High Spatial Resolution of Soil Moisture Using Bagged Regression Trees and Spatio-Temporal Correlations from SMAP L2 Products

Recently, the efforts to obtain Soil Moisture (SM) global measurements, at high spatial resolution, have been increasing several years ago. As a result, there are many techniques to downscale SM from remote observations of different sensors. However, we have focused on the algorithm which was developed in 2018 by Chakrabarti, using spatio-temporal correlations of high-resolution remote sensing products, bagged regression trees (BRT), and in-situ SM measurements. We computed the algorithm to downscale SMAP Level 2 SM products (L2_SM_P_E) at 9 km to 1 km over agricultural fields in Mexico using optical observations such as Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), and Landcover (LC), and precipitation measurements (PPT). We found that the algorithm correctly downscales soil moisture at 1 km over our study area, especially over the corn fields the RMSD is 0.037 m
³
/m
³
. Despite the limitations that we encountered when using optical ancillary products due to adverse weather conditions and the difference of spatial and temporal resolutions between each space-borne mission.