Accounting for dust aerosol size distribution in radiative transfer

作者:Li, Jiangnan*; Min, Qilong; Peng, Yiran; Sun, Zhian; Zhao, Jian-Qi
来源:JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES, 2015, 120(13): 6537-6550.
DOI:10.1002/2015JD023078

摘要

The impact of size distribution of mineral dust aerosol on radiative transfer was investigated using the Aerosol Robotic Network-retrieved aerosol size distributions. Three methods for determining the aerosol optical properties using size distributions were discussed. The first is referred to as a bin method in which the aerosol optical properties are determined for each bin of the size distribution. The second is named as an assembly mean method in which the aerosol optical properties are determined with an integration of the aerosol optical parameters over the observed size distribution. The third is a normal parameterization method based on an assumed size distribution. The bin method was used to generate the benchmark results in the radiation calculations against the methods of the assembly mean, and parameterizations based on two size distribution functions, namely, lognormal and gamma were examined. It is seen that the assembly mean method can produce aerosol radiative forcing with accuracy of better than 1%. The accuracies of the parameterizations based on lognormal and gamma size distributions are about 25% and 5%, respectively. Both the lognormal and gamma size distributions can be determined by two parameters, the effective radius and effective variance. The better results from the gamma size distribution can be explained by a third parameter of skewness which is found to be useful for judging how close the assumed distribution is to the observation result. The parameterizations based on the two assumed size distributions are also evaluated in a climate model. The results show that the reflected solar fluxes over the desert areas determined by the scheme based on the gamma size distribution are about 1Wm(-2) less than those from the scheme based on the lognormal size distribution, bringing the model results closer to the observations.