How do Data Processing Choices Affect Multiple Aperture InSAR (MAI) Precision?

Abstract:

MAI data processing methods report different precisions, sometimes for the same data; we investigate these discrepancies here. MAI, a method to obtain phase-based along-track displacements from InSAR data, has been used for measuring volcanic, glacier, and earthquake displacements. However, different applications show vastly different precisions, ranging from a few millimeters to tens of centimeters per interferogram. While some spatial variability in precision is expected due to the InSAR phase noise dependence on terrain, some reports show different precisions for the same data. Here, I address the following questions: What is the true precision of MAI measurements? What are the MAI noise sources, and how do we mitigate them?

 We use data from 11 Envisat, 2 ERS, and 1 ALOS interferograms to analyze and quantify the MAI noise sources and their dependence on the method’s implementation. To do that, we implement various processing steps with the JPL/Caltech InSAR data processor ROI_PAC. We find that there are several (sometimes surprising) nuances in data processing that together can reduce scatter noise by half, and other corrections that can reduce topography-related noise by up to 25 centimeters. We find that even with all the optimized data processing steps, the actual MAI errors (relative accuracy) are higher then the theoretical ones by a factor that depends on filter window size. We present an updated MAI data processor, which is capable of achieving relative accuracy and precision levels of 6 to 2 centimeters, respectively, in low to medium coherence terrains (correlation coefficient of 0.4 to 0.8), and provides error estimates that better reflect the actual relative accuracy of the measurement.