Freshwater Biology (2009) 54, 1189–1201 doi:10.1111/j.1365-2427.2009.02168.x

Hydroperiod, predators and the distribution of physid snails across the freshwater habitat gradient

ANDREW M. TURNER AND SHARON L. MONTGOMERY Department of Biology, Clarion University, Clarion, PA, U.S.A.

SUMMARY 1. Studies of species distributions across environmental gradients further our under- standing of mechanisms regulating species diversity at the landscape scale. For some freshwater taxa the habitat gradient from small, shallow and temporary ponds to large, deep and permanent lakes has been shown to be an important environmental axis. Freshwater snails are key players in freshwater ecosystems, but there are no compre- hensive studies of their distributions across the entire freshwater habitat gradient. Here we test the hypothesis that snail species in the family Physidae are distributed in a non-random manner across the habitat gradient. We sampled the snails, their predators and the abiotic environment of 61 ponds and lakes, spanning a wide range in depth and hydroperiod. 2. Temporary habitats had the lowest biomass of predators. Shallow permanent ponds had the highest biomass of invertebrate predators but an intermediate fish biomass. Deep ponds and lakes had the highest fish biomass and intermediate invertebrate biomass. Five species of physids occurred in the regional species pool and 60 of the 61 ponds and lakes surveyed contained physid snails. Each pond and lake contained an average of just 1.2 physid species, illustrating limited membership in local communities and substantial among-site heterogeneity in species composition. 3. Physids showed strong sorting along the habitat gradient, with Physa vernalis found in the shortest hydroperiod ponds and Aplexa elongata, P. gyrina, P. acuta and P. ancillaria found in habitats of successively greater permanence. When organised into a site- by-species incidence matrix with sites ordered according to their hydroperiods, we found the pattern of incidence to be highly coherent, showing that much of the heterogeneity in species composition from one pond to another is explained by hydroperiod. We also found that the number of species replacements along this gradient was higher than random, showing that replacement is more important than nesting in describing species compo- sition in ponds of different hydroperiod. 4. Discriminant analysis showed that pond depth, invertebrate biomass and fish biomass were the best predictors of species composition. Analysis of these niche dimensions showed that P. vernalis and A. elongata were most successful in shallow, temporary ponds with few predators. P. gyrina and P. acuta were typically found in ponds of intermediate depth and high predator abundance. P. ancillaria was found in the deepest lakes, which had abundant fish predators but few invertebrate predators. Of the five species considered, P. ancillaria, P. vernalis and A. elongata were relatively specialised with regard to key habitat characteristics, P. gyrina was moderately generalised and P. acuta was remarkably generalised, since it alone occurred across the entire freshwater habitat

Correspondence: Andrew M. Turner, Department of Biology, Clarion University, Clarion, PA 16214, U.S.A. E-mail: [email protected]

2009 Blackwell Publishing Ltd 1189 1190 A. M. Turner and S. L. Montgomery gradient. The exceptional habitat breadth of P. acuta stands in contrast to distributional studies of other freshwater taxa and deserves further attention.

Keywords: gastropoda, hydroperiod, predation, species composition, species turnover

quantitative studies of how other taxa are distributed Introduction across the freshwater habitat gradient, accompanied by It is widely recognised that species diversity is measurements of predator abundance. generated in large part by the process of species Freshwater gastropods are key players in many replacement along environmental gradients (Whittak- freshwater food webs, and given suitable abiotic er, 1956; Huston, 1994). A key gradient in freshwater conditions are ubiquitous inhabitants of lakes and is the environmental continuum from small, shallow, ponds of all sizes (Pip, 1986; Jokinen, 1987; Burch, 1988; temporary ponds to large, deep, permanent lakes Dillon, 2000). A number of investigators, dating back (Wellborn, Skelly & Werner, 1996). Studies of the nearly 100 years, have described the ecological niche of distribution of aquatic organisms suggest that closely freshwater gastropod species (e.g. Baker, 1927; Clam- related species often sort along the freshwater habitat pitt, 1970; Pip, 1986; Jokinen, 1987). Reviews of this gradient, with individual species often occupying a literature suggest that habitat permanence and depth relatively narrow portion of the gradient (Zaret, 1980; play a role in shaping species composition (Lodge et al., Cook & Streams, 1984; Spence, 1989; Wellborn, 1994; 1987; Dillon, 2000; Wethington, 2003), but the distribu- Skelly, 1996; McPeek, Grace & Richardson, 2001; Stoks tional studies are qualitative in nature and unsuitable & McPeek, 2003; Garcia & Mittelbach, 2008). Thus, in a for a rigorous test of the Wellborn et al. (1996) hypoth- landscape with lakes and ponds of various sizes, the esis. To our knowledge, there are no quantitative restricted distributions of individual taxa can generate studies of the distribution of freshwater snail species a level of regional diversity that far exceeds mean and their predators across the entire habitat gradient. local diversity. A leading hypothesis accounting for Focusing on the family Physidae, we tested the the restricted distribution of aquatic taxa invokes hypothesis that freshwater gastropod species have a tradeoffs associated with variation in the predation restricted, non-random distribution along the freshwa- regime along the habitat gradient (Wellborn et al., ter habitat gradient. We focused on the physids because 1996; Skelly, 1997; Wissinger et al., 1999). It is impor- they are among the most abundant and geographically tant to understand the processes responsible for the widespread of North American freshwater inverte- limited distribution of taxa, as these are the processes brates and they inhabit the entire breadth of the habitat that ultimately maintain high levels of regional gradient. Based on the results from an extensive survey, biodiversity (Leibold et al., 2004). we describe the distribution of physid snails in a The hypothesis of Wellborn et al. (1996), that shifts in systematic manner. First, we describe how abiotic species composition are related to changes in predation factors and predator abundances change across the regime, does not yet have extensive empirical support, habitat gradient. Second, we evaluate local and regional as there are few comprehensive studies of species diversity. Third, we ask whether species occurrences distributions along the freshwater habitat gradient. are random with respect to hydroperiod and whether Previous studies have tended to focus on the fauna of non-random distributions are produced by turnover or one habitat type (e.g. lakes or ponds or vernal pools) nesting along the hydroperiod-depth gradient. Finally, and of those studies that have quantified species we describe the realised niche and relative specialisa- distributions across the entire gradient, few have tion of each species. simultaneously measured predator abundance. Thus, the evidence supporting the fundamental tenet of species replacement along the habitat gradient is Methods largely qualitative and anecdotal. One exception is the Study area distributional patterns of amphibians, which has been carefully documented by a number of investigators We surveyed 61 freshwater sites (ponds, marshes, (e.g. Werner et al., 2007). There is a clear need for similar swamps and lakes), ranging in surface area from

