Jani Heino, Ecosystem Change Unit, Natural Environment Centre, Finnish Environment Institute, P.O. Box 413, Oulu, Finland. E-mail: firstname.lastname@example.org
1. The aim of this paper is to review literature on species diversity patterns of freshwater organisms and underlying mechanisms at large spatial scales.
2. Some freshwater taxa (e.g. dragonflies, fish and frogs) follow the classical latitudinal decline in regional species richness (RSR), supporting the patterns found for major terrestrial and marine organism groups. However, the mechanisms causing this cline in most freshwater taxa are inadequately understood, although research on fish suggests that energy and history are major factors underlying the patterns in total species and endemic species richness. Recent research also suggests that not all freshwater taxa comply with the decline of species richness with latitude (e.g. stoneflies, caddisflies and salamanders), but many taxa show more complex geographical patterns in across-regions analyses. These complexities are even more profound when studies of global, continental and regional extents are compared. For example, clear latitudinal gradients may be present in regional studies but absent in global studies (e.g. macrophytes).
3. Latitudinal gradients are often especially weak in the across-ecosystems analyses, which may be attributed to local factors overriding the effects of large-scale factors on local communities. Nevertheless, local species richness (LSR) is typically linearly related to RSR (suggesting regional effects on local diversity), although saturating relationships have also been found in some occasions (suggesting strong local effects on diversity). Nestedness has often been found to be significant in freshwater studies, yet this pattern is highly variable and generally weak, suggesting also a strong beta diversity component in freshwater systems.
4. Both geographical location and local environmental factors contribute to variation in alpha diversity, nestedness and beta diversity in the freshwater realm, although the relative importance of these two groups of explanatory variables may be contingent on the spatial extent of the study. The mechanisms associated with spatial and environmental control of community structure have also been inferred in a number of studies, and most support has been found for species sorting (possibly because many freshwater studies have species sorting as their starting point), although also dispersal limitation and mass effects may be contributing to the patterns found.
5. The lack of latitudinal gradients in some freshwater taxa begs for further explanations. Such explanations may not be gained for most freshwater taxa in the near future, however, because we lack species-level information, floristic and faunistic knowledge, and standardised surveys along extensive latitudinal gradients. A challenge for macroecology is thus to use the best possible species-level information on well-understood groups (e.g. fish) or use surrogates for species-level patterns (e.g. families) and then develop hypotheses for further testing in the freshwater realm. An additional research challenge concerns understanding patterns and mechanisms associated with the relationships between alpha, beta and gamma components of species diversity.
6. Understanding the mechanistic basis of species diversity patterns should preferably be based on a combination of large-scale macroecological and landscape-scale metacommunity research. Such a research approach will help in elucidating patterns of species diversity across regional and local scales in the freshwater realm.
Macroecology is the study of broad-scale ecological patterns, including topics such as gradients in species richness, structure of geographical ranges and species-abundance distributions. It combines information from the fields of ecology, biogeography, palaeontology, evolutionary biology and applied statistics to understand how broad-scale processes affect and constrain the organisation of ecological systems at various scales (Brown, 1995; Gaston & Blackburn, 2000). Although the foundations of macroecology date back to the late 1960s and early 1970s (MacArthur & Wilson, 1967; MacArthur, 1972), macroecological research has experienced a rapid expansion in the last two decades following the publication of influential books on the topic (Brown, 1995; Rosenzweig, 1995; Maurer, 1999; Gaston & Blackburn, 2000) and evidenced by the success of journals devoted to the research of large-scale ecological phenomena (Global Ecology and Biogeography, Journal of Biogeography, Diversity and Distributions, Ecography and others). Although the need to understand large-scale patterns is well recognised, most studies in this field of research concern terrestrial vertebrates and higher plants, whereas many smaller organisms and other systems have thus far received considerably less attention (see also Diniz-Filho, De Marco & Hawkins, 2010). In particular, marine and freshwater systems have been examined less rigorously than their terrestrial counterparts (Raffaelli, Solan & Webb, 2005; Heino, 2009). A comparison of terrestrial, marine and freshwater systems will help in assessing the generality of geographical patterns, testing ecological theories and advancing macroecological research.
A macroecological approach to the problem of scale, pattern and processes in freshwater systems should thus be meaningful. Freshwater ecologists have, however, lagged behind terrestrial ecologists in studying macroecological patterns. This may be because many freshwater ecologists are ecosystem-oriented or are concerned with studying assemblage–environment relationships within a drainage basin. The orientation towards holistic understanding of a few focal ecosystems has dominated the study of freshwater systems for a long time (Forbes, 1887; Minshall, 1988). While undoubtedly important, this holistic within-ecosystem approach has separated much of freshwater ecology from general macroecology that relies on describing phenomena across large numbers of sites and sets of species. Although studies examining variation in assemblage structure across large numbers of freshwater ecosystems have appeared in the last three decades, these data could be utilised more widely in macroecological analyses of diversity patterns.
