Abstract: Synthetic Aperture Radar coherent change detection(SAR CCD) is widely used in environmental monitoring, urban planning and disaster assessment to identify subtle surface variations via low-coherence region analysis of multi-temporal SAR images. Nevertheless, low signal-to-noise ratio areas (e.g., shadows and water bodies) and time-varying surface features induced by vegetation fluctuations will produce massive low-coherence clutter, greatly limiting the detection performance of faint artificial targets such as tire tracks and footprints. To solve this problem, this paper proposes a temporal coherence modeling-based SAR coherent change detection method. Three temporal hidden states are constructed to describe the evolution characteristics of ground objects, and a probabilistic membership soft decision mechanism is introduced to fit complex temporal variations including gradual changes and intermittent disturbances. The temporal adaptive transition matrix and SAR coherence statistics are employed to calibrate model parameters and achieve unsupervised modeling. The Viterbi algorithm is further adopted to decode the optimal temporal state sequence for refined change discrimination. Simulated and airborne experimental results show that the proposed method can effectively suppress natural time-varying clutter interference. It significantly improves the detection accuracy and robustness of SAR coherent change detection in complex field scenarios while preserving high sensitivity to subtle artificial changes.