Study sites The middle Araguaia River floodplain is part of Brazil's complex and highly dynamic Cerrado–Amazonia ecotone (Marques et al., 2020a). The annual flood pulse drastically changes the landscape (wet season: October–April; dry season: May–September), spanning about 88,000 km2 at maximum flood level (February/March) and interconnecting several waterbodies, while only 3.3% (2,930 km2) of water area is present in the low-water period, September–November (Irion et al., 2016). The Araguaia floodplain supports a rich and abundant biota, including many endemic and endangered species, protected by a RAMSAR site (Ilha do Bananal; noo. 624), parks, and indigenous lands (RAMSAR, 2002; SEPLAN, 2012). We carried out fieldwork between July and September of 2016 and 2018 (dry season), a period with a higher abundance of caimans in the waterbodies (Da Silveira, Magnusson & Thorbjarnarson, 2008; Portelinha et al., 2019). We sampled seven localities across the middle Araguaia River floodplain: Fazenda Praia Alta, Lagoa da Confusão, Tocantins (July 27 – Aug 03, 2016), hereafter Lagoa; Cooperativa Agroindustrial Rio Formoso – Cooperformoso, Formoso do Araguaia, Tocantins (July 21 – Aug 12, 2018), hereafter Cooperformoso; Cooperativa Mista Rural Lagoa Grande – Coopergran, Formoso do Araguaia, Tocantins (July 31 – Aug 14, 2018), hereafter Coopergran; Fazenda Xavante, Dueré, Tocantins (Aug 16–23, 2018), hereafter Xavante; Fazenda Bananal-Frutacc, Lagoa da Confusão, Tocantins (Aug 25–28, 2018), hereafter Bananal; Centro de Pesquisas Canguçu, Pium, Tocantins (Aug 30–31, 2018), hereafter Canguçu; Fazenda Santo Antônio, Cristalândia, Tocantins (Sept 02–07, 2018), hereafter Cristalândia. The localities were separated from each other by at least 20 km to maintain a degree of spatial independence, based on migration and movement studies of South American crocodilians that indicate maximum movement distances of 20 km in 1–5 years (Gorzula, 1978; Ouboter & Nanhoe, 1988; Campos et al., 2006). Study design We sampled natural and anthropogenic/artificial habitats in each locality, comprehending 32 sites visited in total. We recorded landscape metrics for circular buffers (500-m, 1-km, and 3-km radius) with land use/cover rasters from the MapBiomas Project (http://mapbiomas.org, collection 4, year: 2016 and 2018). We redefined MapBiomas land-use classes considering five categories: water, forest, savanna (savanna, grassland, non-forest natural formation, and other non-forest natural formation classes), pasture (pasture and other non-vegetated area classes), and agriculture (annual and perennial crop class). We excluded the urban class due to low coverage representation, less than 1%. To improve water coverage, we generated a hydrography raster from a vectorial database acquired from Secretaria do Meio Ambiente e Recursos Hídricos of the State of Tocantins (https://semarh.to.gov.br/car/base-vetorial-digital-tematica-do-car/). Further, we improved MapBiomas land use mapping due to differences between supervised in loco coverage and MapBiomas rasters. We reclassified and redefined the topology guided by Landsat 8 satellite images for the same months of caiman sampling in 2016 and 2018, with a 30-m pixel spatial resolution obtained from the Instituto Nacional de Pesquisas Espaciais–INPE (Brazilian Space Agency; http://www.inpe.br/) using QGIS, version 3.12 (QGIS Development Team, 2020). For each buffer, we calculated landscape metrics for the designated classes with the landscapemetrics package (Hesselbarth et al., 2019): the percentage of landscape (PCLASS) and mean patch area (MPA) for all classes; the mean Euclidean nearest-neighbor distance (ENN), patch cohesion index (COHESION), and largest patch index (LPI) restricted to waterbody class; and the landscape shape index (LSI) and landscape division index (LDI) (McGarigal & Marks, 1995). Field methods We captured 294 caimans during the sampling through nocturnal spotlight surveys with the aid of locking cable snares or by hand after locating the animals by eye-reflection (Fitzgerald, 2012; Brien & Manolis, 2016). From each captured caiman, we recorded the following: habitat type; geographic coordinates, with a Garmin GPSMAP 62sc®; snout-vent length (SVL), from tip of snout to posterior margin of cloaca, and total length (TTL), from tip of snout to end of tail, with a 0.1-cm precision tape; body mass, with a 0.1-kg precision spring scale; and sex, by cloacal examination and palpation of the penis. Moreover, we recorded the presence of ectoparasites (ticks and leeches) and severe injuries (amputations, lesions, fractures, and lacerations) in the animals. We conducted this study under permits SISBIO #13324-6 and #57940-3 (issued by Instituto Chico Mendes de Conservação da Biodiversidade), FUNAI #08620.005147/2018-38 (Fundação Nacional do Índio), and CEUA-UnB #94/2017 (Comissão de Ética no Uso de Animais da Universidade de Brasília). We used the SVL as a body size estimate instead of total length due to bias resulting from tail loss and damage. We estimated the body condition of Caiman crocodilus using the scaled mass index – SMI (Peig & Green, 2009).

Landscape modifications often undermine habitat suitability for species' persistence, with initial effects observed through physiological responses of individuals and populations. However, some landscape features can allow tolerant wildlife species to persist in human-modified landscapes, but they are still overlooked. Across distinct agricultural landscapes, we assessed landscape features affecting the body condition (estimate through scaled mass index - SMI) of Caiman crocodilus (Crocodylia, Alligatoridae) in human-modified landscapes of the Araguaia floodplain, central Brazil. We used a spatial Bayesian model averaging approach to determine the effects of landscape attributes, ectoparasites, tail damage, and severe body injuries on caiman body condition. We found that caimans had higher SMI in anthropogenic (ditches and artificial ponds) than natural habitats (lakes or rivers). Overall, caiman SMI was negatively associated with wetland cohesion (an aggregation and connectivity metric). Otherwise, landscape composition did not influence caiman SMI. Further, ectoparasites and body injuries did not affect SMI, whereas tail damage negatively affected SMI. Our findings underscore that caiman populations can adapt to artificial wetlands and irrigated rice fields, provided they incorporate natural and semi-natural habitat patches that enhance environmental heterogeneity, prey availability, and waterbody availability and connectivity.

Programs able to use: Excel and R plataform.Funding provided by: Conselho Nacional de Desenvolvimento Científico e TecnológicoCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100003593Award Number: 140284/2018-4Funding provided by: Rufford FoundationCrossref Funder Registry ID: http://dx.doi.org/10.13039/100007463Award Number: 23971-1Funding provided by: United States Agency for International DevelopmentCrossref Funder Registry ID: http://dx.doi.org/10.13039/100000200Award Number: AID-OAA-A-11-00012Funding provided by: Programa de Pós-Graduação em Ecologia da Universidade de Brasília*Crossref Funder Registry ID: Award Number: NoneFunding provided by: Coordenação de Aperfeiçoamento de Pessoal de Nível SuperiorCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100002322Award Number: 001Funding provided by: Programa de Doutorado Sanduíche no Exterior - CAPES*Crossref Funder Registry ID: Award Number: 88881.357613/2019-01Funding provided by: Fundação de Apoio à Pesquisa do Distrito FederalCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100005668Award Number: None