Broad-scale data sets spanning a large number of sites across drainage basins have been collected during various environmental assessment and conservation programs. Apart from studies on freshwater fish (e.g. Leprieur et al., 2011), large data bases on algae (e.g. Passy, 2009), zooplankton (e.g. Shurin et al., 2000), macrophytes (e.g. Rørslett, 1991) and macroinvertebrates (e.g. Heino, Muotka & Paavola, 2003) have only rarely been used to address macroecological questions in terms of diversity patterns in the freshwater realm. Such a research agenda should not only be of interest to macroecologists, but also to conservation biologists and environmental managers aiming to understand anthropogenic effects on species diversity across large scales.
The aims of this paper are to review the current status of macroecological research on species diversity patterns in freshwater systems, compare similarities and differences between freshwater, marine and terrestrial systems, and increase our understanding of the processes underlying the patterns. The focus is on large-scale patterns across freshwater systems, and thus within-ecosystem variation in species diversity and effects of various local abiotic factors and biotic interactions on species diversity will not be considered in detail, as these have been reviewed before (Matthews, 1998; Vinson & Hawkins, 1998; Allan & Castillo, 2007; Hugueny, Oberdorff & Tedesco, 2010). This review is an extension of the author’s previous review of aquatic insect biodiversity (Heino, 2009) and deals with topics such as geographical gradients in species richness, beta diversity, nestedness and the local–regional diversity relationship. The aim is also to unite these patterns by considering the relationships between local and regional species richness (RSR) through turnover in species composition among sites. The review is based on literature searches in the ISI Web of Knowledge and personal files, but the coverage of literature is mostly illustrative and not exhaustive.
Geographical diversity gradients in freshwater organisms: patterns and processes
Latitudinal gradients across regions
The foundations of the latitudinal gradient in RSR (also termed as gamma diversity, Table 1) date back to the early 1800s, and it has been appropriately called as the oldest known ecological pattern (Hawkins, 2001). Thus, it is not a very novel observation that RSR typically decreases with latitude. Such a pattern has been found repeatedly in various terrestrial (reviewed by Willig, Kaufman & Stevens, 2003) and marine taxa (reviewed by Hillebrand, 2004a), although there are notable exceptions as well (reviewed by Gaston & Blackburn, 2000). While most groups of terrestrial plants, insects, birds and mammals attain highest levels in the tropics, we have had only a limited amount of knowledge of freshwater taxa in this regard until recently (see the special issue of Hydrobiologia, Balian et al., 2008).
Table 1. Definitions of main terms considered in this review
Locality refers to an ecosystem (e.g. a stream, a river, a lake and a pond) or a site in an ecosystem (e.g. a stream riffle).
A region refers to a drainage basin, an ecoregion, a country or a geographical grid (e.g. 100 × 100 km grid). A region thus incorporates multiple localities.
Number of species in a locality. Mainly determined by local biotic interactions and abiotic environmental factors, as well as the regional species pool.
Variation in species composition among localities in a region. Mainly determined by environmental heterogeneity across the localities within a region.
Number of species in a region. Mainly determined by large-scale climatic, ecological and historical factors.
The relationships between the alpha, beta and gamma components of diversity has been described as multiplicative (γ = αmean × β) or as additive (γ = αmean + β). Such a partitioning has various statistical properties and has a hierarchical extension, where diversity is partitioned into alpha and beta components at successively larger geographical scales (Whittaker, 1975; Lande, 1996; Loreau, 2000; Godfray & Lawton, 2001).
Species co-occurring in a locality. Also called as an assemblage in this review, although assemblage may also refer to a regional set of species.
A set of local communities connected by the dispersal of constituent species.
Taxonomically best known freshwater taxa have traditionally been vertebrates, such as mammals, birds, amphibians and fish (e.g. Leprieur et al., 2011). However, a few quantitative studies have examined latitudinal or other regional patterns in their species richness. For freshwater birds, there are no clear latitudinal gradients at the regional scale, and species richness typically attains highest levels in mountainous regions (e.g. Himalaya; Buckton & Ormerod, 2002). By contrast, fish species richness at the regional scale seems to peak in equatorial regions and decrease sharply towards the poles (Oberdorff, Guégan & Hugueny, 1995; Matthews, 1998). Examples of very high species richness at the regional scale include the fish faunas of the Amazon, Congo and Mekong River basins, while very low species richness can be found in boreal and arctic river basins (Matthews, 1998; Griffiths, 2010). Although our knowledge is limited for many smaller freshwater organisms, it appears that not all taxa obey the general decline of RSR with latitude at the global scale (Crow, 1993; Covich, 2009).
