Statistical analyses
We performed polynomial regression analyses to assess the form of the patterns of species mean range size as a function of elevation along the gradient. Best-fit models were selected based on the corrected Akaike’s information criterion (AICc). To choose the most appropriate method (the Steven’s method, the midpoint method, and the specimen method) to describe the mean elevational range size pattern, we compared the Goodness of Fit (R ²) between the best models. Then we chose the model with the highest R ² and used this method to calculate the species mean elevational range size for subsequent analyses.
Multiple regression analyses were conducted to explain species mean elevational range size. The dependent variable was species mean elevational range size of each 100-m elevational band (according to the Steven’s method). ATR, NDVI, and HH were used as independent variables. Based on the lowest AICc value, the best models (delta AICc<2) were selected from the 7 models representing all possible combinations of the 3 independent variables. OLS linear regression model was also fitted to test the relationship between the area percentage of suitable habitat types and HH and the relationship between inflow intensity and species richness.
The polynomial regression analyses were performed in the PAST 3.0 (http://folk.uio.no/ohammer/past/) (Hammer, Harper, & Ryan, 2001). Correlation analysis, OLS regression models, multiple regression analyses, and model selection were performed in the SAM 4.0 (http://www. ecoevol.ufg.br/sam/) (Rangel, Dinizfilho, & Bini, 2010).