Printable Version

Earthquake Ground Movements

By Marius Necsoiu, Ph.D., David A. Ferrill, Ph.D. and Kevin J. Smart, Ph.D.


Dr. Kevin J. Smart, left, is a structural geologist with cross-training in computational solid mechanics. Dr. Marius Necsoiu, center, is an environmental scientist specializing in remote sensing and Geographic Information Systems. Dr. David Ferrill, right, is a structural geologist and director of the Department of Earth, Material and Planetary Sciences within SwRI’s Geosciences and Engineering Division.


When a damaging earthquake strikes, ground surface displacement is often evident close to the earthquake’s epicenter and along the trace of the fault on which the earthquake occurred. There, the movement along geologic faults can create violent vertical and horizontal disruptions strong enough to topple buildings, destroy homes, open fissures in the ground and misalign highways, railroads and pipelines. However, more subtle displacements also can occur before, during and after an earthquake, at significant distances from the epicenter.

Detecting and measuring these subtle displacements, or monitoring known danger zones for signs of future seismic activity, is limited by the sensitivity of the data-gathering equipment and the sophistication of the analytical and mapping techniques applied. Armed with improved technologies, academic researchers, government agencies and industries can measure or monitor subtle changes in critical areas by generating sensitive before-and-after comparisons of earthquake zones, as well as other areas that experience ground deformation. Such areas could include volcanoes, landslides, glaciers and subsidence zones related to hydrocarbon production, groundwater withdrawal or underground mining.

Building on established satellite-based remote sensing methodology, and supported by internal research and development funding, a team of scientists at Southwest Research Institute (SwRI) has refined synthetic aperture radar-based interferometry (InSAR) methodology to produce line-of-sight displacement and gradient maps with sensitivity to within a few centimeters. The methodology was applied to a series of image pairs acquired by the European Remote Sensing satellites (ERS-1 and ERS-2) for an area near Barstow, California, before and after the 1992 Landers earthquake. The first major application of InSAR to characterize deformation patterns related to an earthquake was performed soon after the Landers earthquake by a team of French researchers from Centre National d’Etudes Spatiales (CNES), led by Didier Massonnet. The research history for this earthquake and the arid setting make this earthquake and its related deformation pattern an ideal opportunity to test refined analytical techniques to expand the utility of InSAR methods. Supported by internal funding, the SwRI team used InSAR analysis to measure the deformation pattern that resulted directly from the Landers earthquake, including near-term aftershocks (independently confirming results of published InSAR investigations). Next, they compared pairs of radar images taken over long (multiple months to years) time intervals to map far-field and near-fault deformations for several years after the Landers earthquake. Finally, the team used displacement gradient analysis to identify and map earthquake-related ground ruptures with displacements as small as 2 cm.

The results demonstrate the tremendous utility of InSAR-based displacement mapping to identify and quantify subtle or otherwise difficult to recognize fault ruptures at the ground surface. This research also demonstrated that long-term deformation patterns surrounding the Landers earthquake continued to evolve for several years after the earthquake, which has important implications for differentiating between elastic (recoverable) strain and permanent deformation accommodated by smaller-scale faulting, fracturing, folding, and grain-scale deformation in rock and sediment.

The Landers Earthquake

The Landers earthquake, with a magnitude of 7.3, occurred April 24, 1992 in the southern Mojave Desert (Figure 1). Field investigations conducted by researchers from the California Institute of Technology and the California Geological Survey showed that maximum right-lateral strike-slip displacement on the Landers surface rupture exceeded six meters, and that this surface displacement maximum occurred approximately 40 kilometers (25 miles) north of the epicenter. The more than 70-km (43.5-mile)-long Landers earthquake rupture included portions of five previously mapped but separate faults that were connected at the surface by new or previously unmapped fault segments. The Landers earthquake also triggered smaller earthquakes as far away as Yellowstone National Park in Wyoming, far beyond the immediate aftershock area.


Figure 1. Map of the Landers earthquake area. The epicenter of Landers earthquake is shown in red. The yellow rectangle represents the boundaries of the April 24, 1992, ERS-1 dataset.


Furthermore, the Landers earthquake served as an optimal subject for interferometric coverage for two reasons. First, it was well covered by Synthetic Aperture Radar (SAR) images as well as global positioning systems (GPS) and other data. Second, the low humidity of the Mojave Desert made for an excellent site for InSAR analysis because low rainfall and scant vegetative cover create an ideal surface for retaining interferometric coherence over time. Such high coherence enables researchers to probe the shortest wavelengths in the interferometric phase to reveal details of the rupture.

The timing of the Landers earthquake was fortuitous, occurring less than three months after ERS-1 began acquiring radar images. Moreover, the precision of orbit determination for both ERS-1 and ERS-2 is optimal over North America because of the accurate tracking made possible by satellite laser ranging stations. These orbits were used in the InSAR process to precisely register the pairs of SAR data.

