small, whereas the distances between statistics for nestedness have been proposed (for a review see ranges of different clusters are large. The variation of distances of Wright et al., 1998). All these test statistics assume that there is a homogeneously distributed set of ranges is expected to be only a single set of nested subsets; however, the number of lower, as there is no clear distinction between ranges that belong supersets is also appropriate as a test statistic, if there is more together and ranges that should be separated. Therefore, we use than one set of nested subsets (see also the discussion in the ratio between the sum of the 25% smallest and the sum of the Hausdorf & Hennig, 2007). This is especially possible if a larger 25% largest geco distances between the ranges of the examined region is considered (as in the present study) and there are taxa as test statistic T. This statistic is expected to be smaller for several separate centres of endemism (e.g. the Pyrenees, the clustered data than for homogeneous data. Alps and the Carpathian Mountains in the study of Hausdorf We investigated whether the observed degree of clustering of & Hennig, 2003a). For n species, all n · (n ) 1)/2 pairs are ranges can be explained completely by the range size distri- screened for supersets. If two ranges are identical, this is not bution, the varying number of taxa per forest, and the spatial treated as if one species range was a superset of the other. The autocorrelation of the occurrences of a taxon using a Monte distribution of the test statistic under the null hypothesis is Carlo simulation to approximate the distribution of the test approximated by a Monte Carlo simulation. Using the null statistic under the null hypothesis. model outlined in the description of the test for clustering, we The null model used in the Monte Carlo simulation should investigated whether the observed degree of nestedness of simulate the case in which all inhomogeneities of the data can ranges can be explained completely by the range size distri- be attributed to varying range sizes, to varying numbers of taxa bution, the varying number of species per forest and the spatial per geographic unit, and to the spatial autocorrelation of the autocorrelation of the occurrences of each species. Conven- occurrences of a taxon. We applied a null model in which all tional null models that have been used for tests for nestedness ranges are generated independently according to the same (for a review see Wright et al., 1998) do not consider well- probabilistic routine. This routine yields ranges in such a way known ecological patterns. Thus, the null hypothesis is often that the distribution of the number of forests per range rejected in tests of nestedness, although the observed degree of approximates the actual distribution of the number of forests nestedness is the result of other well-known ecological patterns per range, the richness distribution of the forests approximates and processes (Hausdorf & Hennig, 2007; Moore & Swihart, the actual richness distribution of the forests, and the tendency 2007). Here we use a null model that considers the spatial to form disjunct areas is governed by a parameter that is autocorrelation of the occurrences of each species, because estimated from the real data set. Computational details have spatial autocorrelation is a prominent pattern (Legendre, been described elsewhere (Hausdorf & Hennig, 2003a; Hennig 1993), especially at a biogeographical scale, and its neglect & Hausdorf, 2004). might result in the rejection of the null hypothesis in cases in If there is significant clustering of distribution areas, biotic which the observed degree of nestedness can be explained by elements – i.e. clusters of taxa with similar ranges – are the spatial autocorrelation alone. determined using model-based Gaussian clustering with a We used the index of nestedness (Hausdorf & Hennig, noise category as implemented in the software mclust (Fraley 2003a) to visualize the spatial distribution of the sets of nested

& Raftery, 1998; see also Hausdorf & Hennig, 2003b), because subsets. The index of nestedness (nj) is defined as the average this method provides decisions about the number of mean- of the pi values of all species present in sample j, where pi is the ingful clusters and the number of points that cannot be number of supersets, that is, the cases in which the occurrences assigned adequately to any biotic element (noise category). of a species i form a subset of the occurrences of another Model-based Gaussian clustering operates on a data set where species. Again, identical ranges are not counted as subsets or

