Cho et al. (2025) Assessing Drought Stress in Legumes using Block-Chain-Assisted Drone-based Spectral Sensors

Identification

• Journal: Legume Research - An International Journal
• Year: 2025
• Date: 2025-12-30
• Authors: Kyung won Cho, Sangmin Lee
• DOI: 10.18805/lrf-869

Research Groups

• Graduate School of AI Policy and Strategy, GIST, Republic of Korea
• Department of Healthcare Data Science and Artificial Intelligence, CHA University, Republic of Korea

Short Summary

This study developed and evaluated a blockchain-assisted drone-based spectral sensing system for secure and accurate drought stress assessment in legume crops. The system successfully classified drought stress levels with high accuracy using machine learning, providing transparent and tamper-proof data for precision agriculture management.

Objective

• To develop and evaluate a blockchain-assisted drone-based spectral sensing system for rapid, non-invasive, and scalable monitoring of drought stress in legume crops, ensuring data security, transparency, and traceability.

Study Configuration

• Spatial Scale: Experimental plots in Maharashtra, India (a semi-arid region with 400-700 mm annual rainfall and 25-38 °C temperature). Soil type ranged from sandy loam to black soil (pH 6.8-7.5, CEC 18-22 cmol/kg). Five legume crops were studied: Chickpea, Soybean, Cowpea, Lentil, and Groundnut. Drone operation altitude was approximately 50 meters above ground level, with an image resolution of 10 cm/pixel.
• Temporal Scale: Data collected twice weekly across critical growth stages: vegetative, flowering, and pod-filling stages.

Methodology and Data

• Models used:
  • Machine Learning: Random Forest (RF), Support Vector Machine (SVM), Decision Tree for drought stress classification.
  • Statistical Analysis: Analysis of Variance (ANOVA) and Critical Difference (CD) test (p≤0.05) for mean comparisons.
• Data sources:
  • Drone-mounted multispectral sensor data (Blue: 450 nm, Green: 550 nm, Red: 650 nm, Red-Edge: 725 nm, NIR: 850 nm).
  • Calculated Vegetation Indices: Normalized Difference Vegetation Index (NDVI), Photochemical Reflectance Index (PRI), Normalized Difference Water Index (NDWI).
  • Ground Truth Data: Leaf water potential (measured with a pressure chamber), Relative Water Content (RWC), and canopy temperature.
  • Soil physicochemical parameters (pH, cation exchange capacity, available nitrogen, phosphorus, potassium, water holding capacity).
  • Blockchain platform for secure, tamper-proof storage and real-time sharing of all collected and processed data.

Main Results

• Drought stress significantly reduced vegetation index values (NDVI, PRI, NDWI) and increased canopy temperatures across all legume crops. For severe stress, mean NDVI dropped to 0.49, PRI to 0.03, and NDWI to 0.42, compared to 0.81, 0.12, and 0.75 respectively in control conditions.
• Strong positive correlations were found between leaf water content and NDVI (r = 0.89) and NDWI (r = 0.91), while all vegetation indices showed strong negative correlations with canopy temperature (r ranging from -0.80 to -0.88).
• The Random Forest machine learning model achieved the highest classification accuracy of 92% in distinguishing drought stress levels, outperforming Support Vector Machine (88%) and Decision Tree (84%).
• The blockchain-assisted system ensured secure, transparent, and immutable storage of all spectral and physiological data, enhancing data integrity and traceability for decision-making.
• Crop-specific responses indicated Chickpea and Lentil were more sensitive to severe drought, showing steeper declines in NDVI, while Cowpea and Groundnut exhibited greater drought adaptability.

Contributions

• Introduces a novel and robust framework for drought stress assessment by integrating drone-based spectral sensing with blockchain technology, ensuring data security, transparency, and traceability from acquisition to analysis.
• Demonstrates high accuracy (92% with Random Forest) in classifying drought stress levels in legumes using machine learning on spectral data, enabling precise identification of affected zones.
• Provides a scalable, non-invasive, and rapid approach for real-time drought monitoring, supporting optimized irrigation and site-specific management in precision agriculture.
• Enhances trust and accountability among stakeholders by providing verifiable and tamper-proof digital records of crop health and environmental conditions.

Funding

• National Research Foundation of Korea (NRF) grant funded by the Korea Government (MSIT) (No. 2022R1F1A107645311).

Citation

@article{Cho2025Assessing,
  author = {Cho, Kyung won and Lee, Sangmin},
  title = {Assessing Drought Stress in Legumes using Block-Chain-Assisted Drone-based Spectral Sensors},
  journal = {Legume Research - An International Journal},
  year = {2025},
  doi = {10.18805/lrf-869},
  url = {https://doi.org/10.18805/lrf-869}
}