Fish species richness is associated with the availability of landscape components across seasons in the Amazonian floodplain

Introduction

Floodplains are key environments for the health of large river ecosystems (Junk et al., 2014), as they regulate water flow and nutrients that are essential for the life cycle of many species (Junk, Bayley & Sparks, 1989; Fernandes, 1997). Floodplains of the Amazon River are highly productive, with estimations of total net primary productivity reaching 300 Tg C year⁻¹ in an area of 1.77 × 10⁶ km² (Melack et al., 2009). The high productivity of the Amazonian floodplain is driven in part by seasonal changes in water level that can exceed 15 m and lead to remarkable spatio-temporal changes in the landscape (i.e., flood pulse, Junk, Bayley & Sparks, 1989). This high spatio-temporal heterogeneity of habitats across the landscape, including open water, macrophyte meadows, flooded shrubs, forests and herbaceous regions, and large extensions of ecotones integrated by a complex chain of connections, influence fishes movement, feeding and reproductive behaviors as well as their growth and survival rates (Petry, Bayley & Markle, 2003; Siqueira-Souza & Freitas, 2004; Freitas et al., 2010a; Siqueira-Souza et al., 2016).

While influences of local habitat features in these ecosystems on fishes have been relatively well studied (e.g., structural or physic-chemical variables measured within lakes) (e.g., Rodriguez & Lewis, 1997; Tejerina-Garro, Réjean & Rodriguez, 1998; Miranda, 2011; Freitas et al., 2014), influences of landscape features have been generally overlooked. The few studies evaluating this issue have found landscape components, such as habitat heterogeneity (e.g., number or size of habitats) and land cover types (e.g., forest cover) influence both fish diversity and fish biomass (Yager, Layman & Allgeier, 2011; Siqueira-Souza et al., 2016; Lobón-Cervia et al., 2015; Arantes et al., 2017; Castello et al., 2017). Given the magnitude of the Amazonian floodplain, its variety of habitats (Siqueira-Souza et al., 2016; Freitas et al., 2010b; Hurd et al., 2016) and increasingly anthropogenic-driven impacts on its landscapes, a continuing understanding of this issue is critically needed. Particularly, there is a need to understand how fish diversity is influenced by the flood-pulse driven availability of habitat within the landscape.

Herein, we tested the hypothesis that landscape components, represented by the spatial extent of four landscape variables (open water, flooded herbaceous, flooded shrub and flooded forest), significantly influenced fish diversity in Amazonian floodplain lakes, with the importance of each landscape variable being dependent on the hydrological period (i.e., high- and low-water). Specifically, we evaluated how taxonomic and functional species richness, measured in high- and low-water periods, responded to the availability of landscape features that surrounded the sampled lakes. This evaluation provided an understanding of the regional species pool and the importance of landscape spatial scales to the fish assemblages and has significant usefulness for the long-term conservation of Amazonian floodplains (Freitas et al., 2014).

Materials and Methods

Study area

Fish assemblages were sampled in 15 lakes on the central Amazon floodplain along the middle and lower Solimões River, 11 located on the margins of the main channel and four within fluvial islands. The Solimões is a whitewater river (the color of heavily creamed coffee) as consequence of the high load of suspended nutrients (Furch, 1984). It flows through a geologically recent system (Hoorn et al., 2010) and is geomorphologically monotonous in its physical structure (Latrubesse & Franzinelli, 2002). The associated floodplain is an active alluvial system still at work, blanketing and reworking the floodplain deposits (Latrubesse & Franzinelli, 2002). All 15 sampled lakes are located within this floodplain and are in general shallow lakes formed behind scroll bars by the overbank deposition of fine material.

