Monthly Natural and Observed Streamflow Predictions for Five Priority Ecosystem Studies Large Landscape Regions in the United States

This data release contains prepared input attributes, monthly streamflow observations (minimum, mean, and maximum of the daily mean streamflow within each month), model outputs (predicted mean), and the code used to fetch and process these data. Code is also provided to generate confidence intervals for the mean, and prediction intervals using the built Random Forest models. Random Forest model predictions are provided for 1981-2020 for about 622,000 National Hydrography Dataset Plus version 2.1 stream segments located in five geographic regions of the Conterminous United States: Mid-Atlantic, South Atlantic West, Great Lakes, Upper Colorado, and the Pacific Northwest. Predictions are made for streamflow in actual observed conditions and also for hypothetical natural reference conditions that are defined by setting anthropogenic input variables to have no influence on model predictions.
This data compilation was funded by the USGS.