Dissecting the Precipitation and Moisture Biases in CMIP5 Models

Precipitation and moisture biases in 4 models participating in CMIP5 (Coupled Model Intercomparison Project Phase5) as well as their corresponding AMIP runs are analyzed over 4 oceanic regions in the tropics. Ten years of monthly mean model outputs are collected to compare with ERA-Interim reanalysis data and Tropical Rainfall Measuring Mission (TRMM) precipitation data. Results show that all models exhibit realistic precipitation distribution pattern except for a common double-ITCZ problem in coupled models. However, the amount of precipitation is overestimated in most areas and the vertical structure of moisture bias is diverse, some models present a dry bias throughout the troposphere while others show a moist bias. The regime-sorting analysis provides a detailed elucidation of these model biases stratified by large-scale circulation. Most excessive precipitation in models are from strong upward motion regimes where the overestimation happens both in the response of precipitation to large-scale circulation and the frequency of occurrence of ω500. On the other hand, in weakly upward motion regimes, insufficient precipitation is found, which is a result of overestimated precipitation-circulation response and underestimated frequency of occurrence of ω500. Unlike precipitation, the column integrated water vapor (IWV) bias pattern mostly follows the distribution of the frequency of occurrence of ω500. These analyses show that model errors in precipitation and moisture are mainly controlled by errors in frequency of occurrence of ω500, while the errors in the response of precipitation to large-scale circulation play a non-negligible role in precipitation bias, those of moisture to large-scale circulation contribute little to moisture bias.