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1191 10 m2 to 34.7 km2 and with depths up to 19.8 m water levels and their depths were measured at other (Appendix S1). Study sites are situated within the times or drawn from published sources (e.g. Butkas upper Ohio River drainage and in Crawford, Mercer & Ostrofsky, 2006). Pond surface areas were esti- and Erie Counties of northwestern Pennsylvania, mated from linear dimensions using a laser range- U.S.A. This area is on the glaciated Allegheny Plateau finder, which was done at the same time that depths and is characterised by moderate relief, elevations were measured. Surface areas of larger lakes (>1 ha) between 300 and 400 m, and soils that are calcareous, were determined from USGS topographical maps or poorly drained, but relatively fertile. Land use in the from Butkas & Ostrofsky (2006). region is approximately evenly divided between Based on hydroperiod and depth, sites were forest and agriculture. About two-thirds of the study divided into three categories: temporary, permanent- sites are human-made (e.g. constructed by excavation shallow and permanent-deep. We categorised perma- or flooded by dams) and a third are of natural origin. nent sites with a midsummer depth less than 150 cm We chose study sites so as to achieve an approxi- as shallow and those with a depth greater than 150 cm mately uniform distribution of habitat sizes from the as deep. We then calculated median pH, conductivity smallest, most temporary vernal pools to the largest, and alkalinity for each habitat type. Differences deepest lakes. The only other selection criteria was among habitat types in pH, conductivity and alkalin- ease of access, and sites were not selected or excluded ity were evaluated with the non-parametric Kruskal– because of presence or absence of physid snails. Wallis ANOVA. Nevertheless, all of the 61 sites contained snails, and all but one contained physid snails. Predator distribution and abundance

Most of the large predatory invertebrates found in Physical and chemical parameters these localities can feed on physid snails to some For each site we measured pH, conductivity, alkalin- degree, and some taxa (notably odonates and hemi- ity, hydroperiod, depth and surface area. pH was pterans) are very effective and important snail pre- measured in situ with a temperature-compensated pH dators (Turner & Chislock, 2007). Therefore, we meter, calibrated daily against buffers. Specific con- measured the abundance of invertebrate predators ductivity was likewise measured with a calibrated in each of our study ponds and lakes. Invertebrate field meter (values standardised to 25 C). Total biomass was sampled with a stove-pipe type sampler, alkalinity was measured by titration with Bromocre- 25-cm in diameter. Large invertebrates were removed ) osol Green color indicator, and is expressed as mg L 1 from the sampler with a small dip net, identified to

CaCO3 equivalents. the level of family, and wet-mass measured. Taxa Sites were ordered along a hydroperiod-depth retained include all Anisoptera, large Zygoptera, all gradient from the shallowest and most ephemeral to Hemiptera, large Coleoptera, and the Hirudinea. the deepest and presumably least ephemeral. Tem- These large invertebrates function as carnivores and, porary habitats were ordered according to hydrope- with the exception of a single taxa (larval Dytiscus riod, and permanent habitats were ordered according verticalis Say), are known snail predators (Turner & to depth. Hydroperiod of temporary habitats was Chislock, 2007). We use biomass instead of density as determined by recording the drying date for each site an index of large invertebrate abundance because in the summer of 2006. Sites were classified as when combining size classes and taxa, biomass of a temporary or permanent based on our observations predator species is a better predictor of its feeding rate over the past 10 years. Even permanent ponds are (Peters, 1983). Dytiscus verticalis do not prey on snails, certain to dry at some longer time scale, so our so we performed an additional calculation of biomass definition of permanent is a drying frequency of with D. verticalis excluded. Small fish were often >10 years. Depths of most permanent habitats were captured in the sampler and their biomass was measured during a 3-day period in July of 2003. By measured as well. Stovepipe samplers perform well taking a snapshot in time, we precluded the con- when sampling the invertebrates of vegetated habitats founding effects of seasonal dry down in the shallow (Turner & Trexler, 1997). Crayfish, another snail habitats. The deeper lakes have relatively stable predator, occur in several of our study lakes and

