Ecosystem-based management of Amazon fisheries and wetlands

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2018 The Authors. Fish and Fisheries Published by John Wiley & Sons Ltd. 1Wildlife Conservation Society (WCS), Bronx, New York 2Universidade Federal do Rio Grande do Norte (UFRN), Natal, Brazil 3Instituto Brasileiro de Geografia e Estatística (IBGE), Brasília, Brazil 4Museu Paraense Emilio Goeldi, Belém, Brazil 5Instituto Nacional de Pesquisas da Amazônia (INPA), Manaus, Brazil 6The Nature Conservancy (TNC), Arlington, Virginia 7Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts 8Centro Nacional de Pesquisa e Conservação da Biodiversidade Amazônica – CEPAM/ Instituto Chico Mendes de Conservação da Biodiversidade – ICMBio, Manaus, Brazil 9Instituto Mamirauá, Tefé, Brazil


| INTRODUC TI ON
The adequate scale of conservation in the Amazon has been of major interest since the 1980s, with most research focusing on upland rainforest and its role in the maintenance of terrestrial biodiversity and the regulation of regional climate (e.g., Laurance et al., 2002;Soares-Filho et al., 2010). In contrast, aquatic conservation in the Amazon has generally focused on floodplain fisheries and rural peoples. These efforts have been successful in managing some non-migratory species, such as the giant pirarucu (Arapaima spp., Arapaimidae) in some floodplain lakes, and developing cooperative actions at the local level in some non-protected areas and sustainable development reserves (McGrath, Castello, Almeida, & Estupiñán, 2015;Queiroz & Crampton, 1999). As more regional data became available and synthesized, however, it became apparent that migratory fish species accounted for most of the commercial catches in the Amazon (Barthem & Goulding, 2007). Considering the large regional scale of the fisheries sector in the Amazon, it also became apparent that isolated community management efforts alone were insufficient to manage commercial fisheries and the diverse wetlands on which they depend (Barthem & Goulding, 1997). Furthermore, widespread overfishing driven by the urban demand for fish (Tregidgo, Barlowa, Pompeub, Rochac, & Parrya, 2017) and large-scale infrastructure impacts (Castello & Macedo, 2015) present major management challenges, especially in an area as large as the Amazon.
Given that overfishing and/or environmental impacts threaten most fisheries around the world, experts now recognize ecosystembased management initiatives as necessary to meet the challenges of scale (e.g., Beard et al., 2011). The objectives of ecosystem-based fisheries might simultaneously attempt, albeit with trade-offs, to optimize the total fish yield of particular species, provide safeguards to overexploitation of species, provide long-term economic viability, conserve wetlands and their biodiversity, maintain a desirable ecosystem state, protect certain species and maintain various ecosystem services (Link, 2002). A first step is to define a convincing spatial context for the ecosystem-based framework. Migratory fish are the logical species to choose when defining aquatic ecosystem size and connectivity in the Amazon because their life cycles encompass various basins and wetlands across large areas.
The concept of fish migration has an ancient history in the Amazon, which is explicit in the commonly used words piracema (fish exit or movement in Tupi) in Brazil (Veríssimo, 1895) and mijano in Spanish-speaking countries (Silva & Stewart, 2017). References to migratory fishes in the Amazon usually pertain to common food species of medium to large size (Barthem & Goulding, 2007;Ribeiro & Petrere-Jr, 1990). Many small species are also recognized as migratory by local peoples, especially near cataracts and in small streams (<20 m width) where their movements are easily observed (Cabalzar, Lima, & Lopes, 2005;Chernela, 1985). Lateral migrations in and out of floodplains are also very important and probably account for the most movement of small (<15 cm as adults) and larger species (Cox-Fernandes, 1997; Goulding, Carvalho, & Ferreira, 1988). Fish also have historically been the most commercially valuable aquatic resources and are critical to food security in the Amazon; thus, they are of interest to a wide array of stakeholders over large areas (Almeida, Lorenzen, & McGrath, 2004). From a geographical and human cultural viewpoint, many fish migrations also include various countries, states/departments, protected areas and indigenous territories. Considering the large size of the Amazon Basin and the overwhelming importance of migratory species in the commercial fisheries, we present an explicit spatial framework designed to integrate flagship species, wetlands and interest groups in order to inform the development of ecosystembased management initiatives for the region.