Until recently, there was virtually no quantitative information on major freshwater taxa except fish with regard to their RSR along latitudinal gradients. Pearson & Boyero (2009) analysed data for the RSR of seven freshwater taxa at the global scale. They found, as expected from above, that fish showed a strong latitudinal decline in species richness, accompanied by similarly strong negative species richness–latitude relationships for frogs and dragonflies. Caddisflies and salamanders showed no latitudinal gradient, while mayflies and stoneflies showed a reversed latitudinal gradient, with higher species richness at high latitudes. These absent or reversed latitudinal gradients may be related to the Linnean and Wallacean shortfalls that are pervasive at low latitudes (Whittaker et al., 2005; Bini et al., 2006). The former refers to the fact that most species are not formally described, and the latter refers to the situation that most species distributions are inadequately known. However, at least stoneflies could also be expected to deviate from the general latitudinal pattern because they are generally adapted to cold- and cool-water environments and thus (most stonefly families) have diversified in mountainous and northerly regions (Ward, 1992; Heino, 2001). However, a comparison of major aquatic insect groups across Europe showed that the species richness–latitude relationships are negative in mayflies, dragonflies and stoneflies, whereas other groups do not exhibit significant latitudinal patterns (Heino, 2009). Thus, patterns may change with regard to the latitudinal range or the spatial extent of the study (e.g. global versus continental versus regional).
The examples mentioned above concerned freshwater animals, but also freshwater plants show interesting latitudinal patterns. Freshwater macrophytes have been known to be relatively diverse at temperate latitudes (Hutchinson, 1975), and it has been suggested that their RSR also diverges from the general pattern of decreasing diversity towards poles (Crow, 1993). Such patterns may well occur in studies where the geographical extent ranges from the tropics to the poles, but different patterns are again likely to be found when studies are conducted at high latitudes only. In the boreal region, for example, the RSR of aquatic macrophytes declines sharply with latitude (Heino, 2001; Heino & Toivonen, 2008).
Very small organisms have been suggested to be ubiquitously distributed (Baas Becking, 1934; Finlay, 2002; Finlay & Fenchel, 2004; de Wit & Bouvier, 2006), and thus may not necessarily show any clear latitudinal gradients in RSR (Hillebrand & Azovsky, 2001; Soininen et al., 2009). Unicellular freshwater algae may belong to this group of organisms, showing no latitudinal gradients. The resolution to this dilemma awaits, however, further studies examining the species status of diatoms using genetic methods. It may well be that there are indeed more species near the equator, but these species cannot perhaps be distinguished based on morphological criteria. However, Vyverman et al. (2007) recently showed that, at least at the genus level, there is a latitudinal decline in regional diatom richness in the Southern Hemisphere, generally high regional richness in the tropics and no clear latitudinal gradient in the Northern Hemisphere.
Similar to overall gradients in species richness, patterns of endemism are best known for fish. In this regard, well-known hotspots of endemic fish species occur in the African Great Lakes and South American drainage basins near the equator, suggesting that endemic species richness shows a latitudinal gradient at the global scale. In the whole Northern Hemisphere, endemic fish richness is high in the river basins, where overall species richness is also high, although these patterns may be partly attributable to different ecological and historical drivers (Oberdorff, Lek & Guégan, 1999). At smaller scales within continents, endemism in the European and North American fish faunas has also been studied rigorously. A general pattern is that the number of endemic species is very low in northerly glaciated areas and increases towards colonisation sources, isolated peninsulas and arid regions. In Europe, endemic species are especially typical in barrier regions around the Mediterranean region, and their species richness declines sharply with latitude (Griffiths, 2006; Reyjol et al., 2007). In North America, high endemism in fish faunas is typical in the drainages of the Mississippi River and the Pacific Northwest south of the glaciated regions (McCallister et al., 1986; Matthews, 1998; Griffiths, 2010). In the Northern Hemisphere, it appears that the regions of high endemic species richness also support high total species richness, although this covariation is not absolute (McCallister et al., 1986; Oberdorff et al., 1999).
Although it appears that at least some freshwater taxa exhibit the classical decline in RSR with latitude, it is of importance to know whether these patterns differ between freshwater, marine and terrestrial realms. Based on a comprehensive meta-analysis, Hillebrand (2004a,b) showed that terrestrial and marine latitudinal gradients were stronger than those in freshwater taxa. The reasons for the weaker freshwater latitudinal gradient can be only speculated, but weaker connectivity of drainage basins and freshwater ecosystems may preclude long-distance dispersal, which may enhance local endemism and prevent freshwater taxa from tracking latitudinal gradients in a way similar to many marine and terrestrial taxa (e.g. Covich, 2009).
Processes behind latitudinal gradients across regions
Among the non-biological mechanisms, the most widely cited is the mid-domain effect, which predicts the highest species richness at the middle of a range bounded by hard boundaries. Across the whole Earth, poles represent the hard boundaries, and species richness should therefore peak near the equator (Colwell & Hurtt, 1994; Colwell, Rahbek & Gotelli, 2004). Although the mid-domain effect seems plausible given the strong latitudinal gradient in widely disparate taxa with highly contrasting biologies, it should be considered only as a null model to which ecological and evolutionary forces may cause deviations (Hawkins, Diniz-Filho & Weis, 2005; Zapata, Gaston & Chown, 2005). Because of the fact that the slopes and strengths of the latitudinal gradient differ among organismal characteristics, habitat types and geographical position, the mid-domain effect may only set a general tendency towards higher species richness near the equator (Hillebrand, 2004a). Furthermore, with regard to freshwater organisms, whose ranges are typically bounded by river basin boundaries (e.g. fishes and amphibians), the basic premise of the mid-domain model may not hold in the context of latitudinal gradients. Thus, the mid-domain model may not, at least alone, account for the latitudinal species richness gradient in such freshwater taxa where it occurs.
