DTSM: A UAV-based framework for temporal rice stress detection and growth analysis

Timely detection of crop stress is important for precision agriculture and agroecosystem resilience, particularly in environments where ground calibration data are limited or environmental conditions change rapidly. Traditional stress detection approaches often rely on single-date imagery or external calibration, which can limit temporal interpretation and transferability across heterogeneous fields. This study presents the Dynamic Temporal Stress Method (DTSM), a data-driven framework for monitoring and quantifying vegetation stress patterns in rice fields using multi-temporal UAV observations. DTSM integrates four vegetation indices, namely NDVI, NDRE, GNDVI, and MSAVI2, derived from UAV imagery acquired at three phenological stages (July 17, August 1, and August 12, 2024). The framework combines binary thresholding for initial stress mapping, temporal differencing to track stress change, and a Stress Persistence Index (SPX) to quantify repeated stress occurrence over time. Spatial patterns of stress were further evaluated using zonal statistics and Getis–Ord Gi* hotspot analysis to identify persistent stress clusters. Results showed clear temporal variation in crop stress, with stress intensity peaking on August 1, when NDRE-based mapping indicated 41.4% of the field under stress. Approximately 15.7% of the study area exhibited persistent stress across all observation dates. Spatial analysis identified Zones Z3, Z4, and Z8 as recurring stress hotspots. These findings indicate that integrating temporal change detection, multi-index analysis, and spatial persistence assessment can improve the interpretation of crop stress dynamics and support precision management in UAV-based agricultural monitoring.