Multi-Model Ensemble Combination and Conditional Stochastic Weather Generation Tool for Improved Streamflow Forecast

Skillful basin wide streamflow forecasts at short (1-2 weeks) and long (seasonal and longer) time scales are important for efficient water resources management. This is particularly so in the Western US, which is semi-arid and its limited water resources are stressed due to unprecedented socio economic growth. The skillful ensemble hydrologic forecasts require (i) skillful hydrometeorological outlooks, (ii) suite of models – physical and statistical that captures the physical and climate features of the basin and provide ensemble forecasts conditioned on the outlooks and, (iii) an optimal combination tool. The outlooks have to be based on the short term weather forecasting information from NOAA/NWS and the seasonal climate forecast from NOAA. Current forecasts are provided by River Forecasting Centers based on a single physical model with limited ensemble generating capability and recent research suggests that a multimodel ensemble forecasting approach provides enhanced skills in the forecast than any single model. To this end this research proposes to develop two key tools – (i) a conditional stochastic weather generator to provide daily weather ensembles based on the NWS short term and NOAA seasonal outlooks and in-situ data including land surface observations to drive the RFC’s physical model to provide ensemble streamflow forecast and, (ii) an optimal multi-model ensemble combination to provide a combined ensemble forecast from physical and statistical models. We will demonstrate the framework by applying it to the Upper Colorado River Basin. The forecasts in this basin are critical for efficient operation and management of major reservoirs and consequently, the impacts on water resources, agriculture, hydropower and aquatic environment in the South Western and Inter mountain region of Western US.

Approach Proposed

The work we propose will involve the following streams:

(1) Development of tools to “translate” short term and seasonal forecasts from NWS and NCEP, respectively, to basin scale ensemble hydrometeorological forecasts.

(2) Drive the physical model with the hydrometeorological ensembles to obtain ensemble streamflow forecasts.

(3) Develop a multi-site statistical ensemble streamflow forecast model

(4) Develop an optimal combination tool to combine these and other available forecasts to provide a multi-model ensemble forecast.

(5) Work with the water managers (USBR) in the basin to implement these forecasts for operations and management.