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 1192 A. M. Turner and S. L. Montgomery ponds, but are not sampled effectively with the Analysis of assemblage structure stovepipe and we did not attempt to estimate their abundance. Survey data yield incidence patterns, with each species Larger ponds and lakes contained fish, most of present in a subset of the sites. With respect to any which are capable of preying on juvenile pulmonates particular environmental variable, these incidence (Brown & DeVries, 1985; Lodge et al., 1987; Martin patterns may or may not differ from a randomly et al., 1992). The small fish of marshes were ade- generated pattern. Here we examine the survey data quately sampled with the stovepipe sampler, but fish and, focusing on the hydroperiod habitat gradient, ask in larger lakes required additional sampling. Fish whether observed species incidence patterns represent were sampled from these habitats by repeated hauls non-random distributions. Following general practice with an 80’ bag seine. Captured fish were identified in community ecology, we evaluated assemblage to species, enumerated and length was measured. structure in two ways. First, we ranked ponds Lengths were then converted into wet biomass based according to hydroperiod and constructed a species on published length–weight relationships (Schneider, by pond incidence matrix. We can then ask whether Laarman & Gowing, 2000). Differences in the bio- the pattern of species incidence across the gradient are mass density of fish and invertebrate predators coherent or random. Second, we used the same matrix across pond types were tested with Kruskal–Wallis to test whether coherent distributions are the result of ANOVA. ‘‘turnover’’ or ‘‘nesting’’. Species turnover describes the degree to which species replace one another along the habitat gradient. Nestedness is the opposite of high Gastropod survey methods turnover and describes the degree to which diversity Snail distributional data are based on 210 sampling accumulates along the gradient by addition of new events, for an average of 3.5 samples per site, species without any corresponding species losses. If a conducted between 1998 and 2007. Each site was large number of species are found in ponds of some sampled at least twice. A D-frame dipnet was repeat- hydroperiod with subsets of these species occurring at edly swept through vegetation and bottom debris with other hydroperiods, the pattern is described as nested approximately 20 min of sampling effort devoted to a and turnover is low. site for each sampling event. Each site was also We used the method outlined by Leibold & sampled on at least one occasion with a stovepipe Mikkelson (2002) to test whether snail assemblages, sampler. The stovepipe sampler, 25-cm in diameter, when ordered along the hydroperiod gradient, exhibit was firmly seated into the bottom sediments in water significant coherence and turnover. Coherence of the that was between 15 and 30 cm deep, and the contents incidence matrix was measured by counting the were removed by repeated sweeps with a small net. number of embedded absences along the hydroperiod The taxonomy of the Physidae has been quite gradient. A highly coherent distribution will have few confused (Dillon et al., 2002), in part because variation embedded absences. Conventional approaches ordi- in food, predators, temperature, or other environmen- nate simultaneously along both axis to minimise the tal factors can induce shifts in shell morphology total number of embedded absences. Since we were (DeWitt, Robinson & Wilson, 2000). Wethington interested only in the environmental gradient repre- (2003); see also Wethington & Lydeard, 2007) recently sented by hydroperiod, we did not ordinate along the revised the classification of species within the Physidae. primary axis. Furthermore, since we did not want our Using molecular data, morphological data and breed- measure of coherence to depend on the ordering of ing experiments, she concluded that the family is physids, we did not count embedded absences represented in North America by a dozen extant occurring along the secondary axis. To test whether species. We followed Wethington (2003) in our identi- the number of embedded absences contained within fications. Species are most reliably distinguished from our matrix was significantly fewer than that expected one another by morphology of the penial complex, and if species were randomly distributed across the we performed dissections of any questionable speci- hydroperiod gradient, we constructed 200 null matri- mens in order to confirm identification of difficult ces with the 0’s and 1s rearranged randomly so as to animals. conserve both row and column totals and compared

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1193 the number of observed embedded absences to the dominant physid, so we described the realised niche null distribution, thereby determining the probability of each species with a focus on these habitat dimen- that the observed frequency of embedded absences sions. We grouped sites by physid species and could be drawn from a random distribution. calculated median parameter values for abiotic habitat To evaluate whether there was significant species variables (depth, surface area, alkalinity, pH and turnover, we filled in the embedded absences along conductivity) and mean predator abundance (inver- the primary axis, and counted the number of times tebrate biomass, fish biomass, total predator biomass). one species was replaced by another between two Niche breadth of each species was examined by sites having different hydroperiods. By filling in the calculating the number of sites dominated, mean CV embedded absences we insured that we were not of abiotic variables of sites dominated, and mean CV simply retesting the hypothesis of coherence. We then of biotic variables for sites dominated. created 1000 null matrices in which the species’ ranges were allowed to float randomly along the hydroperi- Results od gradient and measured the turnover. We then compared the observed turnover to that generated Site characteristics within the null matrices. Our study ponds and lakes (Appendix S1) spanned a large gradient of surface area, depth, and hydroperi- Analysis of realised niche and relative specialisation od. Midsummer surface areas ranged from 0 to 3470 ha and midsummer depths ranged from 0 to The analysis of assemblage structure showed high 19.6 m. Twenty of the 61 study systems were coherence with significant turnover. Therefore, as a observed to dry at some point during the period of final analysis we sought to describe the abiotic and observation and are hereafter categorised as tempo- biotic environment of ponds occupied by each species, rary habitats. There were few systematic changes in i.e. a multivariate description of the species’ realised abiotic parameters along the habitat gradient, other niche. We first constructed a dominance matrix (as than depth and permanence (Table 1). Alkalinity and opposed to a occurrence matrix) listing the physid conductivity were consistently high, reflecting the species that was numerically dominant in each pond along with pond habitat variables. We then deter- Table 1 Abiotic variables and predator biomass of temporary, mined which habitat variables were most useful in shallow and deep lentic habitats predicting the dominant species by performing step- wise discriminant analysis. Using backward stepwise Temporary Shallow Deep elimination, the discriminant analysis retained or Abiotic variables rejected abiotic and biotic habitat variables according Number surveyed 20 25 16 to their ability to predict species identity. The analysis Maximum Depth (cm) 9.5 60 430 Area (m2) 130 360 283,000 also tested for multivariate differences in the habitat ) Alkalinity (mg L 1)727278 variables of ponds dominated by each species (i.e. the pH 7.1 7.1 8.0 ) realised niche). Discriminant analysis assumes that Conductivity (lScm 1) 234 167 210 predictive variables are not highly correlated so we Predator biomass )2 screened habitat variables for correlations and re- Fish biomass (g m ) 0.8 (2.2) 0.6 (1.2) 21.2 (17.9) Invertebrate 1.3 (2.2) 4.0 (5.7) 2.1 (3.0) ) tained depth, alkalinity, pH, conductivity, inverte- biomass (g m 2) brate biomass, and fish biomass. One of the physids, Invertebrate 0.6 (1.0) 3.9 (5.7) 2.1 (3.0) ) Physa acuta Draparnaud, occupied the entire range of biomass (g m 2)* Total predator 2.1 (3.2) 4.6 (5.7) 23.3 (17.5) habitats, which resulted in a violation of the homo- ) biomass (g m 2) geneous variance requirement and made any analysis of niche differentiation relatively meaningless (James Median values are shown for abiotic variables. Predator biomass & Mcculloch, 1990). Thus, we excluded sites domi- density values are means with standard deviations in paren- theses. All four measures of predator biomass differed signifi- nated by P. acuta from the discriminant analysis. cantly among habitat types (P < 0.05). Discriminant analysis yielded a subset of habitat *Invertebrate biomass density calculated after excluding Dytis- variables that were most useful in predicting the cus verticalis.