Techniques for Measuring Subsidence

Rates and directions of crustal deformation can be measured using a variety of geodetic methods, ranging from a simple spirit-level for precisely measuring vertical displacements over a comparatively small area, to GPS for somewhat less precise measurement of both vertical and horizontal displacements over much larger areas. Measurements are spatially limited in general, representing scattered points over broad regions. InSAR provides spatially dense coverage (such as hundreds to millions of samples per survey) and multiple measurements over extended time periods. It measures component displacement within line-of-sight range with a resolution of 5-10 millimeters.

The roots of the InSAR methodology were developed for topographic mapping based on ground-breaking research by Leroy C. Graham under the joint sponsorship of the U.S. Air Force Wright Air Development Center and the U.S. Army Engineer Topographic Laboratories in the 1970s, and subsequent work by a team of researchers including Andrew K. Gabriel, Richard M. Goldstein and Howard A. Zebker at Jet Propulsion Laboratory in the late 1980s. Since 1989, numerous studies have applied the technique to detect changes in the ground surface by removing the topographic component from the radar phase signal. The conventional InSAR technique combines two SAR complex images of a given area to form an interferogram that contains both magnitude and phase channels. The magnitude information corresponds to the intensity of the radar signal reflected back to the antenna from a given ground pixel. Of greatest interest to interferometry is the phase channel corresponding to the fractional wavelength of the echo where each pixel value represents the difference between the phases of the corresponding pixels in the two co-registered SAR images. This phase is expressed as the sum of the contributions by deformation, topographic, atmospheric, and orbital and phase-noise (i.e., thermal disturbance and co-registration error) components of the differences between two SAR images of the same location. From those components, the deformation and topographic phases are used to derive InSAR products. The topographic phase can be estimated via an external digital elevation model (DEM) or by combining two interferograms of the same region (one interferogram, being acquired over a short time period to ensure there is no surface deformation). Once the topographic component is removed, the remaining phase is proportional to surface displacement, atmospheric, orbital and phase noise.

The final steps in generating displacement maps involve phase unwrapping and geocoding. The phase unwrapping calculates the integer coefficients needed for computing the line-of-sight displacement values. The geocoding process converts the resulting displacement map (that is, the unwrapped interferogram) to real world coordinates.

Applying the InSAR Technology

Pairs of SAR scenes with nearly exact repeating passes were needed for this project. The possible candidates within 100 km of the Landers epicenter were selected using the Display Earth Remote Sensing Swath Coverage for Windows (DESCW), public domain software available from the European Space Agency. The time range of interest was from April 24, 1992, just before the Landers earthquake, to October 16, 1999. The latter date was chosen so that deformation related to the Hector Mine earthquake could be avoided.

Methodology to produce the interferograms involved data initialization, co-registration, re-sampling, interferometric generation, phase unwrapping and maps generation. The design of an in-house processing algorithm provided a better understanding of the interferometry technique, the ability to incorporate and test new algorithms in the future, and a tool that potentially could be used on a spectrum of applications involving radar interferometry. The public-domain Delft Object-oriented Radar Interferometric Software (DORIS) with its modular structure provided the framework for developing the method. Other public domain utilities were used to visualize and analyze intermediate products, and commercial-off-the-shelf software was used for final interpolation and map analysis. The entire InSAR algorithm was managed and controlled by in-house developed scripts.

To evaluate the SwRI team’s results, ground GPS and electronic distance meter (EDM) survey data were compared to displacements obtained from InSAR pairs. In all, 17 points were available for the area of overlap between the field survey and the interferogram. For each point, the comparison of the GPS motion vectors in the direction of the projection on the ground of the radar sensor line-of-sight was calculated. The initial correlation coefficient between the GPS/EDM component and the horizontal displacement for observed radar motion was 0.92. The result was further improved by using a variable incidence angle based on the location of each pixel on the scene. The final correlation coefficient was 0.94.


Figure 2. Ground surface disruption is clearly visible looking northwest along the Emerson Fault segment of the Landers earthquake rupture.


Study Results and Future Applications

SAR images were analyzed for the months and years following the Landers event to detect possible long-term deformation. The fundamental pattern reflected by the April to August 1992 displacement map is the large-magnitude, right-lateral slip associated with the main Landers earthquake (Figure 2). This coseismic displacement demonstrates a broad region extending tens of km away from the fault with the area southwest of the Landers epicenter generally moving to the northwest and downward away from the satellite, and the area northeast of the rupture generally moving to the southeast and up toward the satellite (Figure 3A). The fault rupture represents a major discontinuity in displacement associated with the main Landers rupture surface. The SwRI team’s results also revealed that significant deformation continued more than a year after the earthquake. A few months after the Landers earthquake, small-magnitude displacements were still occurring in a broad region to the southwest of the Landers epicenter, and higher-magnitude displacements in a narrower zone extending approximately 5 km to either side of the Landers epicenter (Figure 3B). Nearly a year after the Landers earthquake, motion was still continuing with deformation localized in a zone extending 5 km on either side of the Landers epicenter (Figure 3C). These post-Landers displacement maps illustrate continued deformation that likely reflects permanent deformation in response to stress field perturbation produced by the Landers earthquake.