| Spatial context for the ecosystem-based management of migratory fish species
We developed a new classified river drainage network and scalable river basin hierarchy for the Amazon in order to map commercial fish catches, major river types, fish migrations, wetlands and other biological and limnological phenomena (Venticinque et al., 2016)  these data provide no indication of current or sustainable fishing yields, they nevertheless provide a reasonable spatial indicator of the relative regional distribution of standing biomass.  (Ribeiro & Petrere-Jr, 1990). We thus truncated the tributary basins at 300 km upstream of their confluences with the whitewater river mainstems. In cases where natural or human-made barriers to fish migration exist, such as at major cataracts at the contact zone between the continental shields and alluvial floodplains or dams, sub-basins were also truncated at those points. The resulting blackwater and clearwater sub-basin polygons were then appended to the main commercial fisheries region to define the final interbasin migratory characiform region ( Figure 4).

| Selection of flagship species
With approximately 2,500 species now recognized (Van-der-Sleen & Albert, 2017), the Amazon has the richest freshwater fish fauna in the world and there are many potential candidates F I G U R E 4 (a) The life-history region of dourada represents the largest management area of any interbasin migratory species. (b) The interbasin migratory characiform region that includes the main commercial fishing region and areas outside of it used for feeding for flagship species to target specific interest groups. Here, we focus only on selected migratory species that can inform the ecosystem-based management of fisheries to confront the impacts of infrastructure development and overfishing ( Figure 4).
The selection criteria include species that undertake longdistance interbasin migrations, they are among the historically most captured taxa, they are highly regarded in local cuisine, they command relatively high market prices, and they are currently or potentially vulnerable to overfishing and/or wetland degradation (Tables 1 and 2). Although body size has been emphasized elsewhere as a flagship species criterion (Ebner et al., 2016), we did not include it because the most important Amazonian commercial migratory food fishes are relatively large (>25 cm).
In addition to the criteria listed above, our selection also identifies 14 potential human-interest groups largely adapted from Ebner et al.
(2016) ( Table 1). It is beyond the purview of this paper to appraise the importance and geographical extent of each of these groups, and many of them will become more relevant when a larger selection of flagship species is available beyond just the long-distance migratory species considered here. Our principal target audiences are government resource managers, commercial fishers, water resource managers and environmentalists, as these groups will be essential to implement a pragmatic paradigm shift from local fisheries and wetland management only to a more realistic ecosystem-based perspective that considers the large-scale impacts of overfishing, headwater and wetland deforestation, dams and other far-reaching environmental influences.

| Migratory fish catches and wetlands by subregion
The ecosystem-based management of fisheries requires the conservation of major wetlands critical to fish survival, reproduction and growth at an adequate extent. Our analysis of the relation between migratory fish catches and wetlands focused only on Characiforms because of their overwhelming importance in commercial fisheries and direct connectivity to floodplain productivity. To test the importance of various wetlands for commercial fisheries, we divided the interbasin migratory characiform region into 10 subregions based on major geomorphological areas along the Amazon River mainstem (Dunne, Mertz, Meade, Richey, & Forsberg, 1998) and its major level 2 sub-basin areas (Venticinque et al., 2016). We categorized the fisheries and wetland data based on nine of these subregions ( Figure 5).
We excluded the Javari subregion from the analyses because little commercial fishing takes place there. We used the classification de- we scaled each subregion wetland area as a percentage of the total area of that wetland type in the interbasin migratory characiform region. Finally, stepwise multiple regression analyses of the maximum catches (t/year) of migratory characiform flagship species against major wetland types by subregion used the following models: whitewater river mainstem flooded forests, blackwater/clearwater tributary flooded forests, and whitewater river floodplain lakes that included herbaceous communities (Table 3). Criteria of variable removal were based on the probability of removal of 0.15 backwards.
Blackwater and clearwater tributary lakes were not included in the analyses because they are neither nurseries nor important adult feeding areas for the migratory characiform flagship species. Owing to the fact that the species had different sampling sizes, model adjustment was evaluated by Adjusted Multiple R-squared. In addition, we used a power analysis for each regression (Table 3).

