Using the Atmospheric Radiation Measurement (ARM) Datasets to Evaluate Climate Models in Simulating Diurnal and Seasonal Variations of Tropical Clouds

Abstract

AbstractLong-term Atmospheric Radiation Measurement (ARM) datasets collected at the three tropical western Pacific (TWP) sites are used to evaluate the ability of the Community Atmosphere Model (CAM5) to simulate the various types of clouds, their seasonal and diurnal variations, and their impact on surface radiation. A number of CAM5 simulations are conducted at various horizontal grid spacing (around 2°, 1°, 0.5°, and 0.25°) with meteorological constraints from analysis or reanalysis. Model biases in the seasonal cycle of cloudiness are found to be weakly dependent on model resolution. Positive biases (up to 20%) in the annual mean total cloud fraction appear mostly in stratiform ice clouds. Higher-resolution simulations do reduce the positive bias in ice clouds, but they inadvertently increase the negative biases in convective clouds and low-level liquid clouds, leading to a positive bias in annual mean shortwave fluxes at the sites, as high as 65 W m−2 in the 0.25° simulation. Such resolution-dependent biases in clouds can adversely lead to biases in ambient thermodynamic properties and, in turn, produce feedback onto clouds. Both the model and observations show distinct diurnal cycles in total, stratiform, and convective cloud fractions; however, they are out of phase by 12 h and the biases vary by site. The results suggest that biases in deep convection affect the vertical distribution and diurnal cycle of stratiform clouds through the transport of vapor and/or the detrainment of liquid and ice. The approach used here can be easily adapted for the evaluation of new parameterizations being developed for CAM5 or other global or regional models.