INVESTIGATING THE EFFECTIVENESS OF SUBSEASONAL TO SEASONAL (S2S) PRECIPITATION FORECASTS IN STREAMFLOW PREDICTION AT A LARGE NUMBER OF WATERSHEDS ACROSS CONTIGUOUS UNITED STATES (CONUS)

Abstract

Accurate streamflow prediction at the Subseasonal-to-Seasonal timescale (S2S, i.e., up to 30 days into the future) could greatly benefit water resources planning, flood control, and the management of water infrastructures. Correct estimation of initial hydrological conditions (IHCs) and accurate precipitation forecasts are the two key factors leading to skillful S2S streamflow prediction. Ensemble Streamflow Prediction (ESP) is an operational approach for S2S streamflow predictions. However, ESP has been critiqued by many studies for its use of randomly resampled precipitation as conceptual future forecasts. The S2S precipitation forecasts from the North American Multi-Model Ensemble Phase II (NMME-2) can be an alternative to the randomly resampled precipitation for hydrological forecasting in the ESP framework. But, the effectiveness of the S2S precipitation forecasts from NMME-2 has not been fully understood in terms of its capability of streamflow prediction. In this thesis, the predictive skill of ESP driven by the S2S precipitation forecasts from NMME-2 is evaluated against conventional ESP driven by randomly resampled precipitation at 121 watersheds across the Contiguous United States (CONUS). Additionally, the relative importance of the NMME-2 precipitation forecasts and IHCs in S2S streamflow prediction at different forecast lead times is investigated using the ESP/Reverse Ensemble Streamflow Prediction (RevESP) framework. The results indicate that the integration of the S2S precipitation forecasts from NMME-2 in the ESP framework will lead to more skillful streamflow predictions across CONUS. Furthermore, the results indicate that the S2S precipitation forecasts are relatively more important and informative than the IHCs in streamflow predictions at most of the watersheds, especially at longer forecast lead times. However, certain hydrological properties, such as snow and large drainage areas, may weaken the effectiveness of available S2S precipitation forecasts. In general, this study highlights the broader application of available S2S precipitation forecasts in hydrology, which is likely to benefit a wide range of forecast end-users.