Streamflow Forecasts, Observations, Basin Characteristics, AR Recon Records, and Forecast Performance Metrics for Sites in California, Nevada, and Oregon (Water Years 2014–2025)

This dataset contains all data (raw and processed) used in Allen & Steinschneider (2026) to evaluate operational hydrologic ensemble forecasts from the California Nevada River Forecast Center (CNRFC). It includes: Forecasted streamflow from October 2013 to May 2025 (csv files containing all raw archived CNRFC hourly forecasts used in analysis (CNRFC, 2025); NetCDF files containing the daily aggregated streamflow for lead times of 1 to 14 days – aggregation methods described in manuscript). Observed streamflow from October 2013 to May 2025 from USGS (2025). Cross-reference tables (NCEP, 2025) used to match observed streamflow with forecasted streamflow. AR Recon assimilation records from the Center for Western Weather and Water Extremes (CW3E, 2025), reporting the number of dropsondes assimilated into GEFS. Processed and cleaned records at the 97 sites used in the main analysis. Basin characteristics from the USGS Gages-II dataset (Falcone, 2011); in addition to the raw data, we include an ID table with selected basin characteristics used in analysis (mean annual precipitation, mean elevation, relief ratio, log-transformed drainage area, upstream dams, and watershed permeability). Forecast performance metrics (RMSE, eCRPS, Reliability) computed at daily resolution and aggregated by water year, lead time, and flow threshold. All files are provided as a compressed ZIP archive. These data products enable reproducibility of the analyses in Allen & Steinschneider (2026) and support further research on hydrologic forecast skill trends. Code used to download and process the data provided in this dataset repository can be found in the forecast-skill-trends Github Repository (Allen, 2026), along with the code used in the subsequent Bayesian hierarchical analysis.