Wetlands are important conservation sites due to the extensive food chain and rich biodiversity they support (Getzner 2002). However, the fast degradation of these ecosystems produces an urgent need for ecological studies to develop conservation programs. Almost half of the world’s wetlands have disappeared in the last century due to agricultural and urban development (Shine and Klemm 1999). One of the main hydrological characteristics of South America is the existence of large wetlands (Neiff 2001). Approximately 95% of the inventoried wetlands in South America are in six countries, and Brazil has half of the total wetland area (Naranjo 1995). However, the Brazilian wetland inventory is not updated and it is restricted to only three scientific surveys (Maltchik 2003). South Brazil has approximately 3,441 wetlands and approximately 72% of them are smaller than 1 km² (Maltchik et al. 2003). This pattern is a consequence of strong habitat fragmentation due to agricultural expansion, especially rice plantations (Gomes and Magalhães Júnior 2004). Conservative data indicate that approximately 90% of the wetlands in south Brazil have disappeared in the last century. Understanding the structure of aquatic communities in fragments and natural wetlands is a priority for biodiversity conservation strategies in south Brazil.

The State of Rio Grande do Sul and the Pantanal Region contain most of Brazil’s wetland systems and waterbird diversity (Scott and Carbonell 1986). The Rio Grande do Sul State has 123 species in 20 families (Vélez 1997), corresponding to 14.8% of the world aquatic bird species (Rose and Scott 1994). Part of this diversity is explained by the fact that the State is part of important migratory route, including northern and southern migrants (Antas 1994; Belton 1994; Sick 1983). Nevertheless, the effects of wetland loss and fragmentation on the waterbird assemblages in south Brazil are unknown.

The process of habitat loss and fragmentation is a major cause of loss of species and changes in community structure (Diamond 1976). As the fragmentation proceeds, the remnants tend to became smaller and, at the same time, more isolated (Farina 1998; Sharpe et al. 1981; Wiens 1995). This process affects the movement of individuals through the landscape, and appears to select the species better adapted to the small, isolated wetlands (Brown and Dinsmore 1986; Fahrig and Merriam 1994; Fairbairn and Dinsmore 2001; Whited et al. 2000), thus reducing the local richness of species (alpha-diversity). On the other hand, the turnover of species among sites (beta-diversity) is expected to increase in fragmented landscapes due to isolation effects (Harrison 1997; Kneitel and Chase 2004; Moreno and Halffter 2001). The regional pool of species (gamma-diversity) is determined by alpha and beta components (Cody 1993) affected by habitat loss and fragmentation. Conversely, the regional species pool is also likely to be an important determinant of species richness in fragments (Caley and Schluter 1997).

The understanding of species composition and abundance patterns among sites is a central question in community ecology, but is poorly documented for waterbirds in fragmented wetlands (Cox et al. 2000). This is particularly challenging because wetlands tend naturally to occur as disjoint patches of different sizes scattered in a matrix of upland habitats (Whited et al. 2000) and many wetland species need to use multiple sites to fulfill their need for resources (Boettcher et al. 1995; Gibbs 2000; Haig et al. 1998). Studies are particularly scarce on waterbirds and South American wetlands, where fragmentation is paramount (Caziani et al. 2001).

The aims of this paper are (1) to describe and compare the spatial and temporal patterns of richness, abundance and composition of waterbirds in a set of wetland fragments and natural lagoons and (2) to compare the species richness and assemblage composition with regional pools of species at landscape and regional scales. Our approach is an exploratory analysis to infer the underlying agents of pattern formation in a real landscape.

Methods

Results

The fragments differed in area by several orders of magnitude (Table 1) and most of them were small-50% were less than 10 ha, 90% smaller than 30 ha and only one was over 100 ha.

We recorded 142,000 birds belonging to 66 species, 18 families and 8 orders (Table 1). A total of 59% of the birds were found in the fragments. The families with the highest number of species and individuals were Ardeidae (9 spp.; 2,940 ind.), Anatidae (15 spp.; 73,000 ind.), Rallidae (8 spp.; 42,500 ind.) and Scolopacidae (7 spp.; 256 ind.) (Appendix I). There were eight nearctic and four austral species, all in low numbers and seven were found only in the lagoons. Seven species were recorded for the first time in the peninsula region: Least Grebe (Tachybaptus dominicus), Rufescent Tiger Heron (Tigrisoma lineatum), Comb Duck (Sarkidiornis melanotos), Lake Duck (Oxyura vittata), Masked Duck (Nomonyx dominicus), Giant Wood-Rail (Aramides ypecaha) and the Purple Gallinule (Porphyrula martinica). The Lake Duck is considered a austral winter migrant in Rio Grande do Sul (Belton 1994; Bencke 2001), but the only record of the species was made in summer (December).