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 1194 A. M. Turner and S. L. Montgomery hard water characteristic of the region, and did not were stronger when D. verticalis, an ineffectual pred- differ among habitat types (P > 0.05). pH was also ator of snails (Turner & Chislock, 2007), was excluded generally high, although deep habitats tended to have from the analysis (P < 0.01; Table 1), as most of the a slightly higher pH than did shallow and temporary predatory invertebrates biomass in temporary habitats (P < 0.05, Table 1). habitats was attributable to the presence of larval D. verticalis. There was a strong concordance between mid- Predator abundance summer water depth and the presence of well Our survey results support the hypothesis that pred- developed fish communities. With just two excep- ator abundance changes in a systematic and predict- tions, ponds and lakes deeper than 150 cm contained able manner along the hydroperiod-depth gradient. multiple species of fish and had a biomass density ) Biomass of predatory invertebrates was lowest in >5.0 g m 2, whereas ponds shallower than 150 cm temporary habitats, highest in shallow habitats and lacked fish entirely, or contained two or fewer fish ) intermediate in deep habitats (habitat type effect: species and <5.0 g m 2 fish biomass (Fig. 1 bottom). P = 0.04; Fig. 1 top). Differences among habitat types Differences in fish biomass among habitat types were highly significant (P < 0.001). Because both inverte- brate and fish predators were least abundant in (a) temporary habitats, total predator biomass depended ) 2 25 strongly on habitat type, with deep habitats having the highest total biomass and temporary habitats the 20 Temporary Shallow permanent Deep permanent lowest total biomass (P < 0.001, Table 1).

15 Regional and local diversity 10 Our surveys show that northwest Pennsylvania is home to five species of physid snail: Physa acuta, Physa 5 ancillaria (Say), Physa gyrina (Say), Physa vernalis

Invertebrate biomass density (g m (Taylor and Jokinen), and Aplexa elongata (Say). Local 0 0 10 20 30 40 50 60 diversity averaged 1.2 physid species per site, and Pond hydroperiod or depth thus beta diversity (ratio of regional to local diversity) was 4.1, illustrating limited community membership (b) 60 at local scales and substantial heterogeneity in species )

2 50 composition. The level of local diversity was very Temporary Shallow permanent Deep permanent similar across habitat types, as mean local diversity of 40 temporary habitats was equal to 1.15 species per pond, shallow permanent habitats averaged 1.20 30 species, and deep habitats averaged 1.25 snail species per pond. 20

10 Fish biomass density (g m Assemblage coherence and turnover

0 An examination of the distribution of the five physids 0 10 20 30 40 50 60 Pond hydroperiod or depth across the 61 ponds and lakes shows that there were clear patterns of restricted distribution and species Fig. 1 Invertebrate (top) and fish (bottom) predator abundance replacement along the habitat gradient (Fig. 2). Our along the freshwater habitat hydroperiod gradient. Each data analysis confirmed that species distributional patterns point represents a pond or lake, ranked from shallowest to were non-random with respect to hydroperiod. When deepest. Solid lines showing trends are produced by Loess smoothing. Vertical dashed lines represent the permanence and the data are organised into a species-by-site incidence predator transitions. matrix (e.g. Fig. 2) and sites are ordered according to

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1195 modelled with three habitat variables: maximum depth, invertebrate biomass and fish biomass. The Physa acuta reduced model successfully predicted the dominant Physa ancillaria physid species for 62% of the 21 ponds included in the analysis. Physa gyrina Focusing on those niche dimensions identified in Physa vernalis the discriminate analysis as the best predictors of physid species composition, a simple description of Aplexa elongata niche differentiation emerges. Physa vernalis was dominant in sites with the shortest hydroperiod, a Temporary Fishless Fish low abundance of invertebrate predators and no fish Pond hydroperiod or depth predators (Table 2). Aplexa elongata sites were on Fig. 2 Distributions of five species of physid snails along the average slightly deeper and more persistent than hydroperiod and depth gradient. Vertical dashed lines represent those dominated by P. vernalis. Aplexa sites had the the permanence and predator transitions. lowest invertebrate biomass and also lacked fish (Table 2). Sites dominated by P. gyrina were interme- their hydroperiods, the resulting pattern of incidence diate with respect to depth and fish biomass, but is highly coherent (P < 0.005). The observed matrix biomass of invertebrate predators was highest in P. (Fig. 2) contains 53 embedded absences, whereas null gyrina sites. Physa ancillaria was found only in the matrices conserving row and column totals contained deepest and largest of the lakes surveyed. These lakes an average of 128 ± 22 (±1 SD) embedded absences. are characterised by very high fish biomass and Thus, much of the variation in assemblage structure intermediate invertebrate biomass (Table 2). Physa from one pond or lake to another is explained by acuta occupied a niche with mean values intermediate hydroperiod. We also found that the number of to the other species, but it was so ubiquitous as to species replacements along this gradient was higher make analysis of niche centroids relatively uninfor- that would be expected from a random distribution of mative. species (P = 0.035). Thus species replacement is more With regard to relative habitat breadth, all metrics important than nesting in describing the incidence of (analysis of frequency of occurrence, C.V. of abiotic species along the habitat gradient. variables, and C.V. of predator biomass) yielded the same rank-ordering of specialisation. Physa ancillaria had the narrowest habitat distribution, followed by Realised niches and specialisation P. vernalis, A. elongata, and P. gyrina (Table 3). P. acuta Discriminate analysis showed significant multivariate had by far the broadest habitat distribution, as it was habitat differentiation among species (Wilk’s Lamb- the dominant species in 39 of the 60 sites studied, and da = 0.020, F9,36 = 15.9, P < 0.001). With backward it occupied every habitat from highly ephemeral pools elimination, patterns of physid dominance was best to deep permanent lakes (Table 3).

Table 2 Abiotic and biotic characteristics of ponds dominated by each of the five P. vernalis A. elongata P. gyrina P. acuta P. ancillaria physid species, with species ranked by Mid-summer depth (cm) 3.0 4.5 40 60 1140 depth of the ponds and lakes in which 2 Area (m ) 725 300 770 360 278,500 they are found )1 Alkalinity (mg L ) 131 62.5 84 72.5 164 pH 6.8 6.8 7.1 7.7 8.1 ) Conductivity (lScm 1) 264 260 160 197 260 ) Invertebrate biomass (g m 2) 1.40 0.33 2.4 2.06 0.50 ) Fish biomass (g m 2) 0 0 5.5 6.03 32.6 ) Total predators (g m 2) 1.40 0.33 7.9 8.09 33.1

Median values are shown for abiotic variables and means for predator biomass values. Niche dimensions in bold are those identified by stepwise discriminate analysis as the best predictors of physid species composition.

