Next Article in Journal
Higher-Order Convolutional Neural Networks for Essential Climate Variables Forecasting
Previous Article in Journal
Revealing the Eco-Environmental Quality of the Yellow River Basin: Trends and Drivers
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Advancing the Limits of InSAR to Detect Crustal Displacement from Low-Magnitude Earthquakes through Deep Learning

1
Los Alamos National Laboratory, Intelligence and Space Research Division, Los Alamos, NM 87545, USA
2
Los Alamos National Laboratory, Earth and Environmental Sciences Division, Los Alamos, NM 87545, USA
*
Author to whom correspondence should be addressed.
Current address: Disaster Prevention Research Institute, Kyoto University, Kyoto 611-0011, Japan.
Current address: TerranearPMC, Exton, PA 19341, USA.
Remote Sens. 2024, 16(11), 2019; https://doi.org/10.3390/rs16112019
Submission received: 8 April 2024 / Revised: 28 May 2024 / Accepted: 30 May 2024 / Published: 4 June 2024

Abstract

:
Detecting surface deformation associated with low-magnitude ( M w 5 ) seismicity using interferometric synthetic aperture radar (InSAR) is challenging due to the subtlety of the signal and the often challenging imaging environments. However, low-magnitude earthquakes are potential precursors to larger seismic events, and thus characterizing the crustal displacement associated with them is crucial for regional seismic hazard assessment. We combine InSAR time-series techniques with a Deep Learning (DL) autoencoder denoiser to detect the magnitude and extent of crustal deformation from the M w = 3.4 Gallina, New Mexico earthquake that occurred on 30 July 2020. Although InSAR alone cannot detect event-related deformation from such a low-magnitude seismic event, application of the DL method reveals maximum displacements as small as (±2.5 mm) in the vicinity of both the fault and earthquake epicenter without prior knowledge of the fault system. This finding improves small-scale displacement discernment with InSAR by an order of magnitude relative to previous studies. We additionally estimate best-fitting fault parameters associated with the observed deformation. The application of the DL technique unlocks the potential for low-magnitude earthquake studies, providing new insights into local fault geometries and potential risks from higher-magnitude earthquakes. This technique also permits low-magnitude event monitoring in areas where seismic networks are sparse, allowing for the possibility of global fault deformation monitoring.

1. Introduction

Interferometric synthetic aperture radar (InSAR) is a geodetic technique that uses phase interferometry to produce two-dimensional fields of surface displacement along the line-of-sight (LOS) of the radar. These displacement maps, also referred to as interferograms, typically have footprints on the order of tens to hundreds of kilometers, and the technique itself can resolve deformation at fine resolution (e.g., centimeter scale). These characteristics make InSAR an invaluable geodetic tool for observing geophysical phenomena, including seismic hazards. Since its initial application to measuring crustal deformation from the Landers, California earthquake in 1992 [1], this technique has been applied to well over 100 earthquake studies [2] and this number is currently growing by 20–30 events per year [3]. The actual signal being measured by InSAR, the apparent change in the distance between ground and satellite difference from two synthetic aperture radar (SAR) images, represents a combination of crustal deformation plus noise potentially stemming from changes in atmospheric conditions, topography, soil moisture, or vegetation [4,5,6]. Thus, care must be taken to properly partition out the signal of interest from other non-seismic factors. In particular, the detection of small and transient seismic deformation has remained a challenge, often requiring interpretation of noisy InSAR time series using a priori knowledge of the possible sources of deformation (e.g., [7,8,9]).
Time-series analysis is a common approach for mitigating atmospheric and topographic noise in InSAR datasets as well as effects from soil moisture and vegetation conditions. Such analysis looks for pixel-wise trends in displacement over the time span of an InSAR dataset, smoothing out transient signals and improving the signal-to-noise ratio of the data. Several noise mitigation techniques combined with time-series analysis have been used to extract deformation signals from low-magnitude (Mw < 6) events. Such approaches include reducing atmospheric interference through time-series analysis [10] or accounting for topographic-related tropospheric delays, which was achieved while analyzing a Mw 5.9 earthquake in Tibet [11]. Auxiliary information including Global Positioning System (GPS) data and data from multiple SAR satellites, as well as advanced coregistration techniques, were used to investigate displacement from a Mw 3.9 earthquake in southern Italy [12]. Finally, in at least one case, InSAR was used to investigate four low-magnitude earthquake events (Mw ≤ 5.1) in Turkey where the use of advanced techniques was unnecessary, and therefore absent, due to the events occurring over ideal conditions for InSAR processing [13]. For example, the extent of deformation in this region was small with good coherence due to minimal tropospheric delay. Additionally, most of the study area was flat and the spatial extent of the fringe was large with a regular pattern of deformation.
In the aforementioned studies, InSAR processing techniques were applied to estimate crustal displacement at the centimeter scale. To increase signal detection still further to sub-centimeter scales, other techniques to improve the signal-to-noise ratio of the data must be used. Persistent Scatter (PS) InSAR is one such technique that has been used due to its ability to minimize the effects of temporal and geometric decorrelation, as well as the influence of the atmospheric phase screen (APS) [14]. PS works by considering consistent scatterers, which are generated from the satellite radar signal bouncing off stationary objects [14,15,16]. However, PS performs optimally only when large data stacks are available, high PS density is present, and crustal deformation is minimal. The use of PS is thus problematic in highly vegetated areas or steep terrain that leads to low scattering densities, which often prevent the successful application of this technique. Moreover, PS analysis is computationally very expensive [16] and relies on knowledge of deformation smoothness in time and/or space, making it most suitable for monitoring stable deformation. Accounting for distributed scatterers (DS) (e.g., desert areas, crop fields, etc.) is a related technique that captures typically weaker signals than PS, but still may improve signal detection in InSAR analyses when smaller levels of displacement are expected. By selecting pairs using a minimum spanning tree framework weighted by a measure of image quality, one can obtain similar results using methods that consider both PS and DS (e.g., SqueeSAR; [17]) [18]. However, this method has not been tested using data from ScanSAR/TOPSAR imaging modes (e.g., Sentinel-1), nor has it been applied to fault activity, which may have a more complicated signal than geothermal signatures.
More recent work has shown promise for the use of a deep learning (DL)-based method, which was used to automatically detect and extract ground deformation signals from InSAR time series [18]. This technique relies on a fully convolutional deep auto-encoder to recognize and extract a signal from noisy InSAR time series. Using this technique, deformation on faults can still in principle be recovered from an InSAR signal for signal-to-noise ratios (SNRs) as low as 1%. Therefore, when combined with standard InSAR output, this method can potentially be used to detect millimeter-scale displacement.
In this study, we examine the potential of the aforementioned DL method for detecting sub-centimeter crustal displacement from the 30 July 2020 Mw 3.4 earthquake in Gallina, NM in the vicinity of the Nacimiento –Gallina fault arch. This earthquake occurred less than 70 km from several northern New Mexico communities, as well as the Valles Caldera National Preserve and Bandelier National Monument, which together hold geological and cultural significance. A better understanding of seismicity along this fault system is critical, as a recent study provided evidence that the Gallina fault is currently active [19] and supported earlier, poorly constrained estimates that this fault may be capable of accommodating an Mw 6–7 event in the future [20].

2. Materials and Methods

2.1. Site Background

The Mw 3.4 Gallina, NM earthquake occurred on 30 July 2020 near the southern end of the Gallina fault and the northern end of the Nacimiento fault at 36°13.09′N, 106°50.87′W at a depth of 11.4 km (location determined from Los Alamos Seismic Network (LASN) data, as shown in Figure 1 with faults from the United States Geological Survey (USGS) Geologic Map Database) [21,22]. The Gallina and Nacimiento faults are largely Laramide structures (80–50 Ma) [23,24], although there is evidence for some post-Laramide, Miocene slips [25].
It is not yet resolved whether the Gallina and Nacimiento faults are currently active (i.e., have Quaternary offset). In two probabilistic seismic hazard analyses for the Los Alamos National Laboratory [20,26], and a study on the nearby El Vado Dam [27], both the Gallina and Nacimiento faults were conservatively interpreted to be active, despite no clear evidence of Quaternary offset, largely due to microseismicity observed within and east of these faults [28,29]. More recently, the first clear geologic evidence for Quaternary deposit offset was provided across the northern part of the Gallina fault and also extended the fault 14 km further north [19]. No similar studies have been performed near the southern end of the Gallina fault or the northern end of the Nacimiento fault, near the location of the 30 July 2020 earthquake.
The Nacimiento Fault has been interpreted as a predominantly east-dipping thrust and reverse fault, transitioning northward along strike through vertical to a west-dipping normal fault near its northern end [25]. Its dip is thought to be steep at depth, flattening upwards, with the shallow section accommodating a westward slip over the San Juan Basin sediments [30]. The Gallina fault is mapped as being nearly vertical and has both predominant down-to-the-west and minor down-to-the-east offsets [25]. Detection of ground deformation from the July 2020 event, if it aligns with the known faults, would bolster both the evidence for the Gallina and/or Nacimiento faults being active and provide a means to constrain fault geometry.
Figure 1. Satellite image (Google Earth Pro v7.3, image taken on 19 September 2017 courtesy of Landsat/Copernicus) showing the epicenter of the 30 July 2020 event from LASN, denoted with a yellow circle [21]. White lines are faults from the USGS Geologic Map Database [22]. Map inset in upper right corner depicts the location of the study area in blue relative to the state of New Mexico. Map was created using Generic Mapping Tools v6 [31].
Figure 1. Satellite image (Google Earth Pro v7.3, image taken on 19 September 2017 courtesy of Landsat/Copernicus) showing the epicenter of the 30 July 2020 event from LASN, denoted with a yellow circle [21]. White lines are faults from the USGS Geologic Map Database [22]. Map inset in upper right corner depicts the location of the study area in blue relative to the state of New Mexico. Map was created using Generic Mapping Tools v6 [31].
Remotesensing 16 02019 g001

2.2. Data

We utilize data from the European Space Agency’s C-band Sentinel-1 constellation due to its repeat global coverage and open-source status. Sentinel-1 has an average cadence of 12 days. We work with single-look complex (SLC) data from the Interferometric Wide (IW) mode, which is collected at roughly 5 by 20 m resolution over three subswaths, forming a full image with an overall swath size of 250 km [32]. For our analysis, we use the third subswath of ascending track T49 and the second subswath of descending track T56. The corresponding unit look vectors, considered from a pixel on the ground to the satellite in the global Cartesian coordinate system, are s ^ T 49 = [ 0.70 ,   0.12 , 0.70 ] and s ^ T 56 = [ 0.63 ,   0.11 , 0.77 ] . We focus on data acquired two to three months pre- and post-earthquake event (March to October 2020), excluding months outside of this timeframe to avoid processing scenes affected by snow cover.
Pairs are chosen to maximize the coherence of the dataset by setting thresholds for both the orbital baseline (120 m) and temporal baseline (90 days). In total, we processed 45 interferometric pairs from 16 out of 19 different scenes (i.e., epochs) for ascending track T49 (3 scenes were deemed unfit for interferometric analysis due to poor correlation) and 57 pairs from 16 scenes spanning descending track T56 (Figure A1 and Figure A2 in Appendix A) using the open-source Generic Mapping Tool Synthetic Aperture Radar (GMTSAR) InSAR processing software [33]. We apply a 30 m digital elevation model (DEM) provided by the Shuttle Radar Topographic Mission (SRTM) to remove topographic contributions from each interferogram [34]. Pairs are processed in batch using built-in capabilities from GMTSAR with ionospheric phase correction included via the split-spectrum method (e.g., [35]). The pairs are unwrapped using GMTSAR’s implementation of Statistical-cost, Network-flow Algorithm for Phase Unwrapping (SNAPHU) [36]. This results in pairwise measurements of unwrapped range change or LOS deformation referenced to the radar. In this convention, positive values of range change indicate subsidence while negative values indicate uplift. We note that a common convention for time-series analysis is to reference deformation to the ground, such that negative values of LOS deformation indicate subsidence. For clarity, we hereby refer to ground-referenced LOS deformation as LOS displacement (following time-series analysis convention) and radar-referenced LOS deformation as range change.
A comparison to global positioning system (GPS) data from the Geodetic Facility for the Advancement of Geoscience (GAGE) network suggests our InSAR data have an accuracy on the order of millimeters (Appendix B).

