Fish on Fire: Shifts in Amazonian fish communities after floodplain forest fires

Aquatic and terrestrial biodiversity are highly intertwined in tropical temperate floodplains (Horn et al., 2011; Junk, 2005; Opperman 2017). In Amazonian floodplain forests, many fish species expand their distribution, from main river channels onto inundated during the annual high-water season (e.g. Carvalho 2009; Goulding, 1980; Goulding 1988). These forests provide crucial feeding nursery grounds, where feed extensively on arthropods, fruits, seeds, building up fat reserves that increase survival low-water when food availability is limited (Carvalho Correa & Winemiller, 2018; Gottsberger, 1978; Kubitzki Ziburski, 1994; Saint-Paul 2000). Floodplain also key environments for reproduction rearing Menezes Vazzoler, 1992; van der Sleen 2020). Fish, turn, play an important role as seed dispersers trees. Fish essential upstream dispersal, contrast to exclusive downstream dispersal provided by water, they can scarify increasing germination rates Weiss 2016). tight ecological interactions, between aquatic organisms, be disrupted disturbances affecting either forest or compartments of ecosystems. Severe droughts have impacted Amazon Basin last two decades (Jiménez-Muñoz 2016; Marengo Phillips 2009), a trend predicted climate change progresses (Cai 2014). not only reduce extent flooding, but dry out sufficiently facilitate ignition spread fires (Phillips 2009). show lower resilience drought-induced than non-floodplain (or upland) (Flores Especially along blackwater rivers (known locally igapós), recover very slowly may fail at all after fires, remaining open vegetation state arrested succession This succession, which more susceptible reburning, transition towards white-sand savanna-like with distinct tree composition, higher herbaceous cover sandy soils compared unburnt closed-canopy Holmgren, 2021a). shifts, closed savannas, appear driven both nutrient erosion 2021a) limitation 2021b). Vegetation shifts caused igapó fire scars eventually profound implications communities. Previous studies found loss large impacts abundance diversity (Arantes Bojsen Barriga, 2002; Castello 2018). However, communities been studied. Here, we assess changes composition structure floodplains. We expect fire-driven strongly sources refuges fish. Massive mortality 2016) could fruits invertebrates associated often part diet forests. At same time, exposed burned areas favour forage sifting finely particulate detritus substrate. The woody biomass hiding predators safe (Ferreira 2015; Gois 2012; Wright Flecker, 2004), making good hunting grounds (large) predatory fish, poor dangerous places (smaller) herbivorous omnivorous species. therefore expected these reflected relative different trophic categories well size fires. particular, that: (1) frugivorous those unburned forests; (2) proportion larger sizes (3) white sand savannas benthic substrates scars. studied middle Rio Negro region Central Amazonia, including Mariuá Archipelago Quiuini River, close city Barcelos, Brazil (Figure 1). study area flooded annually May September, 5.5 m water level mean amplitude highest lowest (Goulding 2003). sampled 27 sites three habitats 1): Unburnt (n = 9), based remote sensing images did burn past (this was confirmed interviewing local residents); 9) burnt twice currently persist open-canopy lacks signs recovery; known campinas described al. (1988) Adeney (2016) grass shrublands, stunted stature sclerophyllous leaves comparison adjacent rainforests. soil conditions were previously Flores Holmgren (2021a, seasons (from July) 2018 2019. captured using set nine 20-m long gillnets mesh (40–200 mm) 24 consecutive hours each sampling year. Distance nets nearest edge varied across sites, less 1 km (range 24–710 m). retrieved every 6–8 h. Each collected individual photographed identified level, its standard length total weight recorded. Unidentified preserved 10% formaldehyde solution stored further identification Systematics Ecology Laboratory National Institute Research (INPA) Manaus, Brazil. research conducted under permit ethics committee use animals scientific (CEUA) (INPA, Portuguese acronym), Amazonas state, (Protocol nr. 033/2012) SISBIO/IBAMA permanent Jansen Zuanon (#10199-3). Data years pooled prior statistical analyzes. Differences among number individuals tested ANOVA. Shannon (H) Dominance (D; calculated 1-Simpson index) indexes site differences habitat types assessed Kruskal-Wallis test data follow normal distribution. used Kruskal–Wallis most abundant only: Serrasalmus gouldingi 171). Because differed greatly (5–150 ha), linear regressions effect scar species, individuals, biomass, size. To potential community habitats, first visualized habitats. this end, plotted types, considering five individuals. ‘assigned’ if had ≥50% specific habitat. Next, non-metric multidimensional scaling (NMDS) performed Bray–Curtis similarity index (all included analysis). A PERMANOVA Lastly, classified four categories: carnivore, herbivore, detritivore (4) omnivore, according previous 1988; Röpke online database FishBase (Froese Pauly, 2021). Chi-square types. geographical distance potentially influence similarity, Mantel matrices distances sites. addition, confounding samples open-water body, such lake edges, richness correlation tests. All tests R (R Core Team, 2020), IBM SPSS Statistics Windows (Version 26.0) PAST software (Hammer 2001). 