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 1196 A. M. Turner and S. L. Montgomery

Table 3 Relative specialisation of the five physid species, measured as the mean variation in abiotic characteristics and mean variation in predator biomass of the habitats in which each species is dominant. Abiotic variance is the mean of the coefficient of variation (standard deviation ⁄ mean) for depth, surface area, alkalinity, pH, and conductivity. Predator variance is the mean of the coefficient of variation for biomass of predatory invertebrates and fish

P. vernalis A. elongata P. gyrina P. acuta P. ancillaria

Number of ponds dominated 2 4 13 39 2 Mean C.V. – Abiotic variables 0.41 0.58 1.09 1.33 0.18 Mean C.V. – Predator biomass 1.18 1.21 1.55 1.72 0.62

lated a model which asserts that species turnover is Discussion driven by fitness tradeoffs associated with predation. Further development of a theory of lacustrine com- Physiological, morphological or behavioural traits munity structure depends on a detailed knowledge of that confer protection against the predators of larger species replacement patterns, but comprehensive ponds and lakes often incur costs that put the bearer studies of species turnover along the entire freshwater of those traits at a disadvantage in ponds lacking habitat gradient, including quantitative estimates of predators or with different sorts of predators (e.g. predator abundances, are largely lacking. Here we Werner & McPeek, 1994; Wissinger et al., 1999; Well- show that species belonging to a single freshwater born, 2002; McPeek, 2004; Werner et al., 2007). Con- gastropod family occur across the entire habitat sistent with the Wellborn hypothesis, we found that gradient, but individual species have a restricted turnover in physid species composition coincided and non-random distribution with respect to hydro- with shifts in the type and abundance of snail period (with one important exception). As a result, predators. Also consistent with the predation hypo- there is substantial turnover of gastropod species, and thesis, a number of experimental field studies have regional diversity is much higher than mean local shown that both fish and invertebrate predators diversity. influence gastropod abundance (Kesler & Munns, Descriptive studies aimed at detecting broad-scale 1989; Bro¨nmark, 1992; Bro¨nmark, Klosiewski & Stein, patterns in the species composition of freshwater snail 1992; Martin et al., 1992; Osenberg, Mittelbach & assemblages have traditionally focused on pH and Wainwright, 1992; Lodge et al., 1994; Nystrom & alkalinity (Okland, 1983; Bro¨nmark, 1985; Pip, 1986; Perez, 1998; Chase, 2003; Turner & Chislock, 2007), Jokinen, 1987; Carlsson, 2000). Lodge et al. (1987) but these studies have focused on population regula- reviewed this work and concluded that calcium tion within lakes, not patterns of replacement among availability likely limits snails at a regional scale, but lakes. Ultimately, predation experiments involving hypothesised that competition, predation and pond snail species from across the habitat gradient are drying were important at a local scale. The patterns necessary in order to evaluate the full role of preda- described in this study are consistent with the tors in structuring species composition of freshwater predictions of Lodge et al. (1987), as we found that snail assemblages. variation in pH, conductivity and alkalinity was not Alternatively, snail species turnover along the related to local species composition, but depth and freshwater habitat gradient may be caused by shifts predator biomass did successfully predict species in the type and severity of abiotic stress, independent composition. However, the specific mechanisms of predation. Survival in temporary ponds demands responsible for the restricted distribution of fresh- the ability to aestivate and tolerate desiccation and water gastropods are still poorly understood, and we high soil temperatures. Survival in shallow ponds turn here to a brief evaluation of the evidence for two likely requires an ability to tolerate low winter oxygen processes: predation and stress tolerance. concentrations and high sulphide concentrations in Several studies of aquatic invertebrates suggest that summer. Deep ponds are relatively benign environ- species replacement along the habitat gradient is ments, and tradeoffs associated with adaptations driven by shifts in the abundance and identity of the aimed at dealing with dessication and anoxia may top predators (Cook & Streams, 1984; McPeek, 1990, put the animal at a disadvantage in benign environ- 1998; Wellborn, 2002). Wellborn et al. (1996) formu- ments. The ‘‘stress-tolerance’’ model has received

