Geostatistical analysis of the relationship among geomorphology, land cover, and AOD at the local scale in a South American desert region

Abstract

This work aims to assess the particle matter distribution in Tulum Valley, San Juan, Argentina, during 2019, in relation to geomorphology. Random forest supervised classifications were performed to obtain customized annual landform and seasonal land cover maps using Landsat 8 Surface Reflectance (OLI); Sentinel-1 SAR, SRTM digital elevation images, and a set of spectral indices. Geostatistical analyses, including Linear Mixed Models, were performed considering the ln of the mean, the ln of the standard deviation, and the Angstrom Exponent of the MAIAC Aerosol Optical Depth product as response variables, and landforms, landforms + land cover, Land Surface Temperature, and Moisture Stress Index as explanatory variables. The models allowed statistically significant differentiation of the Aerosol Optical Depth means of the landform and landform + land cover classes. The percentage of residual standard error for the means obtained from the model was up to 4.02%. The Angstrom Exponent values below 0.0858, along with the spatial distribution of the three response variables, indicate that the primary aerosol source in the study area is wind-driven soil/sediment erosion, which shows a strong relationship with emission areas. No long-distance transport of aerosols was observed.