Potential predictability of boreal winter precipitation over central-southwest Asia in the North American multi-model ensemble

Abstract

The potential predictability and skill of boreal winter (December to February: DJF) precipitation over central-southwest Asia (CSWA) is explored in six models of the North American Multimodel Ensemble project for the period 1983–2018. The seasonal prediction data for DJF precipitation initialized at Nov. (Lead-1) observed initial condition is utilized. The potential skill is estimated by perfect model correlation (PMC) method, while observed real skill is calculated by the temporal anomaly correlation coefficient (TCC). The main focus is over the Northern Pakistan (NP: 68°–78°E, 31°–37°N), which is a dominant winter precipitation sub-region in CSWA. All participating models generally capture the observed climatological pattern and variation in winter precipitation over the region. However, there are some systematic biases in the prediction of the climatological mean DJF precipitation, specifically an overestimation of precipitation over the foothills of the Himalayas in all models. The substantial internal atmospheric variability (noise) in the seasonal mean (signal) means that the regional winter precipitation is poorly predictable. The NCEP climate forecast system (CFSv2) and two Geophysical Fluid Dynamics Laboratory models (FLOR-A and FLOR-B) show the lowest potential and real skill. The COLA and NASA models show moderate but statistically significant PMC and TCC values. Each model captures the observed relationship between spatially averaged DJF precipitation over NP, with sea surface temperature (SST) and 200 hPa geopotential height (Z200), in varying details. The COLA and NASA models skillfully matched the observed teleconnection patterns, which could be a reason for their good performance as compared to other models. It also found that SSTs in the tropical oceans are relatively well predicted by NASA model when compared with other models. A critical outcome of the predictive analysis is that the multimodel ensemble (MME: A combination of six models and 79 members) does not show many advantages over the individual models in predicting boreal winter precipitation over the region of interest. Together, these results indicate that reliable prediction of the boreal winter precipitation over CSWA remains a big challenge in initialized models.