A New InSAR Temporal Decorrelation Model for Seasonal Vegetation Change With Dense Time-Series Data

This study proposes an extended temporal correlation model for targets with a noticeable periodic seasonal trend. Several studies have explored the nature of decorrelation in synthetic aperture radar (SAR) interferograms. Specifically, providing a model the decay in interferometric correlation over time between two images remains a challenging task. Initially, it is necessary to assume that the contributions of distributed elements within the same pixel undergo a change, leading to a reduction in correlation with previous acquisitions. The exponential decay model is the simplest and most widely used in the scientific community by considering coherent and incoherent groups of scatterers within a resolution cell. However, the coherence over vegetation canopies with seasonal behavior does not exhibit a monotonic exponential decay with time. Hence, in this study, we introduce a periodic term to account for the nature of this seasonality. The performance of the proposed model is evaluated with a total of nearly 2000 Sentinel-1 interferometric SAR (InSAR) pairs acquired over two test sites located one in Nallamala, India, and the other in Injune, Australia. The proposed model performed significantly better than the exponential model with up to 83% improvement in RMSE in modeling the long-term coherence over vegetation with strong seasonal patterns.