Figure 1 – Flowchart of procedures used for selecting the best Geographically Weighted Regression (GWR) model. W Global - connectivity between sampling sites defined as Euclidian distance between all sites; W Basin - connectivity between sampling sites defined as Euclidian distance between all sites present in a same hydrographic basin and W FEOW - connectivity between sampling sites defined as Euclidian distance between all sites present in the same ecoregion. AIC - Akaike Information Criterion.
Figure 2 – Graph of the AIC and spatial autocorrelation by distance class for the a) Global, b) Basin and c) FEOW matrix.
Figure 3 – Moran scatterplot for the a) Global, b) Basin (b) and c) FEOW matrix.
Figure 4 – Autocorrelation values of fish richness and GWR residuals using the global connectivity (W Global) matrix.
Figure 5 – Global adjustment of the GWR model done using a W Global matrix considering a) total, b) Amazonian, c) Atlantic North/Northeast transect, d) Tocantins, e) São Francisco, f) Atlantic east transect, g) Paraná and h) Atlantic southeast transect data.
Figure 6 – Spatialization of the GWR regression coefficients and classification of sites according to the hydrographic basin. a) Annual Temperature Variation, b) Evapotranspiration in June, c) Terrestrial Primary Production, d) Average Annual Precipitation, e) Annual Precipitation Variation and f) Evapotranspiration in January.