Environmental, socioeconomic, and health risks of thawing permafrost in Tibet

Keywords: Permafrost, Climate Change, Cold Regions
Abstract Type: Paper Abstract

Abstract

Permafrost underlies ~40% of the land area on the Tibetan Plateau and has been undergoing strong warming and fast degradation due to climate change and anthropogenic disturbance. One significant indicator of thawing ice-rich permafrost is the development of thermokarst landforms. These landforms and associated dynamic processes play important roles in permafrost-carbon feedback and have strong impacts on the local and regional ecosystem and infrastructure in cold regions.

This study applies novel geospatial approaches based on deep learning and statistical model to an interdisciplinary and quantitative assessment of thermokarst landforms on the Tibetan Plateau, focusing on active thaw slumps. We applied two deep-learning methods, i.e. a real-time object detection algorithm called YOLOv4 and a semantic segmentation algorithm called DeepLabv3+, to time-lapse high-resolution CubeSat images and obtained a comprehensive inventory containing 2613 thaw slumps across the vast, 1 million square kilometers of permafrost areas on the Plateau. Based on the field monitoring and experimental analysis, we found that thaw slumps resulted in losses of approximately 29.6% and 28.9% for soil organic carbon and total nitrogen, and decreased in total mercury mass of 17.6–30.9% for the top 30 cm layer. We will then integrate statistical learning based on Random Forest with permafrost physics models to predict thermokarst evolution under various shared socioeconomic pathways. The next step will be to assess the risks/impacts posed by accelerated permafrost degradation to human health due to worsening water quality and increased mercury level, to economic loss by destabilized infrastructure, and to reaching carbon neutrality goals.