Long-term Streamflow Forecasting in the Apalachicola-Chattahoochee-Flint River Basin Using Ensemble Streamflow Predictions

Water availability is greatly influenced by climate. In recent years, it is generally acknowledged that the climate is changing. In the Southeast, both flooding and drought are expected to increase. However, because of low precipitations and high evapotranspiration rates during summer months, drought is usually of greater concern. During the past decade, the Southeast has experienced several severe droughts which have adversely impacted water resources of the Apalachicola-Chattahoochee-Flint (ACF) River basin. Since most water resource management challenges (urban sprawl, regulated flows, threatened and endangered species management, oyster production and recreation) are present in the ACF River basin, the basin serves as a test bed for developing drought early warning systems. The National Integrated Drought Information System (NIDIS) program has selected this basin as a pilot for developing Drought Early Warning Systems (DEWS) for the Southeast. Since selecting the ACF as a pilot in July of 2009, the NIDIS program has held a series of meetings and workshops in this basin. The need for long-term (3 to 6 months) streamflow and stage forecasts for coping with drought was highlighted by the stakeholders at the NIDIS workshops in this basin. Such a tool would also help to better manage water resources of this basin during drought.

Because of the availability of the Ensemble Streamflow Predictions (ESP) from the Southeast River Forecast Center (SERFC) at several locations in the basin, there is an opportunity to develop a streamflow and stage forecasting tool to address this critical stakeholder need. Therefore, this project is focused on addressing hydrologic drought in the ACF River basin. The main objective of this study is to evaluate approaches (e.g. precipitation analog) to streamflow forecasting using existing ESP data, and quantify the accuracy of these approaches using past ESP data.

The idea behind this analog method is to identify those years in the historical past where the sequence of observed weekly precipitation spatial patterns over a given number of weeks immediately preceding the forecast season is most analogous to those of the forecast year. Once the analog years are identified, the ESP for those specific analog years can be used to compute the ensemble mean and spread, which can then be used for probabilistic streamflow forecasts for the forecast year. Based on this analog method, we are doing experimental hindcasts for the streamflow prediction in the ACF River basin for the most recent six years period and comparing the forecasted values to the observed stream flows in those years. These forecasts are also being compared against the official forecasts made by the SERFC to find any improvement in predictive skills. The ultimate goal is to develop a streamflow forecasting tool tailored to the needs of the stakeholders of the ACF River basin that can predict onset, evolution, and demise of a hydrologic drought. The presentation will be focused on the results of this study.