Representation of Western Disturbances in CMIP5 Models

Abstract

Abstract Western disturbances (WDs) are synoptic extratropical disturbances embedded in the subtropical westerly jet stream. They are an integral part of the South Asian winter climate, both for the agriculture-supporting precipitation they bring to the region and for the associated isolated extreme events that can induce devastating flash flooding. Here, WD behavior and impacts are characterized in 23 CMIP5 historical simulations and compared with reanalysis and observations. It is found that WD frequency has a strong relationship with model resolution: higher-resolution models produce significantly more WDs and a disproportionately high fraction of extreme events. Exploring metrics of jet strength and shape, we find that the most probable cause of this relationship is that the jet is wider in models with coarser resolution, and therefore the northern edge in which WDs are spun up sits too far north of India. The frequency of WDs in both winter and summer is found to be overestimated by most models, and thus the winter frequency of WDs estimated from the multimodel mean (30 per winter) is above the reanalysis mean (26 per winter). In this case, the error cannot be adequately explained by local jet position and strength. Instead, we show that it is linked with a positive bias in upstream midtropospheric baroclinicity. Despite a positive winter precipitation bias in CMIP5 models over most of India and Pakistan and a dry bias in the western Himalayas, the fraction of winter precipitation for which WDs are responsible is accurately represented. Using partial correlation, it is shown that the overestimation in WD frequency is the largest contributor to this bias, with a secondary, spatially heterogeneous contribution coming from the overestimation of WD intensity.