1762 Journal of Biogeography 35, 1759–1768 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd Distribution patterns of land snails in Uganda supersets of each other. The index of nestedness increases (a) 1.8 towards the geographic centres of the sets of nested subsets. 1.6 1.4 Tests for Rapoport effects 1.2 A correlation between range extension in the direction away 1 from the refugia and the distance from the refugia would be a 0.8 Rapoport effect in the wide sense as defined by Stevens (1996). To test for such an effect, arithmetic means of the range 0.6 extensions in the direction away from the refugia of all species 0.4 occurring in the same distance zone in any direction away Relative species richness 0.2 from the refugia [width 0.2 log10 distance from refugia (km)] are plotted against distance from the refugia, and the range 0 1.5 1.7 1.9 2.1 2.3 2.5 extension of each species in the direction away from the refugia is plotted against the midpoint of the range. Distances and Log10 distance from refugia (km) range extensions have been log-transformed. (b)140

Software 120

Most calculations were made with the statistical software r. 100 The tests for nestedness and clustering as well as the method for the delimitation of biotic elements were implemented in 80 the program package prabclus, which is an add-on package 60 for r. These programs are available at http://cran.r-project.org. 40

Number of species RESULTS 20

Relationship between species richness and 0 environmental parameters 1.5 1.7 1.9 2.1 2.3 2.5 Log distance from refugia (km) In total we recorded 168 land snail species in the 13 10 investigated rain forests in the Albertine Rift Valley and the Figure 2 Relationship between land snail species richness in Lake Victoria forest belt in Uganda (Fig. 1). The most species- Ugandan rain forests and distance from the putative East Con- rich families were Streptaxidae (43 species), Subulinidae (38 golian refugia. (a) Relative species richness (following Prendergast species) and Urocyclidae (29 species). Bivariate correlation et al., 1993) of the 13 investigated forests. (b) Number of land analyses revealed that the relative species richness of Ugandan snail species in zones in any direction away from the refugia forests for land snails significantly increases with mean annual [width of the zones 0.2 log10 distance from refugia (km)]. rainfall and altitude (actually, the relationship with altitude is hump-shaped with an increase up to c. 2000 m altitude and a The best multiple regression model according to the Akaike decrease at higher altitudes) and decreases with mean information criterion (AIC) as well as according to leave- monthly evaporation and distance from the putative East one-out cross-validation (squared loss) includes mean annual Congolian refugia (Table 2, Fig. 2a). Moreover, there is an rainfall and distance from the putative East Congolian refugia only marginally non-significant negative correlation with (R2 = 0.5777, F-statistic: 6.839 on 2 and 10 d.f., P = 0.0134). mean annual maximum temperature and soil pH. Species The AIC of the best model with the two mentioned variables is richness measured as the number of species in zones in a lower ()42.05) than that of the best model with one ()41.26; direction away from the putative refugia is also correlated mean annual rainfall) or three ()41.28) variables. The squared with the distance from the putative East Congolian refugia loss value is also smaller for two variables (0.0443) than for one (Fig. 2b). The monotonous decline in the number of species (0.0448) or three (0.0778) variables. with increasing distance from the refugia cannot be explained by varying sampling intensity. In the five distance zones Tests for clustering of distribution areas considered, 14, 9, 22, 5, 10 plots (in order of increasing distance from the putative refugia) were sampled, respectively. The test statistic T of the tests for the clustering of distribution Thus, we would expect a maximum of the number of species areas, the ratio of the sum of the 25% smallest geco distances at medium distances from the refugia based on sampling to the sum of the 25% largest distances between the species intensity alone, rather than a monotonous decline with ranges, is significantly smaller than would be expected under increasing distance from the refugia. the null model (P = 0.001). The test indicates that the

Journal of Biogeography 35, 1759–1768 1763 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd T. Wronski and B. Hausdorf

Figure 3 Distribution of biotic elements (1–2: montane elements; Figure 4 Distribution of sets of nested subsets in the land snail 3–4: lowland elements; 5: northern element) of land snails in fauna of Uganda as visualized by isolines of the index of nestedness

