Dissecting Simulation Biases In Tropical Precipitation And Water Vapor In Global Climate Models

The reliability of a Global Climate Model?GCM?depends critically on how accurately precipitation and moisture fields are simulated.Thus,the attribution of biases in precipitation and moisture is of great significance for improving model performance.This study focuses on GCM simulations of tropical precipitation and water vapor,as well as their interactions with convection and large-scale circulation.By comparing with reanalysis data and observations,precipitation and moisture in four representative CMIP5?phase 5 of the Coupled Model Intercomparison Project?models are analyzed.It is found that precipitation in these models is overestimated in most tropical areas.However,the vertical distribution of moisture has large intermodel differences.All models overestimate the influence of surface water vapor interannual variations on the water vapor change in the troposphere,especially in the mid-troposphere.This over-coupling between mid-level moisture and surface moisture is exacerbated during El Nino and Southern Oscillation?ENSO?events.Water vapor correlation biases are mainly related to deep convection and water vapor transport in the central and eastern Pacific.Results reveal a close relationship among precipitation,moisture and large-scale circulation.Model biases in precipitation and moisture are further examined in conjunction with large-scale circulation by regime-sorting analysis.500h Pa vertical velocity₅₀₀ is used as a measure of the strength of large-scale circulation.Results show that all models consistently overestimate the frequency of occurrence of strong upward motion regimes.In a given₅₀₀ regime,models produce too much precipitation compared with observations and reanalysis.But for moisture,their biases differ from model to model and also from level to level.Furthermore,error causes are revealed by decomposing contributions to the biases into dynamic and thermodynamic components.For precipitation,the contribution errors in strong upward motion regimes are attributed to the overly frequent strong upward motion.In the weak upward motion regimes,the biases in the dependence of precipitation on the₅₀₀ intensity and probability density function?PDF?of₅₀₀ make comparable contributions,but often of opposite signs.On the other hand,the biases in column-integrated water vapor contribution are mainly due to errors in the frequency of occurrence of₅₀₀,while thermodynamic components contribute little.These findings suggest that errors in the occurrence frequency of₅₀₀ are a significant cause of biases in precipitation and moisture simulation.Finally,to further explore the role that convective parameterization plays in model biases,attention is paid to the NCAR Community Atmospheric Model version 5?CAM5?.Based on the Superparameterized Community Atmospheric Model?SPCAM?,the conventional parameterization scheme in CAM5 is driven by SPCAM large scale forcing.This novel method separates the influence of grid-scale motions on sub-grid processes from the impact of the parameterization scheme on the large-scale environment.Results show that the large-scale forcing from SPCAM can alleviate the double-Intertropical Convergence Zone?double-ITCZ?problem which exists in CAM5 using the convectional parameterization scheme.In addition,the amount of convective precipitation is slightly reduced due to the decreased frequency of convection.However,large scale precipitation is still underestimated.Furthermore,drawbacks still exist in the convective parameterization scheme independent of large-scale forcing,that is,the simulated convection is too frequent,but too weak,using the conventional convective parametrization scheme.This may be related to the trigger function and closure setting in the scheme.To summarize,simulation errors in precipitation and water vapor are attributed to biases in dynamics.Responses of a convective parameterization scheme to large scale forcing are thoroughly discussed.Drawbacks in the convective parameterization scheme are further revealed and their implication is discussed in the context of how to improve model performance.