Comparison of aerosol optical depth between observation and simulation from MIROC-SPRINTARS: Effects of temporal inhomogeneous sampling

Abstract

The global distribution of aerosol optical depth (AOD) is simulated using an aerosol transport model coupled to an atmospheric general circulation model with high spatial and temporal resolution. Daily representative AOD from model simulation is estimated after consideration of observation sampling in daytime (ground-based) and overpass time (satellite) after cloud masking. Large deviations in AOD are found after considering temporally inhomogeneous sampling, with positive differences over desert regions and negative differences over anthropogenic pollution and biomass burning regions. Mean difference in daily AOD of 5.33% (standard deviation of 8.02%), because of temporal inhomogeneous sampling, is identified based on observation time information from the Moderate-Resolution Imaging Spectroradiometer (MODIS). Relative differences in AOD of > 50% and > 30% were found in 7.9% and 22.8% of the data, respectively. Based on the observation time information from the AERONET, relative root mean square error (rRMSE) of AOD due to temporal inhomogeneous sampling is estimated to be 4.30-18.66%. After correcting for temporal sampling inhomogeneity, the simulated global AOD was compared with AODs from MODIS and AERONET. The simulated AOD becomes lower than MODIS AOD because of emission and transport discrepancies related to dust, a limited accounting of nitrate processes, and limitation errors from MODIS AOD retrieval. A regional positive bias in SPRINTARS AOD was found in biomass burning regions, which is due to transport pattern errors related to the initial injection height of emissions. A weak correlation is found over the regions with multiple aerosol sources because of complex interactions of individual aerosol types.