InSAR is a key technique used to constrain contributions of diverse processes to coastal subsidence, also known as vertical land motion (VLM). However, coastal environments can pose major challenges for InSAR due to natural disturbances that degrade interferogram quality. We describe a new multi-primary pairing strategy for persistent scatterer InSAR (PS-InSAR) to estimate subsidence in challenging coastal environments. Our method retains only consistent PS candidates across multi-primary substacks and solves for redundant velocity observations using SVD-based inversion, similar to the conventional small baseline subset (SBAS) method. Through simulations and a case study comparing with single-primary PS-InSAR and conventional SBAS techniques, we show that our pairing strategy reduces temporal and spatial uncertainty in subsidence estimates in the presence of strong but temporary decorrelation loss, even with increased distance from the reference point. Moreover, our method visibly dampens time-series variation and decreases standard error in our time series fit by nearly 2x in our case study. Thus, we find that implementing a multi-primary PS-InSAR configuration is a simple method of increasing the robustness of VLM estimates in challenging coastal environments.

SAR imagery is traditionally produced in range-Doppler geometry. While such algorithms are computationally efficient, they require multiple assumptions regarding signal properties. Time-domain backprojection (TDBP) SAR focusing methods require fewer assumptions and are considerably more robust to perturbations in flight paths and target geometry. Furthermore, TDBP produces single-look complex (SLC) imagery that is directly geocoded, which simplifies subsequent analysis compared to range-Doppler products. With modern improvements in computational power, TDBP SAR-focusing methods have become viable alternatives to range-Doppler. However, TDBP-formed imagery can be sensitive to inaccuracies in global DEMs. Here, we investigate the so-called multi-depth multi-focusing algorithm that has been proposed to correct for DEM errors. We analyze changes in the detection of persistent scatterers (PS) and find that while depth-averaged SLCs yield fewer PS, depth-optimized SLCs yield more PS and produce higher-quality interferograms compared with unoptimized SLCs. We conclude by discussing the implications for improved time-series InSAR analysis.

Multi-temporal InSAR (MT-InSAR) techniques are powerful time-series analysis methods that enable all-weather measurements of surface deformation. However, even advanced MT-InSAR techniques can produce inaccurate results in challenging environments, including many coastal regions. Here, we describe successes and challenges of applying MT-InSAR techniques for imaging post-seismic vertical land motion (VLM) on Tutuila Island in American Samoa. We tested traditional persistent scatterer (PS) and small baseline subset (SBAS) methods and also designed a customized redundant PS method to address observed inconsistencies in both PS and SBAS analysis. The redundant PS method yielded results with lower noise and more realistic spatial variation compared to PS and SBAS workflows. Our case study emphasizes the importance of careful application of MT-InSAR for measuring VLM along coastal regions, where subsidence and uplift rates may be moderate in the presence of numerous noise sources. In particularly challenging areas, customized approaches may be needed to produce useful solutions.

Characterizing diverse contributions to coastal land change is a key step to mitigating the effects of rising sea levels that threaten coastal communities. This task is particularly critical for small island communities in tectonically active regions, which are highly vulnerable to the effects of sea level rise. We highlight here a case study to extract regional estimates of vertical land motion (VLM) over American Samoa, which in recent years has observed increased nuisance flooding. We used persistent scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) to derive high-resolution estimates of crustal deformation in populated regions over Tutuila Island from 2016 to 2021. While the area is small and highly vegetated and poses challenges for InSAR, we were able to construct a regional map of the estimated deformation rate in these areas and validate the time-series with a local GPS station. Our preliminary results suggest that PS-InSAR has the ability to capture local and regional deformation patterns. Further work to refine our VLM estimate will include models to account for more complex atmospheric effects and estimates of error margins, as well as integration of our results into models of sealevel change that can be used by local stakeholders.

It is well-known that the backscatter of high-resolution Synthetic Aperture Radar (SAR) imagery is non-Gaussian in nature. As a result, corresponding heavy-tailed models have been successfully incorporated for the design of improved SAR target detectors. However, Gaussian-based detectors are largely still applied for selection of persistent scatterers (PS) in Interferometric Synthetic Aperture Radar (InSAR) imagery, and implications for the performance of PS techniques have not been well-studied. Here, we extend an existing Gaussian model for PS to incorporate non-Gaussian behavior. We then implement the model for PS detection and compare its performance to its Gaussian counterpart, finding that the non-Gaussian model finds a slightly denser network of PS. Further work will focus on analyzing the characteristics of this disparity, including its relationship with terrain and system parameters such as wavelength and bandwidth, and compare the estimated deformation from the non-Gaussian detector compared to an existing Gaussian-based model. Understanding the limitations of Gaussian models will inform the design of improved PS detectors to produce more complete deformation maps and enable the broader application of InSAR for challenging applications, such as observing small strain rates in natural terrain.

Radar imaging using the cross-correlation of receiver signals possesses several benefits over traditional matched-filter techniques that make it more suitable for imaging fast-moving objects without well-defined flight tracks, such as fast-moving space debris. Correlation-based imaging is implemented with a network of receivers, and requires no knowledge of the imaging pulse or the emitter and receiver locations and is able to compensate for unknown linear and rotational motion with a resolution on the order of the imaging wavelength, comparable to the performance of matched-filter imaging. Here, we compare the performance of a simplified bistatic configuration for correlation-based imaging and the traditional matched-filter method in a representative case study experiment, where a small aircraft is used as a target. We find that the bistatic case is effective even when there are motions that exceed the transmitter wavelength and is more resilient to unaccounted random motion than the monostatic case. However, as expected for a two-receiver configuration, the resolution is lower than the matched-filtering method even in the ideal case. Finally, we comment on extensions as well as implications for multistatic configurations.

With the increasing availability of spaceborne Synthetic Aperture Radar (SAR) imagery, Interferometric Synthetic Aperture Radar (InSAR) has become a prominent technique to study geophysical phenomena. However, its application can be limited in natural terrain due to rapid changes in the surface known as decorrelation. Persistent scatterers (PS) are stable elements that can be exploited in InSAR imagery to allow fine deformation mapping of the Earth's surface even in areas that are highly decorrelated. Despite the widespread use of PS techniques, little research has been dedicated to studying the relationship in between key system parameters and observed PS statistics. Such knowledge is critical in creating effective detection algorithms and informing design choices for satellite missions. Here, we present an analytical expression for the probability density function (PDF) of PS density with respect to system wavelength and bandwidth. This equation could be directly implemented in traditional detection algorithms for PS detection. We offer comparisons with real data, describing how the model could be improved. Further work will extend this expression to the probability of PS incidence and examine performance over different types of terrain.

In InSAR time series analysis, persistent scatterers (PS) are pixels that remain temporally correlated even in areas of otherwise high decorrelation. Using these stable pixels as reference points, we can analyze deformation signals on the Earth's surface down to the millimeter scale. PS possess different physical and statistical properties compared to the background, and analyzing the statistical behavior of PS can help to elucidate radar backscattering behavior from different terrain. Here, we propose a new statistical characterization of SAR image power using a mixture distribution of PS and non-PS points, i.e. clutter. We find that a lognormal distribution fits both pixel types, validated by a scattering model that is a function of bandwidth. We present here from C-band RADARSAT-2 data, and will compare these with additional C-band Sentinel data and L-band data from ALOS-2.

InSAR time series analysis of deformation fields requires a set of single look complex (SLC) radar images whose pixels accurately represent the InSAR phase. In some surface cover environments, such as vegetated regions where plant growth produces echoes unrelated to the underlying surface, reliable phase estimates are difficult to identify. Persistent scattering (PS) methods sort for the most stable points in an image, so that the deformation estimates are as robust as possible. To date the often-used selection algorithms have not always fully incorporated knowledge of PS point distributions or how they vary by terrain type. Here we show estimates of the probability density functions (pdfs) for several types of surface cover and how the added information can be applied to use MAP rather than MLE estimation to improve selection, and also give insights to the radar scattering processes that give rise to PS pixels.