资源环境科技发展态势分析平台

Convective parameterizations are widely believed to be essential for realistic simulations of the atmosphere. However, their deficiencies also result in model biases. The role of convection schemes in modern atmospheric models is examined using Selected Process On/Off Klima Intercomparison Experiment simulations without parameterized convection and forced with observed sea surface temperatures. Convection schemes are not required for reasonable climatological precipitation. However, they are essential for reasonable daily precipitation and constraining extreme daily precipitation that otherwise develops. Systematic effects on lapse rate and humidity are likewise modest compared with the intermodel spread. Without parameterized convection Kelvin waves are more realistic. An unexpectedly large moist Southern Hemisphere storm track bias is identified. This storm track bias persists without convection schemes, as does the double Intertropical Convergence Zone and excessive ocean precipitation biases. This suggests that model biases originate from processes other than convection or that convection schemes are missing key processes.

Plain Language Summary Accurately modeling rainfall is critical for predicting weather and climate. Convective clouds are much smaller than climate models can simulate so the rainfall generated is estimated using a convection scheme. Rainfall is difficult to estimate, and deficiencies in the schemes result in model errors. Using simulations with and without convection schemes, we assess what impact convection schemes have on mean rainfall in state-of-the-art climate models. We find that long-term rainfall patterns are similar regardless of whether the convection schemes are used or not. Without convection schemes the atmosphere is drier and cooler. However, these differences are smaller than might be expected. In this study we show that convection schemes are not required for reasonable long-term rainfall patterns for prescribed SST model runs but are essential for reasonable daily precipitation. We aim to provide guidance for modeling centers on the impact of convection schemes on long-term mean rainfall and on long standing model biases.