Uganda. The lines indicate the areas where more than 30% of the nj (the average number of supersets of ranges of species present species of an element are present. at forest j; see text). nj increases towards the geographic centre of sets of nested subsets in Congo. distribution areas of the Ugandan land snail species are significantly clustered. Biotic elements, groups of species with similar ranges, were (Fig. 5a). There is also a significant positive correlation determined with model-based Gaussian clustering. Although the between the range extension of land snail species in the clustering is significant, 90 of the 168 Ugandan land snail species direction away from the putative East Congolian refugia and were included in the noise component. The remaining species the midpoint of the range in the study area in Uganda (Fig. 5b; were assigned to the following five biotic elements (Fig. 3): an Spearman’s rank correlation coefficient rS = 0.803, two-sided, element restricted to the montane forests above c. 1500 m P = 0.000). altitude (Bwindi, Echuya, Mgahinga; 34 species), another montane element centred in Mgahinga forest (18 species), a DISCUSSION widespread lowland element that occurs up to c. 1700 m altitude (11 species), another widespread lowland element restricted to If the fauna of an area was affected by cycles of contraction to altitudes below 1500 m (seven species), and a northern element refugia and expansion of ranges, diversity should decrease with found only in Semliki, Mabira and Budongo (nine species). increasing distance from refugia, ranges of organisms should be clustered (Hausdorf & Hennig, 2004) and nested (Hulte´n, 1937; Daubenmire, 1975; Hausdorf & Hennig, 2003a), and Tests for nestedness of distribution areas range size should increase with increasing distance from the There are 4495 cases in which the range of a Ugandan land refugia (the Rapoport effect; Pfenninger, 2004; Hausdorf, snail species is a subset of the range of another species. The 2006). observed number of supersets is significantly higher than those Actually, the relative species richness of Ugandan land snails obtained in 1000 Monte Carlo simulations (2978–4899 super- decreases significantly with increasing distance from the sets were observed, mean 3945.97). Thus, the test indicates that putative East Congolian refugia (Table 2, Fig. 2). Such a the ranges of the Ugandan land snail species are significantly correlation has also been found for Ugandan anthropoid nested (P = 0.031). The centre of the sets of nested subsets is in (Struhsaker, 1981) and ferns (Lwanga et al., 1998) the Virunga Mountains (Mgahinga forest) in the south-west of and for East African forest mammals (Rodgers et al., 1982). the study area (Fig. 4), close to the putative East Congolian The distance from the putative East Congolian refugia is also refugia. Nestedness remains significant if the montane ele- included in the best multiple regression model. ments are omitted (P = 0.039; there are 2234 supersets in the The ranges of land snails in the study region in Uganda are real data, whereas 1357–2482 were counted in 1000 Monte significantly clustered, as predicted by the refuge theory. A Carlo simulations, mean 1884.90). biotic element analysis identified five clusters that can be grouped into montane and lowland elements, as is the case for most other African organisms (Hamilton, 1976). A cluster Tests for Rapoport effects analysis of the land snail fauna revealed two montane elements The mean range extension of land snail species increases with centred in the Virunga Mountains in the south-west of the increasing distance from the putative East Congolian refugia study area, two widespread lowland elements, and a northern

1764 Journal of Biogeography 35, 1759–1768 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd Distribution patterns of land snails in Uganda

(a) distribution of land snails in Uganda is less complete than that 2.5 in north-west and the investigated Mediterranean regions. 2 Furthermore, the refuge theory predicts that the biota in regions more distant from a refugium will be subsets of the 1.5 biota closer to the refugium. In fact, the ranges of the Ugandan land snails are significantly nested, and the centre of the sets of

range extension

(km+1) 1 nested subsets is located in the Virunga Mountains (Fig. 4) close to the putative East Congolian refugia. Since the centre of

mean the sets of nested subsets corresponds to the centre of the

10 0.5 montane elements (Fig. 3), the nestedness pattern is probably

Log the result of contraction or expansion of species belonging to 0 the montane element. 1.5 1.7 1.9 2.1 2.3 2.5 Several of the species from south-western Uganda classified Log distance from refugia (km) 10 in the montane elements have widely disjunct distributions (b) (see also Verdcourt, 1984a). For example, Streptostele teres 3 Pilsbry, 1919 is known from the Ruwenzori Mountains and Mount Elgon (Verdcourt, 1983); Elgonocyclus koptaweliensis 2.5 (Germain, 1934) and Punctum ugandanum (E. A. Smith, 1903) are known from the Kenyan highlands (Verdcourt, 1982, 2 1988); and, furthermore, Micromaizania volkensi (Martens, 1895) and Trachycystis lamellifera (E. A. Smith, 1903) are also 1.5 found in the Kenyan and Tanzanian highlands (Verdcourt,