The White-faced Whistling-duck (Dendrocygna viduata) and the Common Moorhen (Gallinula chloropus) were the most abundant and frequent species all over the year, both in lagoons and fragments, corresponding together for 69% of the birds observed. Other frequent and abundant species were the White-faced Ibis (Plegadis chihi), Wattled Jacana (Jacana jacana), Brazilian Teal (Amazonetta brasiliensis), Great Egret (Ardea alba), Southern Screamer (Chauna torquata) and Whispering Ibis (Phimosus infuscatus). Several species showed a pattern of high abundance but low frequency of occurrence (White-winged Coot-Fulica leucoptera, Fulvous Whistling Duck-Dendrocygna bicolor, Rosy-billed Pochard-Netta peposaca, Masked Duck, Greater Yellowlegs-Tringa melanoleuca, Brown-hooded Gull-Larus maculipennis and Wood Stork-Mycteria americana). Few species tended to be widespread but not abundant, such as the Cocoi Heron (Ardea cocoi), Snowy Egret (Egretta thula), Limpkin (Aramus guarauna), Maguari Stork (Ciconia maguari), South American Snipe (Gallinago paraguaiae), Snail Kite (Rostrhamus sociabilis) and Long-winged Harrier (Circus buffoni).

Eleven species were found only in the lagoons and six only in the fragments, all with low abundance and frequency (Appendix I). Fourteen species were found only in lagoons or fragments greater than 300ha: the Least Grebe, Anhinga (Anhinga anhinga), Pinnated Bittern (Botaurus pinnatus), Rufescent Tiger Heron, Kelp Gull (Larus dominicanus), Masked Duck, Black-headed Duck (Heteronetta atricapilla), Coscoroba Swan (Coscoroba coscoroba), Black-necked Swan (Cygnus melanocoryphus), Black Skimmer (Rhynchops niger), Baird’s Sandpiper (Calidris bairdii), Black-bellied Plover (Pluvialis squatarola), Solitary Sandpiper (Tringa solitaria) and the Ringed Kingfisher (Chloroceryle amazona).

The observed and estimated richness of the Chao1 index were about the same for the fragments and for the total census, but the estimated richness was higher than the observed richness for the lagoons (Table 1). More species were registered in the lagoons (60) than in the 42 fragments (55), even though the total area of the fragments was about twice the area of the lagoons studied. At each lagoon approximately the same number of species was found (56 in the Lagoa da Reserva and 51 in the Lagoa dos Gateados).

The number of species observed in the twelve censuses tended to reach an asymptote, suggesting that the effort employed was enough to record the true number of species in the study area (Fig. 2). The majority of species was recorded in the first census—64% in the fragments and 55% in the lagoons. The accumulation of species stabilized slower in the lagoons than in the fragments.

Ordinations of the twelve censuses carried out separately for lagoons and fragments showed a similar pattern of abundance and composition through the year (Fig. 4). The first and the second axes explained respectively 46% and 15% of the variance in the fragments and 45% and 14.9% in the lagoons. Abundances were higher in the wintering period due to the increased abundance of Gruiformes and Anseriformes. The concentration of wintering waterfowls was more pronounced in the fragments, but that of Galliformes was greater in the lagoons, particularly in the Lagoa da Reserva. The first axis separated summer and winter counts, while the second axis separated spring and autumn counts. The abundance of many species increased in winter (Neotropic Cormorant (Phalacrocorax brasilianus), Brazilian Teal, Common Moorhen, White-winged Coot, Coscoroba Swan, South American Snipe, Brown-hooded Gull, Silver Teal (Anas versicolor), the Pied-billed Grebe (Podilymbus podiceps), Rosy-billed Pochard, Spot-flanked Gallinule (Gallinula melanops) and White-tufted Grebe (Rollandia rolland), some were commoner in summer (Snowy Egret, Wood Stork, Collared Plover (Charadrius collaris), Striated Heron (Butorides striatus) and the Black Skimmer), several species were commoner in spring (Roseate Spoonbill (Platalea ajaja), Southern Screamer, Blackish Rail (Pardirallus nigricans), Rufescent Tiger Heron, Greater Yellowlegs, Pectoral Sandpiper (Calidris melanotos), Black-bellied Plover (Pluvialis squatarola), and two species were more common in autumn (Plumbeous Rail, Pardirallus sanguinolentus and Kelp Gull).

There was a strong gradient of abundance and composition across the sites, with all species associated with a single subset of large, species rich lagoons and fragments. The first two axis of the ordination of the 44 wetlands explained cumulatively 72% of the variation between sites. The lagoons were clearly separated from the fragments (Fig. 3). When only the fragments were ordered (graph not shown), the first and second axis explained respectively 56% and 8% of the variation. In both cases, most of the species correlated with the first axis, all of them negatively. The species with low correlations (r < 0.30) were the uncommon ones, such as the Blackish Rail, Ringed Kingfisher and Little Blue Heron (Egretta caerulea). Thirty out of 50 species showed correlations greater than 0.70. The first axis was negatively related to the fragment areas (R² = 0.46; F_{1, 42} = 33.6; P < 0.001).