Among the ecological explanations underlying the latitudinal gradient, the species-energy hypothesis has received much support (Wright, 1983; Currie, 1991). In short, higher energy should lead to higher biomass or abundance and, thereby, higher species richness. However, this hypothesis has been criticised, because it does not provide a mechanistic link between higher energy and higher species richness (Huston, 1999), but only between higher energy and higher biomass (Gaston & Blackburn, 2000). A further problem with the use of energy as a correlate of species richness is that it covaries with a number of other variables, including air temperature, water temperature, evapotranspiration and productivity (Currie, 1991; Hillebrand, 2004a). Thus, it may be difficult to disentangle the relative importance of energy per se and other covarying factors on the latitudinal species richness pattern. Furthermore, it has been recently shown that terrestrial net primary production does not peak in the tropics, and it is thus unknown if terrestrial species diversity, peaking in the tropics, is truly driven by productivity (Huston & Wolverton, 2009). Similar studies are largely lacking from the freshwater realm (Hugueny et al., 2010), hindering further the examination of the mechanistic basis of geographical species richness gradients in freshwater organisms.
Energy and productivity may also both show scale-dependencies, with linear patterns occurring in across-regions analyses and unimodal patterns in across-ecosystems analyses (Currie, 1991; Waide et al., 1999; Mittelbach et al., 2001; Chase & Leibold, 2002). For freshwater systems, support for the species-energy theory has been found in fish at multiple spatial scales (Guégan, Lek & Oberdorff,1998; Romanuk et al., 2009) and in dragonflies at large scales (e.g. Keil, Simova & Hawkins, 2008), but for some other organisms the evidence appears to be more limited even at large scales (e.g. Loyola, 2009). These exceptions are clearly an important aspect for further studies. A possible reason to the deviating findings for some taxa is too low spatial variability in productivity, which is of course related to the geographical extent of a study.
A second ecological hypothesis relates to the effects of area on RSR. This idea suggests that, given the larger geographical area of the tropics than temperate zones, there are more habitats and refuges available, facilitating the occurrence of larger populations, higher speciation rate and lower extinction rate in the tropics (Terborgh, 1973; Rosenzweig, 1995). This idea should also be contrasted to the actual habitat area for freshwater organisms. If there is no relationship between the availability or area of freshwater ecosystems and latitude, this hypothesis can be refuted as the explanation for the latitudinal richness patterns found for frogs, fish, dragonflies and some other taxa. It is actually obvious that there are excess of lacustrine ecosystems in the northerly high latitudes that, nevertheless, support a low diversity of freshwater organisms (Dehling et al., 2010). However, riverine ecosystems in the tropics receive more rainfall, have more run-off and may present a wider range of habitats from headwaters to river mouth than their more northerly counterparts (Hugueny et al., 2010). Thus, at least for riverine taxa, the actual habitat area for freshwater organisms may be higher in the tropics and partly account for latitudinal gradients.
The evolutionary hypothesis underlying the latitudinal species richness gradients is related to the rate and time available for speciation at different latitudes (e.g. Mittelbach et al., 2007). Speciation is assumed to be higher in the tropics than nearer the poles, which should lead to higher species richness near the equator. Higher speciation rate in the tropics can only lead to latitudinal gradient if the large-scale dispersal of species from the tropics to higher latitudes is limited (Hillebrand, 2004a). For freshwater organisms relying on water courses for dispersal, such limited dispersal might potentially account for the latitudinal gradient if speciation rate is higher in the tropics. However, it is at odds with patterns in a number of freshwater taxa that do not exhibit a latitudinal gradient, yet are obviously limited by their large-scale dispersal abilities.
The historical hypothesis relates mainly to glaciations and especially their effects on the biotas at high latitudes (Whittaker, 1977). During the ice ages, high-latitude biotas were totally eliminated, and these regions were subsequently recolonised after the ice sheets receded. Thus, the biotas we see in the northern parts of Eurasia and North America have formed within a relatively short period of time in the last 10 000 years or so (Pielou, 1991). This means that high-latitude regions may not have had enough time to become recolonised by animals and plants, while low-latitude regions have not suffered such profound time lags or large-scale changes in their biotas during the same time period. The degree to which patterns in freshwater biotas agree with this hypothesis has been under considerable recent research, which suggests that historical factors affect latitudinal and other geographical gradients in fish species richness. For example, it has been found that fish species richness increases towards the Ponto-Caspian region in Europe, as these regions have acted as ice-free refugia during the last ice age and from which the recolonisation of high-latitude regions has occurred in Europe (Griffiths, 2006). In North America, high fish species richness (both overall and endemic) has been suggested to occur in areas that have experienced long periods of relatively stable conditions and that have not been glaciated during the last ice age (Griffiths, 2010). Further historical effects affecting the distributions of freshwater organisms include changes in sea level, flooding and river reversals, and all these effects may have imprints on the present-day diversity patterns (Matthews, 1998). Thus, history plays a significant role in determining present-day fish richness gradients (Hugueny et al., 2010). The degree to which other freshwater taxa show similar patterns and are similarly affected by history awaits further analysis.
