InSAR time-series deformation forecasting surrounding Salt Lake using deep transformer models.

Area of Science:

• Geophysics
• Remote Sensing
• Climate Science

Background:

• Sentinel-1 SAR imagery and InSAR enable large-area surface deformation monitoring.
• Permafrost freeze-thaw cycles on the Qinghai-Tibet Plateau are crucial for understanding regional stability.
• Existing deep learning models struggle with complex, long-term dependencies in multivariate time series data for high-resolution deformation prediction.

Purpose of the Study:

• To develop an innovative InSAR deformation prediction algorithm using transformer models for enhanced accuracy in permafrost areas.
• To accurately model complex deformation features and seasonal/non-seasonal signals in InSAR time series data.
• To evaluate the expansion trends of Salt Lake and discuss the impact on permafrost degradation.

Main Methods:

• Utilized Sentinel-1 SAR data and Interferometric Radar (InSAR) for time series surface deformation monitoring.
• Developed a novel InSAR deformation prediction algorithm integrated with transformer models, leveraging self-attention mechanisms.
• Applied the model to predict surface deformation surrounding Salt Lake on the Qinghai-Tibet Plateau.

Main Results:

• The transformer-based InSAR deformation prediction method demonstrated superior performance in capturing complex deformation patterns compared to other models.
• Accurate simulation of seasonal and non-seasonal deformation signals was achieved, enabling effective short-term prediction.
• Analysis revealed Salt Lake's area increased by 57.32 km² between 2015-2019, with a slowing expansion trend from 2019-2022.

Conclusions:

• The proposed transformer-based InSAR deformation prediction framework offers a generic solution for modeling nonlinear deformation processes in permafrost regions.
• The findings highlight the potential impact of Salt Lake outburst events on permafrost deformation and degradation.
• Accurate deformation monitoring and prediction are essential for managing risks associated with climate change and geological events in sensitive environments.