| Context of Amazon fish migrations
Anadromous migrations (i.e., adult spawning migration is in a landward direction followed by the seaward migration of the juvenile in the life cycle) have drawn the most attention worldwide, especially the well-documented salmon migrations in the Northern Hemisphere (Lucas & Baras, 2001). Although most of the Amazon River is low-lying and without topographical barriers to migrations between the Amazon and the Atlantic, there are no known species that make large-scale anadromous migrations, which in general is true of tropical freshwater systems. At least six species of largely marine catfishes of the family Ariidae are common in the Amazon estuary (Barthem, 1985;Marceniuk & Menezes, 2007) or more of these taxa undertake short-distance anadromous migrations, but little is known about their life histories (Barthem, 1985).
The bull shark (Carcharhinus leucas, Carcharhinidae) and the largetooth sawfish (Pristis pristis, Pristidae) have been captured at least 3,000 km upstream in the Amazon, although they are considered to be opportunistic and not obligatorily migratory (Garrick, 1982;Thorson, 1974;Werder & Alhanati, 1981). The presence of newly born individuals (0.6-0.8 m length) of both species in the brackish and freshwater of the Amazon coast indicates they can use the Amazon estuary as a nursery area. Telemetry investigations in the Fitzroy River, Western Australia, also show that large-tooth sawfish can use freshwater as a nursery for large juveniles (0.9-2.5 m length) (Whitty et al., 2017).
If enough were known about the life histories of Amazonian fishes, hundreds of species might be considered migratory under a broad definition of migration (e.g., Chapman et al., 2012). Fish studies that have included some combination of floodplain lakes, river channels and flooded forests show that there are massive seasonal movements of large numbers of species between these major habitats (Cox-Fernandes, 1997;Cox-Fernandes & Petry, 1991;Goulding et al., 1988;Petry, Bayley, & Markle, 2003). There have been no successful long-distance fish tagging experiments for the Amazon.

| Candidate flagship migratory species
Flagship taxa are iconic species used to promote conservation awareness (Caro, 2010). Criteria for the selection of flagship species are often arbitrary and ad hoc, and well-defined candidates for freshwater systems are scarce, although a recent example for Australia at a continental and regional scale is promising (Ebner et al., 2016). The identification of target conservation audiences is crucial for meaningful flagship species recognition (Veríssimo et al., 2014). Five of our candidate migratory flagship taxa are among the six with the highest maximum historical catches (t/year) recorded, and all command relatively high market prices ( Figure 4 and Table 1). Four taxa have been widely exploited, as indicated by their relatively high maximum catches in the first and second quartile regional divisions (Figures 3 and 5 (Batista & Isaac, 2012;Ribeiro & Petrere-Jr, 1990) but subsequent seine mesh agreements led to recuperation of stocks. The lack of data since then, however, leaves the present situation unclear.

| Continental-scale migratory goliath catfish region
Goliath

| Long-distance migratory characiform region
The common nexus among interbasin long-distance migratory Characiforms is that their movements for spawning and upstream dispersal centre on nutrient-rich whitewater rivers (e.g., basins (Melack & Forsberg, 2001;Regarda et al., 2009). Many, if not most, of the long-distance migratory characiform species, however, are not restricted to whitewater rivers but also migrate into and out of nutrient-poor blackwater and/or clearwater rivers that large subadults and adults use as feeding areas (Correa & Winemiller, 2018;Goulding, 1980;Ribeiro & Petrere-Jr, 1990). The available genetic evidence indicates little differentiation in the populations of interbasin migratory Characiforms within the interbasin migration region we defined (Machado, Willis, Teixeira, Hrbek, & Farias, 2016;Santos, Ruffino, & Farias, 2007). In contrast to the interbasin migratory goliath catfishes, interbasin migratory Characiforms spawn throughout the Amazon Basin wherever there are whitewater rivers, and especially near their confluences with blackwater and clearwater confluences (Figure 7). Each annual upstream dispersal migration event places mature fish farther upstream. Some species, such as the jaraquis, become rare near the Andes, at least as indicated by fisheries data (Anderson, Montoya, Soto, Flores, & McClain, 2009).

