Modeling Soil Moisture from Multisource Data by Stepwise Multilinear Regression: An Application to the Chinese Loess Plateau

This study aims to integrate multisource data model the relative soil moisture (RSM) over Chinese Loess Plateau in 2017 by stepwise multilinear regression (SMLR) order improve spatial coverage of our previously published RSM. First, 34 candidate variables (12 quantitative and 22 dummy variables) from Moderate Resolution Imaging Spectroradiometer (MODIS) topographic, properties, meteorological were preprocessed. Then, SMLR was applied without multicollinearity select statistically significant (p-value < 0.05) variables. After accuracy assessment, monthly, seasonal, annual patterns RSM mapped at 500 m resolution evaluated. The results indicate that there a high potential with desired (best fit Pearson’s r = 0.969, root mean square error 0.761%, absolute 0.576%) Plateau. elevation (0–500 2000–2500 m), precipitation, texture loam, nighttime land surface temperature can continuously be used models for all seasons. Including improved both calibration validation. Moreover, SMLR-modeled achieved better than reference almost periods. is finding as method supports use complement and/or replace coarse satellite imagery estimation