资源环境科技发展态势分析平台

The gridding process applied to satellite-retrieved cloud properties results in the loss of certain information. In this study, we analyzed the error associated with using gridded cloud optical depth (tau) in calculating radiative forcing from the perspective of the distribution pattern of tau. Utilizing the simulated results from SBDART (Santa Babara DISORT Atmospheric Radiative Transfer), we calculated this error in ideal probability distribution functions (PDFs) of tau while keeping the average tau constant, and then used the tau retrieved from MODIS (Moderate Resolution Imaging Spectroradiometer) pixel-level observations to simulate real case studies. The results from both the ideal experiments and real case studies indicate that there is a large dependence of the error caused by gridding process on the PDF of tau. The greatest relative error occurs in the cases when tau fits a two-point or uniform distribution, reaching 10-20%, while this error is below 5% when tau follows a binomial distribution. From the analysis of MODIS pixel-level data from June 2016, we found that the PDFs of tau within one grid point (1 degrees x 1 degrees) could not be simply described by a normal distribution. Although using the logarithmic mean of r controls the error effectively, the error can still be up to 4%. Our study suggests that using gridded data (especially the arithmetic mean) to calculate radiative forcing may result in uncertainty to a certain extent, which depends strongly on the distribution pattern of cloud properties within the grid point. The PDF of cloud properties should be comprehensively considered in the gridding process in the future.