Araujo
Based on detailed studies of jaraqui, annual migrations that include spawning and dispersal movements in nutrient-poor tributaries and the Amazon River mainstem can extend for 1,300 km in the Central Amazon, with upstream annual displacements in whitewater rivers of 300 km (Ribeiro & Petrere-Jr, 1990). Numerous ichthyoplankton studies in whitewater river channels also confirm massive downstream displacement of migratory characiform larvae until they enter floodplain nurseries (Araujo- Lima & Oliveira, 1998;Araujo-Lima & Ruffino, 2004;).

| Flagship characiform species and wetlands
Based on previous statistical analyses that showed that commercial fishermen historically targeted the areas with the most productive floodplains (Petrere-Jr, 1983), we expected a priori that the production of regional commercial fisheries would correlate with whitewater river floodplain areas, but with which wetlands in those areas had yet to be determined. Modern satellite imagery has permitted the relatively accurate mapping of major Amazon wetland types (Hess, Melack, Novo, Barbosa, & Gastil, 2003;Hess et al., 2015;Melack & Hess, 2010). Within the migratory characiform region, flooded forests and lakes (including their herbaceous communities) are the dominant floodplain wetlands. We expected a priori that regional fisheries production would correlate with lake areas of whitewater river floodplain because of their known role as nurseries (Bayley, 1988;Leite, Silva, & Freitas, 2006;Mounic-Silva & Leite, 2013;Petry et al., 2003).  (Freitas et al., 2018).
Although the number of subregions (N = 9) we used to detect the regional importance of wetland type in fisheries production ( Figure 5) was low for regression analyses, a power analysis indicated that the sample size effect of all regressions exceeded the minimum value considered to be acceptable (Table 3) (Cohen, 1988 (1.000) and matrinchã (0.958), thus assuring interpretive validity for these four cases. Power analysis for curimatá (0.299) alone indicated the lowest degree of confidence for the various flagship species considered in the fisheries production and wetland regression analyses.
When the four flagship characiform taxa were considered together, the regression model indicated whitewater mainstem flooded forest as the most significant wetland indicator of fisheries production upriver of the estuary, followed by whitewater floodplain lakes (r 2 adj = 0.599; N = 9; F = 6.978; p = 0.027) ( Table 3). The speciesspecific regression results for three of the flagship characiform taxa, however, do not reflect the relative collinearity of floodplain lake and flooded forest area in the western Amazon because the large floodplain lake area of the eastern Amazon is not linearly correlated with fisheries production (Figure 9). A more local study that focused on the eastern subregion indicated that floodplain forest cover was more correlated with fish yield than aquatic macrophytes associated with lakes .
The individual regression models for three of the selected flagship species agree with empirical data studies of wetland use by large subadults and adults. For tambaqui, whitewater mainstem flooded forest explains 68% of the variance in its capture abundance (r 2 adj = 0.684; N = 6; F = 11.808; p = 0.026; filter = 95% of total capture). Tambaqui is one of the most unusual migratory fishes in the Amazon because of its possession of numerous gill rakers, which are associated with a zooplanktivorous diet in whitewater river floodplain lakes and molar-like teeth used for feeding on fruits and seeds in flooded forests after just a few months of age (Araujo- Lima & Goulding, 1997). The detritivorous jaraquis are closely related, have similar life histories and produce hybrids in the wild (Ribeiro, 1984). Blackwater and clearwater river flooded forests explained 95% of the catch variance for these species (r 2 adj = 0.951; N = 9; F = 155.769; p < 0.001) and 72% for matrinchã (r 2 adj = 0.717; N = 9; F = 11.138; p = 0.010). Prior to 1 year of age, jaraqui leave their whitewater river floodplain nurseries and migrate to blackwater and clearwater rivers (Ribeiro & Petrere-Jr, 1990). There they feed on detritus derived from vascular plants and periphyton in flooded forests (Benedito-Cecilio, Araujo-Lima, Forsberg, Bittencourt, & Martinelli, 2002;Forsberg, Araujo-Lima, Martinelli, Victoria, & Bonassi, 1993;Leite, Araújo-Lima, Victoria, & Martinelli, 2002). The migratory pattern of matrinchã is similar to that of jaraquis, although matrinchã is omnivorous as a large subadult and adult with a preference for fruits and seeds and is more dependent on rainforest streams as low-water refuges (Borges, 1986;Lima, 2017). There are no empirical data on the regional detritus composition along the Amazon River floodplain that might indicate spatial differences in curimatá productivity. Maximum catches of curimatá occur in the far western and eastern subregions, the former with relatively small floodplain lakes and the latter with large lakes ( Figure 2).

