Relative Importance of Driving Factors for Aerosol Optical Depth in Hanoi Using Remotely Sensed Imagery and MLP Neural Networks

Air quality, human health, industry activity, and regional sustainable development are all at risk from Aerosol Optical Depth (AOD), which is a reflection of optical attenuation. Understanding these driving factors is essential to the decrease in AOD. The purpose of this paper is to investigate the relative importance of driving factors of AOD in Hanoi city (Vietnam) using remotely sensed imagery, remote sensing and multilayer perceptron (MLP) neural networks. For this purpose, a total of nine driving factors, including natural factors (Digital Elevation Model-DEM, slope, aspect, the Modified Normalized Difference Water Index-MNDWI, and the Soil Adjusted Vegetation Index-SAVI), social factors (population density, distances to roads, and Normalized Difference Built-up Index-NDBI) and meteorological factors (Column Water Vapour-CWV) were used to assess their effects on the AOD variation. The AOD and CWV variables were first retrieved from the MODIS product. Landsat-9 OLI images were used to derive SAVI, NDBI, and MNDWI. The importance of driving factors of AOD variation was finally investigated using the MLP neural networks and Garson’s algorithm. Results show the high importance of population density and DEM in the AOD variation, followed by CWV, slope, distances to roads, MNDWI, and NDBI. The importance of vegetation, approached by the SAVI, appears to be less influential. The results of this investigation provide important insights into how to control the factors that influence AOD variation in urban areas.