In summary, geographical species richness gradients of freshwater taxa, in the cases they exist, may be affected by ecological, evolutionary and historical factors (Fig. 1). There is much scope for research in this field of freshwater macroecology, uniting biogeography, ecology and evolutionary biology. The greatest obstacle for understanding the mechanistic basis of latitudinal richness gradients in freshwater taxa, or lack thereof, is related to the insufficient taxonomic and faunistic/floristic knowledge in the tropics and elsewhere (Mann, 1999; Whittaker et al., 2005; Bini et al., 2006). Only by knowing these basic natural history parameters can we begin to search for causes underlying RSR patterns across latitudinal and other broad geographical gradients. A shortcut for such species-level analyses involves the use of higher-taxon surrogates (e.g. families) that have been utilised successfully at large scales in the terrestrial realm (e.g. Gaston et al., 1995) and in the assessment of local freshwater ecosystems within smaller regions (e.g. Furse et al., 1984). A basic premise of the higher-taxon approach is that the numbers of species and families are highly correlated and that species richness patterns can thus be reliably inferred from family richness patterns.
Geographical gradients across freshwater ecosystems
Patterns detected at the across-regions scale may not hold when species richness gradients across local ecosystems (also termed as alpha diversity) are examined (for definitions below, see Table 1). Nevertheless, there appears to be a rather similar decline in local fish richness along latitudinal gradients as observed in RSR (Griffiths, 1997). However, there is typically wide variation in local fish species richness at given latitude, suggesting a strong role for local limiting factors (Tonn et al., 1990; Griffiths, 1997). Similar to the RSR, these local patterns appear to be poorly understood for a majority of freshwater taxa other than fish at the global scale (Allan & Flecker, 1993; Vinson & Hawkins, 1998).
It has been found, however, that aquatic insects show multiple types of richness–latitude relationships. Based on global literature data on the local genus richness of stream mayflies, stoneflies and caddisflies, it was found that none of these groups showed clear linear latitudinal richness trends (Vinson & Hawkins, 2003). Genus richness in these insect groups peaked at mid-latitudes in North America and South America, and minor peaks occurred in the tropics. While these patterns may be partly attributed to differences in taxonomic knowledge among regions, they also suggest that stream insects may diverge from patterns detected for many major terrestrial taxa in across-ecosystems analyses. A reason for this divergence may be that mayflies, stoneflies and caddisflies as groups contain genera and species showing highly differential temperature preferences (reflecting the latitude where genus radiation may have taken place), with mid-latitude regions possibly providing a wide range of thermal habitats suitable to various kinds of taxa. However, other studies have found either higher (Stout & Vandermeer, 1975; Lake et al., 1994; Jacobsen, Schultz & Encalada, 1997) or lower (Stanford & Ward, 1983; Arthington, 1990; McCreadie, Adler & Hamada, 2005) aquatic insect richness in tropical than temperate streams. These discrepancies suggest that more comprehensive studies especially in tropical streams and spanning several tens of degrees of latitude are needed to determine whether aquatic insects show clear latitudinal gradients at all (Vinson & Hawkins, 2003; Heino, 2009).
In smaller regions, some studies have examined latitudinal patterns in the local species richness (LSR) of freshwater macroinvertebrates. For example, it has been found that the taxa richness of stream macroinvertebrates declines with latitude across Alaska (Wrona et al., 2006), which is in accordance with the decrease in stream insect genus richness from 40°N to 70°N in North America (Vinson & Hawkins, 2003). By contrast, Mori, Murakami & Saitoh (2010) detected a positive relationship between richness and latitude on the Japanese island of Hokkaido. These findings are in disagreement with that of Bêche & Statzner (2009), who found no clear latitudinal gradient in macroinvertebrate genus richness across the United States. Overall, these findings suggest that regional context dependency (e.g. length of the latitudinal gradient examined, region-specific characteristics) is important for varying richness–latitude patterns.
A traditional view on the diversity patterns of microscopic organisms considered them to be ubiquitously distributed with no clear geographical gradients (Baas Becking, 1934; de Wit & Bouvier, 2006). Until recently, this view remained largely unchallenged, but latest research has suggested that not all microscopic organisms are ubiquitously distributed even across small scales (Martiny et al., 2006; Logue & Lindström, 2008), and there are actually signals of geographically structured variation in the local richness of freshwater unicellular diatoms (Vyverman et al., 2007; Soininen et al., 2009). However, the research conducted to date does not suggest a clear linear decline in diatom species richness along latitudinal gradients, but shows more complex spatial structures (Heino et al., 2010a; Passy, 2010). Thus, according to present knowledge at the species level, diatoms do not seem obey the classical latitudinal gradients at local scales, but seem to be restricted more by local resource availability than by climate, evolution and history. By contrast, patterns found at the genus level suggest that local richness is also constrained by historical influences to a considerable degree (Vyverman et al., 2007).