| Scale management challenges
The to nutrient-poor blackwater and clearwater tributaries, a phenomenon that represents not only a major transfer of energy but also dynamic terrestrial-aquatic trophic linkages (Correa & Winemiller, 2018;Ribeiro & Petrere-Jr, 1990;Winemiller & Jepsen, 2004). Some species of migratory Characiforms also represent a high biomass of F I G U R E 9 (a) Flooded forest (feeding areas) and floodplain lake areas (nurseries) are highly correlated in most subregions of the interbasin migratory characiforme region when the Eastern Amazon is not included in the regression. (b) When the Eastern Amazon is included, its extremely large floodplain lake area skews the correlation flooded forest seed predators and dispersers (Anderson, Nuttle, Rojas, Pendergast, & Flecker, 2011;Correa, Costa-Pereira, Fleming, Goulding, & Anderson, 2015;Goulding, 1980). The characiform flagship species thus contribute to several ecosystem services that bridge river types, including regulating (interbasin energy balance), supporting (seed dispersal agents), provisioning (prey for other fishes, dolphins and humans) and cultural (highly valued species).
One of the candidate flagship species, matrinchã, represents connectivity not only between whitewater river floodplain nurseries and blackwater/clearwater tributaries (Leite, 2004), but also to upland rainforest, the streams of which it uses as refuge habitats during the low water season (Borges, 1986).
The migratory flagship species clearly illustrate how spawning, nursery and feeding areas exist along vast environmental connectivity gradients in the Amazon. The migratory flagship species spawn in river channels but very few of these habitats are protected in the Amazon or even given explicit wetland status (Cunha, Piedade, & Junk, 2015), as has been proposed by the Ramsar Convention (Mathews, 1993), to which all Amazonian countries are signatories.