The species richness at the landscape scale was high when compared to regional pools (Table 2). The census captured respectively 76% and 60% of the species listed for the central coastal zone and for the State of the Rio Grande do Sul. The study area also shares more than 90% of the species with the nearby States and 76% with the Pantanal region. The beta-diversity among sites was 94% (52 species) when only fragments are considered, and 38% (25 species) when the lagoons are included in the calculation. The regional beta-diversity was 38% (66 species).

Discussion

The study was carried out in a single annual cycle, what could raise questions about the generality of the patterns found. The patterns are strong and consistent with other studies in Rio Grande do Sul, which also did not find a seasonal richness pattern (Accordi 2003; Vélez 1997). Moreover, the observed richness is above the 80% threshold for comparative analyses (Colwell and Coddington 1995).

We found that large natural lagoons and sets of fragments have waterbird assemblages with similar seasonal patterns of species composition and abundance. The seasonal pattern corroborates the expected fluctuations due to movement and migration. The turnover between winter and summer migrants results in small seasonal variations in the number of species. Huge wintering aggregations are commonplace in waterbird communities in temperate regions (Kershaw and Cranswick 2003; Leopold 1949; Robertson and Cooke 1999). March, the month with the lowest abundance, was also the peak of the rice harvest in the region, suggesting a possible shift of habitat related to food availability. Several studies have shown that waterbirds occupy rice fields in a predictable pattern, related to the rice lifecycle (Colwell and Taft 2000; Czech and Parsons 2002; Day and Colwell 1998; Erwin 2002).

Our data showed that none of the waterbird species favored the small, isolated wetlands, as found in forest and grassland studies (Bellamy et al. 1996; Harrison 1997; Tscharntke et al. 2002). In our study, all species correlated negatively with the first axis of the ordination, indicating that all respond to the fragmentation gradient in the same way. In small forest fragments, the edge effects favor some species, which is not the case in wetlands (Milsom et al. 2000; Pasitschniak-Arts et al. 1998). Area has been shown to be an important factor related to the gradient of species composition and abundance, as found in several other waterbird studies (Brown and Dinsmore 1986; Celada and Bogliani 1993; Paracuellos and Telleria 2004; Riffell et al. 2001). Several factors other than area have been associated with the abundance and richness of waterbirds, such as physicochemical conditions, food resources, vegetation cover and interspersion, and habitat and landscape configuration (Amezaga et al. 2002; Caziani et al. 2001; Elmberg et al. 1994; Fairbairn and Dinsmore 2001; Stickney et al. 2002). All these factors are probably implicated in the gradient found and deserve further attention.

The set of fragments had a waterbird assemblage similar to that found in the natural lagoons, but lacking a small number of uncommon species. The species absent in the fragments are either large-bodied (as the Anatidae), migratory (as the Charadriiformes) or naturally rare over their whole range (as the Ardeidae, Rallidae and the genus Oxyura) (Blake 1977). These traits have been associated with vulnerability to extinction (Bennett and Owens 1997). The small, isolated fragments may not provide adequate habitat for those species. However, it is possible that the absence of those uncommon species is simply a random placement effect (Coleman et al. 1982; Connor and Mccoy 1979).

The study supports the idea that the richness and composition of waterbird species were strongly affected by fragmentation at the local (fragment) scale, but not at the landscape scale (the whole study area), where a greater portion of the species diversity is accommodated as a beta component. Partitioning the diversity in alpha and beta components demonstrated that a greater proportion of waterbird richness is unique to each site than found in studies with plants and insects (Crist et al. 2003; Gering et al. 2003; Summerville et al. 2003). The turnover of species was high (94%) among the fragments, but considerably lower when the lagoons were included in the analysis (38%). At the landscape scale, the species pool was still rich when compared to nearby regions or long-run inventories (Belton 1994). These findings suggest that the lagoons may be acting as core-areas (Hanski 1982) for the waterbirds, which use the fragments to complement or supplement their need for resources (Taylor et al. 1993). The rich regional pool of species could also be important in maintaining the landscape richness, through large scale movements across regions that share most of the species. Even though the importance of beta-diversity studies in regional conservation planning (Crist et al. 2003; Lande 1996; Veech et al. 2002), there are no empirical studies on the way regional richness affects the number of species retained by fragments (Telleria et al. 2003).

The richness, abundance and composition of waterbird assemblages in fragmented landscapes seem to be affected by the interplay of several factors, including the regional pool of species (Gering et al. 2003; Telleria et al. 2003), their particular abundance and range patterns (Murray et al. 1999), the site and landscape structures, specially the area (Brown and Dinsmore 1986; Fairbairn and Dinsmore 2001), the presence of core refuges (Guillemain et al. 2002), and the influence of the surrounding matrix (Czech and Parsons 2002). Therefore, we suggest that a landscape perspective is essential for building sound conservation programs for waterbird assemblages (Caziani et al. 2001; Cox et al. 2000; Erwin 2002; Noss 1996).