(km+1)

range extension 1984a,b). It is likely that they were more widespread during the 1 10 cool, arid glacials, when montane forests descended by 1000– 1500 m (Bonnefille et al., 1990; White, 1993; Maley, 1996, Log 0.5 2001; Flenley, 1998; Kiage & Liu, 2006). Because of the aridity 0 during the glacials, forest patches could develop only in 1.5 1.7 1.9 2.1 2.3 2.5 particularly sheltered locations with sufficient humidity. Such

Log10 distance from refugia (km) isolated forest patches may have existed in the interlacustrine highlands of Uganda in areas of favourable topography (Jolly Figure 5 Rapoport effects with increasing distance from the et al., 1997). The forest patches provided the opportunity for putative East Congolian refugia. (a) Mean range extension of land the stepping-stone dispersal of species (not only snail species {log10 mean[maximum – minimum distance (km) land snails); for example, from the presumably larger montane of the forests where a species has been recorded from the refugia]} forests in Eastern Congo to the smaller refugia in the highlands in zones in the direction away from the refugia [width of the of Kenya and Tanzania. When the isolated patches of montane zones 0.2 log distance from refugia (km)]. (b) Correlation 10 forest at lower elevations disappeared as a result of the between range extension of land snail species {log [maximum – 10 temperature increases in the interglacials or the post-glacial, minimum distance (km) of the forests where a species has been recorded from the refugia +1]} in the direction away from the the ranges of Afromontane species became disjunct. Thus, the refugia and the midpoint of the ranges in the study area in montane regions can be considered contemporary refugia, as Uganda. has been emphasized by White (1993). An alternative expla- nation for the disjunct distribution patterns would be that they are the result of passive chance dispersal by birds or heavy lowland element. The clustering of ranges cannot simply be storms. Although such long-distance dispersal of land snails is explained by current environmental conditions, because the possible (Kirchner et al., 1997; Gittenberger et al., 2006), it is species richness of the sites is considered in the null model. less likely than a stepping-stone dispersal of Afromontane The differentiation of lowland and montane elements fits the species through more densely situated montane forest patches hypothesis that there were separate lowland and montane during the glacials. refugia (Hamilton, 1976; Kingdon, 1990; Maley, 1996, 2001). Lowland rain forest was much more limited during the cool, The percentage of species that could not be classified in biotic arid glacial periods (Maley, 1996, 2001; Flenley, 1998) and elements and that were included in the noise component, that spread post-glacially. In the study region, the centres of the sets is, species that apparently have idiosyncratic distribution of nested subsets are in the montane region at present (Fig. 4), patterns, is higher (54%) than in comparable studies of because the currently ongoing process is the contraction of the north-west European (Hausdorf & Hennig, 2004) and Med- species-rich montane elements, whereas the lowland and lower iterranean (Hausdorf & Hennig, 2006) land snails. This might montane elements have already spread across the study region. be an artefact arising from the fact that knowledge of the Nevertheless, nestedness is still significant if the montane

