Performance of climate reanalyses in the determination of pan-Arctic terrestrial rain-on-snow events

Abstract

Rain-on-snow (ROS) events can cause rapid snowmelt, leading to flooding and avalanches in the pan-Arctic and can also lead to starvation and the death of massive ungulates. Reanalysis products (e.g., ERA-I, ERA5-land, JRA55, MERRA2) are the primary source data for the research about ROS events in the large-scale region. However, the accuracy and reliability of reanalyses have never been evaluated with respect to the determination of terrestrial ROS events. The present study aims to statistically evaluate the performance of reanalysis datasets in identifying ROS events with different criteria based on in-situ rainfall data and MODIS snow cover product. The results show that all reanalysis datasets exhibit poor performance (Recall ≤ 0.16, Kappa coefficient ≤ 0.26, F-score ≤ 0.42, MCC ≤ 0.33) in all criteria in the pan-Arctic, mainly due to the low accuracy of rainfall data (r ≤ 0.56). Nevertheless, the spatial distribution pattern and hot spots of ROS from all reanalysis datasets are essentially close. The hot spots of ROS are mainly located on the coast of Alaska, Norway, and Greenland. All reanalyses demonstrate an increase in rainy days, but there is little overall change in ROS events due to the reduction in snow cover days. This work suggests that none of the current reanalyses are reliable in the determination of ROS events due to the poor representation of the rainfall parameterization scheme. The development of alternative strategies that can investigate ROS events at large-scale is urgently needed in a changing Arctic under rapid warming.

Keywords

• Rain-on-snow
• Pan-Arctic
• Reanalysis datasets
• Rainfall
• MODIS