The PCA performed with average values of the environmental conditions explained 33.26% of variation in fish diversity in the first axis, 18.79% in the second axis, and a total of 52.05% across both axes (Appendix 3). The pattern found by the ranking was: (i) conductivity, pH, dissolved oxygen and channel width positively related to the first axis; (ii) turbidity and water velocity positively to the second axis; and (iii) channel depth negatively to the second axis (Appendix 3). The PCA performed with both average values and standard deviation of environmental conditions explained 21.02% of the variance in the first axis and 16.66% in the second one (Appendix 3). The pattern found by the ranking was: (i) standard deviation of the channel width, standard deviation and average turbidity and average water velocity positively related to the first axis; (ii) standard deviation and average depth of the channel and the standard deviation of pH, conductivity and dissolved oxygen negatively to the first axis; and (iii) average channel width, pH, dissolved oxygen and conductivity and the standard deviation of water temperature positively to the second axis (Appendix 3).
Best models of richness and beta diversity of fish species included average and standard deviation of the local conditions and the spatial eigenvalues maps performed from Local W (Appendix 3). The richness model had a high r-squared value (r² = 0.623), with 56.30% of variance explained by environmental conditions, 6% by spatial maps and 0.3% by interactions between niche and neutral effects (Table 3). Beta diversity had an even higher r-square value (r² = 0.758), with 64.40% of variance explained by environmental conditions, 6.70% by spatial maps, and 4.80% by interactions between environmental and spatial processes (Table 3).
Table 3. Models of linear regression between the axis of PCA performed with the averages and standard deviation of the environmental conditions and the beta diversity and richness of the Cerrado stream fish community. r2 - Correlation coefficient; F -Fisher‘s F; p - Type one error probability; AIC - Information criteria of Akaike; Δ AIC - Akaike variation; CN - Condition Number; Moran’s I - Autocorrelation index of Moran for variable; Res Moran’s I - Autocorrelation index of Moran for residual; A.B - Environmental component; A:B - Shared Component; B.A - Spatial Component;1-(A+B) - Residual.