Machine learning approaches for soil moisture prediction: enhancing agricultural water management with integrated data

Abstract

Effective soil moisture prediction is essential for precision agriculture, enabling optimised irrigation scheduling and sustainable water management. This study investigates the predictive performance of various machine learning (ML) models for soil moisture at two depths (0–30 cm and 0–100 cm) using integrated datasets of climate variables, soil properties, and leaf area index (LAI). We tested models including Random Forest (RF), XGBoost, Support Vector Regression (SVR), and Gaussian Process Regression (GPR), alongside traditional models such as Partial Least Squares (PLS) and Multiple Linear Regression (MLR). The RF and XGBoost models provided the highest accuracy among all the models tested, with the RF model yielding an NSE of 0.81 and R2 of 0.84 at 0–30 cm depth. Notably, RF seemed to perform reasonably well even when forced to rely solely on climate data inputs, with NSE values above 0.75, a performance supporting the robustness of the models in data-scarce conditions. The study further shows that ensemble ML models are better at capturing nonlinear interactions among soil, climate, and crop factors compared to traditional models, by up to 25% improvement in NSE. Soil depth was found to be one of the main factors that influence model performance. The highest prediction accuracy was recorded at shallower depths, which points toward its sensitivity to the dynamics of moisture content within the root zone. Similarly, the feature importance analysis has identified LAI and organic matter in the soil as critical predictors. These findings underline the potential of the use of ensemble ML models in operational soil moisture monitoring in agriculture as a scalable approach to improved water-use efficiency with sustained variable climate conditions.

Document Type

Journal Article

Date of Publication

1-1-2026

Volume

12

Issue

1

Publication Title

Modeling Earth Systems and Environment

Publisher

Springer

School

School of Science

RAS ID

84610

Funders

University of Southern Queensland / Queensland Department of Primary Industries

Comments

Ali, A. (2025). Machine learning approaches for soil moisture prediction: enhancing agricultural water management with integrated data. Modeling Earth Systems and Environment, 12. https://doi.org/10.1007/s40808-025-02607-5

Copyright

subscription content

Share

 
COinS
 

Link to publisher version (DOI)

10.1007/s40808-025-02607-5