| Infrastructure impacts, climate change and flagship species
Considering river impoundment (dams) and deforestation as the major impacts, the main fishing regions we defined have already suffered major impacts from two large dams on the Madeira River in Brazil and deforestation of the Amazon River floodplain downriver of its confluence with the Negro River. There is still no evidence that goliath catfishes use the fish passages constructed around the Santo Antônio Dam on the Madeira; thus, new annual recruits cannot reach headwater-spawning areas in Bolivia and Peru. The long lentic environments now upstream of the dams could also affect the capacity of the larvae and juveniles to move downstream, since these fish depend on currents to support their downstream migration.
The Amazon River floodplain downriver of the mouth of the Negro River has been heavily deforested, with at least 3,500 km 2 removed for agricultural activities since the late 1970s, and likely much more than that since the 1930s due to jute farming and livestock ranching (Goulding, Smith, & Mahar, 1996;Renó et al., 2011). The lower Amazon River floodplain undoubtedly has important nurseries, and wetland deforestation may in part be responsible for the drastic decline in some species, such as the highly frugivorous tambaqui (Isaac & Ruffino, 1996), which was once the most important commercial species. Importantly, however, the remaining forest is still critical to fish production . The blackwater and clearwater sub-basin areas (446,320 km 2 ) upstream of the main commercial fishing area, to which many species migrate partially or permanently after reaching large subadult stages, have suffered relatively little deforestation. The most notable impact is the Balbina Dam reservoir (2,300 km 2 ) near Manaus, which is located on a small blackwater river, but the impoundment has had relatively little impact on regional commercial fisheries.
All of the main Andean tributary basins are experiencing headwater deforestation for agricultural expansion and have booming mining activities in the mountainous regions and hydrocarbon exploitation in the adjacent lowlands (Finer, Jenkins, Pimm, Keane, & Ross, 2008). To date, there are no large dams on major Andean tributaries. Most concern is centred on potential high-walled storage dams for the Marañón, Ucayali and Beni rivers, and their potential downstream impacts on wetlands and fisheries if the hydrological, sediment and nutrient cycles are heavily modified (Anderson et al., 2018;Forsberg et al., 2017;Latrubesse et al., 2017). Proposed channel straightening and/or dredging of the Amazonas, Ucayali, Marañón and Huallaga rivers in Peru are also of concern and will soon reach the environmental impact assessment stage.
In addition to overfishing and wetland degradation, an ecosystembased framework also needs to consider the effects of climate change that could exacerbate direct human-related impacts. Climate models predict wetter conditions (+9-18%) in the next 70-80 years in the western Amazon and drier conditions in the east (Sorribas et al., 2016).
If the predicted conditions prevail, then there would be increased mean and maximum river discharge in the northwestern Andes-Amazon tributaries and an increased inundation extent of western floodplains. The central and eastern Amazon would have decreased river discharges, and a smaller inundation extent is predicted for the central (−15.9%) and lower Amazon (−4.4%) during low water periods.
There is historical evidence for severe decadal drought and flood conditions, with major recent droughts and extreme low water periods along the Amazon River mainstem in 1997, 2005 and 2010 (Marengo, Tomasella, Alves, Soares, & Rodriguez, 2011). Extreme low water seasons reduce floodplain lake areas, and anecdotal data indicate that commercial fishing intensifies, as fish are much easier to catch in smaller and shallower waterbodies (Tomasella et al., 2013). Fishers also correlate intense upstream migrations of the flagship and other migratory species with extreme low water periods; thus, migratory fish in general also become more vulnerable to fisheries and a large number of piscivores in shallower and narrower river channels.

| CON CLUS ION
In this paper, we propose that migratory species are a promising focal point to promote aquatic ecosystem-based conservation in the Species defines transnational migrations as occurring when "the entire population or any geographically separate part of the population of any species… a significant proportion of whose members cyclically and predictably cross one or more national jurisdictional boundaries" (United-Nations-Environment-Programme 1979). Our candidate flagship species all qualify as transnational migratory species. Our results provide key insight, via fish migrations and correlations between fish catch and wetlands, into longitudinal and lateral ecosystem linkage among the Andes, lowland river types and the Amazon River estuary.
Our results demonstrate that Amazon fisheries management needs to consider the life cycle areas of migratory species and the critical importance of whitewater river floodplain nurseries and flooded forests of all river types. Considering the large size of the Amazon and the long-distance fish migrations involved, we suggest that a basin and mainstem spatial context is the most auspicious framework for large-scale ecosystem-based management of the fisheries in the Amazon and the wetlands on which they depend. We defined two major interbasin migratory regions, each with challenges pertaining to specific interest groups, and especially government environmental management agencies charged with fisheries and wetland protection.
We suggest that flagship species provide a means to generate greater interest in fisheries management and conservation at the large scales now needed to confront the challenges of overexploitation and environmental degradation. An "Amazon Basin Fisheries and Wetlands Management Commission," or something similar, is needed to help coordinate and regulate fishery harvests of transboundary migrants and the wetlands on which they depend.

ACK N OWLED G EM ENTS
The Science for Nature and People Partnership (SNAPP) provided funding for this paper. We thank the following partners

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests.