The data is based on multi-source remote sensing images and field investigations of engineering diseases. Machine learning algorithms are used to determine the weights of factors such as terrain, weather, and frozen soil types, achieving systematic identification and mapping of regional scale freeze-thaw disasters. The dataset focuses on reflecting the disaster aggregation characteristics within linear engineering corridors such as highways, railways, and crude oil pipelines (CRCOP), revealing the evolution of frost damage from "patchy distribution" to "point/corridor aggregation" during the transition from large discontinuous permafrost areas in the north to scattered island shaped permafrost areas in the south. This data can provide core data support for the safety evaluation of infrastructure in cold regions and the study of environmental effects of permafrost degradation.
The classification system of this dataset (grid values 1-6) strictly corresponds to typical engineering diseases in permafrost regions:
1: Hot melt lake pond
2: Road cracks
3: Freeze-thaw erosion
4: Uneven settlement of road surface
5: Ice vertebrae
6: Water destroyed.
By integrating multi-source remote sensing images (Landsat/Sentinel/high-resolution series) and on-site disease investigation points, frost damage classification training and validation are carried out through factor weighting and machine learning algorithms. Finally, feature extraction is performed to obtain predicted frost damage types, and GIS spatial analysis technology is used for regional mapping.
Sample collection: Collect field disease samples (GPS positioning+on-site photography) in typical sections such as Mohe, Genhe, Xinlin, and Jiagedaqi.
Feature modeling: Introducing slope, vegetation, ground temperature, and engineering thermal disturbance factors.
Identification and extraction: In large areas of discontinuous permafrost, emphasis is placed on identifying subsidence and thermal melting lakes and ponds; Emphasis is placed on identifying cracks and erosion in sporadic island shaped frozen soil areas.
Refined measurement: Use 1:250000 topographic maps and high-precision Google imagery to logically correct the distribution of frost damage along linear engineering lines.
(1) Data collection and source data: The fusion of high-resolution multi-source remote sensing image data and large-scale linear engineering (highways, railways, China Russia crude oil pipelines) field disease investigation samples ensures the authenticity of disease identification.
(2) Model method: Machine learning algorithms (such as weight determination models) are used to comprehensively calculate various freezing damage factors such as altitude, slope, temperature, and snow cover, effectively reducing the subjectivity of traditional manual identification.
(3) Spatial accuracy: The data is mapped at a scale of 1:250000 to meet the disaster assessment needs of regional scales and large and medium-sized linear engineering corridors; The logical consistency between disaster points, zones, terrain, and permafrost boundaries was ensured through GIS spatial analysis technology.
(4) Validation and verification: Typical validation was carried out on typical areas such as the Mohe Hub, Genhe Itui River section, and Heihe River edge section. The identification results were highly consistent with the field investigation of diseases, and could objectively reflect the engineering response status under the background of warm permafrost degradation.
(5) Scope of application: Suitable for engineering geological research, infrastructure risk assessment, and geological mapping in cold regions.
| # | number | name | type |
| 1 | 2022FY100700 | Survey of Permafrost Conditions and Freeze-Thaw Damage in the High-Latitude Regions of Northeast China | Basic Resource Survey Project |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.NIEER.DB7262.2026
10.12072/ncdc.nieer.db7262.2026
