- Top of page
Over the last two decades the importance of regional processes in determining local species richness has become apparent (Ricklefs & Schluter 1993). Tonn et al. (1990) suggested that at the regional scale factors such as barriers and glaciation events are likely to determine species richness and that species that overcome such factors are subject, at a more local scale, to abiotic and biotic influences. Regional, i.e. historical, processes should become more apparent as vagility decreases. Many of the ecological hypotheses proposed to explain variation in species richness invoke spatial and temporal variation in environmental conditions (Huston 1994; Brown & Lomolino 1998). Southwood (1977, 1988) argued that habitat variation in ecological time and space imposes constraints on the ecological characteristics of species. For example, species cope with fluctuating environments by being generalists or by avoiding unfavourable conditions, either through inactivity, e.g. diapause, or movement, e.g. migration. Dynesius & Jansson (2000; Jansson & Dynesius 2002) examined habitat variation on a much longer, historical, time scale. They also concluded that climatic variation selected for vagility and generalism. Generalists consuming widespread resources will have large range sizes and this and their vagility will result in increased rates of gene flow with reduced rates of speciation and extinction (though vicariant events may become more frequent).
Schlosser (1987) presented a model of stream fish assemblages based on habitat stability and biotic interactions. Greater flow variation and harsh winter conditions in headwater habitats favours small species with high reproductive rates and good colonizing ability while in the more favourable conditions and greater habitat heterogeneity found in the lower reaches biotic interactions are more important. Consistent with this, Poff & Allan (1995), Oberdorff, Hugueny & Vigneron (2001) and Cattanéo (2005) showed that flow variability influenced the functional organization of fish assemblages and/or demonstrated effects on species richness. On a longer time scale, the high levels of endemism in ancient lakes have been attributed to the longer periods of time available for speciation in such habitats compared with most lakes which have lifetimes of less than 20 000 years (Russell-Hunter 1978; Cohen & Johnston 1987).
The historical biogeography of the European freshwater fauna is reasonably well known (Banarescu 1991). In the Neogene much of southern and central Europe was covered by seas: this area was subsequently recolonized from the north and east (Bianco 1990; Economidis & Banarescu 1991). However, between 115 000 and 10 000 years bp the northern European fauna was eliminated by successive glaciations (Andersen & Borns 1994). The biogeographical evidence indicates that these glaciated areas were recolonized mainly from the Ponto-Caspian region, and particularly from the middle and lower sections of the Danube basin. This conclusion has been supported by recent phylogeographical studies (see, for example Durand, Persat & Bouvet 1999; Nesbøet al. 1999; Bernatchez 2001; Kontula & Väinölä 2001; Kotlik & Berrebi 2001; Koskinen et al. 2002). Post-glacial expansions of fish from Iberia and the Adriatic were prevented by mountain ranges (Pyrenees, Alps, Dinarides, Stara Planina) and these isolated southern faunas are faunistically much more disparate than those to the north.
Freshwater fish are ideal for examining the role of ecological factors in regional biogeography. Many species have low vagility because they are not migratory and are confined to freshwater, e.g. cyprinid fishes, whereas others are migratory and can live in the sea, e.g. salmonids. Some species are cold or warm water stenotherms and some are stenotopic, being limited to flowing or to still waters. The relative importance of such characteristics in recolonization following the extensive glaciations that covered much of Europe should be reflected by regional differences in species composition. This paper asks what determines the distribution and species richness of European freshwater fish faunas, by examining ecological patterns at the regional scale. In particular, I consider the roles of habitat preference, migratory ability and body size in determining regional scale patterns following the last glacial period. The emphasis is on the faunas of central and northern Europe because it is here that ecological effects on colonization ability following glaciation should be most apparent.
- Top of page
Most of the analyses presented here are based on presence/absence data for native fish species in 25 different regions of Europe (Fig. 1) obtained from Illies (1978). These data were supplemented by information from Maitland (2000) when estimating total species richness but, because ecological information was not available for all 271 species listed by Maitland, most of the analyses were restricted to the 211 species for which such information was available. There is a strong regional bias to these omissions but, as discussed later, this is likely to obscure rather than enhance many of the patterns examined here. Illies’ (1978) ‘ecoregions’ were loosely defined for a wide range of animal taxa, i.e. were not specific to fish, from spatial and ecological criteria; for example, region 4 consists of the alpine areas of France, Switzerland and Austria. While identification of regions based on criteria of relevance to fish faunas is desirable I do not believe that conclusions based on data delimited by, for example, national boundaries would differ. The Pyrenees (region 2) was omitted because its small size makes it biogeographically unrealistic for fish and because it acted as an outlier in regressions.
Figure 1. Regions used by Illies (1978) (reproduced from Limnofauna Europaea with permission). Regions X and Y were not included in the analyses because species lists appeared to be incomplete.
Download figure to PowerPoint
The taxonomy used in Illies does not always accord with that found in current works. Some species recognized by Illies are now regarded as races/subspecies and their distributions were merged. The taxonomy used here follows Maitland (2000), which is based on Kottelat (1997) but without the extreme splitting suggested, for example, for salmonids and coregonids. The analysis considers cyclostomes (lampreys) and true fishes, which spend a significant fraction of their life in freshwaters: some coastal taxa that enter freshwaters only sporadically (Moronidae, Mugilidae, Pleuronectidae, Syngnathidae) were omitted.
A variety of sources (Ladiges & Vogt 1965; Wheeler 1969; Muus & Dahlstrom 1971; Phillips & Rix 1985; Holcik 1986; Pecl 1990; Greenhalgh 1999) were consulted for information on maximum fish lengths and habitat preferences: note that not all observations of size and habitat recorded in these books will have been made independently. Where works disagreed I placed greater emphasis on Holcik (1986), Maitland (2000) and Ladiges & Vogt (1965) and chose a value or habitat that was consistent with these or the majority of sources. Species were identified as occurring in riverine (R), lacustrine (L), generalist (RL, occurring in both habitats) or coastal (C) habitats. Regions were identified as glaciated or unglaciated depending on whether or not they were covered by ice during the last glaciation: only region 16 fell along this boundary and was omitted from the analysis when this might have affected results. In some analyses areas adjacent to glaciated regions were categorized as periglacial (regions 8–11, 13, 16).
Using information in Maitland (2000) fish were scored as plant feeders (1), invertebrate feeders (2) or piscivores (3): species consuming more than one of these categories as adults were given intermediate scores. Spawning substrates were categorized by the degree of substratum stability as stones (1), stones/plants (2), plants (3) or pelagic (4) and the degree of parental care was scored as exposed (1), hidden (2), or protected eggs/young (3).
Nikolsky (1963), McDowall (1988), Smith (1991), Cowx & Welcomme (1998) and Lucas & Baras (2001) were used to identify whether species are diadromous (migrate between sea and freshwater during their life cycle, score = 3), potamodromous (migrate within freshwaters, score = 2) or resident (species that show no more than local movements, score = 1). Some species can show more than one behaviour, e.g. charr and trout can be diadromous, or potamodromous: I identified such species as diadromous as the evidence for diadromy is more firmly established (though see Dodson 1997), as diadromous individuals normally migrate along river channels and as such fish will experience fewer barriers to movement between regions.
The mean latitude and longitude of species was determined as the average of the latitudinal and longitudinal midpoints for each region, weighted by regional area: values will be incorrect for those species whose ranges extend outside the area covered by Illies (up to 60° east). Only 30 of the 271 species extended further east and of these 14 were predominantly northern in their distributions and only seven southern. Range area was calculated as the sum of the areas of the regions in which a species was recorded: this overestimates true range areas for species confined to Europe, particularly for those with small geographical ranges.
To test what effect distance from the source had on the probability of colonization is far from straightforward (see, for example, Olden, Jackson & Peres-Neto 2001). While most species spread from the Ponto-Caspian region and some information is emerging about colonization routes it is not known how the majority of species arrived at a particular location. In the absence of detailed information the distance travelled from the source was determined as the straight line distance between the source population (arbitrarily taken as the mouth of the Sea of Azov, the approximate midpoint of regions 7, 12, 24, 25) to the midpoint of the destination region.
Numerous analyses have shown that species richness varies with area (see references in Rosenzweig 1995). However, the analyses presented here are based on richness per region because species richness varied only weakly with regional area (r2 = 0·18, n = 24, P = 0·04) and the relation was no longer significant when the outlier (Iceland) was omitted: this region has few fish species for reasons unrelated to area.
All (fork) lengths are in mm, all logarithmic transformations are to the base 10 and, unless otherwise stated, all interval estimates are standard errors and all nonlinear trend lines in the figures are fitted by locally weighted scatterplot smoothing (LOWESS). All statistical tests of percentage data were carried out on arcsine-square root transformed values though the mean percentages in the text and tables are the untransformed values. Compositional similarity of fish faunas was measured as relative euclidean distances, which were then clustered by Ward's method. Differences in proportions were tested following the procedures in Fleiss (1973).
Ideally, hypotheses in biogeography should be tested using phylogenetically independent contrasts. While phylogenies are known for some European fish taxa in others there is extreme confusion. For example, in the cyprinids, comprising 50% of the species examined in this study, the phylogeny is not clear even at subfamily level in some groups while the genus Leuciscus is polyphyletic (Briolay et al. 1998; Hänfling & Brandl 2000). Nested anova was used to determine the taxonomic distribution of variance (Harvey & Pagel 1993), using a composite taxonomy based on papers by Briolay et al. (1998), Gilles et al. (1998), Hänfling & Brandl (2000) and Zaragüeta-Bagils et al. (2002). Most of the variance in body size and shape, the only continuously distributed variables and in habitat preference group, a discrete variable that is approximately normally distributed, occurs at the species level (Table 1). Consequently, I treated species as independent data points.
Table 1. The percentage of variance accounted for at successive taxonomic levels for fish. The results are based on a three-level nested analysis of variance of log-transformed data. 0-values were assigned when the calculated variance component was negative; these values were always small relative to the dominant component
|Freshwater fish||Length||69||8||13||11||This study|
|Shape||78||5|| 0||18||This study|
|Habitat preference group||79||9||12|| 0||This study|
|Marine fishes||Length||87||0||14|| 0||Pauly (1980)|
|Brody growth coefficient||86||0||14|| 0||Pauly (1980)|
|Mortality rate||81||0||19|| 0||Pauly (1980)|