Deformation of Alaskan Volcanoes Measured by Satellite Radar Interferometry

Jeffrey T. Freymueller
Geophysical Institute
University of Alaska, Fairbanks

Summary

Our work this year was hampered significantly by a lack of data. This project will primarily use data from ERS-1 and ERS-2, so the delay was due to the Alaska SAR Facility (ASF) system upgrade. Our funding began officially on April 1 1996, and actually arrived May 21, 1996, just before ASF stopped distributing data while it began what would turn out to be a nearly 6 month system upgrade. On December 6, we finally received our first data requested during this project, a pair of ERS-1 images.

Despite the lack of new data, we have been able to make useful progress. Using existing data from the Katmai area, we investigated the maintenance of phase coherence over time for several different types of surface materials typical of volcanic environments. Environmental conditions in the Katmai area are much more hostile than conditions in Southern California from the point of view of maintaining phase coherence for Interferometric Synthetic Aperture Radar (INSAR), due to seasonal snow and ice, repeated freezing and thawing of the ground surface, and rapid erosion during the spring melt. We find that there are significant seasonal variations in phase coherence for interferograms with a relatively short time interval between passes (35 days), with phase coherence being highest during the summer months (July to September) and declining rapidly on either side of summer. We find that the best surface material overall is fresh lava, which maintains acceptable coherence for at least three years even in this hostile environment. The next best surface material is either weathered lava or stream deposits, depending on the time of year. Sites where the surface materials are composed of loose ash lose coherence very quickly, and within a matter of months there is no coherent signal left.

These results suggest that long-term monitoring of volcanoes that are subject to seasonal snow cover is feasible, although phase coherence is not likely to be maintained over a broad area unless the surface conditions are ideal. In order to maintain coherence over a broad area, a satellite should make a usable pass over the volcano at least every two or three months in the summer, every two or three weeks in spring and fall, and every two or three days in the winter.

Tasks Completed

1. Katmai inflation manuscript in review
Using SAR images for the Katmai group of volcanoes, Alaska, we were able to measure a small area of uplift near the New Trident vent on Trident volcano. An area about 3 km by 3 km in size was uplifted by about 50-70 mm over a two year period from 1993-1995 (Figure 1). Phase coherence was lost outside of this area, so we cannot be sure if the area of uplift is confined to this area. The observed uplift can be explained by a shallow point inflation source. We have completed a manuscript, which is in final review for publication in Geophysical Research Letters.

2. Phase Coherence study
We evaluated the phase coherence of 58 interferograms with time separations ranging from three days to three years. Rather than attempting to evaluate coherence over the entire image, we selected five specific areas where the surface material was known based on field geological experience. The target areas were selected prior to forming the interferograms. We computed a correlation coefficient for each site for each pair of images. The correlation coefficient measures the extent to which the radar signal is consistent from image to image. The interferometric SAR echoes will be correlated with each other if the backscattering characteristics remain unchanged between the two images. This kind of systematic study of phase coherence has not been done before, especially not for volcanic materials. Interferometric coherence is significant because: 1) it determines the precision of topographic or ground surface deformation maps made from an interferogram; 2) it could be a key channel for image classification for land and forest mapping; 3) it determines the feasibility of applying interferometric SAR techniques to geophysical studies; 4) it guides the planning of future SAR missions, by controlling how often a satellite must pass over a site to make future measurements.

We analyzed the interferograms constructed for the Katmai volcano group, Alaska. We selected five sites with typical volcanic surfaces: fresh lava, old weathered lava, ash with strong water re-working, ash with weaker water re-working, and stream deposits. We first investigated seasonal changes in coherence. We used a series of image pairs acquired at different times of year, each pair with a 35 day time interval between images. For all of the sites, we found that coherence is highest for images acquired between the middle of July and the end of September, and phase coherence decreases rapidly outside of this time interval. After the end of November, coherence was completely lost for images separated by 35 days, presumably because of continuous snow cover.

In general, fresh lava has the highest coherence, followed by either weathered lava or stream deposits (surprisingly, these seem to have higher coherence than weathered lava outside of the peak period). The sites where the surface material is ash deposited in 1912 show the poorest coherence at all times. Using images acquired exclusively during the peak of coherence in the summer, we studied the coherence as a function of the time separation between the images. We applied a theoretical correction for the degradation of coherence with increasing baseline length to remove the effect of varying baseline separation. Over periods varying from 35 days to three years, we found that phase coherence for fresh lava and weathered lava decays more slowly than the other three sites. The decrease of coherence can be fit well by an exponential decay with time, with a time constant of order of years. For the best surface materials, we find acceptable phase coherence for image pairs separated by as much as three years. However, for the worst materials phase coherence is degraded within a matter of months to a point where meaningful signals are unlikely to be obtained (Figure 2).

These results suggest that long-term monitoring of volcanoes that are subject to seasonal snow cover is feasible, although phase coherence is not likely to be maintained over a broad area unless the surface conditions are ideal. In order to maintain coherence over a broad area, a satellite must make a usable pass over the volcano at least every two or three months in the summer, and every two or three weeks in spring and fall, and every two or three days in the winter.

Future Plans

1. Deformation associated with the earthquake swarm of March 1996, Akutan volcano.
Akutan volcano, situated in the west-central part of Akutan Island in the east Aleutian Islands, was struck by an intense earthquake swarm on March 10-14, 1996. The largest event in the swarm was a magnitude 5.1 on March 10, and in the first days of the swarm as many as 3000 earthquakes were felt per day. Most of the recorded earthquakes were volcano-tectonic events located to the east of the summit. Alaska Volcano Observatory (AVO) scientists observed fresh northwest-southeast trending en-echelon cracks on both the northwest and southeast flanks of the volcano. The cracks are up to 1 m wide. They are very fresh and are presumed to have formed concurrently with the earthquake swarm. An eight station Global Positioning System (GPS) network was measured with support from AVO in July and August 1996, concurrently with the installation of a permanent seismic monitoring network.

We will measure the ground surface deformation associated with the March 1996 earthquake swarm and ground surface cracking. We have ordered data from ASF from the ERS-1/ERS-2 tandem mission and other data that should allow us to measure the ground surface deformation associated with the ground cracks. Based on interferograms constructed from other data from Akutan, we expect to have acceptable phase coherence over the 70 day time interval of the passes that span the swarm. The anticipated signal would be about 30 fringes based on the ground observations. Using this data we can study the processes at depth that cause the ground cracks to form. We will also study the long-term deformation of this volcano to see if there was any observable deformation that preceded the swarm, or continued afterward.

2. Deformation prior to and during the eruption of 1996, Pavlof volcano.
Pavlof Volcano, located near the southwest end of the Alaska Peninsula, has been since September, 1996. Pavlof is the most active volcano in Alaska, with 40 eruptions since 1760. Seismic monitoring was conducted by Lamont-Doherty Geological Observatory from 1973-1990, and a new network was installed by AVO in the summer of 1996. No previous deformation measurements exist for Pavlof. Eruptions at Pavlof, including the current one, have had only very subtle seismic precursors, so INSAR measurements could be useful for forecasting eruptions if the volcano deforms measurably prior to eruptions. INSAR also offers the possibility of providing constraints on the plumbing system and helping to elucidate the source region for its highly periodic eruptions.

Details of the recent activity of Akutan and Pavlof volcanoes can be found on the AVO web site.

Figure 1. Motion-only interferogram for New Trident vent constructed using ERS-1 images. The area shown is about 3 km by 3 km. One fringe (three color band of red-green-blue) in the interferogram represents 28.3 mm difference in distance from satellite antenna to ground surface between two observations during 1993 and 1995. The observed 70 to 90 mm uplift at New Trident vent over two years can be explained by shallow inflation source located within 2 km of the surface. This inflation source could be a small magma body. Inflation could be due to the arrival of new magma near the surface, or expansion of gases.

<figure 2 not yet available>

Figure 2. Degradation of phase coherence, measured by the correlation coefficient between the two images, over time for five different surface materials. The correlation coefficients are corrected for the decorrelation due to baseline length, so the corrected correlation coefficients all refer to an ideal zero baseline separation. The correlation coefficient for a 200 meter baseline separation, for example, would be about 80% of the value shown. Based on our experience with Katmai interferograms, a correlation coefficient of about 0.3 or higher is required for an useful interferogram.