Evaluation of machine learning models for mapping soil salinity in Ben Tre province, Vietnam

In most tropical climates, one of the most serious natural dangers that negatively impacts agricultural operations in coastal regions is increasing sea levels because of climate alteration-induced soil salinity. This problem has become worse and has been happening more often in Vietnam's Mekong River Delta. Utilizing Sentinel-1 SAR C-band data in conjunction with 5 cutting-edge machine learning models-MLP-NN, RBF-NN, Gaussian Processes, SVR, and RF-the primary goal of the research is to map soil salinity invasion in Ben Tre region that is situated on the Mekong River Delta of Vietnam. In order to do this, 65 soil specimens were gathered in the grassland observation that took place between August 9 and 11, 2022, in accordance with the Sentinel-1 SAR images. The root-mean-square error, mean absolute error, and correlation coefficient were utilized to find and compare the performance of the 5 models. The GP model beat the other machine learning models and produced the best estimation performance (RMSE = 2.116, MAE = 1.247, and correlation coefficient = 0.904), according to the findings. We come to the conclusion that the latest machine learning models may be utilized to map the salinity of the soil in the Delta regions, offering a helpful tool to help farmers and policy makers choose more suitable crop varieties in light of climate change.