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1197 substantial empirical support for other ecosystems Batzer & Wissinger (1996) proposed a two-state (Grime, 1977; Menge & Sutherland, 1987), but has not model in which temporary habitats are dominated by been extensively tested in freshwater ecosystems. insect predators and permanent habitats are domi- There is some evidence that P. gyrina dominates in nated by fish. In contrast, Wellborn et al. (1996) ponds that suffer from low winter oxygen concentra- developed a model with three habitat types: tempo- tions, suggesting that P. gyrina can better tolerate the rary ponds characterised by low predator density; hypoxia of these ponds (A. M. Turner, unpubl. data). permanent, fishless ponds with high invertebrate Other environmental stressors that limit gastropod density; and fish ponds with low invertebrate but species distributions may include parasitism or high fish density. Our predator data show that blooms of toxic algae (Bernot & Lamberti, 2008; shallow, permanent ponds have low fish biomass Gerard, Carpentier & Pallisson, 2008). Future studies relative to deeper ponds and lakes, but a high biomass could address relative tolerance of snail species to of predatory invertebrates relative to temporary environmental stress and integrate the predation and ponds, a pattern consistent with the three-state model stress tolerance models by investigating potential of Wellborn et al. The success of predatory insects and tradeoffs between stress tolerance and anti-predator fish in short hydroperiod ponds may vary with defences. The patterns of species composition shown latitude. At high latitudes, summer drying and a here are probably shaped by both predation and stress short growing season excludes many insects from tolerance, so manipulative approaches will be neces- temporary ponds, and winter anoxia excludes fish sary in order to elucidate the underlying mechanisms. from shallow permanent ponds, allowing insects to Some studies have found that abundance of pred- dominate. Thus, the two-state model proposed by atory invertebrates across the habitat gradient is Batzer and Wissinger may well apply to lower lowest in temporary ponds and reaches its peak in latitude wetlands, especially those that experience long-hydroperiod fishless ponds (Wiggins, MacKay & their wet season in summer. Smith, 1980; Schneider & Frost, 1996; Wellborn et al., Wellborn et al. (1996) asserted that virtually every 1996), but others assert that temporary ponds have a type of animal inhabiting freshwater has a limited high density of predators (Lake, Bayly & Morton, distribution across the lentic permanence – predation 1989). Our quantitative survey showed that there gradient. Our data show that three of the physids are were significant differences among pond types in highly specialised and have narrow habitat distribu- biomass of predatory invertebrates. Average inverte- tions (P. vernalis, A. elongata, and P. ancillaria). One has brate biomass in shallow ponds was about two-fold a moderately broad distribution (P. gyrina), and one, that of deep ponds and lakes and three-fold that of Physa acuta, has an exceptionally broad habitat distri- temporary habitats. bution. Physa acuta occupied more than three quarters Deep ponds and lakes had a mean fish biomass of the ponds studied and were found in the shallow- about 30-fold higher than that of shallow and tempo- est, most temporary ponds as well as the deepest, rary ponds. Water depth was an excellent predictor of largest lakes. Further, Physa acuta is not even restricted whether a pond would contain a well developed fish to the range of systems studied here; it can be community, as all but one of the ponds with a mid- abundant in streams and rivers (Jokinen, 1992), and summer depth >150 cm contained a well developed is also abundant in the benthos of the Laurentian fish community (>2 species and biomass density Great Lakes (Stewart, Miner & Lowe, 1998). Several ) >5 g m 2), and all but one of the ponds with a other studies of aquatic invertebrates have identified midsummer depth <150 cm had a poorly developed species with a broad habitat distribution (Hanson & fish community or no fish at all. These shallow ponds Swanson, 1989; Muthukrishnan & Palavesam, 1992). often contained mudminnows, fathead minnows or Thus, it is clear that in some cases aquatic taxa break sticklebacks, all of which are tolerant of the winter the rule of limited distributions and that the ubiquity anoxia that restricts the development of fish commu- of specialisation in freshwater may have been over- nities. The ‘‘fish’’ transition was fairly abrupt, as stated. ponds with a midsummer depth >150 cm had a much On a broader geographic scale, it appears that higher fish biomass than ponds only slightly shal- habitat breadth is related to geographic range. Physa lower (Fig. 1). vernalis and P. ancillaria, both habitat specialists,

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 1198 A. M. Turner and S. L. Montgomery occur only in northeastern North America and have responsible for limiting the distribution of physids each been reported from only a limited number of remain unclear, with both differential vulnerability to localities (Jokinen, 1983, 1992; Taylor & Jokinen, predators and differential tolerance of abiotic stress 1984). Aplexa elongata and P. gyrina, found in this likely playing a role in producing species turnover. study to be moderately generalised species, both Pulmonate snails provide a powerful model system occupy a broad range across North America (Har- with which to evaluate questions regarding the man & Berg, 1971; Burch, 1988; Martin, 1999). Physa mechanisms responsible for regulation of community acuta, a super-generalist, is found on every continent structure. of the world and is abundant nearly everywhere it occurs (Dillon et al., 2002). Evidence suggests that Acknowledgments Physa acuta is native to North America and has spread around the world in just the past 200 years A number of colleagues helped with the survey work, (Dillon et al., 2002; Bousset et al., 2004). Considering often under adverse conditions. Field assistants the staggering breadth of its geographic distribution include Randy Bernot, Christy Mower, Beth Brokaw, and the high densities typically achieved by Physa Cheri LaFlamme, Rachel Martin, Heidi Lappi, Sarah acuta populations, this species is surely among the Turner, Scott Ray, Mike Chislock, Rebekah Turner, most successful of all freshwater animals. It is and Martin Heben. Jon Chase, Julia Butzler, Jason important to know how frequent such super-gener- Hoverman, and Nancy Schoeppner helped find addi- alists are, as they lower diversity and tend to be tional ponds. This study was supported by the invasive. National Science Foundation (Grant No. 0444939) There have been several studies of phenotypic and is a contribution of the Pymatuning Laboratory plasticity of Physa acuta (Crowl & Covich, 1990; of Ecology. DeWitt et al., 2000; Turner & Montgomery, 2003; Turner, Turner & Lappi, 2006). These show that References P. acuta adjusts its behaviour, morphology and life- history in response to environmental factors, includ- Baker F.C. (1927) Molluscan associations of While Lake, ing predators. We hypothesise that the degree of Michigan: a study of a small inland lake from an plasticity in behaviour and morphology facilitates ecological and systematic viewpoint. Ecology, 8, 353– P. acuta’s wide breadth in habitat and geography. 370. Comparative studies, evaluating the relative plasticity Batzer D.P. & Wissinger S.A. (1996) Ecology of insect communities in non-tidal wetlands. Annual Review of of physid species and relating plasticity to habitat Entomology, 41, 75–100. breadth, have not yet been conducted. Mower & Bernot R.J. & Lamberti G.A. (2008) Indirect effects of a Turner (2004) did compare the morphological traits parasite on a benthic community: an experiment with and behavioural plasticity of P. acuta and Stagnicola trematodes, snails and periphyton. Freshwater Biology, elodes, a habitat specialist, and found that the degree of 53, 322–329. shell development failed to correctly predict habitat Bousset L., Henry P.Y., Sourrouille P. & Jarne P. (2004) distributions, but that a consideration of behavioural Population biology of the invasive freshwater snail plasticity did. By viewing the degree of plasticity as a Physa acuta approached through genetic markers, species trait, we were able to gain insight into patterns ecological characterization and demography. Molecular of species replacement that would not otherwise have Ecology, 13, 2023–2036. been possible. Bro¨nmark C. (1985) Freshwater snail diversity: effects of In sum, our results show that predator abundance pond area, habitat heterogeneity, and isolation. Oeco- logia, 67, 127–131. and identity change in a predictable manner across Bro¨nmark C. (1992) Leech predation on juvenile fresh- the freshwater habitat gradient. The distribution of water snails: effects of size, species, and substrate. physid snails supports the contention that, with Oecologia, 91, 526–529. respect to the freshwater habitat gradient, most lentic Bro¨nmark C., Klosiewski S.P. & Stein R.A. (1992) Indirect species have limited distributions and substantial effects of predation in a freshwater benthic food chain. species turnover, but it also provides an interesting Ecology, 73, 1662–1674. example of a striking exception. The mechanisms