Journal of Biogeography 35, 1759–1768 1765 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd T. Wronski and B. Hausdorf elements are omitted from the analysis, indicating that the Emberton, K.C., Pearce, T.A. & Randalana, R. (1996) Quan- lowland fauna was also affected by cycles of contraction and titatively sampling land-snail species richness in Madagascan expansion of ranges. rainforests. Malacologia, 38, 203–212. We can conclude that the decreasing species richness and Flenley, J.R. (1998) Tropical forests under the climates of the increasing range size of Ugandan land snails with increasing last 30,000 years. Climatic change, 39, 177–197. distance from the putative East Congolian refugia and the Fraley, C. & Raftery, A.E. (1998) How many clusters? Which significant clustering and nestedness of the ranges of Ugan- clustering method? Answers via model based cluster analy- dan land snails indicate that the East African land snail fauna sis. Computer Journal, 41, 578–588. was affected by cycles of contraction and expansion of ranges. Gittenberger, E., Groenenberg, D.S.J., Kokshoorn, B. & Preece, In accordance with previous studies showing that patterns of R.C. (2006) Molecular trails from hitch-hiking snails. endemism can be only partly predicted from contemporary Nature, 439, 409. factors (Linder, 2001; Jetz et al., 2004; Rahbek et al., 2007), Gotelli, N.J. & Entsminger, G.L. (2006) EcoSim: null models our results indicate that the significant clustering and software for ecology, version 7. Acquired Intelligence Inc. & nestedness of land snail species ranges in Uganda cannot be Kesey-Bear, Jericho http://garyentsminger.com/ecosim.htm. explained by current environmental conditions that might be Grubb, P. (2001) Endemism in African rain forest mammals. sufficient to explain the distribution of species richness alone. African rain forest ecology and conservation (ed. by There are distinct montane and lowland elements with W. Weber, L.J.T. White, A. Vedder and L. Naughton- presumably different histories in the Ugandan land snail Treves), pp. 88–100. Yale University Press, New Haven. fauna. Given the environmental history, it can be supposed Haffer, J. (1969) Speciation in Amazonian forest birds. Science, that the lowland elements expanded post-glacially, whereas 165, 131–137. the ranges of the montane species are probably currently Haffer, J. (1982) General aspects of the refuge theory. Biological contracting. diversification in the tropics (ed. by G.T. Prance), pp. 6–24. Columbia University Press, New York. Haffer, J. (1997) Alternative models of vertebrate speciation in ACKNOWLEDGEMENTS Amazonia: an overview. Biodiversity and Conservation, 6, We are grateful to G. Isabirye-Basuta and J. Baranga for their 451–476. support of our project, to A. Apio and P. Barugahare for Hamilton, A.C. (1976) The significance of patterns of distri- assisting in fieldwork, to C. Hennig for statistical advice, and to bution shown by forest plants and animals in M. Griesbach for linguistic corrections. We thank the Uganda for the reconstruction of Upper Pleistocene palaeoenviron- Wildlife Authority, the National Forest Authority of Uganda ments: a review. Palaeoecology of Africa, the surrounding and the Uganda National Council for Science and Technology islands and Antarctica (ed. by E.M. van Zinderen Bakker), for research permission, and the German Research Council pp. 63–97. Balkema, Rotterdam. (445 UGA-18/1/06) for funding this study. Hastie, T., Tibshirani, R. & Friedman, J. (2001) The elements of statistical learning. Springer, New York. Hausdorf, B. (2006) Latitudinal and altitudinal diversity patterns REFERENCES and Rapoport effects in north-west European land snails and Bonnefille, R., Roeland, J.C. & Guiot, J. (1990) Temperature their causes. Biological Journal of the Linnean Society,87, 309–323. and rainfall estimates for the past 40,000 years in equatorial Hausdorf, B. & Hennig, C. (2003a) Nestedness of north-. Nature, 346, 347–349. European land snail ranges as a consequence of differential Byers, S. & Raftery, A.E. (1998) Nearest neighbor clutter immigration from Pleistocene glacial refuges. Oecologia, 135, removal for estimating features in spatial point processes. 102–109. Journal of the American Statistical Association, 95, Hausdorf, B. & Hennig, C. (2003b) Biotic element analysis in 781–794. biogeography. Systematic Biology, 52, 717–723. Cameron, R.A.D. & Pokryszko, B.M. (2005) Estimating the Hausdorf, B. & Hennig, C. (2004) Does vicariance shape species richness and composition of land mollusc commu- biotas? Biogeographical tests of the vicariance model in the nities: problems, consequences and practical advice. Journal north-west European land snail fauna. Journal of Biogeog- of Conchology, 38, 529–547. raphy, 31, 1751–1757. Crowe, T.M. & Crowe, A.A. (1982) Patterns in distribution, Hausdorf, B. & Hennig, C. (2006) Biogeographical tests of the diversity and endemism in Afrotropical birds. Journal of vicariance model in Mediterranean land snails. Journal of Zoology, 198, 417–442. Biogeography, 33, 1202–1211. Daubenmire, R. (1975) Floristic plant geography of eastern Hausdorf, B. & Hennig, C. (2007) Null model tests of clus- Washington and northern Idaho. Journal of Biogeography, 2, tering of species, negative co-occurrence patterns and nest- 1–18. edness in meta-communities. Oikos, 116, 818–828. Diamond, A.W. & Hamilton, A.C. (1980) The distribution of Hennig, C. & Hausdorf, B. (2004) Distance-based parametric forest passerine birds and Quaternary climatic change in bootstrap tests for clustering of species ranges. Computa- Africa. Journal of Zoology, 191, 379–402. tional Statistics and Data Analysis, 45, 875–895.

