The data are comprised of input and output data from Machine Learning models that were developed to predict watershed health (WH) values in HUC-10 sub-watersheds within three major Midwest river basins. The input data included timeseries of hydro-meteorological and reconstructed WQ parameters (sediment, nitrogen, and phosphorus) as well as GIS shape files of watershed attributes (soil, landcover/land use, geomorphology, drainage classes, fertilizer sale data, etc. ). The output data is ensemble-model estimated annual WH values in HUC-10 sub-watersheds within the three river basins. The ensemble-model predicted WH values are derived from WH values obtained from three trained and validated machine learning models.

This dataset is associated with the following publication: Mallya, G., M.M. Hantush, and R.S. Govindaraju. A Machine Learning Approach to Predict Watershed Health Indices for Sediments and Nutrients at Ungauged Basins. WATER. MDPI, Basel, SWITZERLAND, 15(3): 586, (2023).