Oecologia (1994) 98:193-200 © Springer Verlag 1994 H. P6ys~ • J. Elmberg - R Nummi • K. Sj6berg Species composition of dabbling duck assemblages: ecomorphological patterns compared with null models Received: 31 January 1994 / Accepted: l0 March 1994 Abstract Ecomorphological patterns of breeding dabbling duck (Arias spp.) assemblages were studied in six regions in northern Europe. Observed spacings among species in terms of bill lamellar density and body length were compared with expected spacings based on null models incorporating different levels of constraints (regional species pools, species relative abundances, lake size and habitat requirements of species). Deviations of observed spacings from expected ones were compared with prey abundance and prey size diversity in the lakes. Observed spacings in terms of body length, but not in terms of bill lamellar density, were greater than expected on the basis of null models. The most abundant species were generally relatively more different than less abundant species in terms of body length but not in terms of bill lamellar density. Deviations between observed and expected spacings in terms of body length were more like those predicted by the competition hypothesis in lakes with low food abundance than in lakes with high food abundance. Patterns in bill lamellar spacings were not related to food abundance nor to food size diversity. In general, patterns in body length spacings were consistent with the competition hypothesis whether the null model used in comparisons was constrained or not. Key words Species composition - Body length Bill lamellar density • Null models • Competition H. P6ys~(~) Finnish Game and Fisheries Research Institute, Evo Game Research Station, Kaitalammintie75, FIN-16970 Evo, Finland J. Elmberg • K. Sj6berg Department of Animal Ecology, Swedish Universityof Agricultural Sciences, S-90183 Umeft,Sweden R Nummi Department of Applied Zoology,Universityof Helsinki, RO. Box 27, FIN-00014 Universityof Helsinki, Finland Introduction Factors affecting species composition of animal communities have been studied extensively during the last few decades (e.g. Strong etal. 1984; Diamond and Case 1986; Wiens 1989a). An important development in this field occurred when null models were introduced as a tool to examine whether observed species distributions and combinations reflect nonrandom patterns (citations in Harvey et al. 1983; Quinn and Dunham 1983; Jackson et al. 1992). If a difference between observed and expected patterns based on null models is found, we may then proceed by asking why they differ. Null models have been used especially to study the role of interspecific competition in affecting species composition of communities, and for this ecomorphological community patterns have been one important tool (Wiens 1982, 1989a; Brown and Bowers 1984; Colwell and Winkler 1984; Schluter and Grant 1984; Schoener 1984; Tonkyn and Cole 1986 and references therein). Competition theory predicts a nonrandom pattern in which community members differ more than expected if the same number of species were randomly drawn from an appropriate species pool. According to the hypothesis, there should also be a relationship between abundance and morphology among the species so that abundant species should be more different from other species than rare species (Hanski 1982; James and Boecklen 1984; Wiens 1989a). However, the use of null models in community ecology is not without problems, as indicated by the extensive debate on the topic (citations in Harvey et al. 1983; Quinn and Dunham 1983; Strong et al. 1984; Wiens 1989a; Jackson et al. 1992; Wilson 1993). A general consensus is, however, that null models are usable in community ecology if they are appropriately constructed (Harvey et al. 1983; Haefner 1988; Wiens 1989a). The critical question then is what factors should be incorporated when constructing a null model against which a particular pattern is to be compared. Wiens (1989a) presents a hierarchical scheme of factors contributing to the process of community assembly. After 194 defining the species pool for a given community at least the following factors may affect the probability of a given species becoming a member of a local community (see Wiens 1989a): dispersal ability of the species; population size of the species in the source pool; size of the area; habitat and food available in relation to the requirements of the species; predation; and interspecific competition. We are not aware of any study that has considered even most of these factors. It has often been difficult even to define a realistic species pool from which a random null-model community should be constructed. Failing in this basic requirement is dangerous if one attempts to recognize patterns in species composition likely to have been produced by interspecific competitive interactions. Proper tests for competitively structured patterns should be done with ecologically similar species, for instance, within foraging guilds (see Diamond and Gilpin 1982; Colwell and Winkler 1984; Gilpin and Diamond 1984). In this paper we study whether species composition in terms of species ecomorphological spacing in dabbling duck (Arias spp.) assemblages in six regions in northern Europe reflects any deterministic patterns. We focus on testing observed patterns in ecomorphological spacings of local assemblages against random spacings produced using different level of constraints including species distributions, relative abundances as well as area and habitat