2.3. Methods

The overall workflow for our methods is shown in Figure 2. We start by processing our InSAR data into interferometric pairs, the results of which we refer to as observed range changes. We then apply time-series analysis to these pairs to improve the signal-to-noise ratio of our dataset. This process results in cumulative displacements, which we will refer to as TS displacements. We feed these TS displacements to our DL method, which enhances signals of interest. The resulting displacement maps from this method are referred to as DL displacements. Lastly, we apply spatial deformation modeling to these DL displacements using an analytical displacement model for seismic applications. The resulting deformation fields from this model are referred to as modeled slips, which we further temporally analyze in terms of seismic slip.

2.3.1. 2D Time-Series Analysis

After pair formation, the 2D average velocity field and cumulative deformation fields associated with each epoch in the dataset are estimated using pixel-wise time-series analysis. The open-source Python package MintPy [37] is used to perform a weighted least squares inversion informed by coherence (e.g., [38]), with additional tropospheric correction via atmospheric models supplied by PyAPS [39]. Also included are DEM error estimation and orbital phase ramp estimation.

2.3.2. Deep Learning

The TS displacements created using the SBAS time-series analysis procedure described above functioned as input for the DL technique. The DL method applied in this study detects and extracts transient ground deformation signals from the time-series displacements using a fully convolutional autoencoder [18]. More specifically, it has been trained to isolate spatial and temporal patterns of deformation resembling those found along faults following a seismic event from signals due to other modes of surface displacement, such as aquifer depletion or magma intrusion. At the same time, it removes noise associated with variability in atmospheric conditions. The autoencoder uses a 9-timestep window to parse the input TS displacement fields, outputting the cumulative deformation occurring during the time window (excluding the first and last steps, which are used as references by the model). The autoencoder has previously been trained on synthetic data, in which signals of low-magnitude displacement in fault zones were overlain by fields carefully designed to mimic atmospheric noise. The trained tool has shown remarkable success in earthquake detection. Within its training and validation datasets, it recovered displacement from interferograms with SNRs as low as 1%, and, in a case study along the North Anatolian fault in Turkey, it was successfully employed to detect millimeter-scale displacement from a slow slip event [18]. We use our TS displacements and DEM as inputs to this DL model. This results in displacement fields containing refined signals of interest, which we refer to as DL displacements.

2.3.3. Slip Assessment via Spatial Deformation Modeling and Temporal Adjustment

In addition to identifying surface deformation signals amidst background noise in a seismically active area, we aim to identify whether these signals can be reasonably interpreted as seismically related. To do this, we apply spatial deformation modeling by fitting a seismic model to identify signatures of interest found in our DL displacements spanning the event window. We use an open-source package called the General Inversion of Phase Technique (GIPhT) [40]. When supplied with a deformation model and accompanying initial estimates of parameters for such a model, GIPhT uses simulated annealing to solve the nonlinear inversion problem of estimating the model parameters that best describe the deformation observed in the InSAR data. Since we are particularly interested in whether observed signals may be explained seismically, we employ the Okada model [41]. This model calculates the surface deformation generated by the slip of a solid over a rectangular patch at a given depth below the free surface of a homogeneous elastic half-space. We assume a standard Poisson’s ratio of ν = 0.25 following similar studies in the area [20,42]. Parameters of interest that are optimized in this formulation include the dimensions and spatial orientation of the rectangular patch, its depth, and the direction and magnitude of the slip. We first fit the average field of the DL displacements for both descending and ascending tracks to find the best-fitting structural parameters for the slip event. We then fit each individual DL displacement field with the optimized structural parameters constrained to estimate the amount of slip over each temporal pairing.
To analyze the relationship between estimated slip and seismic activity, we apply a time-series analysis technique called temporal adjustment. This technique converts the pairwise estimates of slip to cumulative slip over each epoch of the dataset [43]. Assuming separable temporal and spatial dependence, the displacement field u is then defined by the product of a distinct temporal function f t and spatial function G ( x ) :
u ( t , x ) = f ( t ) G ( x ) .
In our case, G ( x ) is represented by a single Okada or collection of Okada source models. We examine several different temporal representations f ( t ) to best describe temporal trends in the slip. First, we consider a constant rate parameterization with rate a for an epoch t i given starting epoch t 0 :
f 1 t i = a t i t 0 .
We also consider piecewise-linear parameterizations:
f m t i = Σ j = 1 j = m a j D t i ,   where   D t i = 0         if   t i < t j , t j t i           if   t j t i < t j + 1 ,   t j + 1 t j     if   t i t j + 1 .
We examine a two-segment ( m = 2 ) parameterization f 2 t i with a break at the event on 30 July. We also examine a three-segment ( m = 3 ) parameterization f 3 t i with breaks on 1 July and 30 July.
While we also consider nonlinear representations such as exponential decay and polynomial parameterizations, initial analysis indicates these functions are not well suited to characterizing the temporal nature of slip in this study and are thus not discussed further.

3. Results

3.1. InSAR Time Series

The TS cumulative displacement fields derived from time-series analysis on descending track T56 are shown in Figure 3, with the corresponding ascending track T49 analysis having similar results. The sequence of TS displacement fields shows cumulative deformation from the initial data point ( t 0 ), where the date of the final timestamp for each deformation field is shown on the bottom right of each image. There is no apparent deformation surrounding the epicenter of the earthquake (denoted by the outlined star) in either the ascending or descending track. Furthermore, both results are seemingly dominated by an atmospheric gradient effect present from southwest to northeast in TS displacement fields. Consequently, the TS displacement fields derived from our InSAR time-series analysis do not on their own reveal any meaningful deformation present at or in the vicinity of the earthquake epicenter.

3.2. Signal Refinement via Deep Learning

To further extract any subtle signals occurring in the data, we apply the DL autoencoder to our TS displacements. Figure 4 shows the average DL displacement rate fields derived for both tracks along with the LASN-located epicenter and fault traces obtained from the USGS Geologic Map Database [22]. As opposed to the fields in Figure 3, we see noticeable displacement in several sections of the field. We focus our attention on deformation signatures that fall along fault lines and that have paired uplift–subsidence patterns following typical seismic-related deformation. This leads to a site along the Nacimiento Fault system (Figure 4), which we denote as the Nacimiento Fault (NF) site.
When further analyzing the DL results at the Nacimiento Fault site, we find that the model detects surface deformation between the Nacimiento and Gallina faults in both the ascending and descending tracks (Figure A3 in Appendix C). Cumulative displacement measurements within 9-step windows occurring prior to the July 30th event show paired subsidence and uplift signatures that mirror the configuration of the Gallina and Nacimiento faults at this site. Displacement reaches a maximum uplift of ~1.5 mm and maximum subsidence of ~2.5 mm during this time. While cumulative displacements from 9-step windows that encompass the July 30th event maintain the characteristic shape seen in pre-event displacements, they consistently diminish in magnitude as time progresses until cross-sections reveal almost no discernible displacement in the final estimates.
We note that the model has no a priori knowledge of the faults, so its detection of displacement at the Nacimiento Fault site that mirrors the known configuration of the fault system increases our confidence in attributing the signal to the earthquake.

3.3. Spatial Deformation Modeling and Slip Analysis

We further investigate potential causes of the DL displacement signals identified at the Nacimiento Fault site by considering whether the observed displacement could have been caused by co-seismic slip occurring at the time of the event. We investigate this using the Okada model within GIPhT, as described in Section 2.3.3. We start by finding the average DL displacement rate field for each track. We then use simulated annealing via GIPhT to find Okada model parameters that best explain the deformation measured in both tracks’ average displacement rate fields. Once these parameters are identified, we then apply the model to each individual DL displacement field, holding constant the structural parameters of the model (e.g., location and geometry) and inverting only for slip. The similarity between ascending and descending DL displacement fields allowed for one model to describe the structure of the deformation signatures in both tracks.
Based on the deformation pattern at the NF site, we fit two Okada models. The best-fitting model parameters are shown in Table 1. We define a dimensionless misfit of the model to the DL displacement data using the square root of the reduced χ 2 -test statistic [44]. We find the misfit to be χ = 1.6 for both ascending and descending average DL displacement rate fields. Figure 5 shows the corresponding model compared to the descending average DL displacement field.
Once we have estimates of slip for each pairwise DL displacement field at each site, we then perform temporal adjustment to model the resulting epoch-wise slip and determine temporal trends in the slip over time according to our f 1 t , f 2 t , and f 3 t parameterizations. We model each Okada source separately per site. Our results for both sites are described in Table 2.
We find that the piecewise-linear parameterization using three breaks ( f 3 t ) provides the best fit for both Okada sources (Figure 6 and Figure 7), although the first Okada source (O1) is also reasonably described using the two-segment parameterization f 2 t . To determine whether the added complexity of including a break at the start of July is meaningful, we perform an F-test for model complexity at 95 percent confidence [44]. We find an F-test score of F c a l c = 7.8 as compared to a critical value of F α = 0.05 = 7.8 with degrees of freedom d o f 1 = 8 and d o f 2 = 7 . We thus reject the null hypothesis that f 2 t and f 3 t provide similar measures of fit, indicating that f 3 t is significantly better at explaining the temporal slip pattern for Okada 1.
We also consider similarities in temporal trends between the Okada 1 and Okada 2 sources. While there is no significant difference in slip rate estimates for f 1 t based on uncertainty bounds, the best-fitting parameters for the f 2 t and f 3 t parameterizations are markedly different between sources, suggesting that there are more complicated dynamics at this site.

4. Discussion

The detection of displacement associated with the Gallina earthquake is the smallest magnitude earthquake and displacement identified yet using InSAR, with detected displacements as small as 2.5 mm in the area between the Nacimiento and Gallina faults (e.g., Nacimiento Fault site), where deformation signatures were consistently captured amongst co-seismic pairs. Although the DL technique applied does not definitively link the estimated displacement directly to the earthquake, our confidence in this interpretation is supported by the fact that the denoising autoencoder identified maximum deformation in an area between the sections of the Nacimiento and Gallina fault systems that are just West of the reported epicenter, despite having no prior knowledge of the fault location. Moreover, the timing of the displacement we observed correlates with the timing of the earthquake, with deformation peaking just prior to the seismic event, and then gradually relaxing throughout the co-seismic time intervals.
Taking further into consideration only those DL-identified displacements that occur along faults in both the ascending and descending tracks, we focus on the region between the southern end of the Gallina Fault and the Nacimiento (e.g., Figure 4). This area shows complex spatial deformation patterns that suggest two areas of localized slip, which our inverse spatial deformation modeling suggests are related to movement along two normal faults. While both of these slip sources are aligned with the SW-NE trend of the Nacimiento fault bend, the western slip source (Okada 1) has a steeper dip (~80°) as compared to its eastern counterpart (Okada 2, with a dip of ~50°). Additionally, both sources are estimated at shallow depths occurring between 200 and 400 m. Taking the strikes of both sources to be roughly equal, this suggests that the sources’ corresponding faults intersect at a depth of ~2 km roughly 300 m NE of Okada 1’s location. Temporal adjustment on pairwise slips estimated during the span of our dataset suggests that both sources have temporal trends associated with the July 30th event; we see a change in slip for both slip sources starting a month prior to the event and immediately following the event. While both sources show an increase in slip at the start of our dataset and a marked decrease in slip post-event, their slip responses in the month preceding the event differ. For the western deformation (Okada 1), slip starts to decrease, while the eastern deformation (Okada 2) shows a drastic increase in slip. We note here that while we select July 1st as the start of our second time interval for the f 3 t parameterization, the lack of temporal data coverage in June limits us from distinguishing between any significance of a temporal break in June as compared to the start of July.
The timing of the deformation, which starts prior to the seismic event, makes it unlikely that the observed deformation is a direct result of the earthquake. However, the close correlation in both the time and space of the observed deformation and the earthquake suggests that both are related to slip on the same fault system. If so, then we observe subtle precursor slip, leading to the intriguing possibility that this method could potentially be used to identify regions of aseismic slip precursors to earthquakes. This would require additional validation at other sites and time periods.
While we focus our analysis on activity along known fault systems, we note that there are several signals in the NE portion of Figure 4 that occur in both the ascending and descending tracks (further distinguished in Appendix D, Figure A4). While these signals do not occur along a known fault system, their distinguishable presence in both tracks suggests that activity may be occurring here as well. The faulting systems in this area are not well constrained [45], and there is the possibility that our method is detecting movement along faults that have not yet been mapped. While outside the scope of this paper, examining the signatures and geology of this region would better inform on the potential for our method to also help distinguish undetected faults.
One possible approach for constraining the relationship between deformation signatures and geology in this region is to further decompose the InSAR-observed LOS displacements into their vertical and horizontal components. Considering that the primary goal of this study was to assess the potential for sub-centimeter, seismic-related surface deformation detection using the DL technique in ref. [18] and that this technique utilizes LOS deformation, displacement decomposition analysis was not explored in this study. However, such an analysis may be warranted in future studies of this area if there are ascending and descending scenes available with near-coincident temporal pairings.
A comparison of the region of displacement with the highest-resolution geologic map for the area [46] (Figure A6 in Appendix E) indicates that the westernmost boundary between uplift and subsidence in both the ascending and descending track deformation areas follow a line at the headscarp of a mapped landslide. The co-location of the displacement region within a landslide formation raises the possibility that the mechanism of displacement was not seismically related and could perhaps be associated with the instability of the landslide. In order to claim detection of deformation associated with the smallest magnitude earthquake, it is necessary to distinguish between landslide and fault-related deformation.
Given this ambiguity in geologic relations, we performed a field check of the area of observed InSAR deformation on 10 November 2022 and found that the mapped landslide cannot be modern, as the source area for the deposits has been eroded away. In addition, the area of uplift is a topographically high area relative to the western area of subsidence. Neither landsliding nor other geomorphic processes can explain the uplift of a topographically high area, while this is readily explainable by faulting. In addition, the geometry of deformation observed in this study, with 2 uplift–subsidence pairs, is more consistent with fault displacement. Finally, the feature is interpreted to be a landslide scarp on the map in ref. [46] is continuous with a feature interpreted to be the Gallina fault along the strike further east. Thus, the existing geologic data support the interpretation of this InSAR deformation presented here to be related to slip on the Gallina fault, and not landslide-related.
The amount of atmospheric attenuation of a SAR signal is inversely proportional to the wavelength of the satellite [47]. This means that the application of the DL method described here would be most useful for InSAR processing in conjunction with smaller wavelength X- or C-band missions where this technique is likely to be more effective at filtering noise related to atmospheric artifacts from the InSAR image. The utility of the method here is most obvious for the ongoing Sentinel-1 mission, as well as the recent C-band Radarsat Constellation Mission and the X-band family of COSMO-SkyMed, KOMPSAT-5, TerraSAR-X, TanDEM-X, or PAZ [2], where atmospheric delays are often greater than potential signals of interest. Although these other wavelengths were not utilized in this study due to either a lack of availability of open-source data or a lack of data coincident with our study area, they may be useful for follow-on analysis or investigations of other low-magnitude earthquake events where the DL approach is warranted. In addition, there is potential to utilize data from the upcoming L- and S-band NASA-ISRO SAR mission, the joint Earth-observing satellite mission between NASA and the Indian Space Agency [48], which may suffer less from atmospheric delays due to the longer wavelength of the radar. Given the ability of the DL model to extract deformation as a function of the signal-to-noise ratio, its application to less noisy data even from longer wavelength SAR missions may yield even smaller deformation detection capability, thereby still demonstrating its usefulness. We do note that the removal of ionospheric contributions through other techniques (e.g., split spectrum method [35]) would need to be given special consideration for longer wavelength data sources, as the DL technique applied in this study does not explicitly address ionospheric effects.
While the standard InSAR analysis techniques have not found any deformation located at the earthquake site location, the DL technique shows promise for future applications to identify deformation related to low-magnitude earthquakes. This study is at the forefront of low-magnitude, small-displacement earthquake studies from InSAR. To demonstrate this, Figure 8 shows the measured LOS crustal displacement for a number of earthquakes and the associated earthquake magnitudes from previous InSAR processing of Sentinel-1 studies conducted from 2015 until recently. The vast majority of these studies have been on earthquakes with a magnitude greater than 6. The Gallina earthquake that was targeted for this study is not only the smallest magnitude earthquake represented (Mw = 3.4), but the application of the DL technique has improved the LOS displacement detection by more than an order of magnitude relative to other studies.
Through further applications to similar earthquake settings, the technique applied here promises to improve the understanding of earthquake hazard potential, should a higher magnitude earthquake occur, as well as provide information on the local fault geometries through additional modeling. Moreover, the application of this technique to higher-magnitude events holds the potential to reveal more about the extent of crustal displacement by highlighting new signals of displacement that were previously undetected. Such a feat was already demonstrated by the application of this technique to the North Anatolian Fault where the model recovered 2 mm of previously undetected slip along a fault [18]. These new findings of displacement extent uncovered by this method led to the determination that the length of the fault that accommodated slip was twice what was previously thought and new fault lineaments were identified. Further studies on the application of this method to other small earthquakes that have occurred in the Gallina–Nacimiento fault system could be used to investigate similar features.

5. Conclusions

We have demonstrated the effectiveness of combining InSAR time-series analysis with a DL autoencoder denoiser to detect surface deformation associated with low-magnitude ( M w   5 ) seismicity. Applying this technique to Sentinel-1 InSAR pairs spanning the M w = 3.4 Gallina, New Mexico earthquake that occurred on 30 July 2020, we can detect surface deformation signals as subtle as 2.5 mm in the area between the Nacimiento and Gallina faults that are otherwise undetectable with standard time-series analysis techniques. Additionally, the spatial and temporal patterns of the displacement coincide with the seismic event, suggesting a relationship between the observed subsidence and the July 30th event itself. Additionally, the DL autoencoder was able to identify several locations of displacement in both the ascending and descending tracks that were not aligned with known faults, suggesting that this could be a method for fault identification as well. Additional analysis across other low-magnitude events will further determine the extent of this technique for low-magnitude seismic risk characterization and fault identification.

Author Contributions

Conceptualization, K.C.S., E.C.R., C.J.A. and E.M.S.; methodology, E.C.R., C.J.A. and B.R.-L.; software, E.C.R., C.J.A. and B.R.-L.; validation, E.C.R., C.J.A. and E.M.S.; formal analysis, E.C.R., E.M.S., C.J.A. and K.S.; investigation, E.C.R. and C.J.A.; resources, K.C.S., E.C.R., C.J.A. and E.M.S.; data curation, E.E.S., K.C.S., E.C.R. and C.J.A.; writing—original draft preparation, E.C.R. and E.E.S.; writing—review and editing, E.C.R., C.J.A., E.M.S., B.R.-L., E.E.S., K.S. and K.C.S.; visualization, E.C.R., C.J.A. and K.S.; supervision, K.C.S.; project administration, K.C.S.; funding acquisition, K.C.S. and B.R.-L. All authors have read and agreed to the published version of the manuscript.

Funding

Funding for this work was provided by the Los Alamos National Laboratory (LANL) Seismic Hazards program. Bertrand Rouet-Leduc was additionally supported by NASA NSPIRES grant 80NSSC22K1282 and the JSPS LEADER program.

Data Availability Statement

Raw Synthetic Aperture Radar (SAR) data were obtained from the European Space Agency’s (ESA) Sentinel-1 satellite mission free of charge through the Distributed Active Archive Center (DAAC) at the Alaska Satellite Facility (ASF; https://search.asf.alaska.edu/, accessed on 8 April 2024), as well as through the Sentinel Scientific Data Hub. SRTM DEM data for InSAR processing was accessed free of charge through GMTSAR’s DEM retrieval interface (https://topex.ucsd.edu/gmtsar/demgen/, last accessed on 1 October 2021). The GAGE GPS time-series data used for displacement validation is freely available from the Nevada Geodetic Laboratory (http://geodesy.unr.edu/index.php, last accessed on 20 May 2024). The DEM used for the k-means cluster was obtained from the Earth Data Analysis Center of the University of New Mexico (https://edac.unm.edu/, last accessed on 20 May 2024). Fault information is made freely available by the United States Geological Survey (USGS) via the ScienceBase-Catalog (https://www.sciencebase.gov/catalog/item/4f4e496ee4b07f02db5a354e, last accessed on 6 March 2024).

