Two-Pass versus Three-Pass Differential Interferometric Processing Using 1:250,000 Terrain Data

Literature Overview

This 2003 paper by Liu Jiahang, Shan Xinjian, Qin Xuwen, and Zhang Guifang, published in Remote Sensing Information, presents a comparative analysis of two-pass and three-pass differential interferometric SAR (DInSAR) processing using 1:250,000 scale terrain data as an external digital elevation model (DEM). The study uses the M7.6 Mani earthquake in Tibet (November 8, 1997) as a case study, employing two pre-event and one post-event ERS SAR images. The key finding is that while both processing modes produce consistent co-seismic deformation field patterns overall, discrepancies increase with distance from the fault due to residual topographic effects.

DInSAR Processing Modes

Differential Interferometric SAR (DInSAR) is a technique for measuring surface deformation by removing the topographic phase contribution from interferometric phase measurements. Two primary processing modes are used:

Two-Pass Differential Interferometry

In the two-pass mode, a single interferogram is formed from two SAR acquisitions (typically pre- and post-event), and the topographic phase is removed using an external DEM. The process involves:

1. Forming the interferogram from the two SAR images.
2. Computing the expected topographic phase from the external DEM using the known radar geometry.
3. Subtracting the topographic phase from the interferogram to isolate the deformation phase.

Three-Pass Differential Interferometry

In the three-pass mode, two interferograms are formed: one from two pre-event images (the "reference" interferogram, which contains only topographic and atmospheric phase) and one from one pre-event and one post-event image (the "deformation" interferogram). The reference interferogram is subtracted from the deformation interferogram to remove the topographic contribution.

Case Study: Mani Earthquake

The Mani earthquake (M7.6, November 8, 1997) in Tibet provides an excellent case study for DInSAR analysis due to the availability of multiple ERS SAR acquisitions and the significant co-seismic deformation. The study area covers a region of complex topography, which makes the comparison of two-pass and three-pass processing particularly informative.

Data Characteristics

Comparative Analysis Results

The authors' comparative analysis reveals several important findings:

1. Overall consistency: Both processing modes produce co-seismic deformation field patterns that are consistent in their overall distribution, with good agreement near the fault zone.
2. Fault-adjacent accuracy: Near the fault, where the deformation signal is strong, both methods yield similar results, with the deformation patterns matching well.
3. Distance-dependent divergence: As the distance from the fault increases, the residual topographic phase becomes the dominant contributor to the interferometric fringes, and the differences between the two processing modes increase.
4. DEM adequacy: The 1:250,000 scale terrain data is adequate for two-pass processing when the ERS SAR resolution is considered and no special requirements are imposed.

Quantitative Comparison

Implications for DEM Selection

A key conclusion of this paper is that the 1:250,000 scale Chinese terrain data can serve as an adequate external DEM for two-pass differential interferometric processing under the following conditions:

• The SAR system resolution is comparable to or coarser than the DEM resolution.
• The primary interest is in the near-fault deformation field.
• No special requirements for far-field accuracy are imposed.

This finding has practical significance for regions where high-resolution DEMs are not readily available. It demonstrates that topographic maps at the 1:250,000 scale, which are widely available in China, can be digitized and used as external DEMs for DInSAR processing, reducing the dependency on expensive satellite-derived DEM products.

Study Insights and Engineering Practice

From an engineering perspective, this paper provides valuable guidance on the selection of DInSAR processing modes and DEM sources for deformation monitoring applications. The findings suggest a practical decision framework:

1. When three pre-event images are available: Three-pass processing is preferred, as it does not depend on external DEM quality and provides more accurate results over the entire study area.
2. When only two images are available: Two-pass processing with an appropriate external DEM is viable, provided the DEM resolution is compatible with the SAR system resolution and the analysis focuses on the near-fault region.
3. When DEM quality is uncertain: Three-pass processing is the more robust choice, as it is inherently immune to DEM errors.

The paper also highlights the importance of understanding the limitations of each processing mode. The two-pass method's dependence on DEM quality means that systematic errors in the DEM will manifest as systematic errors in the deformation field, particularly in areas of steep terrain or where the DEM resolution is insufficient to capture the true topography.

In conclusion, this paper provides a practical and well-supported comparison of two-pass and three-pass DInSAR processing, demonstrating that the choice between these modes depends on data availability, DEM quality, and the specific requirements of the deformation analysis, with the 1:250,000 scale terrain data proving adequate for many practical applications.