Investigating the causes of cloud diurnal variation biases in global climate models using the TaiESM1

Global climate models (GCMs) have significantly advanced in simulating mean cloud properties, yet challenges persist in accurately capturing cloud diurnal variations (CDVs), especially over land. This study explores the underlying causes of CDV biases by leveraging the Taiwan Earth System Model version 1 (TaiESM1), satellite observations, and reanalysis data. An assessment of CMIP6 models, including TaiESM1, indicates that these models perform well in simulating high and middle clouds over land and ocean. However, low clouds exhibit the weakest correlation, highlighting ongoing difficulties in their representation. Our analysis shows that TaiESM1 underestimates the diurnal amplitude of cloud fraction over land and predicts peak cloud coverage earlier than observed. While the simulated convective cloud fractions from shallow and deep convection schemes align well with observations, the stratiform cloud fractions produced by macrophysical schemes show a phase misalignment with observed CDVs. Moreover, the diurnal variations of stratiform clouds and their associated moisture tendencies deviate notably from observations, revealing limitations in the current parameterizations for stratiform clouds. We recommend enhancing the macrophysical and microphysical parameterization schemes to address these biases. Specifically, integrating stratiform cloud macrophysics, microphysics, and turbulent boundary layer processes within GCMs, such as incorporating the Cloud Layers Unified By Binormals framework, might improve the simulation of diurnal variations in cloud fraction over land. Moreover, shifting from bulk moisture budget approaches to models emphasizing variations in convective detrainment and subgrid vertical velocities associated with turbulent mixing will enhance the realism of stratiform cloud simulations. Implementing these adjustments is anticipated to improve the accuracy of CDV simulations in TaiESM1 and other GCMs, leading to better climate simulations. • TaiESM1 shows CDV biases, with underestimated amplitude and peak cloud fraction too early over land. • Biases stem from stratiform low clouds due to inadequate parameterizations. • Enhancing CDV simulations requires integrating cloud processes and adding turbulent mixing.