Acknowledgments

We are grateful to Romain Jolivet of École Normale Supérieure for his subject matter expertise. The interferograms from this study were created using open-source GMTSAR processing software and visualized in part using Generic Mapping Tools. The MintPy time-series package and PyAPS atmospheric model package used to calculate our TS displacements are freely available under GitHub’s MintPy repository (https://github.com/insarlab/MintPy, accessed on 14 March 2023) and PyAPS repository (https://github.com/insarlab/PyAPS, accessed on 14 March 2023). The spatial deformation modeling and temporal adjustment scripts used in our analysis were openly provided under GitHub’s General Inversion of Phase Technique repository (San Emidio branch, https://github.com/feigl/gipht/tree/SanEmidio, accessed on 11 May 2022). K-means plotting was supported by the open-source software QGIS (http://www.qgis.org/, last accessed on 4 April 2024). Funding for this work was made possible through a generous grant from the Los Alamos National Laboratory (LANL) Seismic Hazards program.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A. Baseline Maps for InSAR Pair Formation

Figure A1. Baseline map for the ascending Sentinel-1 track T49 data that were used in this study. The lines connecting the data points illustrate the interferogram pairs selected as the optimal dataset maximizing coherence, plotted according to the time span and orbital separation of the associated epochs in each pair. The dates of each image are as follows. A: 03-05-2020, B: 03-17-2020, C: 03-29-2020, D: 04-10-2020, E: 05-04-2020, F: 05-16-2020, G: 05-28-2020, H: 06-09-2020, I: 06-21-2020, J: 07-15-2020, K: 07-27-2020, L: 08-08-2020, M: 08-20-2020, N: 09-01-2020, O: 09-13-2020, P: 09-25-2020, Q: 10-07-2020, R: 10-19-2020, S: 10-31-2020. We note that although pictured here, pairs with epochs Q, R, and S were not included in any further analysis due to low signal quality.
Figure A1. Baseline map for the ascending Sentinel-1 track T49 data that were used in this study. The lines connecting the data points illustrate the interferogram pairs selected as the optimal dataset maximizing coherence, plotted according to the time span and orbital separation of the associated epochs in each pair. The dates of each image are as follows. A: 03-05-2020, B: 03-17-2020, C: 03-29-2020, D: 04-10-2020, E: 05-04-2020, F: 05-16-2020, G: 05-28-2020, H: 06-09-2020, I: 06-21-2020, J: 07-15-2020, K: 07-27-2020, L: 08-08-2020, M: 08-20-2020, N: 09-01-2020, O: 09-13-2020, P: 09-25-2020, Q: 10-07-2020, R: 10-19-2020, S: 10-31-2020. We note that although pictured here, pairs with epochs Q, R, and S were not included in any further analysis due to low signal quality.
Remotesensing 16 02019 g0a1
Figure A2. Baseline map for the descending Sentinel-1 track T56 data that were used in this study. The lines connecting the data points illustrate the interferogram pairs selected as the optimal dataset maximizing coherence, plotted according to the time span and orbital separation of the associated epochs in each pair. The dates of each image in MM-DD-YYYY are as follows. A: 03-11-2020, B: 03-23-2020, C: 04-04-2020, D: 04-16-2020, E: 04-28-2020, F: 05-10-2020, G: 05-22-2020, H: 06-03-2020, I: 06-15-2020, J: 06-27-2020, K: 07-09-2020, L: 07-21-2020, M: 08-02-2020, N: 08-14-2020, O: 08-26-2020, P: 09-07-2020.
Figure A2. Baseline map for the descending Sentinel-1 track T56 data that were used in this study. The lines connecting the data points illustrate the interferogram pairs selected as the optimal dataset maximizing coherence, plotted according to the time span and orbital separation of the associated epochs in each pair. The dates of each image in MM-DD-YYYY are as follows. A: 03-11-2020, B: 03-23-2020, C: 04-04-2020, D: 04-16-2020, E: 04-28-2020, F: 05-10-2020, G: 05-22-2020, H: 06-03-2020, I: 06-15-2020, J: 06-27-2020, K: 07-09-2020, L: 07-21-2020, M: 08-02-2020, N: 08-14-2020, O: 08-26-2020, P: 09-07-2020.
Remotesensing 16 02019 g0a2

Appendix B. InSAR-Measured Displacement Accuracy Assessment Using GPS

We use GPS data to perform a calibration check on our InSAR measurements. There are two GAGE network continuous GPS stations, RG09 (36.305°N, 107.056°W) and RG10 (36.451°N, 106.539°W), within the coverage of our ascending T49 and descending T56 tracks [49,50]. Both stations have been analyzed according to standard procedures, with their resulting time series of relative positions made freely available [51]. However, both stations only provide temporal coverage until mid-2019, which is outside the range of our 2020 analysis. To perform our validation, we instead use InSAR pairs spanning the same months used in our 2020 analysis but from T49 and T56 scenes acquired in 2018. This resulted in 46 pairs derived from 15 ascending track T49 scenes spanning from 4 March 2018 to 24 September 2018 and 36 pairs from 12 descending track T56 scenes spanning from 10 March 2018 to 1 August 2018.
We perform our analysis similar to previous studies [52,53,54]. In order to account for the relative nature of InSAR measurements, we assign GPS station RG09 as a reference station and treat RG10 as being within a deforming region. To arrive at estimates of InSAR-observed range change within the deforming region of RG10 with respect to RG09 for each InSAR pair, we compare InSAR-observed range change averaged over a 1 km2 area centered around station RG10 to InSAR-observed range change averaged over a 1 km2 area centered around station RG09. To compare our findings with the corresponding GPS measurements, we first convert GPS-observed displacements to range change using the unit pointing vectors described in Section 2.2. The GPS measurements (now in terms of range) are then compared at each station across time, corresponding with the epochs of each InSAR pair, to calculate each GPS station’s measure of relative range change over the span of each InSAR pair. We are then able to find the difference of the GPS-measured range change estimates between the stations to derive GPS-measured range changes from RG10 with respect to RG09, giving a direct comparison to our InSAR-observed range changes at RG10 with respect to RG09. We find a mean difference of 5 ± 3   m m in the range changes derived from GPS and InSAR measurements from both our T49 pairs and T56 pairs. Moreover, using uncertainties for our InSAR measurements calculated using standard deviation and uncertainties for our GPS measurements calculated using error propagation, we find overlap between μ ± 1 σ ranges between all GPS-derived and InSAR-derived range change measurements per pair, suggesting there is no significant difference in the measurements obtained from each modality. This analysis was repeated using 250 m2 and 500 m2 areas for InSAR-observed range change sampling, with no difference in the results.

Appendix C. DL Displacement Results at Nacimiento Fault Site

Figure A3. Pairwise LOS displacement results from DL analysis on ascending track T49 and descending track T56 for the Nacimiento Fault site (e.g., region denoted by the square in Figure 4). Each plot shows resulting cumulative displacement per 9-step window as estimated after applying our DL method to TS displacements within said time window. Star denotes the fault junction.
Figure A3. Pairwise LOS displacement results from DL analysis on ascending track T49 and descending track T56 for the Nacimiento Fault site (e.g., region denoted by the square in Figure 4). Each plot shows resulting cumulative displacement per 9-step window as estimated after applying our DL method to TS displacements within said time window. Star denotes the fault junction.
Remotesensing 16 02019 g0a3

Appendix D. K-Means Analysis of DL Displacements

To identify additional areas of potential interest for future analysis, we apply an unsupervised k-means clustering algorithm [55] to the pairwise DL displacement fields obtained from the DL autoencoder (Figure A4). The k-means clustering method identifies distinct clusters in a dataset by minimizing the distance between each data point in a cluster and its centroid. The pre-seismic and co-seismic DL displacement fields from both the ascending and descending tracks are converted to rates, co-located, and then stacked together for joint analysis. We then assign each pixel a weight based on its magnitude of displacement rate, such that the pixel with the greatest magnitude was assigned the highest weight. We determine the number of clusters by calculating the inertia of clusters [55], which is defined as the sum of the squared distance between each data point and its assigned cluster. We find the optimum number of clusters to be 4, as the decrease in inertia was not significant beyond this point (Figure A5). Additionally, while we cluster based on magnitude alone, results in Figure A4 suggest a relationship between cluster assignment and displacement direction (e.g., subsidence or uplift), although the comparative variance in each cluster limits any further meaningful interpretation. Cluster characteristics are described as follows:
  • Cluster 0 (not shown in Figure A4)—low-magnitude noise with a weighted mean of 0.02 mm and a standard deviation of 0.018 mm. This cluster did not exhibit distinct uplift or subsidence.
  • Cluster 1—high-magnitude deformation with a weighted mean value of −0.56 mm and a standard deviation of 0.73 mm. Overall, this cluster favors subsidence.
  • Cluster 2—overall uplift with a weighted mean value of 0.08 mm and a standard deviation of 0.14 mm.
  • Cluster 3—overall subsidence with a weighted mean of −0.14 mm and a standard deviation of 0.22 mm.
Interestingly, Clusters 2 and 3 appeared in tandem, with proximity (Figure A4), suggesting that these two clusters might be related to a fault movement. These clusters also include the Nacimiento Fault site, providing additional support for our focused analysis of this region. While further analysis is needed to identify their source and increase confidence in distinctions between these two clusters, this preliminary clustering work indicates that signals within these clustering regions may be of potential interest for future studies.
Figure A4. Relevant clusters obtained from k-means clustering with k = 4 . Cluster 1 consists of high-magnitude subsidence (mean = −0.56 mm; std = 0.73 mm). Cluster 2 consists of uplift (mean = 0.08 mm; std = 0.14 mm). Cluster 3 consists of subsidence (mean = −0.14 mm; std = 0.22 mm). Cluster 0 (considered to contain noise) is not shown. Clusters are overlaid on a DEM obtained from the Earth Data Analysis Center of the University of New Mexico as visualized using the open-source software QGIS. Bounding region for analysis is shown with a dark green rectangle.
Figure A4. Relevant clusters obtained from k-means clustering with k = 4 . Cluster 1 consists of high-magnitude subsidence (mean = −0.56 mm; std = 0.73 mm). Cluster 2 consists of uplift (mean = 0.08 mm; std = 0.14 mm). Cluster 3 consists of subsidence (mean = −0.14 mm; std = 0.22 mm). Cluster 0 (considered to contain noise) is not shown. Clusters are overlaid on a DEM obtained from the Earth Data Analysis Center of the University of New Mexico as visualized using the open-source software QGIS. Bounding region for analysis is shown with a dark green rectangle.
Remotesensing 16 02019 g0a4
Figure A5. Elbow plot used to determine the optimum number of clusters [55].
Figure A5. Elbow plot used to determine the optimum number of clusters [55].
Remotesensing 16 02019 g0a5

