Optimizing DEM error mitigation in multi-temporal InSAR: A detection-and-estimation strategy based on phase gradient direction consistency

Abstract

Accurate topographic phase removal in Differential InSAR (DInSAR) processing relies on Digital Elevation Models (DEMs), yet limitations in DEM accuracy and currency hinder precise surface displacement measurement. Although modern SAR satellites feature a relatively narrow orbit tube, the phases induced by DEM errors cannot be safely ignored especially in areas under rapid urbanization. Current Multi-Temporal InSAR (MT-InSAR) methods, which estimate DEM errors alongside deformation, suffer from potential biases due to inaccurate deformation models and high computational cost from per-point processing. We present here a novel detection-and-estimation strategy for efficient DEM error mitigation. Our key innovation is a phase gradient direction consistency (GDC) criterion, which provides a direct and intuitive visualization of pixels affected by DEM errors (PEEs)—a capability not previously available. This is a significant advancement as it allows targeted correction instead of exhaustive estimation. We further develop a generalizable framework for DEM error retrieval applicable to various scenarios. Validation with simulated and real-world data from urban and mountainous environments demonstrates effective separation of DEM errors from various spatiotemporal deformation signals. In addition, the proposed method achieves an order-of-magnitude improvement in processing efficiency compared to conventional approaches. By directly identifying and estimating DEM errors from wrapped phases, our approach streamlines deformation retrieval and is readily integrated into existing MT-InSAR workflows.