Liu et al. (2026) Monitoring the river ice phenology along the Inner Mongolia reach of the Yellow River using time-series images from landsat and Sentinel-2

Identification

• Journal: Journal of Hydrology Regional Studies
• Year: 2026
• Date: 2026-01-14
• Authors: Bin Liu, Honglan Ji, Haifeng Xu, Yu Deng, Hongchun Luo, Zhongshu Xue, Wenhao Ren
• DOI: 10.1016/j.ejrh.2026.103140

Research Groups

• State Key Laboratory of Water Engineering Ecology and Environment in Arid Area, Inner Mongolia Agricultural University, Hohhot, China
• Inner Mongolia Key Laboratory of Ecohydrology and High Efficient Utilization of Water Resources, Hohhot, China
• Inner Mongolia section of the Yellow River Basin Water Resources and Water Environment Comprehensive Management Autonomous Region Collaborative Innovation Center, Hohhot, China
• Yellow River Institute of Hydraulic Research, Zhengzhou, China

Short Summary

This study developed a segment-based area-ratio framework using multi-source Landsat-8/9 and Sentinel-2A/B imagery to automatically extract and analyze river-ice phenology along the Inner Mongolia reach of the Yellow River for six ice seasons (2018–2023). The research revealed distinct downstream-to-upstream freeze-up and upstream-to-downstream break-up patterns, with ice-covered duration increasing downstream, primarily driven by thermal conditions with secondary influences from flow dynamics and reservoir regulation.

Objective

• To develop an automated and transferable segment-based area-ratio framework for extracting river-ice phenology from multi-source optical satellite data at the reach scale.
• To characterize the spatiotemporal patterns of freeze-up dates, break-up dates, and ice-covered duration along the entire Inner Mongolia reach of the Yellow River (IMYR).
• To investigate how variations in river-ice phenology are associated with key meteorological and hydrological factors.

Study Configuration

• Spatial Scale: The Inner Mongolia section of the Yellow River (IMYR) in northern China, approximately 830 km long, spanning 39°–41°N and 106°–112°E, with elevations around 1000 m. The river was partitioned into five sub-sections (A–E) and analyzed using standardized 10 km long by 3 km wide river segments.
• Temporal Scale: Six complete winter ice seasons, from 2018 to 2023, were analyzed for river-ice phenology. Satellite imagery was acquired from 2013 to 2023 (Landsat) and 2018 to 2023 (Sentinel-2).

Methodology and Data

• Models used:
  • Random Forest models for ice-water classification.
  • Segment-based area-ratio framework for river-ice phenology extraction.
  • Modified Normalized Difference Water Index (MNDWI) combined with OTSU threshold segmentation for river mask delineation.
  • Spearman rank correlation analysis to assess relationships between phenological indicators and environmental variables.
• Data sources:
  • Satellite Imagery: Landsat-8/9 (30 m spatial resolution, 8-day revisit interval) and Sentinel-2A/B (10 m spatial resolution, 5-day revisit interval). Data processed on Google Earth Engine (GEE).
  • Hydrological Data: Daily discharge and water level records from Shizuishan, Bayangaole, Sanhuhekou, Baotou, and Toudaoguai hydrological stations (Yellow River Conservancy Commission, YRCC).
  • Meteorological Data: Daily mean air temperature and solar radiation from the ERA5-Land reanalysis dataset (0.1° spatial resolution, approximately 9 km).
  • Ancillary Data: Watershed vector data from the National Cryosphere Desert Data Center.

Main Results

• Ice-Water Classification Accuracy: Both Random Forest models demonstrated high accuracy. The Landsat model achieved an Overall Accuracy (OA) of 0.980 and a Kappa coefficient of 0.954. The Sentinel-2 model showed slightly superior performance with an OA of 0.984 and a Kappa coefficient of 0.968.
• Phenology Validation: Remote sensing-derived freeze-up and break-up dates showed strong agreement with in-situ hydrological station observations. For freeze-up, R² = 0.72 and Mean Absolute Error (MAE) = 3.92 days. For break-up, R² = 0.70 and MAE = 4.20 days.
• Spatial-Temporal Variation (2018–2023 averages):
  • Freeze-up Date: Exhibited a distinct downstream-to-upstream progression. The earliest average freeze-up was in sub-section E (Toudaoguai–Qingshuihe) on December 6, while the latest was in sub-section A (Wuhai–Bayangaole) on January 1.
  • Break-up Date: Showed an upstream-to-downstream progression. The earliest average break-up was in sub-section A (Wuhai–Bayangaole) on February 24, while the latest was in sub-sections C (Sanhuhekou–Baotou) and E (Toudaoguai–Qingshuihe) on March 16.
  • Ice-Covered Duration: Increased progressively downstream. Sub-section A had the shortest average duration (55 days), while sub-section E had the longest (100 days).
• Influencing Factors (for natural middle reaches B, C, D):
  • Freeze-up: Moderately negatively correlated with Freezing Degree Days (FDD) (R = -0.52, P = 0.04) and positively correlated with mean air temperature (R = 0.48, P = 0.06), indicating earlier freeze-up with rapid cold accumulation.
  • Break-up: Moderately negatively correlated with Thawing Degree Days (TDD) (R = -0.54, P = 0.03) and mean air temperature (R = -0.55, P = 0.03), suggesting accelerated warming leads to earlier ice decay.
  • Ice-Covered Duration: Negatively correlated with mean air temperature (R = -0.51, P = 0.06) and moderately negatively correlated with river discharge (R = -0.33, P = 0.25).
  • Overall, river-ice phenology is primarily controlled by thermal conditions, with flow dynamics and reservoir operations exerting secondary effects that amplify spatial heterogeneity.

Contributions

• First systematic monitoring of river-ice phenology along the Inner Mongolia reach of the Yellow River using multi-source optical remote sensing data (Landsat-8/9 and Sentinel-2A/B).
• Development of a novel segment-based area-ratio method for automated, transferable, and scalable extraction of river-ice phenology, which overcomes limitations of traditional single-pixel or station-based approaches by providing spatially continuous and temporally consistent characterization at the reach scale.
• Provides new hydrological insights into the spatiotemporal dynamics of river-ice formation and decay in a partially regulated mid-latitude river system, highlighting the combined influence of climatic conditions and human activities (reservoir regulation).
• Establishes a robust and transferable workflow for large-scale river-ice monitoring, offering a conceptual and methodological framework applicable to other seasonally frozen rivers for comparative and cross-basin studies.

Funding

• National Natural Science Foundation of China (Grant No. 52379014)
• Joint Fund of the National Natural Science Foundation of China (Grant No. U23A2012)
• Inner Mongolia Natural Science Foundation (Grant No. 2023QN05026)

Citation

@article{Liu2026Monitoring,
  author = {Liu, Bin and Ji, Honglan and Xu, Haifeng and Deng, Yu and Luo, Hongchun and Xue, Zhongshu and Ren, Wenhao},
  title = {Monitoring the river ice phenology along the Inner Mongolia reach of the Yellow River using time-series images from landsat and Sentinel-2},
  journal = {Journal of Hydrology Regional Studies},
  year = {2026},
  doi = {10.1016/j.ejrh.2026.103140},
  url = {https://doi.org/10.1016/j.ejrh.2026.103140}
}