Future projections of winter cold surge paths over East Asia fromCMIP6models

Previous studies using observed data suggested a decreasing trend in the frequency of cold surge (CS) over East Asia (EA) under global warming. However, the future change of CSs in different paths is relatively understudied. In this study, after evaluating the classification of CSs paths and the trend of CSs in different paths in historical simulation, the changes of CS paths in EA in future warming scenario have been investigated using the latest 21 models from the phase six of the Coupled Model Intercomparison Project (CMIP6) under the shared socioeconomic pathways (SSP245) scenario. First of all, the observed paths of winter CS in EA from 1979 to 2013 are tracked by the Flexible Particle (FLEXPART) model using ERA-Interim daily reanalysis datasets, and the CS paths are divided into the northwestern path, northern path, and western path by hierarchical clustering algorithm. Random forest algorithm is introduced to match the future CSs according to the observed classification results. The projected results indicate that the frequency of total CSs will decrease significantly in the 21st century with a trend of -0.47 times/10a. Compared with the late 20th century, the total CSs, the CSs in the northwestern path and the western path decrease by 49, 61, and 60%, respectively, in the late 21st century, while the CSs in northern path do not show a significant trend. The substantial decrease of the northwestern path CSs and the western path CSs make the proportion of northern path CSs in total CSs rise from 19.5% in the late 20th century to 37.3% in the late 21st century. It means the strong CSs are still frequent in the future, probably due to the stability of the northern path CSs.

Automated summary

This study investigated the future changes of cold surge (CS) paths in East Asia using 21 models, indicating a significant decrease in the frequency of total CSs in the 21st century with a larger reduction in CSs in the northwestern and western paths.