Appendix E. Geologic Relationship of Observed Displacement

Figure A6 shows that the observed deformation is more consistent with faulting than landslides or other surficial deformation.
Figure A6. Overlay of deformation between 4 and 15 to 7 and 20 near the Nacimiento Fault site, with geology from the New Mexico Bureau of Mines & Mineral Resources [46]. Also shown are fault locations from the USGS Quaternary Fault Database [56] and the New Mexico USGS Fault Database [22].
Figure A6. Overlay of deformation between 4 and 15 to 7 and 20 near the Nacimiento Fault site, with geology from the New Mexico Bureau of Mines & Mineral Resources [46]. Also shown are fault locations from the USGS Quaternary Fault Database [56] and the New Mexico USGS Fault Database [22].
Remotesensing 16 02019 g0a6

Appendix F. Comparison Studies for InSAR-Observed Seismic Displacement

Table A1. Summary of previous studies measuring crustal displacement from seismic events using the Sentinel-1 InSAR analysis, as visualized in Figure 8.
Table A1. Summary of previous studies measuring crustal displacement from seismic events using the Sentinel-1 InSAR analysis, as visualized in Figure 8.
EQ DateEQ LocationMwLOS Disp. [cm]Study
24 August 2014Napa, California6.16Yang et al., 2017 [57]
6 February 2018Hualien, Taiwan6.430Yang et al., 2018 [58]
11 January 2018Mandali, Iran5.253Barnhart et al., 2018 [59]
17 November 2017Nyingchi, China6.59Yu et al., 2018 [60]
12 November 2017Darbandikhan, Iran7.380Barnhart et al., 2018 [59]
8 September 2017Tehuantepac, Mexico8.120Chen et al., 2018 [61]
21 August 2017Ischia, Italy4.24De Novellis et al., 2018 [12]
8 August 2017Jiuzhaigou, China6.521Nie et al., 2018 [62]
5 April 2017Sangsefid, Iran6.112Su et al., 2019 [63]
6 February 2017Gulpinar, Turkey5.16.2Ganas et al., 2018 [13]
24 August 2016Amatrice, Italy6.218.5Xu et al., 2017 [64]
25 December 2016Melinka, Chile7.645Moreno et al., 2018 [65]
25 November 2016Aketao, China6.610Bie et al., 2018 [66]
13 November 2016Kaikoura, New Zealand7.8130Hamling et al., 2017 [67]
30 October 2016Norcia, Italy6.635Papadopoulos et al., 2017 [68]
17 October 2016Zaduo, China5.93Jiang et al., 2018 [11]
3 September 2016Pawnee, Oklahoma5.83Grandin et al., 2017 [10]
26 June 2016Nura, Kyrgyzstan6.512.1He et al., 2018 [69]
21 May 2016Petermann Ranges, Australia4.113Polcari et al., 2018 [70]
16 April 2016Pedernales, Ecuador7.860Bejar-Pizarro et al., 2018 [71]
5 February 2016Meinong, Taiwan6.412Qu et al., 2017 [72]
20 January 2016Menyuan, China5.98Liu et al., 2018 [73]
7 December 2015Murghab, Tajikistan7.256Metzger et al., 2017 [74]
15 November 2017Lefkada, Greece6.528.6Avallone et al., 2017 [75]
3 July 2015Pishan, China6.411.8Ainscoe et al., 2017 [76]
25 April 2015Gorkha, Nepal7.8110Castaldo et al., 2017 [77]
24 August 2014South Napa, California610Guangcai et al., 2015 [78]
26 November 2019Durres City, Albania6.35Panuntun, 2021 [79]
5 July 2019Ridgecrest, California7.15Wang and Burgman, 2020 [80]
29 December 2020Petrinja, Croatia6.240Bjelotomić Oršulić et al., 2021 [81]
16 October 2019Mindanao, Philippines6.45Li et al., 2020 [82]
30 November 2018Anchorage, Alaska7.15He et al., 2020 [83]
17 June 2019Changning, China66Wang et al., 2020 [84]
20 March 2019Acipayam basin, Turkey5.74.8Yang et al., 2020 [85]
9 August 2019Lombok Island, Indonesia6.239Wibowo et al., 2021 [86]
16 April 2016Kumamoto, Japan720He et al., 2019 [87]
24 September 2019Mirpur, Pakistan5.49.5Vaka et al., 2020 [88]
11 November 2019Le Teil, France4.916Marconato et al., 2021 [89]
19 January 2020Jiashi, China65.3Yu et a., 2020 [90]
26 December 2016Chiloe, Chile7.550Xu, 2017 [91]
12 August 2016Xiangyang Lake, China4.91Tian et al., 2018 [92]
28 September 2018Sulawesi, Indonesia7.512Bacques et al., 2020 [93]
21 March 2020Greece5.75Bignami et al., 2021 [94]
22 March 2020Zagreb, Croatia5.43Bignami et al., 2021 [94]

