Remote sensing and machine learning models to assess and predict turfgrass performance and soil water stress under deficit irrigation for creeping bentgrass

Read full text Summarise with VIRAT
Authors: Haoguang Yang, Devan Gladden, Stephanie Rossi, Bingru Huang
Cited by: 0
References: 41

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract Turfgrass management faces challenges related to water scarcity, making it essential to explore innovative technologies, such as remote sensing and machine learning (ML), to monitor turf quality (TQ) and soil moisture levels effectively. The objective of this study was to develop ML models using unmanned aerial vehicle and imagery technology to assess and predict TQ and soil water stress in creeping bentgrass ( Agrostis stolonifera ; cv. “007”) under variable deficit irrigation levels. Experimental plots were maintained under golf course fairway conditions and irrigated using a linear gradient irrigation system to create different levels of water stress. Vegetative indices (VIs) were generated from multispectral images collected during the experimental period of 2023 and 2024. TQ and soil water content (SWC) data were collected. Four ML models were developed from remote‐sensed and ground data using algorithms, such as random forest, extreme gradient boosting, light gradient‐boosting machine (LGBM), and multilayer perceptron (MLP). All four ML models significantly outperformed linear regression between VIs and TQ or SWC, with R 2 for both TQ and SWC exceeding 0.76. The MLP was most effective for predicting TQ, while the LGBM model performed best for SWC. Analysis of the feature importance analysis revealed that VIs were the primary drivers for both predictions, though their relative importance varied among models. The stacked model further improved prediction accuracy. These findings demonstrate the potential of using MLP and LGBM models for monitoring and predicting turfgrass responses to drought stress and model stacking method for prediction improvement.

Fields of Study & Subjects

Turfgrass Adaptation and Management Remote Sensing in Agriculture Smart Agriculture and AI

Publication details

Journal / source
Crop Science
Volume
66
Issue
2
Publication date
1 March 2026
Language
EN