There are at least three reasons why there should not necessarily be strong latitudinal gradients in LSR. First, many of the influential environmental features of freshwater ecosystems do not typically show clear latitudinal variation (e.g. substratum particle size heterogeneity, velocity, nutrients and acidity; but see Passy, 2010). According to the view that species are distributed along multiple environmental gradients through environmental filtering (Keddy, 1992; Poff, 1997), high within-region variability in environmental conditions is very likely to produce lows and highs of species richness among sites. Second, local biotic and abiotic environmental factors may override the influences of large-scale historical and climatic factors on local communities even across broad geographical gradients (Hillebrand, 2004b; Field et al., 2009). This would be the case when local environmental conditions strongly limit species richness across most of the sites in a region. Third, the importance of climatic gradients for freshwater systems differs from that on terrestrial systems. For example, unlike for terrestrial systems, precipitation may not necessarily limit the productivity of freshwater systems, but is more likely to affect species richness through hydrologic regimes (Vinson & Hawkins, 2003; Bêche & Statzner, 2009). However, high precipitation may increase nutrient inputs to freshwater ecosystems and thus also affect productivity (Correll, 1999; Wetzel, 2001).
Although the findings above suggest that LSR is often under regional control, they also underscore the importance to study simultaneously historical processes, drainage basin characteristics and local environmental conditions in the context of understanding variation in LSR (Fig. 1). Although linear patterns have been most common, these findings also point out that there is no single type of the LSR–RSR relationship in the freshwater realm. Likely reasons to the varying findings are related to levels of LSR and RSR, regional environmental features and differences in regional delineations (Angermeier & Winston, 1998; Heino et al., 2003; Marchant et al., 2006; Passy, 2009; Soininen et al., 2009).
Species turnover along geographical and environmental gradients
At the within-region extent, a large number of studies have examined the distance decay relationships along environmental and spatial gradients (e.g. Tuomisto, 2010). In a meta-analysis of distance decay patterns, Soininen et al. (2007b) found that community similarity at large scales was decreasing a little faster at high latitudes than low latitudes, whereas community similarity at small scales showed the opposite, suggesting that patterns of beta diversity are strongly scale-dependent. They also found that terrestrial systems showed higher small-scale turnover than marine and freshwater systems. However, a few studies have examined distance decay in freshwater environments (Shurin et al., 2009). For example, pond phytoplankton community similarity decreased with geographical distances among sites, while zooplankton community similarity did so along both spatial distances and environmental dissimilarities (Soininen et al., 2007a). This finding is in agreement with those from stream invertebrates (Thompson & Townsend, 2006). These findings also suggest that freshwater communities are often structured by both dispersal limitation (e.g. Hubbell, 2001) and species niches differences (e.g. Chase & Leibold, 2003).
Another approach to understanding variation in ecological communities along environmental and spatial gradients is concerned with the use of constrained ordination methods (e.g. Legendre, Borcard & Peres-Neto, 2005). This approach has also been combined with understanding the relative fit of the four metacommunity paradigms (Table 2) to field data of ecological communities (e.g. Cottenie, 2005), although it has also been criticised (e.g. Smith & Lundholm, 2010). Specifically, major criticism of the approach concerns the fact that if influential and spatially structured environmental variables are not included in the constrained ordination model, the significance of pure spatial variables can be misleading. However, this approach is generally considered an important tool for distinguishing the relative importance of the competing metacommunity paradigms.
Table 2. Metacommunity paradigms and support for them in freshwater studies
A perspective that emphasises that resource or environmental gradients affect strongly the local demography of species and the outcome of species interactions and that site quality and dispersal jointly affect local community structure. In this scheme, some degree of dispersal is important for community structure, as it allows species to track variation in local site conditions across a region. There is much support from freshwater systems at the landscape to regional scales.
A perspective that emphasises the effects of immigration and emigration on local population dynamics. Based on this scheme, species can be rescued from local competitive exclusion in communities, where they are poor competitors by immigration from communities in which they are good competitors. This perspective can also be extended to abiotic environmental gradients, and sites that are highly suitable for a species act as sources, while sites that are suboptimal for a species are sinks. Thus, high rates of dispersal from source sites to sink sites maintain species also in the latter, where they would go extinct without continuous immigration. There is some support from freshwater systems within small landscapes.
A perspective in which all species are similar in their competitive ability, movement and fitness. Interactions among populations include random walks that alter relative frequencies of species. The dynamics of local communities thus results both from probabilities of species loss (extinction, emigration) and species gain (speciation, immigration). Species do not respond to environmental heterogeneity, but dispersal limitation mostly determines community structure. There is little support from freshwater studies, although possible especially at very large scales.
A perspective that assumes that patches are identical and that each patch can contain populations. Patches may be occupied or unoccupied. Local communities are limited by dispersal, and spatial dynamics are dominated by local extinction and colonisation. There is some support from freshwater systems at small spatial scales (grain and extent).
