Improvement of operational atmospheric parameters observation: increasing spatial resolution of aerosols optical depth maps by a data fusion process

Accurate aerosol data at global scale are a key product for climate system monitoring. Maps of optical parameters of the atmosphere often result from application of numerical models or satellite images processing. Such maps usually exhibit very large cell sizes, of the order of 1 degree of arc angle. There is a need to increase their spatial resolution to obtain accurate atmospheric parameters data sets as input for climate change detection models. To improve the spatial resolution of such maps, an oversampling by linear or bi-cubic interpolation can be used. Even if such process increases the resolution, no additional information is brought and the accuracy of these maps is actually not improved. Other methods are based on a fusion process between maps of the parameter under concern and higher resolution maps of a correlated parameter. Maps resulting from this process are more accurate than the latter. Such fusion processes have already been performed on satellite images over urban areas and provided fruitful results. We propose in this paper an application of such data fusion process to improve the spatial resolution of climatological aerosols maps taken from the Global Aerosol Data Set (GADS). These maps have been constructed from different measurements and more extensive models. They present a very low spatial resolution (5 degrees of arc angle). We fused these data with Linke turbidity maps presenting a spatial resolution of 5 minutes of arc angle. Linke turbidity is a convenient factor taking into account water vapor and aerosols properties; it is mainly used to estimate solar irradiance at ground level. Linke turbidity can be linearly correlated with aerosols optical depth. The fusion process consists in extracting spatial details of Linke turbidity maps by wavelet decomposition. These details (or high spatial frequencies) are injected into GADS maps. We obtained aerosols maps with a spatial resolution of 0.5 degrees of arc angle. These maps are much more detailed than the original maps. Their accuracy is higher over areas where Linke turbidity and aerosols optical depth were particularly well correlated. Our study demonstrates the benefits of data fusion process for climate monitoring by the potential improvement of low-resolution gridded atmospheric parameters data sets.