The data fusion of aerosol optical thickness using universal kriging and stepwise regression in East China

Recently, aerosol optical depth (AOD) study has become more important in the field of atmosphere sciences. AOD datasets retrieved from satellites are widely used in multiple fields because of their wide coverage and low cost. However, the integrity of AOD spatial coverage can be easily influenced by clouds, rain, haze and other weather phenomena. Fortunately, the full coverage AOD images are producible by employing the data fusion algorithm and ancillary methods. Based on AOD data derived from MODIS and OMI with meteorological parameters on November 18, 2013 over the East China, this study combined the universal kriging with stepwise regression and second-order polynomial fitted to extend the coverage of MODIS AOD at 550 nm. Results showed that stepwise regression method is efficient to infer the MODIS AOD by using the OMI AOD and meteorological parameters. The wind speed, relative humidity, pressure and solar radiation have significant impacts on the spatial and temporal distributions of AOD. The mean prediction error of universal kriging prediction model is 0.0047 in this paper, indicating that the universal kriging is an effective and accurate interpolation method for AOD data fusion. The methods employed in this paper can provide the data source of AOD for studies in climate and other related fields, effectively compensating the non-full coverage shortcoming of satellite AOD datasets.