Comparing the forecast skills of four S2S models to predict Iran's southwest and central south precipitation and capturing the signals of ENSO, MJO, and atmospheric flows over the Middle East

Abstract

Abstract To improve the forecast skills of Subseasonal to Seasonal (S2S) models in predicting Iran's southwest precipitation from 1 to 4 weeks ahead, the characteristics of observed precipitation and atmospheric variables were compared with the corresponding hindcasts generated by the CMA, UKMO, ECWMF, and Meteo France models. This comparison was performed by utilizing several deterministic and probabilistic metrics. Precipitation data at 176 rain gauge stations and the NOAA-based data of atmospheric flows for Dec-April 1995–2014 constructed our observed datasets. While almost all models underestimated wet events over the southern and eastern districts, these events were overestimated in the western and northern regions. Moreover, all models overforecasted the frequency of wet events in all leads. Except for Meteo-France, the over-forecasting was usually more pronounced in eastern drylands than in western wet areas. The correlation scores were high during the first week and decreased with the increase in lead times. The ECMWF yielded the highest correlation scores in all regions and provided the more significant deterministic and probabilistic forecast skills in all leads over western districts. As a most conservative representative of other models, the UKMO efficiently captured signals of the El-Niño Southern Oscillation (ENSO) and Madden Julian Oscillation (MJO) over the study area and the Middle East. Also, this model competently predicted the characteristics of the near-surface (mid-atmosphere) moisture transport (air vertical velocity) over these areas, particularly during the MJO's rainy phases. Our findings, presented for the first time, enhance the quality of operational S2S precipitation forecasts in Iran and the Middle East.