Analysis of the biases in the downward shortwave surface flux in the GFDL CM2.1 general circulation model

摘要

Simulations of downward shortwave surface fluxes by the coupled Geophysical Fluid Dynamics Laboratory (GFDL) CM2.1 general circulation model are compared against climatology derived from the Baseline Surface Radiation Network (BSRN), Global Energy Balance Archive, and International Satellite Cloud Climatology Project ISCCP-FD data sets. The spatial pattern of the model's biases is evaluated. An investigation is made of how these relate to accompanying biases in total cloud amount and aerosol optical depth and how they affect the surface temperature simulation. Comparing CM2.1's clear-sky fluxes against BSRN site values, for European, Asian, and North American locations, there are underestimates in the direct and overestimates in the diffuse, resulting in underestimates in the total flux. These are related to overestimates of sulfate aerosol optical depth, arising owing to the behavior of the parameterization function for hygroscopic growth of these aerosols at very high relative humidity. Contrastingly, flux overestimate biases at lower latitude locations are associated with underestimates in sea-salt and carbonaceous aerosol amounts. All-sky flux biases consist of underestimates for North America, Eurasia, southern Africa, and northern oceanic regions and overestimates for the Amazon region, equatorial Africa, off the west coast of the Americas, and southern oceanic regions. These biases show strong correlations with cloud amount biases. There are modest correlations of the flux biases with cool surface temperature biases over North America and Eurasia, warm biases over the Amazon region, and cool (warm) biases over the northern (southern) oceanic regions. Analyses assuming nonhygroscopicity illustrate that there is a reduction of surface temperature biases accompanying a reduction of sulfate aerosol optical depth biases, whereas a more significant improvement in the temperature simulation requires refining the model's simulation of cloudiness.