requirements. Our approach avoids several of the pitfalls of previous null-model analyses of community assembly (see Harvey et al. 1983; Quinn and Dunham 1983; Wiens 1989a). First, dabbling ducks form a compact foraging guild in waterfowl communities (P6ysfi 1983a, b; see also Nudds 1992), justifying a test for deterministic patterns. Secondly, with the exception of predation, we consider all the factors listed by Wiens (1989a) as potential determinants of species composition of local dabbling duck assemblages. As dabbling ducks have similar food requirements at the beginning of the breeding season (protein-rich animal food, Krapu and Reinecke 1992), we do not use food as a constraint in null models, but rather consider it as a resource to be competed for (see below). Compared with P6ys~i (1984) the present study is also a step forward because he did not consider regional species pools, relative abundances of the species or species area and habitat requirements when constructing null models. Finally, we also study whether differences found between observed patterns and null models are associated with prey abundance and prey size diversity in the environment. This has rarely been done in community ecology studies (see Wiens 1989a, b). We use bill morphology and body length to reflect ecological similarity of the species. These are relevant ecomorphological characters since Nudds and Bowlby (1984) and Nummi (1993) identified an association between lamellar density and prey size in dabbling ducks and P6ys~i (1983a, b, 1986, 1987; see also Thomas 1982; Nummi 1993) found that species differing in body length use different feeding methods and depths. Nudds and Bowlby (1984) suggested that dabbling ducks partition food resources according to prey size whereas P6ys~i (1983a, b) suggested that differences in feeding method and feeding depth also may be important (see also Thomas 1982). Three main questions to be answered here are: (1)Do observed ecomorphological spacings among species in local assemblages differ fom those expected on the basis of null models? (2) Are there any differences between bill lamellar density and body length in terms of the deviation of observed spacings from expected ones? (3)Are deviations between observed and expected spacings related to the abundance and size diversity of food resources in the lakes? Materials and methods Data All field data were gathered by identical methods in six regions in Finland and Sweden in 1990 and 1991 (all data from a given region were gathered in one year). Latitudes of the study regions are given in Table 1 (a map of study regions is given in Elmberg et al. 1993). Details of methods used in bird censuses, invertebrate sampling and habitat measurements are given in Elmberg et al. (1993; see also Elmberg et al. 1992, 1994; Nummi et al. 1994). In brief, in each region, ten lakes were selected to represent the regional gradient of luxuriance of emergent vegetation. For each of the 60 lakes we have data on species composition and pair number of breeding dabbling ducks, as well as area, habitat diversity, and food resource abundance and size diversity. We express the habitat diversity of each lake as its score on the 1st axis of a principal component analysis comprising 18 vegetation variables measured in the field (taxonomic composition, width and height of shoreline vegetation, abundance of floating vegetation and vegetation heterogeneity; see Elmberg et al. 1992, 1994, 1993; Nummi et al. 1994). The 1st axis represents a gradient from lakes with low and narrow belts of sparse emergent vegetation (negative scores on 1st component axis) to lakes with tall, wide and heterogeneous emergent vegetation and abundant floating vegetation (positive scores on 1st component axis). Spearman rank correlations revealed that the occurrence (presence/absence) and breeding density of all species excluding A n a s strepera (found in one lake only) increased with habitat diversity (Spearman's rank correlation, rs>0.273, P<0.05, n=60 in all cases). Prey abundance here is the abundance (weighted by size class) of nektonic and benthic invertebrates and amphibian larvae in the littoral zone at a water depth of 0.25-0.75 m (details in Elmberg et al. 1993; see also Elmberg et al. 1992, 1994). The measure thus gives the abundance of food within reach for dabbling ducks (availability s e n s u Wiens 1984, 1989b). The taxonomic composition of the invertebrate samples (Table 4 in Elmberg et al. 1992) corresponds well with the actual diet of dabbling ducks in the breeding season. Prey were assigned to four size classes (0-2.5 ram, 2.6-7.5 ram, 7.6-12.5 ram, and >12.5 ram), and food size diversity was calculated using Simpson's diversity index. Log-transformed diversity values were used in statistical tests (Elmberg et al. 1993). As the index of food abundance varied more between regions than between lakes within regions (Elmberg et al. 1993), we used food abundance indices that were standardized within each region (Y=0, s2--1). Regions did not differ with respect to food size diversity (Elmberg et al. 1993). We used bill lamellar density (data from Nudds et al. 1994, Palearctic dabbling ducks) and body length (mean of the range values of total length minus tail length; data from Cramp and Simmons 1977) to indicate species spacing in ecomorphological niche space. An association between bill lamellar density and prey size has been found in other studies (references given in introduction). Body length indicates the use of different feeding methods and