Windows of Opportunity for Seasonal Prediction of droughts: the case of the Middle East

Drought warnings are vital to sectors like agriculture and water management, especially at the seasonal time scale. Identifying the sources of drought predictability in regions where a prediction system demonstrated potential for useful applications of the forecasts, represents an important step toward building confidence in the predictions and refining the seasonal predictions. To better identify higher forecast skill in this context, one possible approach is to focus on specific “windows of opportunity”. The approach aims to identify periods when persistent anomalies occurring in the ocean, the atmosphere or the land surface may positively precondition the predictive ability of the seasonal forecast. In the case of SPI3, a high potential for preconditioned predictive skill is identified in the Middle East region, as suggested by a robust relationship with large-scale climate modes. Building on these results, this study explores the contributions of individual years to the skill for the region during the autumn season and in the hindcast period 1993-2016. We used a Multi Model Ensemble (MME) of eight seasonal prediction systems (SPSs) provided by the Copernicus Climate Data Store (CDS) and observations from the Climate Research Unit (CRU) to calculate the SPI3 time series and the values of the Pacific and Indian teleconnection indices, the Oceanic Nino Index (ONI) and the Dipole Mode Index (DMI), respectively. A novel methodology is implemented to cluster the year-by-year MME contributions to the Pearson correlation coefficient (PCC) that are preconditioned by the large-scale teleconnections. 

Results indicate that years with extreme high or low values of ONI and DMI are the main contributors to the forecasting skill of the MME drought predictions over the Middle East. In particular, a window of opportunity is identified in four (out of 24) years that show significantly high contribution to overall skill. These years are robustly preconditioned by El Niño or La Niña events. Among the years with higher contributions, 1994 stands out as being more influenced by the DMI, thus driven primarily by SST anomalies in the Indian Ocean rather than the Pacific Ocean.  The methodological approach developed in this study successfully highlighted the potential windows of opportunity for seasonal prediction in the Middle East region, and could be applied extensively to develop early warnings for the coming seasons to serve agriculture and water management operations.