Yellow River ice monitoring based on hyperspectral technology

Ice floes are ice blocks that move along with the river water before the river freezes over. Since moving ice is called "ice floes," ice moving on the water surface or in a body of water is called ice floe. Ice floes can be divided into winter ice floes and spring thawing ice floes. The former is mainly composed of ice flowers and ice clumps, while the latter is mainly composed of broken ice. This refers to ice floes formed before the river freezes over.

△ Figure 1. Monitoring ice flow through ice holes

△ Figure 2 Ice jam disaster

Why monitor icicles?

1. Flood control and disaster reduction:Ice flooding or blockage of river channels can cause backflow, leading to ice jams and threatening downstream areas. Monitoring ice floes allows for timely warnings and measures to reduce the occurrence and impact of floods.

2. Protect water conservancy facilities:Ice flow accumulation can damage water conservancy facilities such as dams, bridges, and dikes, affecting the normal operation and safety of water conservancy projects. Monitoring ice flow can help identify potential hazards in a timely manner and take appropriate maintenance and reinforcement measures.

3. Water resource management:The formation and evolution of ice floes reflect changes in water temperature and hydrological characteristics, providing valuable insights for water resource management and hydrological forecasting. Monitoring ice floes allows for a better understanding of river ice conditions, offering data support for water resource management.

4. Scientific Research:Monitoring and studying ice floes helps to gain a deeper understanding of the formation mechanism, evolution, and relationship with climate change of ice floes in rivers or streams, which is of great significance for climate change research and the development of hydrology.

△ Figure 3 Location map of the Yellow River ice monitoring study area

Introduction to the research area

The study area is located in the Inner Mongolia section of the Yellow River, at the northernmost point of the river, at an elevation of over 1,000 meters. Far from the ocean, warm and temperate air currents rarely reach it, and it is frequently controlled by the Mongolian High. It exhibits significant continental climate characteristics, with long, harsh winters where temperatures remain below freezing for 4-5 months, reaching as low as -35°C. The river freezes for 4-5 months in this section, with most of it being a stable, frozen section. In winter, the river flow in the Inner Mongolia section generally exceeds 400 cubic meters per second. The low temperatures and thick ice layers in this section, combined with the impact of flood peaks causing ice floes to break and create large ice floes, along with the river's morphology, lead to dense ice formations that can trap and dam the ice, causing a sharp rise in river levels and making the area highly susceptible to breaches, flooding, and other disasters.

The study area is a 103-kilometer-long river section located in the southern part of Baotou City, running east-west. The river channel area is approximately 70 square kilometers.

△ Figure 4 Satellite image of the Yellow River ice monitoring area

△ Figure 5 Types of river features

△ Figure 6 Reflectance map of river channel land cover types

Xi'an Institute of Optics and Precision Mechanics CAS Aerospace Science and Technology Group Co.,LTD - Flowing water Monitoring Research Methods

1. Data Acquisition: 150-band hyperspectral images of the Baotou area in Inner Mongolia were used, taken by the XIGUANG003 hyperspectral satellite on January 5, 2024.
2. Study area determination: The original raster data was masked using river boundary vectors to extract hyperspectral images within the study area.
3. Determine the number of categories: Before[Hyperspectral Applications Column] Snow and Ice Cover Monitoring in Dalad Banner, Inner Mongolia Based on Hyperspectral Mixture Spectral DecompositionBased on the classification of ice and snow, three types of land features in the river channel were identified: flowing water surface (clear ditch*), ice and snow mixture, and snow-covered ice surface. As shown in Figure 5, in the 450-850nm band, the reflectivity of flowing water surface is the lowest, averaging less than 10%, the reflectivity of ice and water mixture is between 10% and 30%, while the reflectivity of snow-covered ice surface is as high as 65%.
4. Classification: Pixel-based unsupervised classification of land features within the river channel.
5. Statistical analysis: Statistics on the area and area percentage of each feature in different regions.

*In frozen river sections, there are exposed water surfaces of varying lengths, called clear channels. Clear channels that form during the freezing period are called primary clear channels; clear channels that form during the thawing period are called secondary clear channels.

△ Figure 8 Distribution map of ice floes in the Baotou section of the Yellow River

Research Conclusions and Applications

By extracting and analyzing ice floes under different mixing conditions in the Yellow River channel of the study area, an ice floe distribution map was created, which intuitively shows the distribution location of ice floes in different states. At the same time, the distribution area and area ratio were statistically analyzed, which can help determine the freezing status of the river channel at the current time, identify risk areas, and provide a usable means for large-scale ice floe monitoring.

Wenkaihe: The ice melts and the river opens up mainly due to thermal action. The situation is relatively stable, but it takes a long time.

Wukai River: The ice-breaking and opening of the river is mainly formed by water power. The upstream river channel thaws first, while the ice in the downstream section of the river channel remains solid. When the amount of water flowing down from the upstream suddenly increases and encounters the ice cover in the downstream, the phenomenon of ice and water flowing down the river channel at the same time and water surging and ice breaking is formed under the action of water. 

Meanwhile, ice floe monitoring based on large-scale satellite remote sensing provides accurate data on ice floe distribution within a specific time period, providing a data foundation for the establishment of prediction models. By combining meteorological data, monitoring data from fixed-point equipment in the river channel, and remote sensing data from UAVs, comprehensive information on river freezing and ice floe can be obtained, reducing labor costs while achieving continuous, cross-basin monitoring and risk prediction.

Furthermore, the combined features of hyperspectral imagery and atlas support further quantitative inversion and sub-pixel-level classification, helping researchers and water conservancy practitioners to more accurately determine the type and state of ice floes. This allows for a deeper understanding of river freezing conditions (flat or vertical) and thawing processes (gradual or rapid thawing), such as monitoring the ratio of ice to water in the river channel, ice crystal morphology, ice cap type (static or dynamic), and the increase or decrease of river channel clearing. For example, changes in clearing can be used to predict the type of thawing (increased clearing is a sign of gradual thawing), enabling proactive measures to reduce losses caused by ice jams.