Machine learning algorithms for soil moisture estimation using Sentinel-1: Model development and implementation

The present study provided the first-time comprehensive evaluation of 12 advanced statistical and machine learning (ML) algorithms for Soil Moisture (SM) estimation from dual polarimetric Sentinel-1 radar backscatter. ML namely support vector (SVM) with linear, polynomial, radial sigmoid kernel, random forest (RF), multi-layer perceptron (MLP), basis function (RBF), Wang Mendel’s (WM), subtractive clustering (SBC), adaptive neuro fuzzy inference system (ANFIS), hybrid interference (HyFIS), dynamic evolving neural (DENFIS) were used. Extensive field samplings performed collection in-situ SM data other parameters selected sites seven different dates at two locations (Varanasi Guntur District, India), concurrent to overpasses. backscattering coefficients considered as input variables output variable training, validation testing algorithms. site Varanasi was used models. On hand, an independent checking model performance, before finalizing performances trained evaluated in terms correlation coefficient (r), root mean square error (RMSE) (in m3/m3) bias m3/m3). identified RF, SBC ANFIS top three best performing models comparable promising estimation. In order test robustness these (RF, ANFIS), further performance analysis datasets sites, which indicates that consistent can be recommended among all