Gaussian process based spatial modeling of soil moisture for dense soil moisture sensing network

Abstract

Agricultural practices by Wireless sensor networks (WSN) together with precision irrigation systems facilitate efficient use of water resources to maintain soil water balance and crop water requirement. In situ soil moisture measurements are expensive, point-based and cannot be scaled spatially over a field. In this work, to provide reasonable soil moisture maps across the site, Gaussian process regression (GPR) is used. Furthermore, soil moisture semivariograms are modeled by GPR using Matérn covariance function to generate interpolated surfaces of soil moisture.