References

  1. Zebker, H.A.; Rosen, P.A.; Goldstein, R.M.; Gabriel, A.; Werner, C.L. On the derivation of coseismic displacement fields using differential radar interferometry: The Landers earthquake. J. Geophys. Res. Solid Earth 1994, 99, 19617–19634. [Google Scholar] [CrossRef]
  2. Elliott, J.; Walters, R.; Wright, T. The role of space-based observation in understanding and responding to active tectonics and earthquakes. Nat. Commun. 2016, 7, 13844. [Google Scholar] [CrossRef]
  3. Biggs, J.; Wright, T.J. How satellite InSAR has grown from opportunistic science to routine monitoring over the last decade. Nat. Commun. 2020, 11, 3863. [Google Scholar] [CrossRef]
  4. Bürgmann, R.; Rosen, P.A.; Fielding, E.J. Synthetic aperture radar interferometry to measure Earth’s surface topography and its deformation. Annu. Rev. Earth Planet. Sci. 2000, 28, 169–209. [Google Scholar] [CrossRef]
  5. Zebker, H.; Shankar, P.; Hooper, A. InSAR remote sensing over decorrelating terrains: Persistent scattering methods. In Proceedings of the 2007 IEEE Radar Conference, Waltham, MA, USA, 17–20 April 2007; pp. 717–722. [Google Scholar]
  6. Kirui, P.K.e.; Reinosch, E.; Isya, N.; Riedel, B.; Gerke, M. Mitigation of atmospheric artefacts in multi temporal InSAR: A review. PFG–J. Photogramm. Remote Sens. Geoinf. Sci. 2021, 89, 251–272. [Google Scholar] [CrossRef]
  7. Chaussard, E.; Bürgmann, R.; Shirzaei, M.; Fielding, E.J.; Baker, B. Predictability of hydraulic head changes and characterization of aquifer-system and fault properties from InSAR-derived ground deformation. J. Geophys. Res. Solid Earth 2014, 119, 6572–6590. [Google Scholar] [CrossRef]
  8. Rousset, B.; Jolivet, R.; Simons, M.; Lasserre, C.; Riel, B.; Milillo, P.; Çakir, Z.; Renard, F. An aseismic slip transient on the North Anatolian Fault. Geophys. Res. Lett. 2016, 43, 3254–3262. [Google Scholar] [CrossRef]
  9. Khoshmanesh, M.; Shirzaei, M. Episodic creep events on the San Andreas Fault caused by pore pressure variations. Nat. Geosci. 2018, 11, 610–614. [Google Scholar] [CrossRef] [PubMed]
  10. Grandin, R.; Vallée, M.; Lacassin, R. Rupture process of the Mw 5.8 Pawnee, Oklahoma, earthquake from Sentinel-1 InSAR and seismological data. Seismol. Res. Lett. 2017, 88, 994–1004. [Google Scholar] [CrossRef]
  11. Jiang, G.; Wen, Y.; Li, K.; Fang, L.; Xu, C.; Zhang, Y.; Xu, X. A NE-trending oblique-slip fault responsible for the 2016 Zaduo earthquake (Qinghai, China) revealed by InSAR data. Pure Appl. Geophys. 2018, 175, 4275–4288. [Google Scholar] [CrossRef]
  12. De Novellis, V.; Carlino, S.; Castaldo, R.; Tramelli, A.; De Luca, C.; Pino, N.A.; Pepe, S.; Convertito, V.; Zinno, I.; De Martino, P. The 21 August 2017 Ischia (Italy) earthquake source model inferred from seismological, GPS, and DInSAR measurements. Geophys. Res. Lett. 2018, 45, 2193–2202. [Google Scholar] [CrossRef]
  13. Ganas, A.; Kourkouli, P.; Briole, P.; Moshou, A.; Elias, P.; Parcharidis, I. Coseismic displacements from moderate-size earthquakes mapped by Sentinel-1 differential interferometry: The case of February 2017 Gulpinar earthquake sequence (Biga Peninsula, Turkey). Remote Sens. 2018, 10, 1089. [Google Scholar] [CrossRef]
  14. Rocca, F. Modeling interferogram stacks. IEEE Trans. Geosci. Remote Sens. 2007, 45, 3289–3299. [Google Scholar] [CrossRef]
  15. Ferretti, A.; Savio, G.; Barzaghi, R.; Borghi, A.; Musazzi, S.; Novali, F.; Prati, C.; Rocca, F. Submillimeter accuracy of InSAR time series: Experimental validation. IEEE Trans. Geosci. Remote Sens. 2007, 45, 1142–1153. [Google Scholar] [CrossRef]
  16. Crosetto, M.; Monserrat, O.; Cuevas-González, M.; Devanthéry, N.; Crippa, B. Persistent scatterer interferometry: A review. ISPRS J. Photogramm. Remote Sens. 2016, 115, 78–89. [Google Scholar] [CrossRef]
  17. Ferretti, A.; Fumagalli, A.; Novali, F.; Prati, C.; Rocca, F.; Rucci, A. A new algorithm for processing interferometric data-stacks: SqueeSAR. IEEE Trans. Geosci. Remote Sens. 2011, 49, 3460–3470. [Google Scholar] [CrossRef]
  18. Rouet-Leduc, B.; Jolivet, R.; Dalaison, M.; Johnson, P.A.; Hulbert, C. Autonomous extraction of millimeter-scale deformation in InSAR time series using deep learning. Nat. Commun. 2021, 12, 6480. [Google Scholar] [CrossRef] [PubMed]
  19. Thompson Jobe, J.A.; Chupik, C. Low-Rate Faulting on the Margin of the Colorado Plateau and Rio Grande Rift in North-Central New Mexico. Tectonics 2021, 40, e2021TC006860. [Google Scholar] [CrossRef]
  20. Wong, I.; Silva, W.; Olig, S.; Dober, M.; Gregor, N.; Gardner, J.; Lewis, C.; Terra, F.; Zachariasen, J.; Stokoe, K. Update of the Probabilistic Seismic Hazard Analysis and Development of Seismic Design Ground Motions at the Los Alamos National Laboratory; URS Corporation: Oakland, CA, USA, 2007; p. 25. [Google Scholar]
  21. Roberts, P.; Ten Cate, J.; House, L.; Greene, M.; Morton, E.; Kelley, R. The Los Alamos Seismic Network (LASN): Improved Network Instrumentation, Local Earthquake Catalog Updates, and Peculiar Types of Data. In Proceedings of the AGU Fall Meeting Abstracts, San Francisco, CA, USA, 9–13 December 2013; p. S51A-2287. [Google Scholar]
  22. Stoeser, D.B.; Green, G.N.; Morath, L.C.; Heran, W.D.; Wilson, A.B.; Moore, D.W.; Van Gosen, B.S.; New Mexico Faults, from the USGS Geologic Map Database 2005. USGS ScienceBase-Catalog. Available online: https://www.sciencebase.gov/catalog/item/4f4e496ee4b07f02db5a354e (accessed on 19 January 2024).
  23. Baltz, E.H. Stratigraphy and Regional Tectonic Implications of Part of Upper Cretaceous and Tertiary Rocks, East-Central San Juan Basin, New Mexico; 2330-7102; US Government Publishing Office: Washington, DC, USA, 1967. [Google Scholar]
  24. Kelley, S.A.; Chapin, C.E.; Corrigan, J. Late Mesozoic to Cenozoic Cooling Histories of the Flanks of the Northern and Central Rio Grande Rift, Colorado and New Mexico; New Mexico Bureau of Mines & Mineral Resources: Socorro, NM, USA, 1992; Volume 145. [Google Scholar]
  25. Woodward, L.A.; Hultgren, M.C.; Crouse, D.L.; Merrick, M.A.; Lucas, S.; Kues, B.; Williamson, T.; Hunt, A. Geometry of Nacimiento-Gallina fault system, northern New Mexico. San Juan Basin IV New Mex. Geol. Soc. Guideb. 1992, 43, 103–108. [Google Scholar]
  26. Wong, I.; Kelson, K.; Olig, S.; Kolbe, T.; Hemphill-Haley, M.; Bott, J.; Green, R.; Kanakari, H.; Sawyer, J.; Silva, W.; et al. Seismic hazards evaluation of the Los Alamos National Laboratory; Unpublished Consulting Report Prepared for Los Alamos National Laboratory; Woodward-Clyde Federal Services: Oakland, CA, USA, 1995; Volume 3. [Google Scholar]
  27. Derouin, S.; Su, F. Revised Probabilistic Seismic Hazard Analysis for El Vado Dam, New Mexico. Technical Memorandum 86-68330-2013-20; Bureau of Reclamation: Denver, CO, USA, 2013. [Google Scholar]
  28. Sanford, A.R.; Jaksha, L.H.; Cash, D.J.; Slemmons, D.; Engdahl, E.; Zoback, M.; Blackwell, D. Seismicity of the Rio Grande rift in New Mexico. Neotectonics N. Am. 1991, 1, 229. [Google Scholar]
  29. House, L.; Hartse, H. Seismicity and faults in northern New Mexico: New Mexico Geological Society. In Proceedings of the 46th Field Conference, Guidebook, Santa Fe, NM, USA, 27–30 September 1995; pp. 135–137. [Google Scholar]
  30. Woodward, L.A.; Kaufman, W.H.; Anderson, J.B. Nacimiento fault and related structures, northern New Mexico. Geol. Soc. Am. Bull. 1972, 83, 2383–2396. [Google Scholar] [CrossRef]
  31. Wessel, P.; Luis, J.F.; Uieda, L.a.; Scharroo, R.; Wobbe, F.; Smith, W.H.; Tian, D. The generic mapping tools version 6. Geochem. Geophys. Geosystems 2019, 20, 5556–5564. [Google Scholar] [CrossRef]
  32. Geudtner, D.; Torres, R.; Snoeij, P.; Davidson, M.; Rommen, B. Sentinel-1 system capabilities and applications. In Proceedings of the 2014 IEEE Geoscience and Remote Sensing Symposium, Quebec City, QC, Canada, 13–18 July 2014; pp. 1457–1460. [Google Scholar]
  33. Sandwell, D.; Mellors, R.; Tong, X.; Wei, M.; Wessel, P. GMTSAR: An InSAR Processing System Based on Generic Mapping Tools; Technical Report; Scripps Institute of Oceanography: La Jolla, CA, USA, 2011. [Google Scholar]
  34. Farr, T.G.; Rosen, P.A.; Caro, E.; Crippen, R.; Duren, R.; Hensley, S.; Kobrick, M.; Paller, M.; Rodriguez, E.; Roth, L. The shuttle radar topography mission. Rev. Geophys. 2007, 45, 1–33. [Google Scholar] [CrossRef]
  35. Brcic, R.; Parizzi, A.; Eineder, M.; Bamler, R.; Meyer, F. Estimation and compensation of ionospheric delay for SAR interferometry. In Proceedings of the 2010 IEEE International Geoscience and Remote Sensing Symposium, Honolulu, HI, USA, 25–30 July 2010; pp. 2908–2911. [Google Scholar]
  36. Chen, C.W.; Zebker, H.A. Network approaches to two-dimensional phase unwrapping: Intractability and two new algorithms. JOSA A 2000, 17, 401–414. [Google Scholar] [CrossRef] [PubMed]
  37. Yunjun, Z.; Fattahi, H.; Amelung, F. Small baseline InSAR time series analysis: Unwrapping error correction and noise reduction. Comput. Geosci. 2019, 133, 104331. [Google Scholar] [CrossRef]
  38. Perissin, D.; Wang, T. Repeat-pass SAR interferometry with partially coherent targets. IEEE Trans. Geosci. Remote Sens. 2011, 50, 271–280. [Google Scholar] [CrossRef]
  39. Jolivet, R.; Grandin, R.; Lasserre, C.; Doin, M.P.; Peltzer, G. Systematic InSAR tropospheric phase delay corrections from global meteorological reanalysis data. Geophys. Res. Lett. 2011, 38, L17311. [Google Scholar] [CrossRef]
  40. Feigl, K.L.; Thurber, C.H. A method for modelling radar interferograms without phase unwrapping: Application to the M 5 Fawnskin, California earthquake of 1992 December 4. Geophys. J. Int. 2009, 176, 491–504. [Google Scholar] [CrossRef]
  41. Okada, Y. Surface deformation due to shear and tensile faults in a half-space. Bull. Seismol. Soc. Am. 1985, 75, 1135–1154. [Google Scholar] [CrossRef]
  42. Newton, C.A.; Cash, D.J.; Olsen, K.H.; Homuth, E.F. LASL Seismic Programs in the Vicinity of Los Alamos, New Mexico; Los Alamos National Lab (LANL): Los Alamos, NM, USA, 1976. [Google Scholar]
  43. Reinisch, E.C.; Cardiff, M.; Feigl, K.L. Graph theory for analyzing pair-wise data: Application to geophysical model parameters estimated from interferometric synthetic aperture radar data at Okmok volcano, Alaska. J. Geod. 2017, 91, 9–24. [Google Scholar] [CrossRef] [PubMed]
  44. Strang, G.; Borre, K. Linear Algebra, Geodesy, and GPS; Wellesley-Cambridge Press: Wellesley, MA, USA, 1997. [Google Scholar]
  45. Woodward, L.A.; Karlstrom, K.E.; Daniel, C.G. Restoration of Laramide right-lateral strike slip in northern New Mexico by using Proterozoic piercing points: Tectonic implications from the Proterozoic to the Cenozoic: Comment and Reply. Geology 1994, 22, 862–864. [Google Scholar] [CrossRef]