Figure 3. This three-image series illustrates displacement from the Landers earthquake at the time of the event (A) and at intervals afterward (B and C). Higher values represent line-of-sight displacements in the direction to the satellite (i.e., red = up). (A) shows displacements directly associated with the Landers earthquake. (B) shows deformation continuing two to seven months after the earthquake, probably reflecting permanent deformation in response to stress field perturbation produced by the earthquake. (C) shows that deformation appears to be ongoing nearly a year afterward, generally localized in a zone extending 5 km on either side of the Landers rupture.


More detailed analysis focused on an area east of Barstow, where displacement and displacement gradient, calculated from the line-of-sight displacement, demonstrated that the InSAR data recorded small displacement faults with displacements of 1 to 10 cm. Although many of these ruptures previously had been mapped in the field, other ruptures either had not been mapped or did not have measured displacements (Figure 4). The InSAR technique appears very promising for use in identifying locations and orientations of previously unmapped fault ruptures.

Because the InSAR maps record relative displacement changes on the order of centimeters, the distribution of slip can be accurately mapped along a selected fault from the area of maximum slip to near the tip. In addition to a better understanding of the relationship between fault propagation and displacement accumulation, our analyses provided insights into slip rates, including both spatial and temporal variations, along several individual faults; geometry and kinematics of fault linkage and relationships between slip rates for nearby faults; and the relationship between fault growth and earthquake recurrence intervals.


Figure 4. Image (A) acquired from Landsat 5 TM (RGB 743) on May 7, 1990 shows mapped traces of small fault ruptures associated with the Landers earthquake. The resolution of satellite data was enhanced by fusion with an orthophoto image. (B) Detail of the displacement gradient map. Gradiometric changes from cultural and agricultural features are outlined in white. Suspected fault ruptures are indicated by black ellipses. The pistachio orchard and center-pivot irrigation system shown below illustrate agricultural activities of the sort detected with remote sensing investigations in this analysis.


Conclusions

This project explored a technical area that is on the cutting edge of crustal deformation analysis and seismic hazard assessment. It represents a step forward in the capabilities of InSAR to map ground ruptures associated with fault slip. Depending on the application and data availability, this methodology could also use data from other radar satellites that monitor the environment, such the Japanese Earth Resources Satellite (JERS), RADARSAT, ENVIronment SATellite (ENVISAT) and the Advanced Land Observing Satellite (ALOS).

Based on experience gained through this project, the SwRI team is exploring future applications of InSAR technology that include refinements to techniques for monitoring subsidence related to hydrocarbon production for the oil and gas industry; analyzing ground subsidence and aquifer damage related to groundwater withdrawal for aquifer management authorities; analyzing and investigating earthquake-related damage or hazards; and monitoring active volcanoes and assessing their hazards for local, national, and international agencies and private companies. This technique also could be useful in making site decisions for construction of critical structures, such as power plants, in areas with both rapid and slow rates of subsidence or uplift.

Comments about this article? Contact Necsoiu at (210) 522-5541 or marius.necsoiu@swri.org.

Acknowledgments

Financial support for this research was provided by the Southwest Research Institute Advisory Committee on Research. The SAR SLC data used in this study were produced by the European Space Agency (ESA) and distributed by Eurimage. Landsat data are courtesy of Global Land Cover Facility, University of Maryland. The interferometric processing was performed using the freely available DORIS software package developed by the Delft Institute for Earth-Oriented Space Research (DEOS), Delft University of Technology. Our investigation strongly benefited from discussions with and information from Dr. Evelyn Price (Institute of Geophysics, University of Texas at Austin), Dr. Earl Hart (California Department of Mines and Geology, retired), Dr. Falk Amelung (University of Miami), Dr. Enrique Cabral (University of Miami/Instituto de Geofisica, Universidad Nacional Autonoma de Mexico), Dr. Paul Rosen (Jet Propulsion Laboratory), and Dr. Bert Kampes (DLR — The German Aerospace Center). We thank Dr. Alan Morris and Dr. Larry McKague for their suggestions and careful reviews of the manuscript.

Published in the Spring 2007 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.

Spring 2007 Technology Today
SwRI Publications SwRI Home