Novel indices and multi-source data fusion for monitoring plant moisture stress in winter wheat fields.

Abstract

Drought is a significant challenge to winter wheat production. Its impact can be mitigated by preventing plant moisture stress through precision agriculture. Remote sensing and machine learning have proven effective for managing moisture stress in winter wheat. This study highlights the potential of new indices that combine visible (VIS) and near-infrared (NIR) bands along with canopy temperature (Tc), to monitor plant moisture content (PMC) and leaf moisture content (LMC) in winter wheat under irrigation treatments: W0 (no irrigation), W1 (45-65%), W2 (55-75%), W3 (65-85%), W4 (75-95%) of field capacity, and Z (irrigation and rainfall). Our findings show that the ratio stress index (RSI), with band combinations such as RSI7 (650, 428) , RSI8 (663, 422) , and RSI9 (671, 450) , performs better in tracking PMC and LMC, demonstrating high correlation and improved average prediction metrics for vegetation index (VI) models with R 2 , RMSE, and MAE of 0.838, 2.791, and 2.093 respectively, for LMC and VI-Tc input models with 0.850, 2.731, and 2.105 for PMC. Incorporating Tc into RSI models enhances prediction accuracy, increasing R² by up to 13.82% in the RSI-Tc-SVM-PMC model and decreasing RMSE and MAE by 15.89% and 18.33%, respectively. Therefore, a combination of RSI-Tc-SVM-ANN is recommended to monitor winter wheat moisture stress.

Plant phenotyping relevance

冬小麦の植物・葉の含水量および水分ストレスを、VIS/NIRと冠層温度のデータ融合および機械学習で推定する手法が研究の中心であり、植物生理状態の定量的フェノタイピングに該当する。

Code and data availability

The paper's phenotype data (spectroradiometer canopy reflectance, canopy temperature, PMC/LMC measurements) and R-based analysis are not publicly deposited; the authors state data are available only upon reasonable request. No public code or dataset URL is provided.

No evidence-backed public reproduction asset is currently recorded.