Author(s): Xiang Zhang; Shuzhe Huang; Tailai Huang

Linked Author(s):

Keywords: Deep learning; Point-surface data fusion; Surface soil moisture

Abstract: Surface soil moisture (SSM) has considerable impact in land-atmosphere exchanges of water and energy. However, due to the inherent deficiencies of remotely sensed data (e. g., cloud contamination), none of the current algorithms alone can provide daily and seamless SSM at field scale (i. e., 30 m). To explore this research gap, we proposed a novel SSM fusion framework of Generating high Resolution, Accurate, Seamless data using Point-Surface fusion (GRASPS) based on multiple remotely sensed, reanalysis, and in-situ datasets. First, 30 m seamless continuous SSM correlated variables (land surface temperature, NDVI, Albedo) were generated by enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM). Then, downscaled auxiliary variables and other background variables were input into a deep learning model to produce 30 m daily and seamless SSM. The average Pearson correlation coefficient (PCC), root mean squared error (RMSE), unbiased RMSE (ubRMSE), Bias, and mean absolute error (MAE) over all validation sites for the downscaled SSM achieved 0.78,0. 048 m^3 m^ (-3), 0.033 m^3 m^ (-3), -0.001 m^3 m^ (-3), and 0.041 m^3 m^ (-3), respectively. After bias correction, the RMSE, ubRMSE, Bias, and MAE at validation sites further decreased by 13%, 7%, 22%, and 18%, respectively.

DOI: https://doi.org/10.3850/iahr-hic2483430201-502

Year: 2024