Investigation of Seasonal Predictability and Forecast Skill of Arabian Peninsula Wet Season Rainfall using Recalibrated Multimodel Ensembles

This work assessed the performance of the dynamical seasonal predictions to forecast wet seasonal rainfall over Arabian Peninsula. The deterministic and probabilistic metrics are used to drive extensive information about the forecast quality in terms of association, discrimination, accuracy and reliability of the predictions in the period 1991 to 2020. This study employed a set of five operational state-of-the-art seasonal forecasting models participating in the North American Multi-Model Ensemble (NMME) project. A calibration method based on extended logistic regression (ELR) is applied to ensemble mean of seasonal mean precipitation data of the individual models from NMME against observed precipitation data CPC-CMAP-URD. After calibration, a multi-model ensemble (MME) is constructed to produce reliable MME summer rainfall forecasts for the Arabian Peninsula. The skill evaluation of the recalibrated MME forecasts shows moderate deterministic and probabilistic skill for wet seasonal rainfall across Arabian Peninsula. Real time probabilistic forecasts of the 2023-24 AP rainy season were also created using the recalibrated MME approach in tercile (below-normal, normal, and above-normal) format. The forecasts show above-normal seasonal rainfall in the northern parts of AP region. A mechanism for wet seasonal rainfall is further explored and it is found that ongoing El Nino conditions predicted to strengthen during the boreal fall and winter will shift the odds towards the above-normal precipitation over the AP.