In a thorough meta-analysis of terrestrial, marine and freshwater systems, Cottenie (2005) found that community composition varied mostly along environmental gradients (corroborating species sorting, see Leibold et al., 2004), secondarily along both environmental and spatial gradients (corroborating species sorting and mass effects, see Holyoak et al., 2005), and rarely with spatial gradients only (corroborating neutral theory, see Hubbell, 2001). Following that meta-analysis, various case studies of freshwater systems have tried to disentangle the relative roles of spatial and environmental gradients for variation in community composition. The main message from these studies appears to be that if study sites are distributed across a very wide geographical region, spatially structured variation in community composition increases and environmental control decreases (Mykrä, Heino & Muotka, 2007; Bennett et al., 2010; Heino et al., 2010a). This suggests that dispersal limitation increases with increasing geographical extent and species sorting increases with decreasing regional extent. However, Pinel-Alloul, Niyonsenga & Legendre (1995) found that environmental variables were strongly spatially structured and that the combination of environmental and spatial variables accounted for most variability in zooplankton community structure. Because spatial location itself did not appear influential, Pinel-Alloul et al. (1995) also suggested that dispersal was fairly uniform across a relatively large geographical region (c. 236 000 km2). By contrast, in two regional studies of phytoplankton, no environmental or spatial variables were significantly related to community composition (Beisner et al., 2006; Nabout et al., 2009). This suggests that phytoplankton communities show either a high degree of unpredictability or that some important environmental factors were not measured (e.g. grazer abundances). By contrast, at the landscape scale (e.g. 10–100 km2), wetland phytoplankton and zooplankton communities were related as strongly to environmental variables as spatial location (Soininen et al., 2007a). This suggests that these communities are jointly structured by dispersal limitation and species sorting. Within even smaller landscapes (e.g. 1–10 km2), dispersal limitation should be of less importance, and community composition should track changes in environmental conditions among sites or be affected by mass effects. This is what was actually suggested for pond zooplankton communities that were structured by both spatial location and environmental conditions (Cottenie et al., 2003), and for pond phytoplankton communities that were structured by environmental variables (Vanormelingen et al., 2008) across a small landscape.
Despite that a large amount of research has assessed the degree of nestedness, very few freshwater studies have considered the mechanistic basis underlying this pattern. In general, it could be assumed that the same factors driving species richness patterns, including area, isolation and habitat heterogeneity (Huston, 1994; Rosenzweig, 1995), should also drive patterns in nestedness at the local scale (Wright et al., 1998; Heino & Muotka, 2005). This has indeed been found to be the case, although there may be regional differences in the particular drivers of nestedness. For example, Heino et al. (2010b) found that nestedness in stream insects correlated with different factors in each of the eight drainage basins studied, although variables related to habitat size, heterogeneity and adversity ranked the best. Similarly, McAbendroth et al. (2005) found that nestedness was primarily driven by habitat size, habitat features and isolation in pond macroinvertebrates. Thus, it seems that nestedness in freshwater systems is mainly driven by area and habitat characteristics.
Uniting different patterns of species diversity through metacommunity processes
It now appears that the combinations climate-energy and speciation-history largely determine the RSR gradients, which are subsequently modified by more localised ecological factors (Fig. 1). Local communities are thus likely to be under some regional influences, although local biotic interactions and abiotic environmental gradients mainly filter species from the regional pool to coexist in an ecosystem (Keddy, 1992; Poff, 1997; Zobel, 1997). This environmental filtering process within a region is closely related to the idea of species sorting, emphasising that species have different niches and are thus able to occur in a certain limited part of environmental gradients (Chase & Leibold, 2003; Leibold et al., 2004; Holyoak et al., 2005). However, structuring of local communities is more than species sorting along environmental gradients (Table 2), and theoretical and empirical research has shown that local communities may also be structured by mass effects (e.g. Shmida & Wilson, 1985), dispersal limitation (e.g. Hubbell, 2001) or patch dynamics (e.g. Leibold et al., 2004). The relative importance of these four types of mechanisms in accounting for patterns shown by local communities in a region may be hard to distinguish using field data (e.g. Cottenie, 2005), and they should rather be understood as parts of a continuum of possibilities (e.g. Gravel et al., 2006).
Most metacommunities are likely to be structured by species sorting, mass effects, dispersal limitation and patch dynamics in combination, yet their relative importance may be related to the spatial extent of a region (Mykräet al., 2007; Bennett et al., 2010; Heino et al., 2010a). Thus, dispersal limitation may be highly important within a large region (e.g. 105 km2 to continental), and its importance should decrease with decreasing region size. Species sorting should be most important at intermediate and small regional extents (e.g. 102 to 105 km2), where environmental gradients are large enough, dispersal is adequate for species to track changes in environmental conditions and spatial gradients do not lead to excessive dispersal limitation. Mass effects should increase in commonness in smaller regions (e.g. 1–102 km2), where the exchange of individuals among sites is efficient enough to overcome the effects of environmental gradients on community structure. Patch dynamics should be evident at even smaller scales.
