Investigating MAI's Precision: Single Interferogram and Time Series Filtering

Multiple aperture InSAR (MAI) is a technique to obtain along-track displacements from InSAR phase data. Because InSAR measurements are insensitive to along-track displacements, it is only possible to retrieve them using none-interferometric approaches, either pixel-offset tracking or using data from different orbital configurations and assuming continuity/ displacement model. These approaches are limited by precision and data acquisition conflicts, respectively. MAI is promising in this respect as its precision is better than the former and its data is available whether additional acquisitions are there or not. Here we study the MAI noise and develop a filter to reduce it. We test the filtering with empirical noise and simulated signal data. Below we describe the filtered results single interferogram precision, and a Kalman filter approach for MAI time series. We use 14 interferograms taken over the larger Los Angeles/San Gabrial Mountains area in CA. The interferograms include a variety of decorrelation sources, both terrain-related (topographic variations, vegetation and agriculture), and imaging-related (spatial and temporal baselines of 200-500m and 1-12 months, respectively). Most of the pixels are in the low to average coherence range (below 0.7). The data were collected by ESA and made available by the WInSAR consortium. We assume the data contain “zero” along-track signal (less then the theoretical 4 cm for our coherence range), and use the images as 14 dependent realizations of the MAI noise. We find a wide distribution of phase values σ = 2-3 radians (wrapped). We superimpose a signal on our MAI noise interferograms using along-track displacement (-88 - 143 cm) calculated for the 1812 Wrightwood earthquake. To analyze single MAI interferograms, we design an iterative quantile-based filter and test it on the noise+signal MAI interferograms. The residuals reveal the following MAI noise characteristics: (1) a constant noise term, up to 90 cm (2) a displacement gradient term, up to 0.75cm/km (3) a coherence dependent root residuals sum of squares (RRSS), down to 5 cm at 0.8 coherence In the figure we present two measures of the MAI rmse. Prior to phase gradient correction the RRSS follows the circled line. With the correction, the RRSS follows the solid line. We next evaluate MAI's precision given a time series. We use a Kalman Filter to estimate the spatially and temporally correlated components of the MAI data. We reference the displacements to a given area in the interferograms, weight the data with coherence, and model the reminder terms of the spatially correlated noise as a quadratic phase gradient across the image. The results (not displayed) again vary with coherence. MAI single interferogram precision