A Remote Sensing Driven Soil Moisture Estimator: Uncertain Downscaling With Geostatistically Based Use of Ancillary Data.

In recent years, remote sensing satellites, including Soil Moisture Active Passive (SMAP), continuously monitor SM of the earth. However, the spatial resolution of the remotely sensed SM is insufficient for some hydrological applications. In this paper, a data‐driven model is proposed based on a hybrid approach called artificial neural network kriging (ANNK) to capture the interaction between SM and ancillary data (AD), including normalized difference vegetation index (NDVI), altitude, slope, Antecedent Moisture Condition, AMC (5‐day cumulative precipitation), and daily air temperature (DAT)/land surface temperature (LST). By co‐regionalization of the estimated SM from the ANNK and SMAP 36‐km product, SM in any part of the study area is estimated more accurately and on a finer resolution (1‐km). Finally, through a linear relationship between in‐situ SM with the up‐to‐date AD including AMC (previous 5‐day cumulative precipitation), LST or DAT, slope, NDVI, and the downscaled satellite SM, the soil moisture estimates are adjusted by also adding an error term. Estimation of the final SM product in terms of the Unbiased Root Mean Square Error of less than 0.04 m3/m3 meets the accuracy requirements of the SMAP SM retrieval. Moreover, the results of SM validation show that the proposed technique has a potential for real‐time SM estimations in developing areas. Key Points: A combined data‐driven geostatistically based model is developed for uncertain downscaling Soil Moisture Active Passive (SMAP) soil moisture (SM) data utilizing ancillary data (AD)Proposing a platform for uncertain SM estimation with 5‐day antecedent moisture condition that utilizes land cover and an error termA remarkable estimation of 1 km downscaled SM data by estimating missing data in 36 km SMAP readings using cokriging and AD [ABSTRACT FROM AUTHOR]