Subseasonal to seasonal streamflow forecasting in a semiarid watershed

Authors: BROXTON, P.D. - University Of Arizona; VAN LEEUWEN, W.J.D. - University Of Arizona; SVOMA, B.M. - Salt River Project; WALTER, J. - Salt River Project; Biederman, Joel

Submitted to: Journal of the American Water Resources Association
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/10/2023
Publication Date: 7/25/2023
Citation: Broxton, P., van Leeuwen, W., Svoma, B., Walter, J., Biederman, J.A. 2023. Subseasonal to seasonal streamflow forecasting in a semiarid watershed. Journal of the American Water Resources Association. 59(6):1493-1510. https://doi.org/10.1111/1752-1688.13147.
DOI: https://doi.org/10.1111/1752-1688.13147

Interpretive Summary: River managers rely on accurate streamflow forecasts to make reservoir operations and water delivery decisions affecting water supplies, hydropower and flood protection. Current streamflow predictions are done mostly at the seasonal (winter/spring) scale and updated as precipitation and snowpack conditions change throughout the winter. In this study, we use increasingly accurate near-term weather forecasts (> 2 weeks in advance) to improve the model skill of the existing seasonal forecasting approach. We train and evaluate this combined Seasonal/subseasonal approach in three watersheds in central Arizona, US. Given skillful weather forecasts, the combined system can forecast streamflow better than the seasonal model alone during the early winter months, However, this improvement is minor later in the winter and spring, when streamflows are already predictable based on prior precipitation and snowpack conditions. Combined seasonal-subseasonal systems, like the one used here, represent the next major advance in operational streamflow forecasting, and they should prove useful for a wide variety of streamflow forecasting applications.

Technical Abstract: Operational streamflow forecasting is critically important to managers of river basins that supply water, hydropower, and flood protection. While seasonal-scale forecasting is important for long-term water resources planning operations, shorter-term, subseasonal forecasts are critical for balancing water conservation with flood risk during wet periods. In this study, we designed a combined seasonal / subseasonal streamflow forecasting system with the water resources group at the Salt River Project (SRP), a provider of water and power to millions of customers in and around Phoenix, Arizona (AZ) to provide daily seasonal / subseasonal streamflow forecasts for a diverse and operationally important set of watersheds in central AZ. The forecast system integrates a machine learning-based seasonal streamflow model and a shorter-term ensemble weather forecast-driven rainfall-runoff model. This hybrid design leverages increasingly accurate sub-seasonal (> 2 week) weather forecasts to improve longer-term seasonal streamflow forecasts. The system updates once daily, providing SRP with state-of-the-art forecast tools to assess potential streamflow responses to an ensemble of subseasonal weather forecasts (out to 35 days) and to quantify how different weather forecasts would affect seasonal streamflow forecasts. We find that given skillful weather forecasts, the combined system can forecast streamflow better than the seasonal model alone during the early winter months, but that this improvement is minor later in the winter and spring, when streamflows are already predictable based on antecedent conditions. Our results underscore the value of accurate subseasonal weather forecasts in late fall / early winter to improve streamflow forecasting. Combined seasonal-subseasonal systems, like the one used here, represent the next major advance in operational streamflow forecasting, and they should prove useful for a wide variety of streamflow forecasting applications.