14B.6 An Empirical Method to Generate Probabilistic Drought Intensification Forecasts over Sub-Seasonal Time Scales

Thursday, 26 January 2017: 4:45 PM
602 (Washington State Convention Center )
Jason Otkin, University of Wisconsin, Madison, WI; and D. J. Lorenz, M. D. Svoboda, M. C. Anderson, C. Hain, and Y. Zhong

In this presentation, we will describe a new empirical method that has been developed to produce probabilistic drought intensification forecasts over sub-seasonal time scales across the U.S.  With this method, probabilistic drought intensification forecasts over 2, 4 and 8 week time periods as depicted by changes in the U.S. Drought Monitor (USDM) are developed based on recent anomalies in precipitation, soil moisture, and evapotranspiration.  These statistical forecasts are computed using logistic regression with cross validation.  While use of recent precipitation, evapotranspiration (ET) and soil moisture anomalies was shown to provide skillful forecasts, it was also found that using additional information about the current state of the USDM adds significant skill to the forecasts.  The current USDM state information takes the form of a metric that quantifies the “distance” from the next higher drought category using a non-discrete estimate of the current USDM state.  This continuous version of the current USDM state adds skill to the forecasts because USDM states that are close to the next higher drought category are more likely to intensify than states that are further from this threshold.

When assessed over a 14-year period, the probabilistic drought forecasts show skill over most of the US, with the highest skill located over the central US.  The 2- and 4-week forecasts were the most skillful in most locations, with very good reliability.  The 8-week probabilities, on the other hand, were noticeably over-confident.  For individual drought events, the method shows the most skill when forecasting high amplitude flash droughts and when large regions of the U.S. are experiencing intensifying drought conditions.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner