Variable importance plots for random forest regressions predicting abundances of Cyrtonaias tampicoensis , Lampsilis bracteata , and Utterbackia imbecillis using modeled hydraulic variables during a low (0.42 m 3 s -1 ), moderate (5.32 m 3 s -1 ), moderate-high (32.28 m 3 s -1 ), and high (361.89 m 3 s -1 ) flow in a 20 km segment of the San Saba River, TX, USA.
Variable importance was determined using conditional permutation importance (CPI) because some predictor variables were found to be highly correlated [ 85 ]. Note: CPI should not be compared across different RF model. (TIF)
