Inter-comparison of ERA-5, ERA-interim and GPCP rainfall over the last 40 years: Process-based analysis of systematic and random differences

• Published in: Journal of hydrology (Amsterdam) Vol. 583; p. 124632
• Main Author: Nogueira, Miguel
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2020
• Subjects: Climate; Dynamics; Rainfall datasets; Rainfall variability; Reanalysis; Reanalysis; Climate; Rainfall variability; Dynamics; Rainfall datasets
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2020.124632

•ERA-5 had lower systematic and random errors than ERA-Interim, mainly over the tropics.•Convective rainfall and moisture convergence patterns are better represented in ERA-5.•Rainfall trends had large spread between datasets, without improvement in ERA-5. This study presents a comprehensive inter-comparison between two widely used rainfall datasets – the Global Precipitation Climatology Project (GPCP) and the ERA-Interim reanalysis – and the recently released ERA-5, which will replace ERA-Interim as the main European Centre for Medium-Range Weather and Forecasting (ECMWF) reanalysis by 2020. Systematic and random error components were computed for the reanalysis datasets over the 1979–2018 period using GPCP as reference. The analysis was taken at monthly timescale at 2.5° spatial resolution, limited by GPCP. Despite its resolution limitations and complex algorithms, GPCP has been extensively used for model evaluation due to its global and long-term (over 40 years) coverage. Over most of the tropics, and over localized mid-latitude regions, ERA-5 showed lower bias and unbiased root-mean squared error (ubRMSE), as well as higher correlations, compared to ERA-Interim. Throughout the rest of the globe, these error metrics displayed similar values between both reanalysis, except over localized regions of the eastern tropical Pacific, the Andes and the Himalayas, where ERA-Interim outperformed ERA-5. A process-based analysis revealed that ERA-Interim tended to overestimate deep convection and moisture flux convergence over the tropical oceans and land, leading to excessive rainfall. Similarly, ERA-Interim showed significant rainfall underestimation over the mid-latitude oceans due to underestimation of deep convection and moisture flux convergence. Both cases were significantly improved in ERA-5, likely due to its improved parameterizations and higher resolution. Indeed, the differences in monthly rainfall between the two reanalysis were mainly due to improved dynamical (circulation) rather than thermodynamical (Clausius-Clapeyron) processes or surface evaporation changes. Nonetheless, the results also revealed improved representation of the moisture sink/source patterns over the tropical oceans in ERA-5. Finally, there were significant differences in the long-term rainfall trend patterns amongst the three datasets (with differences reaching ~30%/decade), that can extend even to the sign of the trend, with the most notorious differences occurring over tropical Africa. Furthermore, ERA-5 didn’t show improved representation of these trends. In fact, the trend of global-mean rainfall in ERA-Interim was closer to GPCP than ERA-5.