Abstract
this parameter is of the utmost importance. Point-based measurements of soil moisture while accurate, are expensive in terms of time and effort, not to mention that their inability to depict spatial variability of SMC accurately on a large scale. Soil moisture retrieval methods using remote sensing technologies show great promise but suffer from numerous limitations. To minimize the effects of those limitations, a novel decision level data fusion algorithm for SMC estimation is proposed in this research. Initially, individual estimations are determined from 3 different methodologies; the inversion of Empirically Adapted Integral Equation Model (EA-IEM) which is semi-empirically calibrated using a parameter Lopt for Sentinel-1, the Perpendicular Drought Index (PDI), and Temperature Vegetation Dryness Index (TVDI) for LANDSAT-8. Then, three feature level fusions using novel combinations of salient features extracted from each of the method mentioned above are performed using an Artificial Neural Network (ANN). The latter is characterized by the modification of its performance function from absolute error to Root Mean Square Error. Finally, all estimations including the feature level fusions estimation are fused at the decision level using a novel weights-based estimation, which is implemented through a novel Matlab code. The performance of the proposed system is validated and tested using measurements collected from three study areas, an agricultural field in Blackwell farms, Guildford, United Kingdom, and two different agricultural fields in Sidi Rached, Tipasa, Algeria. Those measurements consisted of SMC level, and surface roughness parameters which were extracted using a newly designed laser profilometre. The proposed SMC estimation system produces stronger correlations and lower RMSE values than any individual SMC estimation in the order of at least 0.38%, 1.4%, and 1.09% for Blackwell farms, Sidi Rached 1 and Sidi Rached 2 datasets respectively.