Vulnerability of the biota in riverine and seasonally flooded habitats to damming of Amazonian rivers

Asian School of the Environment and Earth Observatory of Singapore (EOS), Nanyang Technological University (NTU), Singapore National Institute of Amazonian Research (INPA), Manaus, Brazil Institute of Floodplain Ecology, Karlsruhe Institute of Technology, Rastatt, Germany National Institute of Education (NIE), NanyangTechnological University, Singapore Bren School of Environmental Science and Management, University of California (UCSB), Santa Barbara, California, USA Department of Geography and the Environment, University of Texas at Austin, Austin, Texas, USA Nicholas School of the Environment, Duke University, Durham, North Carolina, USA


| INTRODUCTION
The Amazon River system comprises Earth's most complex network of fluvial channels connected to some of the largest, most hydraulically intricate, and most productive wetlands on the planet. The river basin discharges ca. 6,600 km 3 y −1 (16-18% of the planet's freshwater flow) to the Atlantic (Filizola & Guyot, 2009;Meade, Dunne, Richey, Santos, & Salati, 1985). The scales of the Amazon basin's fluvial features are extreme. For example, four of the world's 10 largest rivers are in the Amazon basin (the Amazon, Negro, Madeira, and Japurá), and 20 of the 34 largest tropical rivers are Amazonian tributaries (Latrubesse, 2008(Latrubesse, , 2015. The Amazon system transfers water, sediments, and solutes across continental distances, constructing and sustaining Earth's most massive continuous belt of floodplains and a mosaic of continental wetlands encompassing more than 1,000,000 km 2 . The flood-pulse (Junk, Bayley, & Sparks, 1989), the fluvial styles (fluvial channel and floodplain morphologies) (Latrubesse, 2012), and their spectra of morphodynamic conditions in space and time provide predictable disturbance regimes that result in high habitat diversity for aquatic and non-aquatic organisms within the alluvial landscape (Salo et al., 1986). This is evident in the high α and β biological diversity found in and among these habitats. The Amazon Basin harbours the highest diversity of freshwater fishes in the world, with more than 2,700 species and a still unknown number of undescribed forms (Dagosta & De Pinna, 2019;Oberdorff et al., 2019). This remarkable fish diversity is heterogeneously distributed in the basin, with the species richness by basins being influenced by historical factors such as climatic stability, as well as by current factors such as temperature and energy availability (Oberdorff et al., 2019). Ter Steege et al. (2013) recognized 4,962 tree species in the Amazon basin, of which 2,166 flood-tolerant tree species occur in river floodplains. Between 10 and 30% of all floodplain tree species are estimated to be endemic (Wittmann et al., 2013). Floodplain trees play a major role in the carbon cycle. It has been estimated that methane emissions from Amazon floodplain trees are equivalent to the whole Arctic CH 4 source, and represent 15% of the global wetland CH 4 source (Pangala et al., 2017).
Amazonia also hosts the highest number (in absolute and percentage terms) of vertebrates specialized on or dependent upon flooded habitats. More than 150 species of non-aquatic birds are also restricted to these environments or are highly dependent on them (Cohn-Haft, Naka, & Fernandes, 2007;Laranjeiras, Naka, & Cohn-Haft, 2019;Remsen & Parker, 1983). The majority of Amazonian primate species exhibit some level of dependence on flooded forests, and some are highly dependent (e.g. Cacajao spp.) (Haugaasen & Peres, 2005). This high biodiversity, combined with significant carbon storage (Abril et al., 2014) and multiple uses by humans, such as food, timber, and non-timber forest products, including medical uses, ensures that Amazonian large-river wetlands provide more ecosystem services than almost any other large landscape feature worldwide (Castello & Macedo, 2016;Richey, Melack, Aufdenkampe, Ballester, & Hess, 2002;Wittmann & Oliveira Wittmann, 2010).
In a recent article, we provided an analysis of the irreversible consequences for hydrophysical features of Amazon valley environments to be expected at different scales from the more than 400 dams that exist already or are under consideration .
Other recent papers also have drawn attention to the potential impacts of dams at a regional scale, with emphasis on Andean basins and specific biotic groups (Anderson et al., 2018;Castello et al., 2013;Forsberg et al., 2017;Winemiller et al., 2016), or have pointed out more specific impacts (Fearnside, 2013(Fearnside, , 2014(Fearnside, , 2015(Fearnside, , 2016 Latrubesse et al. (2017) compared vulnerabilities between tributary basins and emphasized the need for a more efficient and integrative legal framework involving all nine of the basin countries for anticipatory assessments of how socio-environmental and ecological impacts of hydropower production can be better managed. To quantify the current and potential impacts of dams within tributary basins, a Dam Environmental Vulnerability Index (DEVI) was developed and applied, based on a multidisciplinary analysis at the basin scale, including geomorphological, hydrological, and land-cover features. It was demonstrated that many rivers of the Amazon basin and the coastal zone of South America are vulnerable to the cumulative and synergistic effects of large dams, and a set of actions was recommended within the existing legal and institutional framework for a transparent, multinational, inclusive decision-making process .
The extent and intensity of impacts of multiple dams on specific biological groups are potentially significant, but still loosely documented and need to be better understood. River disruption and regulation by dams may affect sediment supplies, river channel migration, floodplain dynamics and, as a major adverse consequence, are likely to decrease or even suppress ecological connectivity among populations of aquatic organisms and of organisms dependent upon seasonally flooded environments, with detrimental consequences for regional human populations. For instance, the reduction in the flood pulse amplitude and consequently in the flooded area along the Amazon River main stem resulting from the construction of six large dams in the Andean Amazon is expected to result in a dramatic decrease in fisheries yield in the Brazilian Amazon, with potential consequences for food security of human riverine populations (Forsberg et al., 2017). This article complements the previous results  by assessing the relationships between dams, DEVI, and the threatened biota. The impact of Amazonian dams is mostly focused on the biota of rivers and the periodically inundated environments that border them, so here DEVI values are contrasted with patterns of diversity and distribution of fish, flooded forest tree species, and birds associated with periodically flooded environments. Because of the cartographic representation of DEVI, it is a useful tool to compare the potential hydrophysical impacts of proposed dams in the Amazon basin with the spatial distribution of biological diversity.

| METHODS
Information on planned and constructed dams in the Amazon basin was compiled from multiple sources, designated by the Brazilian government classification system . It is based on energy generation capacity and differentiates small (1 ≤ MW < 30) and large (30 ≤ MW < 1,000 MW) hydroelectric power plants (Agência Nacional de Energia Elétrica: ANEEL, 2015). An additional category of megadams (≥1,000 MW) was incorporated.
The DEVI was created to assess the vulnerability of rivers to dams . It incorporates threats to the basins that support natural river and floodplain activity, and ecological services (sediment supplies, channel mobility, and the flood pulse) into three sub- Thus, the calculated the normalized PBD (NPBD) and normalized PUD (NPUD) for basin i is: Where PBD and PUD denote the percentage of the basin that is at present deforested, and the percentage of the basin that is deforested but located upstream of the dam that is furthest downstream, respectively. Normalized variables range from 0-1. The normalization of protected area variables PBP and PUP requires the inversion of each element i because higher percentage values indicate lower vulnerability. This is obtained by switching the min/max values.
Consequently, for the variables PBP and PUP, the normalized value for basin i is: where PBP and PUP refer to the percentage of the basin within protected areas, and the percentage of the protected area upstream of the dam that is furthest downstream, respectively. The BII is calculated as the sum of each normalized variable, weighted equally, ranging from 0-1 expressed as: The FDI is an indicator of the fluxes of sediment transported by the river flow (as sediment yield, SY), the morphodynamic activity of the rivers (represented here by the average channel migration rates), and the height range of the flood pulse (as mean water stage variability of maximum and minimum stages, WSV). It is calculated as: where NSY, NMR, and NWSV are the normalized mean SY (Mt km −2 yr −1 ), normalized mean channel migration rates (ch-w yr −1 ), and normalized average water stage annual variability (m), respectively, for basin i. Migration rates are calculated at a multi-temporal scale from remote sensing imagery (in our case, we used Landsat TM) by generating erosional-depositional polygons, and then dividing the polygons by the average channel width for inter-basin comparison.
Water level data were supplied from the Hydrogeodynamics of the Amazon Basin and Brazilian National Agency of Water.
The DII is calculated for each basin as: Each term -PLU, PTA, and PNUdenotes a ratio of river length directly affected by dams, a ratio between the number of major tributaries with dams and the total number of major tributaries, and the number of dams (planned and existing) per basin, respectively. The river length directly affected by dams is calculated using the percentage of the total river length affected by dams. It is an indicator of how much 'free' river is available upstream of the uppermost dams and how much is affected downstream of the uppermost dam. The third parameter concerns the percentage of affected tributaries. It is the number of major tributaries with dams (planned and existing) divided by the total number of major tributaries within the basin.
The DEVI for basin i is calculated as the sum of all three indices: DEVI ranges from 0-3, with higher values indicating higher vulnerability of the basin.
The DEVI and DII were compared with the Dendritic Connectivity Index (DCI), developed by Cote, Kehler, Bourne, and Wiersma (2009) and applied by Anderson et al. (2018) to Andean rivers. To assess the correlations among DCI, DEVI, and DII, a Spearman's rank correlation was used.
To understand the spatial aspects of threats to biodiversity indicated by DEVI, species richness maps for fish, trees, and birds associated with periodically flooded habitats were generated. These maps allow the analysis of the vulnerability of the biota, based on qualitative comparisons between the DEVI and richness patterns of species associated with aquatic and flooded habitats. Owing to the heterogeneity of the data available for these groups, different approaches to producing the richness maps were adopted. For fishes, published sources were used (Winemiller et al., 2016)

