Dynamic monitoring of soil moisture in cotton fields using electromagnetic induction and extended Kalman filters

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Dynamic monitoring of soil moisture (βg) is crucial for optimizing agricultural water use.Non-invasive electromagnetic induction (EMI) presents a promising alternative, adopting measurements of apparent electrical conductivity (ECa) to establish inversion models for predicting βg across different soil layers. However, the accuracy of these static models is limited. This study aimed to develop and validate a dynamic monitoring framework by integrating EMI with an Extended Kalman Filter (EKF) algorithm. The research was conducted in cotton fields using drip irrigation under plastic films in the Alaer Reclamation Area. First, ECa data were acquired using an EM38-MK2 ground conductivity meter in both vertical and horizontal dipole orientations during different growth periods. The optimal modeling factor was determined by combining βg values at various periods and soil depths. Subsequently, the EKF algorithm was implemented to assimilate the time-series ECa data, thereby refining the accuracy of the βg estimations. Finally, natural neighbor interpolation (NNI) was used to generate and analyze three-dimensional (3D) spatiotemporal distributions of soil moisture. The results show that the EKF improved the model accuracy. For the soil depth of 60-80 cm in October, the coefficient of determination (R2) is increased from 0.36 to 0.83; at a soil depth of 60-80 cm in March, the Root Mean Square Error (RMSE) is decreased from 12.37 to 1.65. This research validates a novel methodology for monitoring high-accuracy, real-time soil moisture in arid regions. The proposed framework provides critical guidance for precision irrigation management in cotton cultivation, contributing to more sustainable agricultural water use.

This publication has 21 references indexed in Scilit: