REOPTRAM: a reconstructed optical trapezoidal model that utilizes drone spectra to optimize satellite feature spaces for enhanced soil moisture mapping in irrigation areas

The current OPtical TRApezoid Model (OPTRAM) has received much attention in estimating soil moisture content (SMC). However, the inherent correlation between satellite shortwave infrared (SWIR) mixed pixels and SMC is relatively weak in vegetated areas. On the other hand, the dry/wet edges in the normalized difference vegetation index (NDVI)-shortwave infrared transformed reflectance (STR) feature space do not always exhibit easily fitted linear patterns, particularly in vegetation-soil mixed areas. This directly results in low SMC estimation accuracy. To this end, this study innovatively introduced upscaled drone spectral data to optimize the satellite NDVI-STR characteristic space and proposed a REconstructed OPtical TRApezoid Model (REOPTRAM). Meanwhile, to extract spectral information most relevant to SMC, the effect of different aggregation scales (step = 1 m) on the satellite SWIR reconstruction performance was investigated specifically during the upscaling of drone spectra. Finally, the REOPTRAM was determined using the reconstructed NDVI-STR space and tested for its accuracy in estimating SMC in irrigation districts. The result showed that (1) the SWIR reconstruction accuracy showed a trend of first increasing and then decreasing with expanding aggregation scales. There existed an optimal aggregation scale (13 m) between the drone resolution and satellite resolution that maximizes reconstruction accuracy. (2) The Pearson correlation coefficient ( r ) between the satellite SWIR bands (including the SWIR1 and SWIR2) reconstructed based on the optimal aggregation scale and the observed SMC increased from 0.479 to 0.680 on average, demonstrating a significant improvement. After reconstructing the NDVI-STR space, the dry/wet edges also became easier to fit with a linear regression. (3) Compared with OPTRAM, the SMC estimation accuracy based on REOPTRAM was significantly improved. The R 2 , RMSE, and MAE between SMC predicted and observed values increased from 0.503 to 0.763, reduced from 0.029 to 0.022, and decreased from 0.022 to 0.017, respectively. Moreover, REOPTRAM generated more reasonable SMC estimated values spatially, particularly showing decreases in building areas and increases in water bodies. The REOPTRAM offers the advantages of low cost, good applicability, and high accuracy, which provides a new direction and theoretical support for improving SMC estimation.