Shi (2026) Spatiotemporal Dynamics and Drivers of Potential Winter Ice Resources in China (1990–2020) Using Multi-Source Remote Sensing and Machine Learning

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-01-13
• Authors: Dunfa Shi
• DOI: 10.3390/rs18020250

Research Groups

• Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

Short Summary

This study developed a multi-source remote sensing and machine learning framework to map and analyze the spatiotemporal dynamics of potential winter river and lake ice resources in China from 1990 to 2020. It found that despite a significant northwestward shift in the freezing-zone boundary, the total ice-covered area increased by approximately 1.1% per year, while the average ice season slightly shortened, highlighting asynchronous responses to climate change driven by hydrological-thermal conditions and urbanization.

Objective

• To develop a simple and transferable "surface water + land surface temperature (LST)" framework on Google Earth Engine to map potential winter ice area across China from 1990 to 2020.
• To characterize the spatiotemporal distribution and evolution of potential winter ice resources and identify the main natural and anthropogenic drivers influencing these patterns under climate change.

Study Configuration

• Spatial Scale: National scale (China), provincial level, covering large-scale river and lake systems.
• Temporal Scale: 30-year period (1990–2020), with daily and monthly data aggregated annually.

Methodology and Data

• Models used:
  • LSTM-Autoencoder for time-series feature learning and pattern extraction.
  • K-means clustering for grouping provinces based on ice resource trends.
  • Linear regression and Theil–Sen slope estimation for trend quantification.
  • Random Forest regression for analyzing influencing factors and variable importance.
• Data sources:
  • MOD11A1.061 Terra Land Surface Temperature and Emissivity Daily Global 1 km dataset (NASA LP DAAC) for LST.
  • Climate Hazards Center InfraRed Temperature with Stations (CHIRTS) daily temperature data product (University of California) for air temperature.
  • Global Surface Water Explorer (JRC Monthly Water History) dataset (Google) for permanent surface water bodies (30 m spatial resolution).
  • Various socioeconomic, topographic, eco-environmental, and surface characteristic indicators (e.g., population, GDP, nighttime lights, built-up area, elevation, slope, NDVI, NPP, forest/grassland cover).
  • All data processed and analyzed on the Google Earth Engine (GEE) platform.

Main Results

• The freezing-zone boundary in China exhibited a pronounced northwestward shift from 1990 to 2020, with a maximum displacement exceeding 400 km in the Northeast Plain and an average rate of approximately 13 km per year along the eastern coast.
• Despite the warming-induced boundary shift, the total potential winter ice-covered area nationwide increased from 50,016.77 km² in 1990 to 68,955.81 km² in 2020, representing an average annual growth rate of approximately 1.1%.
• Concurrently, the national average ice-covered period showed a slight decreasing trend of about -0.005 month per year, indicating asynchronous spatial expansion and temporal compression of ice resources.
• A persistent "Northwest–Northeast dual-core" spatial pattern of ice resources was identified, characterized by strong positive spatial autocorrelation (Global Moran’s I values of 0.5647 in 1990 and 0.5677 in 2020).
• Provincial-level analysis revealed that most ice-rich regions (e.g., Tibet, Qinghai, Xinjiang, Inner Mongolia, Heilongjiang, Jilin, Liaoning) showed an increasing trend in ice area, while decreases were localized to Beijing and Yunnan, attributed to intense urbanization and low-latitude warming.
• Random Forest modeling identified water area fraction as the leading predictor, followed by human-activity proxies (nighttime lights, built-up area) and topographic factors (elevation, topographic exposure), as well as land surface temperature, as the main controls on potential winter ice area.

Contributions

• Developed a novel, simple, and transferable "surface water + LST" framework on Google Earth Engine for consistent, large-scale, and long-term monitoring of freshwater ice resources across diverse climate zones in China, overcoming limitations of complex SAR models and region-specific thresholds.
• Revealed the asynchronous spatiotemporal responses of China's river and lake ice resources to climate warming, showing an overall increase in ice area despite a shortening of the ice season and a northwestward shift of the freezing-zone boundary.
• Identified and characterized a stable "Northwest–Northeast dual-core" spatial structure of ice resources and classified provincial-level trends into increasing and decreasing types using an integrated deep learning and statistical approach.
• Quantified the relative importance of various natural and anthropogenic factors, highlighting the combined influence of hydrological-thermal conditions and urbanization in shaping winter ice patterns.

Funding

• General Program of the National Natural Science Foundation of China (Grant No. 42301349).

Citation

@article{Shi2026Spatiotemporal,
  author = {Shi, Dunfa},
  title = {Spatiotemporal Dynamics and Drivers of Potential Winter Ice Resources in China (1990–2020) Using Multi-Source Remote Sensing and Machine Learning},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18020250},
  url = {https://doi.org/10.3390/rs18020250}
}