Yang et al. (2026) From ESWB to CDI-ESWB: A novel daily drought warning and monitoring method based on Eigen small water bodies using sentinel-1&2 and machine learning
Identification
- Journal: Journal of Hydrology
- Year: 2026
- Date: 2026-11-09
- Authors: Zhen Zhang, Linguang Miao, Yangyang Ma, Linyi Li, Wen Zhang, Lingkui Meng
- DOI: 10.1016/j.jhydrol.2026.136796
Research Groups
- School of Chemistry and Geographical Sciences, Chuxiong Normal University, Chuxiong 675000, China
- School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China
- School of Environmental Arts & Architectural Engineering, Heilongjiang University of Technology, Jixi 158100, China
Short Summary
This study introduces CDI-ESWB, a novel daily drought warning and monitoring method based on Eigen Small Water Bodies (ESWB) extracted from Sentinel-1 and Sentinel-2 imagery using machine learning. The method effectively captures drought dynamics and provides earlier warnings compared to official meteorological alerts.
Objective
- To develop and validate a novel daily drought warning and monitoring method, CDI-ESWB, utilizing Eigen Small Water Bodies (ESWB) derived from remote sensing and machine learning, to provide sensitive and early signals of drought conditions.
Study Configuration
- Spatial Scale: Chuxiong City, Yunnan Province, southwestern China.
- Temporal Scale: February 15, 2017 to June 1, 2025 (for data acquisition); daily scale for drought warning and monitoring.
Methodology and Data
- Models used: Random Forest (for ESWB extraction), UNet-based methods (for comparison), Z-score standardization.
- Data sources:
- Sentinel-1 images (982 images)
- Sentinel-2 images (982 images)
- Dynamic World V1
- Digital Elevation Model (DEM)
- Daily precipitation data
- Daily soil moisture data
- SDWI (Sentinel-1 derived water index)
- NDWI (Sentinel-2 derived water index)
Main Results
- Random Forest achieved high Intersection over Union (IoU) for ESWB extraction: 0.967 for Sentinel-1 and 0.996 for Sentinel-2, outperforming four UNet-based methods.
- The ESWB area difference between same-day Sentinel-1 and Sentinel-2 images was only 0.05%.
- The mean Eigen Small Water Bodies Index (ESWBI) exhibited an annual "V"-shaped pattern, reaching its minimum around mid-July each year.
- Drought warnings based on ESWBI matched 74.6% (53 cases) of official meteorological warnings and provided earlier warnings by approximately 5 days.
- The daily Composite Drought Index based on ESWB (CDI-ESWB) successfully captured severe drought events in 2019 and 2023, consistent with historical records.
- Five drought categories (Severe Wet, Mild Wet, Normal, Mild Drought, and Severe Drought) were defined using CDI-ESWB thresholds of 1.5, 0.5, -0.5, and -1.5, respectively.
Contributions
- Introduces a novel framework for daily drought warning and monitoring based on Eigen Small Water Bodies (ESWB).
- Develops a robust machine learning-based method for ESWB extraction using multi-source Sentinel satellite data.
- Proposes the ESWBI and CDI-ESWB, integrating ESWB dynamics with precipitation and soil moisture for comprehensive drought assessment.
- Demonstrates the capability of providing earlier drought warnings (approximately 5 days) compared to official meteorological warnings.
- Offers a new perspective for drought warning and monitoring applications using remote sensing.
Funding
No specific funding projects, programs, or reference codes are mentioned in the provided text.
Citation
@article{Yang2026From,
author = {Yang, Yunyuan and Wang, Zhe and Zhang, Zhen and Wang, Hui and Wang, Xi and Miao, Linguang and Ma, Yangyang and He, Ping and Li, Linyi and Zhang, Wen and Meng, Lingkui},
title = {From ESWB to CDI-ESWB: A novel daily drought warning and monitoring method based on Eigen small water bodies using sentinel-1&2 and machine learning},
journal = {Journal of Hydrology},
year = {2026},
doi = {10.1016/j.jhydrol.2026.136796},
url = {https://doi.org/10.1016/j.jhydrol.2026.136796}
}
Original Source: https://doi.org/10.1016/j.jhydrol.2026.136796