“…An initial set of environmental variables was depicted to the study area (Table S2) and, for each species individually, predictor variables were selected based on their importance and correlation and visual inspection of response curves resulting from univariate models, as proposed by da . We chose variables that are widely used in modelling freshwater species (e.g., Filipe et al, 2002;Low et al, 2020;Yiwen et al, 2016), describing climatic conditions (Filipe et al, 2013;Hastie et al, 2003), hydrology and topography (Filipe et al, 2002;Friedrichs-Manthey et al, 2020), geology and soil (Mcrae et al, 2004), land cover and anthropogenic impacts (Amoatey & Baawain, 2019;Gillis et al, 2017;Maceda-Veiga, 2013), and biotic interactions (da Silva et al, 2023) that are known to influence the distribution of mussels and fish. A maximum of n/5 environmental variables (where n is the number of presence records) were selected, to avoid overfitting that could result from the ratio between many predictors and a small sample size (Fielding & Bell, 1997).…”