| Patterns of species richness
There is a consistent relationship between higher DEVI values and the patterns of higher species richness of fish and floodplain trees and Despite some differences in the diversity patterns observed for different biological groups, there is a general trend towards greater diversity associated with white-water river basins (Figure 1), and a higher rate of fish endemism in the sub-basins of the western portion of the Amazon basin (Oberdorff et al., 2019). Among other factors, the diversity patterns are determined by high habitat heterogeneity and high productivity of these ecosystems (Oberdorff et al., 2019).
Andean-foreland rivers have high bio-and geo-diverse fluvial habitats, related to the high values of FDI (Figures 1 and 2).

DEVI and river connectivity
River connectivity describes the degree to which matter and organisms can move among spatially defined units over diverse temporal and spatial scales (Amoros & Roux, 1988;Wohl, 2017). Riverfloodplain connectivity is described in longitudinal, lateral, and vertical dimensions, and also has a temporal dimension (seasonal, annual, decadal, and beyond). Thus, connectivity is related to water and sediment fluxes, floodplain hydrogeomorphological characteristics, channel pattern style and mobility, shaping habitats for different types of organisms, including fish (Pouilly & Rodríguez, 2004;Rodríguez & Lewis, 1997). In large rivers, connectivity links a dominant agent However, the combined inclusion of the seasonal flow range and channel-floodplain mobility and connectivity in FDI, and thus in DEVI, provides a more integrative, multi-dimensional approach for anticipating these dam-related threats to habitat values ( Figure 3).

Dams, DEVI, and habitats
River fragmentation by dams, combined with an interruption of sediment supply, modification of channel migration rates, and alteration of the hydrological regime will trigger loss of habitat (β-diversity), cause disconnection among populations, and put endemic species of fishes, birds and the riverine vegetation at risk, especially in the Madeira, Ucayali, and Marañon Rivers, in addition to compromising fisheries yield in the Amazon main stem (Forsberg et al., 2017).
The major issue for floodplain trees is that they are adapted to the predictable duration and timing of the flood pulse over evolutionary time scales and that their leaf physiology, flowering and fruiting phenology, and growth are linked to the seasonality of flooding (Parolin, Lucas, Piedade, & Wittmann, 2010;Schöngart, Wittmann, Piedade, Junk, & Worbes, 2005). Once downstream flood regimes are modified in amplitude or timing, the highly flood-adapted species either lose their ecological niche (when low water regimes are higher than before) or are outcompeted by terrestrial species (when high water regimes are lowered). Where loss of unique floodplain habitat leads to the extinction of its specialist tree community, it is not clear how floodplain forests and their ecosystem services might be restored: no other tree species on Earth are likely to be capable of filling these niches (Wittmann & Householder, 2017). In addition, the loss of floodplain forest is likely to cascade down to planktonic communities, benthic organisms, food webs, and fish communities, as many Amazonian fish species depend on arboreal fruits during the high water-levels (Correa et al., 2015;Gottsberger, 1978;Goulding, 1980). Thus, the fishery yields based on floodplain-forest specialist fishes are likely to decrease (Araujo-Lima, Goulding, Forsberg, Victoria, & Martinelli, 1998).
In the Madeira basin, it has been shown that dams disrupt migratory routes of many fish species (e.g. Brachyplatystoma, Brycon, F I G U R E 3 Spearman's rank correlation between Dendritic Connectivity Index (DCI) and Dam Environmental Vulnerability Index (DEVI), and DCI and Dam Impact Index (DII), for both existing and planned (including existing) dam scenarios. P-values for each case are reported to assess the significance of the trend, as well as r s values showing the negative trends for all cases. Trend line derived from linear regression is given in each case. For abbreviations, refer to Figure 1 caption and Prochilus), as these species would be losing their access to breeding sites in the western Amazon (Hauser et al., 2018). This disruption has consequences for human populations, as the migratory fish constitute important sources of food and income (Cella-Ribeiro et al., 2015;Lima, Carvalho, Nunes, Angelini, & da Costa Doria, 2020;Santos, Pinto-Coelho, Fonseca, Simões, & Zanchi, 2018).
In cratonic rivers, such as the Xingu, Tapajós, and Trombetas, lateral migration rates and sediment supply are low; thus, the most sensi- Although not as diverse as Andean-foreland and Madeira basins, the Tapajós and Xingu rivers exhibit high diversity, including many endemic species of fish (65 endemics in Tapajós basin and 47 in Xingu basin) (Dagosta & De Pinna, 2019), trees (Ferreira & Prance, 1998;Salomão et al., 2007), and birds (Laranjeiras et al., 2019). Thus, the potential impact of disturbing the highly diverse flooded environments of these basins is enormous. For the cratonic basins, the major threats are fragmentation of the aquatic/alluvial habitats, the elimination of rapids (Winemiller et al., 2016;Zuanon, 1999), and the disturbance of the flood pulses/hydrological regime. Large zones of shallow rapids harbour a high diversity of strictly rheophilous fishes and high rates of fish endemicity, and also support some species of birds and bats that depend on rapids. For example, from the 61 species threatened by hydroelectric dams as identified by the Ministry of the Environment of Brazil (MMA, 2014), the Xingu supports 28% (17 spp) and the Tapajós is the habitat for 20% (12 species; Table 1). Furthermore, it is important to note that the extinction risk assessed according to the International Union for Conservation of Nature (IUCN) criteria does not include the pervasive and potentially synergistic detrimental effects of the continuing climate change on the survival odds of those endangered species. This points to even more dramatic effects of fish diversity loss resulting from the combined disturbances generated by dam construction and environmental modifications derived from climate change (Frederico, Olden, & Zuanon, 2016).
The Belo Monte run-of-river dam causes a huge reduction of The situation is also critical in the Tapajós, which has the largest DII in the whole Amazon basin (0.95). If the planned dams are constructed, the rapids mentioned above will disappear, and the hydrological regime and flood pulse will be greatly altered. The Tapajós and main tributaries would be regulated and transformed into a megalake system through a cascade of large dams extending more than 1,000 km. This pressure on the Tapajós River basin is a consequence of a combination of two main interests: energy production and commercial navigation for soy and meat export. This sequence of dams is equivalent to damming the Mississippi River from Saint Louis to New Orleans, or creating a cascade of reservoirs as long as the distance from Madrid to Paris. Any project like that would raise serious concerns and be considered infeasible in many parts of the world.
Regarding the relationships between vegetation cover/land use, and DEVI, the vulnerability of the Tapajós basin is further enhanced by the high BII (0.87), the highest value for the whole Amazon basin.
It reflects widespread land-use changes and limited conservation of forests, particularly upstream of the planned dams. A critical point is that the upper Tapajós extends beyond the Amazon forest, and 22.4% (110,000 km 2 ) of the Tapajós basin was originally covered by the Cerrado biome, where the agricultural frontier continues to expand aggressively. This scenario will become even more severe owing to the stimulus the waterway is likely to give to the expansion of agriculture, livestock rearing, and mining. Only fragmented natural Cerrado patches remain because 50,000 km 2 of Cerrado are already deforested and fragmented, and only 1,901 km 2 are officially protected areas, with 49,880 km 2 of available remnant natural area without specific legal protection . Moreover, the effectiveness of the current protected areas may be low for certain groups of organisms such as Amazonian stream-dwelling fishes (Frederico, Zuanon, & De Marco, 2018), which implies the need for innovative strategies for effective conservation of biodiversity (Azevedo-Santos et al., 2018).
In Brazil, although conservation units have been used to protect water springs, to maintain water quality near large urban centres, and to preserve marine biomes and waterscapes of aesthetic value (e.g. rapids and waterfalls; Dean, 1996;Drummond, Franco, & Oliveira, 2010), few were explicitly designed to protect limnological The upper Xingu is also in a critical condition, as there is no con- The environments that make up the Amazon landscapes are deeply connected. For example, many vertebrates considered typical upland species also show high seasonal dependence on flooded habitats (Haugaasen & Peres, 2007), and are essential to floodplain ecological processes such as pollination and dispersal that maintain the biological diversity in flooded and non-flooded environments (Terborgh et al., 2008). The terra firme forests and their drainage systems are intrinsically interconnected, and the disruption of river dynamics will affect Amazonian biota as a whole, with great potential to bring about a massive loss of diversity in these systems.

| RECOMMENDATIONS
Our assessment of vulnerability at the tributary basin scale, the assessment of biodiversity patterns, and DEVI, indicate that the recent construction of dams is already affecting the Amazon basin and its biota. The index reflects field research and experience in other dammed river basins indicating that if the planned dams are constructed, their cumulative effects will have further impacts on extensive parts of the river-related ecosystems in the Amazon basin. As noted by Latrubesse et al. (2017), society has to become aware of the magnitude and complexity of the Amazon basin; there is no imaginable mitigation technology to reverse the cumulative impact caused by hundreds of dams.
Our recommendations go further. Countries such as Guyana, Suriname, and Francethrough the department of French Guianaand Brazilian states such as Amazonas and Amapá are, as yet, only indirectly threatened by the construction of upstream dams. However, they are not involved at present in the discussions of potential