  46. Merrick, M.A.; Woodward, L.A. Geology of Regina Quadrangle, Rio Arriba and Sandoval Counties, New Mexico; New Mexico Bureau of Mines & Mineral Resources: Socorro, NM, USA, 1982. [Google Scholar]
  47. Doerry, A.W. Atmospheric loss considerations for synthetic aperture radar design and operation. In Proceedings of the Radar Sensor Technology VIII and Passive Millimeter-Wave Imaging Technology VII, Orlando, FL, USA, 14–15 April 2004; pp. 17–27. [Google Scholar]
  48. Rosen, P.A.; Kim, Y.; Kumar, R.; Misra, T.; Bhan, R.; Sagi, V.R. Global persistent SAR sampling with the NASA-ISRO SAR (NISAR) mission. In Proceedings of the 2017 IEEE Radar Conference (RadarConf), Seattle, WA, USA, 8–12 May 2017; pp. 0410–0414. [Google Scholar]
  49. Sheehan, A. Rio Grande Rift GPS Network—RG09-RG09LndrthNM2006 P.S; The GAGE Facility operated by EarthScope Consortium and UNAVCO: Boulder, CO, USA, 2007. [Google Scholar] [CrossRef]
  50. Sheehan, A. Rio Grande Rift GPS Network—RG10-RG10CbollaNM2006 P.S; The GAGE Facility operated by EarthScope Consortium and UNAVCO: Boulder, CO, USA, 2007. [Google Scholar] [CrossRef]
  51. Blewitt, G.; Hammond, W. Harnessing the GPS data explosion for interdisciplinary science. Eos 2018, 99. [Google Scholar] [CrossRef]
  52. Reinisch, E.C.; Cardiff, M.; Feigl, K.L. Characterizing volumetric strain at Brady Hot Springs, Nevada, USA using geodetic data, numerical models and prior information. Geophys. J. Int. 2018, 215, 1501–1513. [Google Scholar] [CrossRef]
  53. Reinisch, E.C.; Cardiff, M.; Kreemer, C.; Akerley, J.; Feigl, K.L. Time-Series Analysis of Volume Change at Brady Hot Springs, Nevada, USA, Using Geodetic Data from 2003–2018. J. Geophys. Res. Solid Earth 2020, 125, e2019JB017816. [Google Scholar] [CrossRef]
  54. Reinisch, E.C.; Henderson, B.G. Spatio-temporal analysis and volumetric characterization of interferometric synthetic aperture radar-observed deformation signatures related to underground and in situ leach mining. J. Appl. Remote Sens. 2023, 17, 044511. [Google Scholar] [CrossRef]
  55. Kodinariya, T.M.; Makwana, P.R. Review on determining number of Cluster in K-Means Clustering. Int. J. Adv. Res. Comput. Sci. Manag. Stud. 2013, 1, 90–95. [Google Scholar]
  56. Valley, H.W.; Quaternary Fault and Fold Database of the United States. US Geological Survey. 2010. Available online: http://earthquakes.usgs.gov/hazards/qfaults (accessed on 6 March 2024).
  57. Yang, Y.; Chen, Q.; Xu, Q.; Zhang, Y.; Yong, Q.; Liu, G. Coseismic surface deformation of the 2014 Napa earthquake mapped by Sentinel-1A SAR and accuracy assessment with COSMO-SkyMed and GPS data as cross validation. Int. J. Digit. Earth 2017, 10, 1197–1213. [Google Scholar] [CrossRef]
  58. Yang, Y.-H.; Hu, J.-C.; Tung, H.; Tsai, M.-C.; Chen, Q.; Xu, Q.; Zhang, Y.-J.; Zhao, J.-J.; Liu, G.-X.; Xiong, J.-N. Co-seismic and postseismic fault models of the 2018 Mw 6.4 Hualien earthquake occurred in the junction of collision and subduction boundaries offshore eastern Taiwan. Remote Sens. 2018, 10, 1372. [Google Scholar] [CrossRef]
  59. Barnhart, W.D.; Brengman, C.M.; Li, S.; Peterson, K.E. Ramp-flat basement structures of the Zagros Mountains inferred from co-seismic slip and afterslip of the 2017 Mw7. 3 Darbandikhan, Iran/Iraq earthquake. Earth Planet. Sci. Lett. 2018, 496, 96–107. [Google Scholar] [CrossRef]
  60. Yu, C.; Li, Z.; Chen, J.; Hu, J.-C. Small magnitude co-seismic deformation of the 2017 Mw 6.4 Nyingchi earthquake revealed by InSAR measurements with atmospheric correction. Remote Sens. 2018, 10, 684. [Google Scholar] [CrossRef]
  61. Chen, K.; Feng, W.; Liu, Z.; Song, Y.T. 2017 Mw 8.1 Tehuantepec earthquake: Deep slip and rupture directivity enhance ground shaking but weaken the tsunami. Seismol. Res. Lett. 2018, 89, 1314–1322. [Google Scholar] [CrossRef]
  62. Nie, Z.; Wang, D.-J.; Jia, Z.; Yu, P.; Li, L. Fault model of the 2017 Jiuzhaigou Mw 6.5 earthquake estimated from coseismic deformation observed using Global Positioning System and Interferometric Synthetic Aperture Radar data. Earth Planets Space 2018, 70, 55. [Google Scholar] [CrossRef]
  63. Su, Z.; Yang, Y.-H.; Li, Y.-S.; Xu, X.-W.; Zhang, J.; Zhou, X.; Ren, J.-J.; Wang, E.-C.; Hu, J.-C.; Zhang, S.-M. Coseismic displacement of the 5 April 2017 Mashhad earthquake (Mw 6.1) in NE Iran through Sentinel-1A TOPS data: New implications for the strain partitioning in the southern Binalud Mountains. J. Asian Earth Sci. 2019, 169, 244–256. [Google Scholar] [CrossRef]
  64. Xu, G.; Xu, C.; Wen, Y.; Jiang, G. Source parameters of the 2016–2017 Central Italy earthquake sequence from the Sentinel-1, ALOS-2 and GPS data. Remote Sens. 2017, 9, 1182. [Google Scholar] [CrossRef]
  65. Moreno, M.; Li, S.; Melnick, D.; Bedford, J.; Báez, J.C.; Motagh, M.; Metzger, S.; Vajedian, S.; Sippl, C.; Gutknecht, B. Chilean megathrust earthquake recurrence linked to frictional contrast at depth. Nat. Geosci. 2018, 11, 285–290. [Google Scholar] [CrossRef]
  66. Bie, L.; Hicks, S.; Garth, T.; Gonzalez, P.; Rietbrock, A. ‘Two go together’: Near-simultaneous moment release of two asperities during the 2016 Mw 6.6 Muji, China earthquake. Earth Planet. Sci. Lett. 2018, 491, 34–42. [Google Scholar] [CrossRef]
  67. Hamling, I.J.; Hreinsdóttir, S.; Clark, K.; Elliott, J.; Liang, C.; Fielding, E.; Litchfield, N.; Villamor, P.; Wallace, L.; Wright, T.J. Complex multifault rupture during the 2016 Mw 7.8 Kaikōura earthquake, New Zealand. Science 2017, 356, eaam7194. [Google Scholar] [CrossRef] [PubMed]
  68. Papadopoulos, G.; Ganas, A.; Agalos, A.; Papageorgiou, A.; Triantafyllou, I.; Kontoes, C.; Papoutsis, I.; Diakogianni, G. Earthquake triggering inferred from rupture histories, DInSAR ground deformation and stress-transfer modelling: The case of central Italy during August 2016–January 2017. Pure Appl. Geophys. 2017, 174, 3689–3711. [Google Scholar] [CrossRef]
  69. He, P.; Hetland, E.A.; Niemi, N.A.; Wang, Q.; Wen, Y.; Ding, K. The 2016 Mw 6.5 Nura earthquake in the Trans Alai range, northern Pamir: Possible rupture on a back-thrust fault constrained by Sentinel-1A radar interferometry. Tectonophysics 2018, 749, 62–71. [Google Scholar] [CrossRef]
  70. Polcari, M.; Albano, M.; Atzori, S.; Bignami, C.; Stramondo, S. The causative fault of the 2016 Mwp 6.1 Petermann Ranges intraplate earthquake (Central Australia) retrieved by C-and L-band InSAR data. Remote Sens. 2018, 10, 1311. [Google Scholar] [CrossRef]
  71. Béjar-Pizarro, M.; Álvarez Gómez, J.A.; Staller, A.; Luna, M.P.; Pérez-López, R.; Monserrat, O.; Chunga, K.; Lima, A.; Galve, J.P.; Martínez Díaz, J.J. InSAR-based mapping to support decision-making after an earthquake. Remote Sens. 2018, 10, 899. [Google Scholar] [CrossRef]
  72. Qu, C.; Zuo, R.; Shan, X.; Hu, J.-c.; Zhang, G. Coseismic deformation of the 2016 Taiwan Mw6. 3 earthquake using InSAR data and source slip inversion. J. Asian Earth Sci. 2017, 148, 96–104. [Google Scholar] [CrossRef]
  73. Liu, Y.; Zhang, G.; Zhang, Y.; Shan, X. Source parameters of the 2016 Menyuan earthquake in the northeastern Tibetan Plateau determined from regional seismic waveforms and InSAR measurements. J. Asian Earth Sci. 2018, 158, 103–111. [Google Scholar] [CrossRef]
  74. Metzger, S.; Schurr, B.; Ratschbacher, L.; Sudhaus, H.; Kufner, S.K.; Schöne, T.; Zhang, Y.; Perry, M.; Bendick, R. The 2015 Mw7. 2 Sarez strike-slip earthquake in the Pamir interior: Response to the underthrusting of India’s western promontory. Tectonics 2017, 36, 2407–2421. [Google Scholar] [CrossRef]
  75. Avallone, A.; Cirella, A.; Cheloni, D.; Tolomei, C.; Theodoulidis, N.; Piatanesi, A.; Briole, P.; Ganas, A. Near-source high-rate GPS, strong motion and InSAR observations to image the 2015 Lefkada (Greece) Earthquake rupture history. Sci. Rep. 2017, 7, 10358. [Google Scholar] [CrossRef] [PubMed]
  76. Ainscoe, E.; Elliott, J.; Copley, A.; Craig, T.; Li, T.; Parsons, B.; Walker, R. Blind thrusting, surface folding, and the development of geological structure in the Mw 6.3 2015 Pishan (China) earthquake. J. Geophys. Res. Solid Earth 2017, 122, 9359–9382. [Google Scholar] [CrossRef]
  77. Castaldo, R.; De Novellis, V.; Solaro, G.; Pepe, S.; Tizzani, P.; De Luca, C.; Bonano, M.; Manunta, M.; Casu, F.; Zinno, I. Finite element modelling of the 2015 Gorkha earthquake through the joint exploitation of DInSAR measurements and geologic-structural information. Tectonophysics 2017, 714, 125–132. [Google Scholar] [CrossRef]
  78. Guangcai, F.; Zhiwei, L.; Xinjian, S.; Bing, X.; Yanan, D. Source parameters of the 2014 Mw 6.1 South Napa earthquake estimated from the Sentinel 1A, COSMO-SkyMed and GPS data. Tectonophysics 2015, 655, 139–146. [Google Scholar] [CrossRef]
  79. Panuntun, H. Geodetic slip model of the November 26, 2019 Albania earthquake estimated from Sentinel-1 TOPS interferometry. Tectonophysics 2021, 807, 228814. [Google Scholar] [CrossRef]
  80. Wang, K.; Bürgmann, R. Co-and early postseismic deformation due to the 2019 Ridgecrest earthquake sequence constrained by Sentinel-1 and COSMO-SkyMed SAR data. Seismol. Res. Lett. 2020, 91, 1998–2009. [Google Scholar] [CrossRef]
  81. Bjelotomić Oršulić, O.; Markovinović, D.; Varga, M.; Bašić, T. Coseismic ground displacement after the Mw6. 2 earthquake in NW Croatia determined from Sentinel-1 and GNSS CORS Data. Geosciences 2021, 11, 170. [Google Scholar] [CrossRef]
  82. Li, B.; Li, Y.; Jiang, W.; Su, Z.; Shen, W. Conjugate ruptures and seismotectonic implications of the 2019 Mindanao earthquake sequence inferred from Sentinel-1 InSAR data. Int. J. Appl. Earth Obs. Geoinf. 2020, 90, 102127. [Google Scholar] [CrossRef]
  83. He, P.; Wen, Y.; Chen, Y.; Xu, C.; Ding, K. Coseismic rupture geometry and slip rupture process during the 2018 Mw 7.1 Anchorage, south-central Alaska earthquake: Intraplate normal faulting by slab tear constrained by geodetic and teleseismic data. Earth Space Sci. 2020, 7, e2019EA000924. [Google Scholar] [CrossRef]
  84. Wang, S.; Jiang, G.; Weingarten, M.; Niu, Y. InSAR evidence indicates a link between fluid injection for salt mining and the 2019 Changning (China) earthquake sequence. Geophys. Res. Lett. 2020, 47, e2020GL087603. [Google Scholar] [CrossRef]
  85. Yang, J.; Xu, C.; Wang, S.; Wang, X. Sentinel-1 observation of 2019 Mw 5.7 Acipayam earthquake: A blind normal-faulting event in the Acipayam basin, southwestern Turkey. J. Geodyn. 2020, 135, 101707. [Google Scholar] [CrossRef]