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1199

Brown K. & DeVries D. (1985) Predation and the Harman W.N. & Berg C.O. (1971) The freshwater snails distribution and abundance of a pond snail. Oecologia, of central New York. Search (Cornell University Agri- 66, 93–99. cultural Experiment Station), 1, 1–68. Burch J.B. (1988) North American Freshwater Snails. Huston M.A. (1994) Biological Diversity: The Coexistence of Walkerana, 2, 1–80. Species on Changing Landscapes. Cambridge University Butkas K.J. & Ostrofsky M.L. (2006) The status of unionid Press, Cambridge. and dreissenid mussels in northwestern Pennsylvania James F.C. & Mcculloch C.E. (1990) Multivariate inland lakes. The Nautilus, 120, 106–111. analysis in ecology and systematics: panacea or Carlsson R. (2000) Species-area relationships, water pandora’s box? Annual Review of Ecology and System- chemistry, and species turnover of freshwater snails atics, 21, 129–166. on the Aland islands, southwestern Finland. Journal of Jokinen E.H. (1983) The Freshwater Snails of Connecticut. Molluscan Studies, 67, 17–26. State Geological and Natural History Survey of Con- Chase J.M. (2003) Strong and weak trophic cascades necticut, Hartford, Bulletin 109. along a productivity gradient. Oikos, 101, 187–195. Jokinen E. (1987) Structure of freshwater snail commu- Clampitt P.T. (1970) Comparative ecology of the snails nities: Species-area relationships and incidence cate- Physa gyrina and Physa integra (Basommatophora: gories. American Malacological Bulletin, 5, 9–19. Physidae). Malacologia, 10, 113–151. Jokinen E.H. (1992) The Freshwater Snails (Mollusca: Cook W.L. & Streams F.A. (1984) Fish predation on Gastropoda) of New York State. New York State Museum, Notonecta (Hemiptera): relationship between prey risk Albany, Bulletin 482. and habitat utilization. Oecologia, 64, 177–183. Kesler D.H. & Munns W.R. Jr (1989) Predation by Crowl T.A. & Covich A.P. (1990) Predator-induced life- Belastoma flumineum (Hemiptera): an important cause history shifts in a freshwater snail: a chemically of mortality in freshwater snails. Journal of the North mediated and phenotypically plastic response. Science, American Benthological Society, 8, 342–350. 247, 949–951. Lake P.S., Bayly I.A.E. & Morton D.W. (1989) The DeWitt T.J., Robinson B.W. & Wilson D.S. (2000) Func- phenology of a temporary pond in western Victoria, tional diversity among predators of a freshwater snail Australia, with special reference to invertebrate suc- imposes an adaptive tradeoff for shell morphology. cession. Archiv fur Hydrobiologie, 115, 171–202. Evolutionary Ecology Research, 2, 129–148. Leibold M.A. & Mikkelson G.M. (2002) Coherence, Dillon R.T. Jr (2000) The Ecology of Freshwater Mollusks. species turnover, and boundary clumping: elements Cambridge University Press, New York, NY. of meta-community structure. Oikos, 97, 237–250. Dillon R.T. Jr, Wethington A.R., Rhett J.M. & Smith T.P. Leibold M.A., Holyoak M., Mouquet N. et al. (2004) (2002) Populations of the European freshwater pulmo- The metacommunity concept: a framework for nate Physa acuta are not reproductively isolated from multi-scale community ecology. Ecology Letters, 7, American Physa heterostropha or Physa integra. Inverte- 601–613. brate Biology, 121, 226–234. Lodge D.M., Brown K.M., Klosiewski S.P., Stein R.A., Garcia E.A. & Mittelbach G.G. (2008) Regional coexis- Covich A.P., Leathers B.K. & Bro¨nmark C. (1987) tence and local dominance in Chaoborus: species Distribution of freshwater snails: spatial scale and the sorting along a predation gradient. Ecology, 89, 1703– relative importance of physicochemical and biotic 1713. factors. American Malacological Bulletin, 5, 73–84. Gerard C., Carpentier A. & Pallisson J.-M. (2008) Long- Lodge D.M., Kershner M.W., Aloi J.E. & Covich A.P. term dynamics and community structure of fresh- (1994) Effects of an omnivorous crayfish (Orconectes water gastropods exposed to parasitism and other rusticus) on a freshwater littoral food web. Ecology, 75, environmental stressors. Freshwater Biology, 53, 470– 1265–1281. 484. Martin S.M. (1999) Freshwater snails (Mollusca: Gastro- Grime J.P. (1977) Evidence for the existence of three poda) of Maine. Northeastern Naturalist, 6, 39–88. primary strategies in plants and its relevance to Martin T.H., Crowder L.B., Dumas C.F. & Burkholder ecological and evolutionary theory. American Natural- J.M. (1992) Indirect effects of fish on macrophytes in ist, 111, 1169–1194. Bays Mountain Lake: evidence for a littoral trophic Hanson B.A. & Swanson G.A. (1989) Coleoptera species cascade. Oecologia, 89, 476–481. inhabiting prairie wetlands of the Cottonwood Lake McPeek M.A. (1990) Determination of species composi- Area, Stutsman County, North Dakota. Prairie Natu- tion in the Enallagma damselfly assemblages of ralist, 21, 49–57. permanent lakes. Ecology, 71, 83–98.

