Supporting data and model outputs for hydrologic alteration modeling in the Pearl and Pascagoula river basins

Anthropogenic hydrologic alteration threatens the health of riverine ecosystems. This study assesses hydrologic alteration in the Pearl and Pascagoula river basins using modeled daily streamflow. Machine learning was used to identify locations that have undergone statistically significant streamflow alteration, quantify the volume of the alteration, and predict alteration using cubist models. Statistically significant alteration was determined by hypothesis testing. The pre- and post-alteration flow duration curves were used to calculate the net change across 60 years. Cubist models were developed for both basins to predict hydrologic alteration and to identify important basin characteristics. This data release includes 3 input files used in the development of the models: 2 shapefiles of the Hydrologic Unit Code level 12 (HUC12) pour points in the Pearl and Pascagoula river basins and the master .csv file of covariates used in the cubist model. The data release also includes 6 model output files: the estimates of the cubist model variable coefficients for each basin, the variable importance ranked from most to least important for each basin, and observed p-values versus cubist model predicted p-values at HUC12 pour points for each basin. The observed versus predicted p-value files indicate the basins were essentially identical with respect to the amount alteration and the coefficient estimates and variable importance files demonstrate that the importance of basin geomorphology and land cover differs between the basins.