The spatial extent of landscape features is substantially different between high and low-water periods (Figs. 1 and 2). The central Amazon floodplain is a dynamic sedimentary formation that includes both marginal plains and isolated islands that are continuously re-worked by fluvial erosion and sedimentation (Dunne et al., 1998). The geomorphological processes result in variations in elevation and inundation, which have a fundamental effect on the distribution and dynamics of the floodplain vegetation and habitats (Junk, Bayley & Sparks, 1989; Schöngart et al., 2002). As the elevation declines and inundation period increases, floodplain vegetation and habitats change from alluvial forest to shrubs, herbaceous vegetation and finally, to open water. These habitats are only available to fish when flooded and the extent of flooded habitats may vary by a few orders of magnitude as the Amazon main channel undergoes its annual 10–12 m flood cycle (Hess et al., 2003).

Results

A total of 178 species was collected. Characiformes was the most diverse group with 73 species, followed by Siluriformes with 64 species. A total of 32 and 52 species were present in more than 75 and 50% of lakes, respectively (Supplemental Information 3). Species richness varied between 52 in Santo Antonio Lake and 89 in Maracá Lake. Species richness was consistently higher during low than in high-water periods. Trophic guilds richness was generally constant among lakes in high and low-water seasons, but the number of resident species was generally higher than the number of migrant species in all lakes (Table 1).

During the high-water season, flooded shrub area was the only significant landscape variable (Table 2). Total species richness tended to be greater where shrub cover was greater in different buffer scales (500 m and 1,000 m) (Table 2; Figs. 3A and 3B). This pattern also was observed for resident species (Table 2; Figs. 3C and 3D). Omnivorous species richness was, again, positively related with shrub cover but just in the small-scale (buffer of 500 m) (Table 2; Fig. 3E).

During the low-water season, flooded herbaceous had the strongest relationship with fish diversity (Table 2). Flooded herbaceous was positively related to total species richness for the larger (5,000 m) and medium (1,000 m) buffers (Table 2; Figs. 4A and 4B). A second influential landscape variable was open water that positively affected total species richness and resident species richness (Table 2; Figs. 4C and 4D). The number of fish sampled positively influenced total species richness at the largest buffer, and showed a slightly positive, but not statistically significant, relation with carnivorous richness (p < 0.10) (Table 2; Fig. 4E).

The best models with respect to the scale of analysis varied depending on the season and fish group analyzed. During high-water, the small and medium sized buffers (500 m and 1,000 m buffers, respectively) explained more variability in the relationship between total species richness and landscape variables (pseudo-R² > 0.4, Table 2). During this season, models with omnivorous and herbivorous species richness as response variables, showed better fits for the small (500 m) buffer (pseudo-R² > 0.45) than larger buffer sizes, although the results for herbivores was not significant at any scale. Models for carnivorous and migrant species richness during high-water showed weak relationships with the landscape variables at all scales of analysis (pseudo-R² < 0.40, Table 2).

During low-water, best fits were observed for the 500 and 1,000 m buffers (Table 2). During this season, the best-fitted model was observed for the medium buffer when using resident species as the response variable (1,000 m buffer, pseudo-R² = 0.66).

Discussion

Our results demonstrated that fish assemblages in the Amazon basin respond to changes in the spatial extent of landscape components and that these responses vary depending on the ecological strategies of the fish and on the stage of the hydrological cycle. It is evident that fish species richness is related to the extent of shrub vegetation found in surrounding floodplain lakes during high-water, and to the extent of herbaceous and open water regions during low-water periods. Within this framework though, omnivorous and resident species, regardless of season, showed stronger correlations with these categories than carnivorous and migratory species. The results support the view that degradation of Amazonian floodplain landscapes can impact fish diversity (Lobón-Cervia et al., 2015), with certain groups being more vulnerable than others (Arantes et al., 2017).