1766 Journal of Biogeography 35, 1759–1768 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd Distribution patterns of land snails in Uganda

Hennig, C. & Hausdorf, B. (2006) A robust distance coefficient Nichol, J.E. (1999) Geomorphological evidence and Pleisto- between distribution areas incorporating geographic dis- cene refugia in Africa. Geographical Journal, 165, 79–89. tances. Systematic Biology, 55, 170–175. Noonan, B.P. & Wray, K.P. (2006) Neotropical diversification: Howard, P.C., Davenport, T.R.B., Kigenyi, F.W., Viskanic, P., the effects of a complex history on diversity within the Baltzer, M.C., Dickinson, C.J., Lwanga, J., Matthews, R.A. & poison frog genus Dendrobates. Journal of Biogeography, 33, Mupada, E. (2000) Protected area planning in the tropics: 1007–1020. Uganda’s national system of forest nature reserves. Conser- Pfenninger, M. (2004) Comparative analysis of range sizes in vation Biology, 14, 858–875. Helicidae s.l. (Pulmonata, Gastropoda). Evolutionary Ecology Hulte´n, E. (1937) Outline of the history of and Boreal Research, 6, 1–18. biota during the Quaternary period. Thule, Stockholm. Pomeroy, D. (1993) Centers of high biodiversity in Africa. Jetz, W., Rahbek, C. & Colwell, R.K. (2004) The coincidence of , 7, 901–907. rarity and richness and the potential signature of history in Prendergast, J.R., Wood, S.N., Lawton, J.H. & Eversham, B.C. centres of endemism. Ecology Letters, 7, 1180–1191. (1993) Correcting for variation in recording effort in Jolly, D., Taylor, D., Marchant, R., Hamilton, A., Bonnefille, analyses of diversity hotspots. Biodiversity Letters, 1, 39–53. R., Buchet, G. & Riollet, G. (1997) Vegetation dynamics in Rahbek, C., Gotelli, N.J., Colwell, R.K., Entsminger, G.L., since 18,000 yr BP: pollen records from the Rangel, T.F.L.V.B. & Graves, G.R. (2007) Predicting conti- interlacustrine highlands of Burundi, and western nental-scale patterns of bird species richness with spatially Uganda. Journal of Biogeography, 24, 495–512. explicit models. Proceedings of the Royal Society B: Biological Kiage, L.M. & Liu, K.-b. (2006) Late Quaternary paleoenvi- Sciences, 274, 165–174. ronmental changes in : a review of multiproxy Rodgers, W.A., Owens, C.F. & Homewood, K.M. (1982) evidence from palynology, lake sediments, and associated Biogeography of East African forest mammals. Journal of records. Progress in Physical Geography, 30, 633–658. Biogeography, 9, 41–54. Kingdon, J. (1990) Island Africa: the evolution of Africa’s rare Smith, T.B., Calsbeek, R., Wayne, R.K., Holder, K.H., Pires, D. animals and plants. Collins, London. & Bardeleben, C. (2005) Testing alternative mechanisms of Kirchner, C., Kra¨tzner, R. & Welter-Schultes, F.W. (1997) evolutionary divergence in an African rain forest passerine Flying snails – how far can Truncatellina (Pulmonata: Ver- bird. Journal of evolutionary Biology, 18, 257–268. tiginidae) be blown away over the sea? Journal of Molluscan Stevens, G.C. (1996) Extending Rapoport’s rule to Pacific Studies, 63, 479–487. marine fishes. Journal of Biogeography, 23, 149–154. Kruskal, J.B. (1964) Multidimensional scaling by optimizing Struhsaker, T.T. (1981) Forest and conservation in goodness of fit to a nonmetric hypothesis. Psychometrika, 29, East Africa. African Journal of Ecology, 19, 99–114. 1–27. Tattersfield, P. (1996) Local patterns of land snail diversity in a Langdale-Brown, I., Osmaston, H.A. & Wilson, J.G. (1964) The Kenyan rain forest. Malacologia, 38, 161–180. vegetation of Uganda and its bearing on land-use. Uganda Uganda Department of Land and Surveys (1967) Atlas of Government Printer, Entebbe. Uganda, 2nd edn. Uganda Government Printer, Entebbe. Legendre, P. (1993) Spatial autocorrelation: trouble or new Verdcourt, B. (1982) A new operculate genus from East Africa paradigm? Ecology, 74, 1659–1673. – Elgonocyclus (Mollusca, Cyclophoridae), a preliminary Linder, H.P. (2001) Plant diversity and endemism in sub- diagnosis. Basteria, 46, 77–78. Saharan tropical Africa. Journal of Biogeography, 28, 169–182. Verdcourt, B. (1983) A list of the non-marine Mollusca of East Lwanga, J.S., Balmford, A. & Badaza, R. (1998) Assessing fern Africa (Kenya, Uganda, Tanzania, excluding Lake Malawi). diversity: relative species richness and its environmental Achatina, 11, 200–239. correlates in Uganda. Biodiversity and Conservation, 7, 1387– Verdcourt, B. (1984a) Discontinuities in the distribution of 1398. some East African land snails. World-wide snails. Biogeo- Maley, J. (1996) The Africa rain forest – main characteristics of graphical studies on non-marine Mollusca (ed. by A. Solem and changes in vegetation and climate from the Upper Creta- A.C. van Bruggen), pp. 134–155. Brill and Backhuys, Leiden. ceous to the Quaternary. Proceedings of the Royal Society of Verdcourt, B. (1984b) Trachycystis lamellifera (Smith) in Edinburgh, 104B, 31–73. Tanzania (Pulmonata: Endodontidae sensu lato). Journal of Maley, J. (2001) The impact of arid phases on the African rain Conchology, 31, 373–374. forest through geological history. African rain forest ecology Verdcourt, B. (1988) The rediscovery of Punctum ugandanum and conservation (ed. by W. Weber, L.J.T. White, A. Vedder (E.A. Smith) (Pulmonata: Punctidae). Archiv fu¨r Mollus- and L. Naughton-Treves), pp. 68–87. Yale University Press, kenkunde, 118, 149–151. New Haven. White, F. (1993) Refuge theory, ice-age aridity and the history Mayr, E. & O’Hara, R.J. (1986) The biogeographic evidence of tropical biotas: an essay in plant geography. Fragmenta supporting the Pleistocene forest refuge hypothesis. Evolu- Floristica et Geobotanica, Supplement, 2, 385–409. tion, 40, 55–67. Wright, D.H., Patterson, B.D., Mikkelson, G.M., Cutler, A. & Moore, J.E. & Swihart, R.K. (2007) Toward ecologically Atmar, W. (1998) A comparative analysis of nested subset explicit null models of nestedness. Oecologia, 152, 763–777. patterns of species composition. Oecologia, 113, 1–20.

Journal of Biogeography 35, 1759–1768 1767 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd T. Wronski and B. Hausdorf

BIOSKETCHES

Torsten Wronski is a post-doctoral researcher at the Zoological Museum of the University of Hamburg. His main research interests are the biogeography, ecology and systematics of African land snails and the behavioural ecology of ungulates.

Bernhard Hausdorf is curator for malacology at the Zoological Museum of the University of Hamburg. His main research interests are the systematics, phylogeny, biogeography and ecology of land snails.

Editor: Melodie McGeoch

1768 Journal of Biogeography 35, 1759–1768 ª 2008 The Authors. Journal compilation ª 2008 Blackwell Publishing Ltd

Top View