Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal MTH0768

TitleMonitoring Critical Infrastructure with TerraSAR-X DifInSAR: The Howard Hanson Dam
Investigator Jones, Cathleen - Jet Propulsion Laboratory, Radar Science and Engineering Section
Team Member
Dr. Hensley, Scott - Jet Propulsion Laboratory, Radar Science and Engineering
SummaryWe propose to collect and analyze repeat pass TerraSAR-X data over the Howard Hanson Dam in Washington State, USA, for nine months beginning in Sept. 2010 and extending though the end of the potential winter/spring flood period in 2011. During this time, data will be collected in High Resolution Single Polarization (HH) Spotlight mode every 22 days. The SLC data product will be used at JPL to generate repeat pass interferograms. Funding for the TerraSAR-X data is available through a NASA-funded grant to Dr. Cathleen Jones, entitled "Radar Imaging of the Howard Hanson Dam for Change Detection."
Final ReportThe Howard A. Hanson Dam on the Green River in Washington State has had a history of seepage along the right embankment since it first came into operation in 1962 (Smith et al. 2010a). Although the earthen dam’s foundation and left abutment are founded on bedrock, the topography of the region precluded similar founding of the right abutment, which instead is founded on a combination of bedrock and partially unconsolidated alluvial and landslide material (Smith et al., 2010a, 2010b). Following record pool levels in January 2009, two depressions developed on the upstream face of the right abutment and turbid water discharge from the right abutment drainage tunnel was observed. In response, the dam rating was reclassified to 1 (unsafe) and mitigation measures were implemented in the latter half of 2009, including installation of a double grout curtain in the right abutment and additional drains in the existing drainage tunnel (Smith et al., 2010a, 2010b). Radar remote sensing is a standard tool for a wide variety of scientific investigations and practical applications such as measurements of subsidence/uplift, landslides, aquifer level change through the expression of surface deformation, and including, more recently, usage during the response to natural and technological disasters, such as the Deepwater Horizon oil spill. SAR has not gained wide usage within the broader end-user community that includes managers and engineers responsible for critical infrastructure because in most cases its imagery, which does not look like a photograph, requires technical interpretation, as compared to images from optical sensors, whose easily generated, true-color products can often be interpreted by non-experts. A notable exception is the use of radar to identify water and oil slicks, which show up clearly as radar-dark areas in the imagery and does not require special interpretation. However, radar remote sensing has one overarching advantage over instruments operating in the optical through thermal bands that makes it of great use to the emergency response community: Radar can see through clouds and operate in any ambient light conditions, so it can be used to image the surface under all weather conditions, day or night, anywhere, any time of year. This advantage is so striking that radar’s other advantages are often overlooked, although they are in many ways as or more compelling. DINSAR can be used to measure surface deformation with sub-centimeter accuracy, detecting subsidence, motion along faults, and slow creep slumps, all from instruments collecting data across image swaths from tens to hundreds of kilometers wide. SAR remote sensing can inform ground-based dam and levee health assessment through identification of areas that changed between image collections. Change detection where new data is compared to a baseline data set is much simpler and more accurate than interpretation of a single radar scene to determine absolute surface conditions and furthermore allows advanced algorithms for quantitative measurement of surface deformation to be applied. Change detection algorithms can use either or both the amplitude and phase of the measured radar return signal, with the following general applications and limitations: 1. Changes in the radar signal return power • Detects changes in the type of surface, e.g., land to water (flooding) or water to mud flat (sediment transport), or major disruptions to the surface, e.g., digging or plowing. • Detects gross changes in surface properties, so generally will not be confused by changes in vegetation senescence or soil moisture. • Measurement of polarization-dependent changes improves surface differentiation. 2. Changes in the correlated amplitude and phase of the radar return (interferometric coherence) • Sensitive to more subtle changes in the scattering surface and changes at scales smaller than the radar resolution, i.e., vegetation growth, precipitation effects. • Indicates whether the surface has changed, but does not indicate the cause of the change. Combine with 1. and 3. to quantify the nature of the change. 3. Changes in the interferometric phase of the radar return (DINSAR) • Provides highly accurate measurement of surface movement under conditions where the surface’s physical and dielectric properties do not change much. • Is sensitive to small changes in soil moisture, which change the depth to which the radiation penetrates below the surface. • Combine with high correlation (2 above) to identify areas where accurate measurement of deformation can be made. All of these types of change detection could be of value to infrastructure health assessment. The project to study the Howard A. Hanson Dam with UAVSAR and TerraSAR-X was undertaken to explore the use of state-of-the-art radar instruments and processing techniques to monitor dams for change following major repair. We have evaluated the potential utility of high resolution SAR for change detection using the three methods outlined above through their application to a baseline UAVSAR data set collected during the grout wall installation and a second image collected a year later, and to a TerraSAR-X data series covering the winter-to-spring interval overlapping the last UAVSAR acquisition. For this study, we acquired ~15 TerraSAR-X scenes of the Howard A. Hanson Dam on the Green River in Washington state, USA, with 11 or 22 day intervals between acquisitions. The first acquisition, which was concurrent with the 2nd UAVSAR acquisition over the site, occurred on November 9, 2010, with acquisitions of TerraSAR-X imagery continuing through the winter and spring until the end of the spring 2010 fill of the pool upstream of the dam. TerraSAR-X has an image ground resolution of 2-3 m, which is comparable to the UAVSAR single-look resolution but the shorter wavelength causes much more rapid temporal decorrelation between repeat pass pairs. We collected pairs at 11 day and 22 day intervals to quantify the effect of temporal decorrelation in the environment of this particular dam and during the winter-to-spring time period, which is typically when the dam experiences its maximum load. The TerraSAR-X data collected over the dam was acquired on an ascending orbit and at a look angle of ~38°. This look direction views the downstream face of the dam nearly directly and has a reasonable amplitude return from both the upstream and downstream sides. The TerraSAR-X scenes were processed to determine how well the dam face retained phase coherence between November 2010 and May 2011. The minimum time between repeats for TerraSAR-X is 11 days, during which all phase coherence is lost in the surrounding forests near the dam, but both sides of the dam show high correlation (>= 0.5) as do most of the road surfaces in the area. This is an example of how more frequent repeat intervals can compensate for phase noise from other sources, making it possible in some cases to measure the health of a slope that is not viewed with optimal geometry. Although reducing the time interval between imaging helps, it cannot compensate for all imaging geometries or all decorrelation mechanisms, particularly for short wavelength radar instruments for which coherence is more rapidly lost. When forming an interferogram from the Nov. 9, 2010 data set and the May 4, 2011, data sets (176 days temporal baseline) only the downstream face retains high enough correlation to enable deformation measurements using DINSAR. Retaining high correlation appears to depend upon the season, however, because a 22 day interferogram formed from data collected on Feb. 16, 2011, and March 10, 2011 showed much poorer coherency, probably because of precipitation during that time interval or changes in snow cover. Satellite SAR monitoring of dams can be very cost effective compared to deploying aircraft to cover the same area and can be used to reliably obtain data at a regular (and relatively short – 1-2 week) time interval. This study showed that the 11-day repeat cycle is adequate for dam monitoring even in poor weather conditions. For long-term health monitoring, satellite remote sensing is the more economic option. The most significant disadvantage of satellite SAR is that there is little choice in the viewing geometry beyond some choice in the look angle, generally in the range from 20°-50°, with the look angle often tied to the specific operating mode of the instrument. The orbit inclination is not selectable and the user has only the option to image on an ascending or descending orbit. This limitation is not so significant for POLSAR studies, but can have a drastic effect on the ability to monitor motion along a failure direction using DifInSAR, since only movement along the radar line-of-sight can be measured. The satellite orbits are generally 5°-10° from polar, so structures are viewed from the east or from the west. This means that DInSAR imaging of a north or south-facing slope will have only minor sensitivity to failure along the gravity gradient. The viewing geometry limitation means that some dams, depending upon their orientation, cannot be monitored with satellite-based SARs in such a way that takes full advantage of the capabilities of the instruments, particularly displacement measurement. Airborne SARs do not have this limitation so optimal viewing geometry can be achieved. A very practical solution that will facilitate coherent imaging with satellite SAR of linear structures such as dams and levees that have a definite preferred direction for failure is to install hard radar targets on the face that can be oriented to maximum backscatter. We are preparing a follow-on study that will do this.

Back to list of proposals

© DLR 2004-2016