Furthermore, our results support the contention that seasonal variations in landscape components impact different groups of fishes at different times, and thus drive cross-habitat migration within floodplains driven by flood-pulses (Fernandes, 1997; Castello, 2008). The observed rise in resident species (e.g., Osteoglossum bicirrhosum, Cichla monoculus) richness with increasing shrub cover, could reflect the lateral migration of these species into these flooded habitats. It is likely that the flooded shrubs represent suitable regions for reproduction, feeding and/or refuge. An observation that is also true for omnivorous species richness (e.g., Triportheus spp). It is also possible that there is a direct association between the spatial extent of shrub cover and the extent of flooding during the high-water periods, that is, the observed extent of this habitat class is maintained by the annual flooding regime (Silva, Costa & Melack, 2010; Hess et al., 2015). Should this be true, then we can hypothesize that the regional abundance of shrubs together with their structural complexity are responsible for seasonality in fish diversity. Studies have demonstrated that during low-water, many fish species migrate from flooded habitats to lakes and secondary channels (Fernandes, 1997). Our observation that during this season species richness is related to the amount of open water, is consistent with previous studies showing that fish seek out lakes with greater water volumes (e.g., deep lakes) possibly to avoid the effects of extreme droughts (Arantes et al., 2013).

A strong positive correlation is often expected between sample number and species richness (Angermeier & Schlosser, 1989). However, in our results the number of individuals was only related to species richness in the low-water model. It is possible that this result is derived from high populational densities and consequently higher catch rates of our sampling gears during these periods of water retraction.

Our results showing differential influences of landscape components on fish diversity across scales are consistent with previous studies in the Amazon floodplain areas (Freitas et al., 2014; Siqueira-Souza et al., 2016) and support the view that spatial scale of investigation on the processes affecting biodiversity matters (Chase & Leibold, 2002; Willis & Whitaker, 2002; Rahbek, 2004). The medium scale buffer (500 m) explained high variability in the relationships between fish diversity and landscape components for both seasons (low and high-water) and, therefore, may represent an appropriate scale of analysis. However, further understanding of the scales of processes driving spatial fish diversity patterns could be achieved by exploring other scales, including different buffer sizes along with a local catchment (i.e., lake system sensu Arantes et al., 2017; Castello et al., 2017).

The large amount of variation that remained unexplained by the landscape variables in our results is likely related to the high spatial-temporal heterogeneity and variability of these floodplain habitats and associated local environmental variables (Freitas et al., 2014; Siqueira-Souza et al., 2016; Junk, Bayley & Sparks, 1989; Röpke et al., 2016; Hurd et al., 2016). The large dimensions and heterogeneity of the Amazon floodplain coupled with strong temporal variations result in a complex ecosystem whose structure and dynamics are governed by deterministic and stochastic mechanisms operating across a broad range of temporal and spatial scales (Freitas et al., 2014; Hess et al., 2015). Yet a large set of variables including several that we did not include in our models, have been found to influence populations and assemblage dynamics in these floodplains (e.g., depth, transparency, dissolved oxygen, connectivity, Rodriguez & Lewis, 1997; Freitas et al., 2014; Kemenes & Forsberg, 2014; Siqueira-Souza et al., 2016; Hurd et al., 2016; Lobón-Cervia et al., 2015). Habitat variations between high- and low-water may, therefore, represent only part of the variations affecting fish community structure in this system. Including these variables along with landcover components may reveal stronger spatial patterns of fish diversity across the hydrological cycle.

Understanding the importance of landscape components in structuring freshwater fish assemblages is a core question in biological conservation, especially in light of the increasing loss of aquatic habitats and the threat it poses to freshwater fish diversity globally (Dudgeon et al., 2006). Our results demonstrate that fish species richness is closely linked to habitat composition at the landscape scale and suggest that increasing losses of aquatic habitats in the Amazon due to deforestation and river impoundment (Kahn, Freitas & Petrere, 2014; Lees et al., 2016; Lobón-Cervia et al., 2015; Arantes et al., 2017; Castello et al., 2017; Forsberg et al., 2017) could disrupt these complex relationships with unpredictable consequences on Amazonian fish diversity.

Conclusions

We conclude that the total species richness was more sensitive to variations in the landscape areas than number of species within guilds and the spatial extent of the wetland class of shrubs was consistently the more influential on fish species diversity. In synthesis, our results highlight the importance of the geospatial extent of landscape variables surrounding Amazonian lakes systems in the maintenance of their fish diversity.

Supplemental Information