Electrical Conductivity and Calcium Carbonate Mapping Combining Prisma Imagery and Machine Learning Techniques

Soil salinization and calcification in arid lands facilitate the desertification process, reduce the nutrient availability for plants, and disturb ecological equilibrium. The achievement of the perspectives of spatial variation in electrical conductivity (EC) and calcium carbonate equivalent (CCE) and monitoring based on imaging spectroscopy sensors ensures sustainable and reliable management over large areas. In this study, to evaluate the capability of the PRISMA imagery to predict spatial variations of EC and CCE using machine learning (ML) regression algorithms, 252 soil samples from the Sirjan region in Iran, which is mainly under pistachio cultivation, were used. The highest accuracy for CCE estimation using PRISMA imagery was acquired by the Gaussian Process Regression (GPR) algorithm (R
²
=0.75, RMSE=4.09), and for EC, it was acquired by the partial least squares regression (PLSR) algorithm (R
²
=0.64, RMSE=44.8). Moreover, results demonstrate that gypsum abundance is the main confounding parameter impacting the capability of PRISMA for EC estimation.