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 1200 A. M. Turner and S. L. Montgomery

McPeek M.A. (1998) The consequences of changing the Spence J.R. (1989) The habitat template and life-history top predator in a food web: a comparative experimen- strategies of pond skaters (Heteroptera: Gerridae): tal approach. Ecological Monographs, 68, 1–23. reproductive potential, phenology, and wing dimor- McPeek M.A. (2004) The growth ⁄predation risk tradeoff: phism. Canadian Journal of Zoology, 67, 2432–2447. so what is the mechanism? American Naturalist, 163, Stewart T.W., Miner J.G. & Lowe R.L. (1998) Macroin- 88–111. vertebrate communities on hard substrates in western McPeek M.A., Grace M. & Richardson J.M.L. (2001) Lake Erie: structuring effects of Dreissena. Journal of Physiological and behavioral responses to predators Great Lakes Research, 24, 868–879. shape the growth ⁄predation risk trade-off in damsel- Stoks R. & McPeek M.A. (2003) Predators and life flies. Ecology, 82, 1535–1545. histories shape Lestes damselfly assemblages along a Menge B.A. & Sutherland J.P. (1987) Community regu- freshwater habitat gradient. Ecology, 84, 1576–1587. lation: variation in disturbance, competition, and Taylor D.H. & Jokinen E.H. (1984) A new species of predation in relation to environmental stress and freshwater snail (Physa) from seasonal habitats in recruitment. American Naturalist, 130, 730–757. Connecticut. Freshwater Invertebrate Biology, 3, 189– Mower C. & Turner A.M. (2004) Behavior, morphology 202. and the coexistence of two pulmonate snails with Turner A.M. & Chislock M.F. (2007) Dragonfly predators molluscivorous fish: a comparative approach. American influence biomass and density of pond snails. Oecolo- Malacological Bulletin 19, 39–46. gia, 153, 407–415. Muthukrishnan J. & Palavesam A. (1992) Secondary Turner A.M. & Montgomery S.L. (2003) Spatial and production and energy flow through Kiefferulus barb- temporal scales of predator avoidance: experiments itarsis (Diptera: Chironomidae) in tropical ponds. with fish and snails. Ecology, 84, 616–622. Archiv fur Hydrobiologie, 125, 207–226. Turner A.M. & Trexler J.C. (1997) Sampling aquatic Nystrom P. & Perez J.R. (1998) Crayfish predation on the invertebrates from marshes: evaluating the options. common pond snail (Lymnaea stagnalis): the effect of Journal of the North American Benthological Society, 16, habitat complexity and snail size on foraging effi- 694–709. ciency. Hydrobiologia, 368, 201–208. Turner A.M., Turner S.L. & Lappi H.A. (2006) Learning, Okland J. (1983) Factors regulating the distribution of memory, and predator avoidance by freshwater snails: freshwater snails (Gastropoda) in Norway. Malacologia, effects of experience on predator recognition and 24, 277–288. defensive strategy. Animal Behaviour, 72, 1443–1450. Osenberg C.W., Mittelbach G.G. & Wainwright P.C. Wellborn G.A. (1994) Size based predation and the (1992) Two-stage life histories in fish: the interaction evolution of prey life histories: a comparative study between juvenile competition and adult performance. of freshwater amphipod populations. Ecology, 75, Ecology, 73, 255–267. 2104–2017. Peters R.H. (1983) The Ecological Implications of Body Size. Wellborn G.A. (2002) Trade-off between competitive Cambridge University Press, Cambridge. ability and antipredator adaptation in a freshwater Pip E. (1986) The ecology of freshwater gastropods in the amphipod species complex. Ecology, 83, 129–136. central Canada region. The Nautilus, 100, 56–66. Wellborn G.A., Skelly D.K. & Werner E.E. (1996) Mech- Schneider D.W. & Frost T.M. (1996) Habitat duration and anisms creating community structure across a fresh- community structure in temporary ponds. Journal of the water habitat gradient. Annual Review of Ecology and North American Benthological Society, 15, 64–86. Systematics, 27, 337–363. Schneider J.C., Laarman P.W. & Gowing H. (2000) Werner E.E. & McPeek M.A. (1994) Direct and indirect Chapter 17: length-weight relationships. In: Manual of effects of predators on two anuran species along an Fisheries Survey Methods II: with Periodic Updates (Eds environmental gradient. Ecology, 75, 1368–1382. J. Schneider & C. James ), pp. 1–16. Michigan Werner E.E., Skelly D.K., Relyea R.A. & Yurewicz K.L. Department of Natural Resources, Ann Arbour, Fish- (2007) Amphibian species richness across environmen- eries Special Report 25. tal gradients. Oikos, 116, 1697–1712. Skelly D.K. (1996) Pond drying, predators, and the Wethington A.R. (2003) Phylogeny, Taxonomy, and Evolu- distribution of Pseudacris tadpoles. Copeia, 1996, 509– tion of Reproductive Isolation in Physa (Pulmonata: Phys- 605. idae). PhD Thesis, University of Alabama, Tuscaloosa. Skelly D.K. (1997) Tadpole communities: pond perma- Wethington A.R. & Lydeard C. (2007) A molecular nence and predation are powerful forces shaping the phylogeny of physidae (Gastropoda: Basommatopho- structure of tadpole communities. American Scientist, ra) based on mitochondrial DNA sequences. Journal of 85, 36–45. Molluscan Studies, 73, 1–17.

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201 Species turnover along the freshwater habitat gradient 1201

Whittaker R.H. (1956) Vegetation of the Great Smoky Appendix S1. Abiotic and biotic data for the 61 Mountains. Ecological Monographs, 26, 1–80. study ponds and lakes. T, S, and D denote temporary, Wiggins G.B., MacKay R.J. & Smith I.M. (1980) Evolu- shallow, and deep ponds and lakes. Latitude and tionary and ecological strategies of animals in annual longitude are degrees and minutes. Species codes are temporary ponds. Archiv fur Hydrobiologie, 58, 97–206. as follows: A = Physa acuta,E=Aplexa elongata, Wissinger S.A., Whiteman H.H., Sparks G.B., Rouse G.L. V=Physa vernalis,G=Physa gyrina, and Anc = Physa & Brown W.S. (1999) Foraging trade-offs along a ancillaria. ‘‘TPred’’ is total predator biomass, ‘‘Inv predator-permanence gradient in subalpine wetlands. Pred’’ is biomass of invertebrate predators. Ecology, 80, 2012–2116. Zaret T.M. (1980) Predation and Freshwater Communities. Please note: Wiley-Blackwell are not responsible for Yale Univ. Press, New Haven, CT. the content or functionality of any supporting mate- rials supplied by the authors. Any queries (other than missing material) should be directed to the corre- Supporting Information sponding author for the article. Additional Supporting Information may be found in the online version of this article: (Manuscript accepted 2 January 2009)

2009 Blackwell Publishing Ltd, Freshwater Biology, 54, 1189–1201