  86. Wibowo, S.B.; Hadmoko, D.S.; Isnaeni, Y.; Farda, N.M.; Putri, A.F.S.; Nurani, I.W.; Supangkat, S.H. Spatio-temporal distribution of ground deformation due to 2018 lombok earthquake series. Remote Sens. 2021, 13, 2222. [Google Scholar] [CrossRef]
  87. He, P.; Wen, Y.; Xu, C.; Chen, Y. Complete three-dimensional near-field surface displacements from imaging geodesy techniques applied to the 2016 Kumamoto earthquake. Remote Sens. Environ. 2019, 232, 111321. [Google Scholar] [CrossRef]
  88. Vaka, D.S.; Rao, Y.; Singh, T. Surface deformation of the 2019 Mirpur earthquake estimated from Sentinel-1 InSAR data. In Proceedings of the 2020 IEEE India Geoscience and Remote Sensing Symposium (InGARSS), Virtual, 1–4 December 2020; pp. 130–133. [Google Scholar]
  89. Marconato, L.; Leloup, P.-H.; Lasserre, C.; Caritg, S.; Jolivet, R.; Grandin, R.; Cavalié, O.; Métois, M.; Audin, L. Insights on fault reactivation during the 2019, Mw4. 9 Le Teil earthquake in southeastern France, from a joint 3D geology and InSAR study. In Proceedings of the EGU General Assembly Conference Abstracts, Online, 19–30 April 2021; p. EGU21-4388. [Google Scholar]
  90. Yu, P.; Qiao, X.; Xiong, W.; Chen, W.; Nie, Z.; Wang, D.; Jia, Z.; Li, J. Source model for the M w 6.0 earthquake in Jiashi, China on 19 January 2020 from Sentinel-1A InSAR data. Earth Planets Space 2020, 72, 169. [Google Scholar] [CrossRef]
  91. Xu, W. Finite-fault slip model of the 2016 Mw 7.5 Chiloé earthquake, southern Chile, estimated from Sentinel-1 data. Geophys. Res. Lett. 2017, 44, 4774–4780. [Google Scholar] [CrossRef]
  92. Tian, Y.; Liu-Zeng, J.; Luo, Y.; Li, Y.; Zhang, J. Deformation Related to an M < 5 Earthquake Sequence on Xiangyang Lake-Burog Co Fault in Central Xizang, China, Observed by Sentinel-1 Data. Bull. Seismol. Soc. Am. 2018, 108, 3248–3259. [Google Scholar]
  93. Bacques, G.; de Michele, M.; Foumelis, M.; Raucoules, D.; Lemoine, A.; Briole, P. Sentinel optical and SAR data highlights multi-segment faulting during the 2018 Palu-Sulawesi earthquake (Mw 7.5). Sci. Rep. 2020, 10, 9103. [Google Scholar] [CrossRef] [PubMed]
  94. Bignami, C.; Antonioli, A.; Atzori, S.; Kiratzi, A.; Polcari, M.; Svigkas, N.; Tolomei, C. Mapping InSAR deformation of low and moderate earthquakes. Procedia Comput. Sci. 2021, 181, 214–219. [Google Scholar] [CrossRef]
Figure 2. Flowchart showing the major steps for our displacement analysis and the names of LOS displacement resulting from each step.
Figure 2. Flowchart showing the major steps for our displacement analysis and the names of LOS displacement resulting from each step.
Remotesensing 16 02019 g002
Figure 3. Example of TS displacement fields from time-series analysis on InSAR pairs from descending track T56. (a) LOS cumulative displacement from the start of the descending track T56 dataset (11 March 2020) to 2 August 2020, just after the event on 30 July (LASN-located epicenter denoted with a triangle). The TS deformation field is overlaid on topography from an SRTM DEM. (b) Full TS displacement fields for the descending track analysis. Geographic coverage and color scale as in (a).
Figure 3. Example of TS displacement fields from time-series analysis on InSAR pairs from descending track T56. (a) LOS cumulative displacement from the start of the descending track T56 dataset (11 March 2020) to 2 August 2020, just after the event on 30 July (LASN-located epicenter denoted with a triangle). The TS deformation field is overlaid on topography from an SRTM DEM. (b) Full TS displacement fields for the descending track analysis. Geographic coverage and color scale as in (a).
Remotesensing 16 02019 g003
Figure 4. Average DL-derived LOS displacement rate fields for ascending track T49 and descending track T56. The Nacimiento Fault site is denoted with a black square, the LASN-located epicenter is denoted with a black star, and faults from the USGS are denoted with black lines.
Figure 4. Average DL-derived LOS displacement rate fields for ascending track T49 and descending track T56. The Nacimiento Fault site is denoted with a black square, the LASN-located epicenter is denoted with a black star, and faults from the USGS are denoted with black lines.
Remotesensing 16 02019 g004
Figure 5. Best-fitting model for NF Site using two Okada sources, as fitted on DL-derived average displacement rate fields from ascending track T49 and descending track T56. Pictured is the observed average displacement rate field from descending track T56, the modeled displacement rate field derived from the best-fitting model parameters for both Okada sources (O1 and O2), and the residual field (observed minus modeled).
Figure 5. Best-fitting model for NF Site using two Okada sources, as fitted on DL-derived average displacement rate fields from ascending track T49 and descending track T56. Pictured is the observed average displacement rate field from descending track T56, the modeled displacement rate field derived from the best-fitting model parameters for both Okada sources (O1 and O2), and the residual field (observed minus modeled).
Remotesensing 16 02019 g005
Figure 6. Results of temporal adjustment on NF Okada 1 slip estimates occurring at ~207 m depth using f 3 t parameterization. Modeled cumulative slip is shown with the black line, with corresponding 68% confidence intervals indicated by the dashed black lines. Temporal breaks in the parameterization are indicated with dashed, vertical green lines. Red segments represent pairwise measurements of slip with black dots distinguishing individual epochs per pair. Vertical blue bars indicate 1 σ uncertainty in the pairwise slip measurements post-scaling using the square root of the variance.
Figure 6. Results of temporal adjustment on NF Okada 1 slip estimates occurring at ~207 m depth using f 3 t parameterization. Modeled cumulative slip is shown with the black line, with corresponding 68% confidence intervals indicated by the dashed black lines. Temporal breaks in the parameterization are indicated with dashed, vertical green lines. Red segments represent pairwise measurements of slip with black dots distinguishing individual epochs per pair. Vertical blue bars indicate 1 σ uncertainty in the pairwise slip measurements post-scaling using the square root of the variance.
Remotesensing 16 02019 g006
Figure 7. Results of temporal adjustment on NF Okada 2 slip estimates occurring at ~370 m depth using f 3 t parameterization. Plotting conventions as in Figure 6.
Figure 7. Results of temporal adjustment on NF Okada 2 slip estimates occurring at ~370 m depth using f 3 t parameterization. Plotting conventions as in Figure 6.
Remotesensing 16 02019 g007
Figure 8. Comparison of LOS crustal displacement and moment magnitudes studied using Sentinel-1 InSAR pairs. The red circle represents the Gallina earthquake targeted for this study, which is an order of magnitude smaller in LOS displacement than typical studies. Associated references used to obtain this information are provided in Appendix F (Table A1).
Figure 8. Comparison of LOS crustal displacement and moment magnitudes studied using Sentinel-1 InSAR pairs. The red circle represents the Gallina earthquake targeted for this study, which is an order of magnitude smaller in LOS displacement than typical studies. Associated references used to obtain this information are provided in Appendix F (Table A1).
Remotesensing 16 02019 g008
Table 1. Best-fitting Okada model parameters for the NF Site. Following standard Okada conventions [40,41], dip is measured downward relative to the top edge of the fault plane, strike is measured clockwise from North, and rake is measured in-plane and counterclockwise from the strike.
Table 1. Best-fitting Okada model parameters for the NF Site. Following standard Okada conventions [40,41], dip is measured downward relative to the top edge of the fault plane, strike is measured clockwise from North, and rake is measured in-plane and counterclockwise from the strike.
ParameterEstimateUncertainty
Okada 1
(O1)
Length [m]1668.8484.4
Width [m]210.893.8
Centroid Depth [m]207.046.9
Dip [°]79.220.6
Strike [°]192.95.6
Centroid Easting [m]328,415.4150.0
Centroid Northing [m]4,008,606.1500.0
Rake [°]−74.119.4
Okada 2
(O2)
Length [m]146.4281.3
Width [m]228.0 67.5
Centroid Depth [m]370.562.5
Dip [°]49.66.6
Strike [°]216.44.4
Centroid Easting [m]328,856.741.3
Centroid Northing [m]4,008,818.762.5
Rake [°]−107.517.5
Table 2. Results from temporal adjustment on slip estimated at NF site from the best-fitting Okada 1 (O1) and Okada 2 (O2) sources. O1 is associated with slip at a centroid depth of ~207 m and O2 is associated with slip at a centroid depth of ~370 m.
Table 2. Results from temporal adjustment on slip estimated at NF site from the best-fitting Okada 1 (O1) and Okada 2 (O2) sources. O1 is associated with slip at a centroid depth of ~207 m and O2 is associated with slip at a centroid depth of ~370 m.
ParameterizationPlanar Slip Rate at Depth [cm/yr] χ
17 March to 1 July1 July to 30 July30 July to 25 August
O1 f 1 t 6.9 ± 1.2 4.3
f 2 t 8.5 ± 0.5 30.3 ± 5.1 1.6
f 3 t 10.1 ± 0.1 3.8 ± 4.4 17.0 ± 6.1 1.2
O2 f 1 t 6.0 ± 0.9 3.7
f 2 t 6.1 ± 1.1 3.6 ± 10.9 3.8
f 3 t 2.0 ± 1.3 36.5 ± 7.9 29.4 ± 10.8 2.4
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Reinisch, E.C.; Abolt, C.J.; Swanson, E.M.; Rouet-Leduc, B.; Snyder, E.E.; Sivaraj, K.; Solander, K.C. Advancing the Limits of InSAR to Detect Crustal Displacement from Low-Magnitude Earthquakes through Deep Learning. Remote Sens. 2024, 16, 2019. https://doi.org/10.3390/rs16112019

AMA Style

Reinisch EC, Abolt CJ, Swanson EM, Rouet-Leduc B, Snyder EE, Sivaraj K, Solander KC. Advancing the Limits of InSAR to Detect Crustal Displacement from Low-Magnitude Earthquakes through Deep Learning. Remote Sensing. 2024; 16(11):2019. https://doi.org/10.3390/rs16112019

Chicago/Turabian Style

Reinisch, Elena C., Charles J. Abolt, Erika M. Swanson, Bertrand Rouet-Leduc, Emily E. Snyder, Kavya Sivaraj, and Kurt C. Solander. 2024. "Advancing the Limits of InSAR to Detect Crustal Displacement from Low-Magnitude Earthquakes through Deep Learning" Remote Sensing 16, no. 11: 2019. https://doi.org/10.3390/rs16112019

APA Style

Reinisch, E. C., Abolt, C. J., Swanson, E. M., Rouet-Leduc, B., Snyder, E. E., Sivaraj, K., & Solander, K. C. (2024). Advancing the Limits of InSAR to Detect Crustal Displacement from Low-Magnitude Earthquakes through Deep Learning. Remote Sensing, 16(11), 2019. https://doi.org/10.3390/rs16112019

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop