Automatic Mapping of Thermokarst Landforms from Remote Sensing Images Using Deep Learning: A Case Study in the Northeastern Tibetan Plateau

• Published in: Remote sensing (Basel, Switzerland) Vol. 10; no. 12; p. 2067
• Main Authors: Huang, Lingcao, Liu, Lin, Jiang, Liming, Zhang, Tingjun
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.12.2018
• Subjects: Aerial photography; Algorithms; Artificial neural networks; Carbon; Case depth; Case studies; Deep learning; DeepLab; Digitization; Digitizing; Geodetics; Global positioning systems; GPS; Gullies; Gully erosion; High resolution; Image resolution; Laboratories; Landforms; Machine learning; Mapping; Mountains; Neural networks; Permafrost; permafrost degradation; Post-production processing; Remote sensing; semantic segmentation; Sinkholes; Soil degradation; Studies; Thawing; thermo-erosion gullies; thermokarst landforms; Tibetan Plateau; Training; Unmanned Aerial Vehicle Images; Unmanned aerial vehicles; Vegetation; Heihe River; Beijing China; Tibetan Plateau; Switzerland; China; Arctic region
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs10122067

Thawing of ice-rich permafrost causes thermokarst landforms on the ground surface. Obtaining the distribution of thermokarst landforms is a prerequisite for understanding permafrost degradation and carbon exchange at local and regional scales. However, because of their diverse types and characteristics, it is challenging to map thermokarst landforms from remote sensing images. We conducted a case study towards automatically mapping a type of thermokarst landforms (i.e., thermo-erosion gullies) in a local area in the northeastern Tibetan Plateau from high-resolution images by the use of deep learning. In particular, we applied the DeepLab algorithm (based on Convolutional Neural Networks) to a 0.15-m-resolution Digital Orthophoto Map (created using aerial photographs taken by an Unmanned Aerial Vehicle). Here, we document the detailed processing flow with key steps including preparing training data, fine-tuning, inference, and post-processing. Validating against the field measurements and manual digitizing results, we obtained an F1 score of 0.74 (precision is 0.59 and recall is 1.0), showing that the proposed method can effectively map small and irregular thermokarst landforms. It is potentially viable to apply the designed method to mapping diverse thermokarst landforms in a larger area where high-resolution images and training data are available.