The dispersal abilities of the organism group should also be important. For example, dispersal limitation should be highly important for strongly dispersing lake diatoms at a continental scale only, and its importance apparently declines with decreasing areal extent of a region (Bennett et al., 2010). By contrast, even within relatively small regions, lake fish communities may be more strongly structured by dispersal limitation than environmental heterogeneity among lakes (Drakou et al., 2009). Furthermore, the spatial organisation of freshwater ecosystems (especially running waters) as networks tends to limit the dispersal of freshwater organism to a great degree, and different groups of organisms (e.g. fish and mussels versus aquatic insects and diatoms) may be differently affected by the characteristics of such network systems.
These mechanisms operating within a metacommunity can be associated with the relationships between alpha (LSR), beta and gamma (RSR) diversity (Fig. 2). First, if either dispersal limitation along spatial gradients or species sorting along environmental gradients is very strong, there should be high degrees of beta diversity within a region (Harrison, Ross & Lawton, 1992; Clarke et al., 2008). If beta diversity is high in high-richness regions, there should also be an asymptotic relationship between LSR and RSR (Loreau, 2000). Such an asymptotic relationship may thus also result from mechanisms other than strong biotic interactions that were traditionally conceived as main agents of saturation (Cornell & Lawton, 1992). Second, if dispersal limitation or species sorting is weak, there should not be much variation among local communities within a region. In such cases, there should be low degrees of beta diversity and no appreciable variation in LSR. These conditions should allow the existence of linear LSR–RSR relationships, where large-scale drivers determine strongly both RSR and LSR. An approach to study simultaneously the effects of regional factors, RSR and local factors on LSR should thus be pursued (Fig. 3).
A challenge for future studies also lies in determining at which levels of environmental heterogeneity and geographical distances the pattern changes from no variation in assemblage composition to nestedness and from nestedness to high species turnover (see also Baselga, 2010). This challenge could be approached with species-by-sites data, for example, from multiple drainage basins differing in environmental heterogeneity and spatial extent. To my knowledge, such studies have not been performed in the freshwater realm.
Theoretical simulations have shown that variation in alpha, beta and gamma diversity should bear a relationship to dispersal rates (Loreau & Mouquet, 1999; Mouquet & Loreau, 2003; Fig. 4). First, at very low dispersal rates, alpha diversity should be low and beta diversity high. Second, at intermediate dispersal rates, alpha diversity should attain the highest levels, but beta diversity is low. Third, at very high dispersal levels, alpha, beta and gamma diversity should be low, because very high dispersal leads to one single community. These patterns are mainly because of homogenisation of community composition across a region with high degrees of dispersal. The degree to which dispersal affects gamma diversity in field systems is more difficult to judge, but both low (through higher beta diversity) and intermediate (through generally increased alpha diversity) dispersal rates may sustain high-richness regional biotas. The effects of dispersal on alpha, beta and gamma diversity in field systems at large scales could be examined by comparing taxa with different dispersal ability along the same geographical gradient or the same taxon across different regions differing in the connectivity of sites. Understanding the effects of dispersal on species turnover among sites within a region and associating that with the metacommunity processes is obviously a key to understanding the LSR–RSR relationship (e.g. Hugueny, Cornell & Harrison, 2007) and possibly large-scale species richness gradients in general (e.g. Jocque et al., 2010).
Recently, Jocque et al. (2010) suggested that dispersal and ecological specialisation are related to climatic conditions at large scales and that metacommunity processes could thus explain the latitudinal diversity gradient. They hypothesised that climate affects the continuity of habitat availability in space and time, thus mediating a trade-off between dispersal ability and ecological specialisation at the metacommunity level. This was further hypothesised to result from latitudinal gradients in climate favourability (i.e. degree of seasonality, harshness and predictability), underlying the latitudinal richness gradient. Thus, the higher climate favourability and species richness in the tropics should be associated with weaker dispersal ability, greater ecological specialisation, smaller range size and higher speciation rates than at higher latitudes. Some support for these suggestions has already been obtained (see Jocque et al., 2010), but tests with freshwater organism would be worthwhile (see Griffiths, 2010).
As freshwater organisms live in aquatic ‘islands’ within the ‘sea’ of terrestrial landscapes and are generally limited by drainage basin boundaries, they can be studied relatively easily as local communities embedded in the metacommunity of a drainage basin. Freshwater organisms also portray a range of dispersal abilities from weak (e.g. fish, mussels and others relying on riverine connections for dispersal) and intermediate (e.g. aquatic insects with flying adult stages) to high (e.g. bacteria, diatoms and others relying on passive overland as well as aquatic dispersal). Thus, knowing the degree to which different organism groups differing in dispersal ability show inconsistent latitudinal richness gradients and, more importantly, if the metacommunities at different latitudes harbour species with differing dispersal abilities and ecological specialisation should be a prerequisite for understanding the generation of geographical diversity gradients in the freshwater realm.
The writing of this paper was supported by a grant from the Academy of Finland. Comments on earlier drafts of this paper were provided by David Allan, Luis Mauricio Bini, Musa Mlambo, Timo Muotka, Steve Ormerod, Sophia Passy, Colin Townsend and anonymous reviewers.