Application of ARIMA and Annual Variations Models in InSAR for Permafrost Studies

Temperature fluctuations are expected to drive nonlinear patterns in annual permafrost variations. The InSAR technique effectively captures these changes by tracking surface deformation over time. However, atmospheric phase noise often requires spatiotemporal filtering, which reduces temporal resolution for nonlinear signals. Additionally, the influence of temperature on annual freeze-thaw cycles remains underrepresented in InSAR models. This study presents a method to improve InSAR time-series analysis in permafrost regions. Rather than relying on traditional filtering that limits temporal resolution based on convolution kernel size, we utilize the autoregressive integrated moving average (ARIMA) model to isolate nonlinear deformation signals. Furthermore, we extract annual variation parameters from the InSAR time-series deformation data. Synthetic experiments under varying noise delays reveal accuracy gains between 27.8% and 55.3% in nonlinear deformation analysis. Applying Sentinel-1 datasets from 2017 to 2021 and ground data from northern Alaska, we achieve a time-series deformation accuracy improvement of over 22%.