Multi-Sensor and Multi-temporal High-Throughput Phenotyping for Monitoring and Early Detection of Water-Limiting Stress in Soybean

TL;DR

This study presents a time-series, multi-sensor UAV phenotyping pipeline to monitor and detect drought-induced canopy wilting in soybean. By integrating RGB, multispectral, hyperspectral, and thermal data with vegetation indices and machine learning, the authors identify red-edge bands, especially the Red-Edge Chlorophyll Index (RECI), as key discriminators for tolerant versus susceptible accessions and for pre-visual stress detection. The results show that multispectral data and VI-based features yield strong classification performance, with time-series analysis enabling early stress detection up to 0.74 accuracy across time points. The work demonstrates a practical, high-throughput approach for screening and selection in breeding programs and has potential applications in production for timely drought mitigation.

Abstract

Soybean production is susceptible to biotic and abiotic stresses, exacerbated by extreme weather events. Water limiting stress, i.e. drought, emerges as a significant risk for soybean production, underscoring the need for advancements in stress monitoring for crop breeding and production. This project combines multi-modal information to identify the most effective and efficient automated methods to investigate drought response. We investigated a set of diverse soybean accessions using multiple sensors in a time series high-throughput phenotyping manner to: (1) develop a pipeline for rapid classification of soybean drought stress symptoms, and (2) investigate methods for early detection of drought stress. We utilized high-throughput time-series phenotyping using UAVs and sensors in conjunction with machine learning (ML) analytics, which offered a swift and efficient means of phenotyping. The red-edge and green bands were most effective to classify canopy wilting stress. The Red-Edge Chlorophyll Vegetation Index (RECI) successfully differentiated susceptible and tolerant soybean accessions prior to visual symptom development. We report pre-visual detection of soybean wilting using a combination of different vegetation indices. These results can contribute to early stress detection methodologies and rapid classification of drought responses in screening nurseries for breeding and production applications.

Figures (5)

• Figure 1: Class categorization of drought stress symptoms in soybean utilized in analysis. Visual ground-truth data was collected on 450 diverse accessions in water-limited screening nursery via traditional wilt score rating scale of 1-6 adapted from previous studies king2009differentialkaler2017genomechamarthi2021identification. Visual plot scores of 1-6 were re-classified into multi-class (3-class) setting and binary (2-class setting) as specified for two analysis pathways.
• Figure 2: Accuracy of random forest method for soybean drought stress using all possible combinations of bands of the multispectral sensor. Mean accuracy for different number of bands for (A) 3 classes and (B) 2 class drought classification.
• Figure 3: Mean spectra of tolerant and susceptible soybean plots of time points (A) Time point 1: 46 DAP, (B) Time point 2: 65 DAP, (C) Time point 3: 81 DAP, at each band of the Micasense RedEdge MX Dual Camera. The symbol * indicates significant differences, with *, **, and *** indicating differences at $p < 0.05$, $p < 0.01$, and $p < 0.001$, respectively.
• Figure 4: Boxplots showing changes in Modified Soil-Adjusted Vegetation Index (MSAVI), Photochemical Reflectance Index (PRI), Red-Edge Chlorophyll Index (RECI), and Ratio Analysis of Reflectance Spectra B (RARSb) derived from soybean plot spectra across time points 1, 2, and 3. The symbol * indicates significant differences, with * and ** indicating differences at $p < 0.05$ and $p < 0.01$ levels, respectively.
• Figure 5: Classification accuracy for distinguishing tolerant and susceptible soybean plots by using different types of multispectral-based features over time points 1 (46 DAP), 2 (65 DAP), and 3 (81 DAP).