Assessment of Arctic Cloud Cover Anomalies in Atmospheric Reanalysis Products Using Satellite Data

Abstract

loud cover is one of the largest uncertainties in model predictions of the future Arctic climate. Previous studies have shown that cloud amounts in global climate models and atmospheric reanalyses vary widely and may have large biases. However, many climate studies are based on anomalies rather than absolute values, for which biases are less important. This study examines the performance of five atmospheric reanalysis products?ERA-Interim, MERRA, MERRA-2, NCEP R1, and NCEP R2?in depicting monthly mean Arctic cloud amount anomalies against Moderate Resolution Imaging Spectroradiometer (MODIS) satellite observations from 2000 to 2014 and against Cloud?Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) observations from 2006 to 2014. All five reanalysis products exhibit biases in the mean cloud amount, especially in winter. The Gerrity skill score (GSS) and correlation analysis are used to quantify their performance in terms of interannual variations. Results show that ERA-Interim, MERRA, MERRA-2, and NCEP R2 perform similarly, with annual mean GSSs of 0.36/0.22, 0.31/0.24, 0.32/0.23, and 0.32/0.23 and annual mean correlation coefficients of 0.50/0.51, 0.43/0.54, 0.44/0.53, and 0.50/0.52 against MODIS/CALIPSO, indicating that the reanalysis datasets do exhibit some capability for depicting the monthly mean cloud amount anomalies. There are no significant differences in the overall performance of reanalysis products. They all perform best in July, August, and September and worst in November, December, and January. All reanalysis datasets have better performance over land than over ocean. This study identifies the magnitudes of errors in Arctic mean cloud amounts and anomalies and provides a useful tool for evaluating future improvements in the cloud schemes of reanalysis products.