Monitoring Wheat Crop Biochemical Responses to Random Rainfall Stress Using Remote Sensing: A Multi-Data Approach

Abstract

Random rainfall poses a significant threat to wheat production in India, adversely affecting grain quality and yield. Precise assessment of rain-induced stress responses on wheat is crucial for implementing timely preventive measures to mitigate the damage. While traditional methods involve toxic chemicals and laborious processes, satellite data offers a quicker, easier, and eco-friendly alternative. This study explores the potential of Sentinel-2 data to monitor biochemical changes in wheat crops caused by random rainfall events. Over three years, satellite, drone, weather, and agronomic data were collected for wheat crops. Four datasets were integrated through a multi-step processing workflow: agronomic data were derived from laboratory analyses of leaf samples collected at different growth stages of the wheat crop; drone imagery was processed via mosaicking, orthorectification, and geo-registration; and Sentinel-2 data underwent atmospheric correction, resampling, and sub-setting to produce reflectance maps. Weather data, including rainfall patterns, were directly sourced from the Agromet Field Unit at IIT Roorkee, India. Vegetation indices were extracted from the satellite and drone data, and these indices were used to develop a partial least square regression (PLSR) model to predict eight key biochemical variables. During rainfall stress, the study observed significant biochemical shifts, with declines in chlorophyll, carbohydrate, protein, and lipid levels, and increases in phenol, proline, hydrogen peroxide, and superoxide dismutase. The PLSR model accurately captured these changes, achieving R2-values $\ge 0.7$ and root mean square error (RMSE) <0.12 for all variables except lipids. The results highlight the effectiveness of this approach in monitoring rain-induced stress on wheat crops, providing a non-invasive and reliable method for protecting grain quality and minimizing yield losses. This technique offers a promising solution for large-scale, eco-friendly agricultural monitoring and stress management.

Keywords

• Biochemical variables
• non-invasive techniques
• random rainfall
• sentinel-2 data
• stress monitoring
• vegetation indices