Qi et al. (2025) Monitoring River–Lake Dynamics in the Mid-Lower Reaches of the Yangtze River Using Sentinel-2 Imagery and X-Means Clustering

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-10-13
• Authors: Zhanshuo Qi, Shiming Yao, Xiaoguang Liu, Bing Ding, Hongyang Wang, Yuqi Jiang, Jinpeng Hu
• DOI: 10.3390/rs17203421

Research Groups

• Changjiang Water Resources Committee, Changjiang River Scientific Research Institute, Wuhan, China
• Key Laboratory of River and Lake Regulation and Flood Control in the Middle and Lower Reaches of the Changjiang River, Ministry of Water Resources, Wuhan, China
• River Research Department, Changjiang River Scientific Research Institute, Wuhan, China
• Spatial Information Technology Application Department, Changjiang River Scientific Research Institute, Wuhan, China

Short Summary

This study developed a robust Sentinel-2 and X-means clustering-based method to monitor river-lake dynamics in the Mid-Lower Reaches of the Yangtze River (MLRYR) from 2018-2023, finding overall surface water area (SWA) stability but significant declines in major lakes (Poyang, Dongting, Shijiu) and an increase in Danjiangkou Reservoir, with river networks buffering climatic impacts.

Objective

• To employ a precise and efficient water extraction method based on the Vegetation Red Edge-based Water Index (VREWI) and a modified multidimensional X-means clustering algorithm with Sentinel-2 imagery, and to construct and validate a new dataset for large-scale river–lake system research in the MLRYR.
• To analyze the spatiotemporal dynamics of the river–lake system in the sub-basins of the MLRYR over six years (2018–2023), identifying change characteristics for rivers, lakes, and reservoirs.
• To analyze how large lakes and reservoirs in the MLRYR sub-basins respond to climatic events, particularly the buffering role of river networks.

Study Configuration

• Spatial Scale: Mid-Lower Reaches of the Yangtze River (MLRYR), encompassing six major second-level basins (Hanjiang River, Middle Main Stream, Lower Main Stream, Taihu Lake, Dongting Lake, and Poyang Lake Basins), covering an area of 7.95 × 10^5 km^2. Focus on lakes/reservoirs with an area greater than 10 km^2 and the river network system.
• Temporal Scale: Six-year period from 2018 to 2023, analyzed seasonally (Spring: March–May, Summer: June–August, Autumn: September–November, Winter: December–February of the following year).

Methodology and Data

• Models used:
  • Multidimensional X-means clustering algorithm (main water extraction method).
  • Vegetation Red Edge-based Water Index (VREWI) for water body differentiation.
  • Mathematical morphology methods (dilation and erosion algorithms) for river structure extraction and noise reduction.
  • Normalized Sen’s slope calculations for annual mean SWA trend analysis.
  • Pearson correlation coefficient for analyzing relationships between SWA and climatic factors.
  • Standardized Precipitation Evapotranspiration Index (SPEI) for drought assessment.
  • Comparison methods: Support Vector Machine (SVM), NDWI-B12 clustering, and a multi-index water extraction method (AWEI, MNDWI, NDVI, EVI).
• Data sources:
  • Sentinel-2 imagery products (2018–2023) with cloud coverage below 30%, processed on Google Earth Engine (GEE).
  • Cloud Score+ S2_HARMONIZED V1 dataset for cloud contamination removal.
  • Joint Research Centre (JRC) Global Surface Water (GSW) dataset for comparison.
  • HydroLAKES lake polygons dataset and National Major Lakes Distribution dataset for lake/reservoir boundaries.
  • Global River Widths from Landsat (GRWL) data for river channel distribution.
  • ERA5-Land monthly dataset (0.1° × 0.1° spatial resolution) for monthly climate information (precipitation, temperature, evaporation).
  • High-resolution Google Earth historical imagery and field-surveyed river channel boundary data from the Changjiang Water Resources Commission for validation and boundary refinement.

Main Results

• The developed water extraction method (VREWI + X-means clustering) achieved high accuracy: Overall Accuracy (OA) of 97.98%, Producer’s Accuracy (PA) of 98.02%, User’s Accuracy (UA) of 96.01%, Matthews Correlation Coefficient (MCC) of 0.954, and Kappa coefficient of 0.954, outperforming comparison methods.
• The overall Surface Water Area (SWA) in the MLRYR remained relatively stable from 2018 to 2023, with an annual mean of 35,514.82 km^2 in 2018, declining to 31,661.52 km^2 by 2023. The maximum annual mean SWA was 39,489.92 km^2 in 2020, and the minimum was 29,992.64 km^2 in 2022.
• Seasonal SWA variations showed a clear cyclical trend, with maximum inundation predominantly in spring (peak of 53,445.23 km^2 in spring 2020) and minimum in winter (minimum of 25,784.81 km^2 in winter 2022).
• Significant declines in water area were observed in major lakes: Poyang Lake (decreased by 279.21 km^2), Dongting Lake (decreased by 326.46 km^2), and Shijiu Lake.
• Danjiangkou Reservoir recorded the most significant increase in water area, expanding by 33.29 km^2 to 716.49 km^2. Honghu Lake and Liangzi Lake also showed increases.
• SWA exhibited the strongest correlations with precipitation and drought index (SPEI) in most sub-basins.
• In sub-basins containing large lakes and reservoirs, the presence of dense river networks played a buffering role by regulating and storing water, thereby reducing the direct influence of climatic factors on lake and reservoir water extent.

Contributions

• Developed and validated a robust, precise, and efficient water extraction method for large-scale river-lake systems using Sentinel-2 imagery, VREWI, and X-means clustering, demonstrating superior performance over traditional methods.
• Generated a new, high-resolution (10 m) seasonal SWA product for the MLRYR (2018-2023), providing enhanced detail for monitoring aquatic ecosystem dynamics.
• Provided a comprehensive spatiotemporal analysis of river, lake, and reservoir dynamics at sub-basin scales in the MLRYR, distinguishing changes across different water body types and revealing regional heterogeneity.
• Identified and quantified the buffering role of dense river networks in mediating the impact of climatic factors (precipitation, drought) on large lakes and reservoirs, advancing the understanding of complex hydrological-climatic interactions.
• Offers critical insights into drought impacts and informs effective water resource management and conservation strategies, providing a valuable framework applicable to other basins facing similar environmental challenges.

Funding

• National Natural Science Foundation of China (grant number No. U2240224)
• National Key R&D Program of China (grant number 2022YFC3202601)
• Special Fund of the Chinese Central Government for Basic Scientific Research Operations in Commonwealth Research Institute (grant numbers CKSF2024326/HL, CKSF2023343/HL, CKSF2023328/HL)

Citation

@article{Qi2025Monitoring,
  author = {Qi, Zhanshuo and Yao, Shiming and Liu, Xiaoguang and Ding, Bing and Wang, Hongyang and Jiang, Yuqi and Hu, Jinpeng},
  title = {Monitoring River–Lake Dynamics in the Mid-Lower Reaches of the Yangtze River Using Sentinel-2 Imagery and X-Means Clustering},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17203421},
  url = {https://doi.org/10.3390/rs17203421}
}