Functional responses of fisheries to hydropower dams in the Amazonian Floodplain of the Madeira River

1. Tropical river fisheries support food security for millions of people but are increasingly threatened by hydropower development. How dams affect these fisheries remains poorly known in most regions. Here, we used a functional traits approach to evaluate the extent to which the composition of fishery yields in the Madeira River Basin the largest sub-basin in the Amazon, respond to dam construction. We also explored how dams affected the monetary value of yields and fishing-based income of the communities. 2. We collected fishing data in 17 locations distributed more than 300 km across upstream, reservoir, and downstream zones during pre-and post-dam construction periods. We interviewed 711 fishers from 13 communities to assess fishing income during pre- and post-dam periods. 3. Catch-per-unit effort (CPUE) declined significantly, i.e., by 37%, after dam construction. Multivariate analysis yielded six species clusters according to trait syndromes related to life history, migration, swimming-performance/microhabitat-use and economic value that were associated with the environmental data characteristic of pre- and post-dam periods. Comparison of CPUE of each cluster indicated that large species with periodic life-history strategy and regional or long-distance migratory behavior were most affected by dam construction, with CPUE declining by, on average, 31%. Declines in yields and shifts in functional composition of the fishery yields resulted in average decline of 21% in the monetary of functional clusters and 30% in fishing income. 4. Synthesis and applications. Our study indicates that the implementation of the dams and associated changes in environmental conditions affected the functional composition of yields and reduced catches, negatively affecting the fishing-based income of the studied communities in the Madeira River.Whereas catches of all functional clusters declined after the dams, species possessing large body sizes, traits related to a periodic life history strategy, either with long-distance or regional migratory behaviors, and the greatest economic importance appeared to be particularly vulnerable. These results imply that hydropower expansion will cause detrimental effects for fisheries and the livelihoods they sustain. Our results underscore the urgent need for considering alternative sources of renewable energy (e.g., solar power and in-stream turbines) to avoid irreversible socio-environmental damages of large dam projects. In river reaches where dams are already in operation or under construction, minimizing impacts will require improving operational protocols to reduce hydrological alterations and developing research and technology to improve the functionality of fish passages. In these locations, addressing losses in fishery value and fishing-based income will also require the implementation of fair compensation measures. Maintaining fish production requires conserving flow pulses and free-flowing rivers and tributaries critical for completing life cycles of fish species with vulnerable traits.This dataset contains the environmental data (seasonal hydrology and flooded area at buffer or floodplain scales, and the pre- or post-dam periods) (R table), and the CPUE of each taxon (L table).Funding provided by: National Institute of Food and AgricultureCrossref Funder Registry ID: http://dx.doi.org/10.13039/100005825Award Number: 1026124Funding provided by: National Science FoundationCrossref Funder Registry ID: http://dx.doi.org/10.13039/100000001Award Number: 1639115Fishery data: The fishery data were collected by a monitoring program coordinated by the Ichthyology and Fisheries Laboratory of the Federal University of Rondônia that included 17 locations from November 2003 to February 2005, and again from April 2008 to November 2013 (Fig. 1, Table S2). This dataset consists of standardized interviews conducted with fishers on a daily basis after their return from fishing trips. Data from before closure of the dam, Nov- 2003 to Nov- 2011, represented pre-dam period and data from after closure of the dams, Dec- 2011 to 2013, represented post-dam construction, following Cella-Ribeiro et al. (2017). The data recorded included yield (total kg/fishing trip), taxon (common names of species, or of species group, Table 1), and effort (fishers/day) of fishing trips. We calculated catch-per-unit effort (CPUE) for every month (CPUE = monthly yields of fish divided by monthly fishing effort (Petrere et al., 2010) for each taxa group and period (pre- and post-dam). Environmental data: Water level data (stage) were obtained from gauges of the Brazilian water agency (Agência Nacional de Águas ANA, www.hidroweb.ana.gov.br). Gauges were selected according to their proximity to the fishery monitoring locations. Obtained data was aggregated at a monthly scale to get a 'seasonal hydrology' indicator for the study area. We used land cover data from the European Space Agency Climate Change Initiative's Land Cover project (ESA-CCI; http://maps.elie.ucl.ac.be/ CCI/) as described in Chaudhari et al. (2019). The data comprised an annual time series of land cover mapped at a 300 m spatial resolution based on combinations of map from several remote sensing instruments. The classification follows the LULC classes defined by the UN Land Cover Classification System (LCCS). Based on these classes, we calculated annual areas of forest cover, urbanization, and flooding (flooded area) at two scales: the floodplain (locally called 'várzea') and a 20 km buffer to each side of the river. This buffer size was shown to encompass the expansion of the reservoir (Cochrane et al., 2017). We used Pearson correlation tests to evaluate collinearity among land cover variables. Flooded area was negatively correlated (−0.97) with forest cover and positively correlated with urban areas (0.94). These correlations were expected because during and after the construction of the dams, forests were submerged (see Introduction and Cochrane et al., 2017), and the population as well as urbanization have increased substantially. To avoid data multicollinearity and assure visualization of results in biplots, we used flooded area (water_all_buffer, water_all_flood) and excluded forest and urban areas from the analyses