938 80 flood seasons: 144 specimens 41 794 76 About 37% (30 80) largest unique (21.3%; 17 followed (16.3%; 13 (8.8%; 7 80). abundance, dominance indices significantly differ variability replicates other 2). Yet, 3). There significant assemblage (Chi-squared 42.989, df 6, p-value < 0.001). showed omnivores, functional group became savannas. contrast, detritivores slight detected (PERMANOVA, F: 1611; p 0.005); (p 0.021) 0.018), while latter 4). Mean (19.23 cm ± 0.27 SE) (17.94 0.42 SE; Test, 0.02). (18.74 0.41 seem mostly piranha 5). Excluding S. dataset resulted no fresh 160 kg, 39% 32% scars, 28% overall difference Interestingly, strong negative relationships 0.013, adjusted r2 0.56), 0.024, 0.47) 6). verified our results affected spatial (i.e. nearby similar regardless type) body (as closer bodies), geographic (Mantel index, 0.57), nor Fire perturbations relatively recent, increasingly pressing, threat integrity covering vast basin assemblages groups remain characterized diverse communities, carnivores become common Our assessment Amazonia. Remarkably, patterns comparable reported cause turnover resembles earlier birds (Ritter 2012) Negro. interconnected season, move freely search adequate environmental conditions. So, observed suggest there processes driving consistent expectations partially explained availability. Smaller insectivorous 1988), depending terrestrial, allochthonous material, falls canopy into water. Since kill 90% trees 2016), depend organisms find resources Increased predation risk contribute burn. distribution (18% caught), shows shift smaller-sized larger-sized 5b,c). likely offer smaller 2012). light penetration environment visibility success diurnal, visually oriented fishes, gouldingi, turn (Rodríguez Lewis Jr, 1997). Higher has suggested explanation increased natural newly created reservoirs hydropower dam constructions (Melo Rodrigues Lower thus smaller-size fishes It possible factors, temperature oxygen availability, (Arrington 2006; Ceneviva-Bastos 2017; Willis 2005). decreases steadily suggests differently Indeed, throughout low one caught one. general make scares unfavourable particular given high energy demands (Weitzman Vari, Alternatively, temperatures unshaded limit large-sized tolerance (Anjos 2008). worrisome, because presence due combination slow recovery rate increases occurrence changes. lack pronounced heterogeneity within clearly variance terms density, openness height creates patchiness clustered savanna structurally homogeneous, few isolated scattered largely areas, covered plants dead wood died fire. High explain values. Such accentuated fact effective forested density submerged lead incomplete unfolding (Knight Bain, 1996). method underestimate actual some small located well-preserved matrix. Consequently, cases. productivity exceptionally basin, need several centuries late-successional 2021a; Junk 2015). lost pervasive effects include population densities Changes conservation itself persistent biological compromise capacity perturbations. Indeed proved major barrier regrowth Future should contribution sensitivity better understand interactions terrestrial-aquatic plant–animal groups. Persistent ultimately impact human welfare region. fundamental source (Barthem 2007), income through ornamental (Ladislau 2019; Zehev 2015) sport fishing (Holley 2008; Sobreiro, Conservation efforts focus prevention One tool development educative campaigns consequences drought events providing alternatives burning agriculture purposes. Preventing will easily frequent. Milena Peter conceived ideas designed methodology; Arnold Lugo data; analysed led writing manuscript. authors contributed critically drafts gave final approval publication. thank Daniel Brito Porto his help field campaigns. Bernardo Flores, Ethan Householder, Florian Wittmann insightful suggestions. funded Geographic Society (NGS-165R-18; P.v.d.S.; M.H.); Wageningen University (PvdS; M.H.), Fundação de Amparo à Pesquisa do Estado (A.L.-C.); MSCA-IF grant European Commission (746181; P.v.d.S; MH.); Brazilian CNPq (313183/2014-7; J.Z.). conflicts interest. available Dryad Digital Repository https://doi.org/10.5061/dryad.vdncjsz0h (Lugo-Carvajal 2023).