Assessing sources of skill in forecasts of meteorological drought on seasonal to interannual time scales

With the recent, heightened interest in the predictability and early warning of drought there has emerged a need to objectively evaluate the predictive skill of various drought forecast methodologies. As drought is a location and sector-specific phenomenon no unique definition exists, so in the present work multiple meteorological drought indicators are considered in the context of their prediction. We first examine the contribution of the recursive nature of the indices, such as the standardized precipitation index (SPI), to predictive skill as measured by the temporal correlation between forecast and observed conditions at different lead times. Differences in the correlation structures of indicators as a function of season, lead time and initial condition will also be considered. We then consider the potential added skill derived from the use of a bias-corrected GCM forced with observed sea surface temperatures. An example of a prototype web-based drought analysis, prediction and display tool being developed for drought assessment in the US and Mexico will be briefly introduced.