While many volcanoes experience periods of unrest and deformation that do not culminate in major eruptions, it is likely that many (perhaps most) of the world's volcanoes that actually do erupt experience significant preeruption surface deformation (Newhall and Dzurisin, 1988; Thatcher, 1990; Shimada et al., 1990). Thus, precise monitoring of surface deformation could lead to accurate predictions of volcanic eruption. However, very few of the world's active volcanoes are adequately monitored for ground deformation at the present time. Lack of resources is an important factor but, even when substantial resources are available, it is often difficult or impossible to acquire the needed information with available techniques without putting field crews at significant risk.
SAR interferometry has the potential to provide the requisite data. Detection of volcanic deformation at Mt. Etna with SAR interferometry (Massonnet et al., 1995) demonstrates that the technique can work at least under some circumstances. However, many questions remain before routine monitoring of the world's active volcanoes with SAR interferometry can be considered. In this chapter, we address how surface deformation data can be used for the very practical benefit of eruption prediction and to address even more fundamental scientific problems bearing on understanding magmatic systems. We also make some recommendations aimed at improving the utility of SAR interferometry for volcano studies.
The existence of a large surface-strain signal presaging eruption suggests that accurate measurements of surface strain can be inverted to study details of critical preeruption magmatic processes. Experience in Hawaii, Iceland, and Long Valley illustrates the "state of the art" in using high-precision geodetic data to track the motion of magma at depth. At basaltic volcanoes, such as Kilauea in Hawaii, decades of study have succeeded in mapping their shallow magmatic plumbing systems, including the location, size, and capacity of reservoirs only a few kilometers beneath the surface (e.g., Yang et al., 1992; Dvorak, 1993). By tracking the progressive tumescence of these reservoirs, volcanologists can anticipate most eruptions. Linde et al. (1993) document in great detail the evolution of an eruption at Hekla volcano in Iceland over several days. They used borehole strainmeters to track a packet of magma, detecting the deflation of a deep reservoir and the corresponding upward motion of the magma in a conduit, culminating in a surface eruption.
Reservoirs comparable to those detected in Hawaii and Iceland have not been identified beneath most other volcanoes, and we understand much less about their eruption processes compared to these classic basaltic systems. The more petrologically evolved volcanoes, such as calc-alkaline island arc volcanoes or caldera-forming continental rhyolitic centers, tend to generate explosive eruptions--the kind we would most like to predict. The andesitic-to-sili-ceous character of these volcanoes leads to higher viscosity magmas and higher concentrations of volatiles (e.g., water, carbon dioxide, and sulphur dioxide) making them much more prone to erupt catastrophically. At Long Valley, a modern eruption has not yet occurred, but leveling and trilateration over the last two decades have defined the locus of deformation associated with recent magma movement and permitted modeling of the size, shape, orientation, and depth range of the magmatic body including dyke-like intrusions coming from the main magma chamber (Savage and Cockerham, 1984; Denlinger and Riley, 1984; Rundle and Whitcomb, 1984; Savage, 1988; Vasco et al., 1988; Langbein et al., 1989, 1993). Newer GPS measure-ments have begun to provide high-accuracy three-dimensional measurement of surface defor-mation (Dixon et al., 1993). Continuous GPS measurements from two permanent stations installed in early 1993 and late 1994 on the resurgent dome are providing data relevant for rapid assessment of a volcanic hazard in the region (Mao et al., 1994; Dixon et al., 1995; Fig. 4.1).
Key Questions
At any active volcano, some key questions need to be addressed: Where is the magma today that will explode to the surface during the next eruption? How can its ascent be tracked to further our understanding of the magmatic plumbing system, the physical and chemical pro-cesses that result in eruptions, and provide advance warning that an eruption may be imminent? What causes temporal variations in strain and episodic eruptions? Is it mainly related to variable rates of magma recharging into the base of the magma chamber, or does it reflect episodic trapping and release of volatiles exsolving from the magma (e.g., Dzurisin et al., 1990). For most volcanoes, even for ones that are fairly well monitored, we are a long way from answering these questions.
A satellite-based SAR system holds a number of potential benefits over existing ground-based deformation-monitoring techniques for addressing some of these questions. First, preexisting ground control is not necessary for useful SAR measurements to be made. Ground-based geodetic methods require baseline measurements to be made against which future observations are compared. This requires volcanologists to anticipate which volcanoes will become active in the near future, a task that is virtually impossible. Given that there are about 1,000 potentially active volcanoes in the world, it is not possible in the foreseeable future to monitor them all with ground-based methods. In addition, some of the most dangerous volcanoes are those with very long repose intervals that can lull inhabitants and volcanologists alike into a false sense of security. With existing technology it is likely that dangerous volcanoes will continue to go unmonitored. SAR interferometry can provide the necessary background coverage.
Fig. 4.1. Time evolution of the vertical component of the baseline vector between station CASA at Long Valley caldera in east-central California and station Goldstone, 305 km south-southeast, between April and November of 1994. Error bars are one standard error. The best fit line through the data points indicates uplift of CASA at a rate of 30.7+/-5.5 mm/yr with respect to Goldstone, in agreement with other measures of deformation within uncertainties. The weighted root-mean-square (wrms) scatter of indi-vidual data points (representing 24 hours of data) is 12.3 mm. From Dixon et al. (1995).
Secondly, SAR provides contiguous coverage over a broad area, whereas existing methods provide data only at a limited number of points on the ground. Existing techniques grossly undersample the deformation field, while SAR interferometry is capable of mapping the entire deforming region around an active volcano. This broad and detailed spatial coverage provides important information on magma migration and other processes at depth, thereby improving our understanding of volcanic processes (Fig. 4.2) and perhaps our ability to predict and understand eruptions.
Third, volcanic crises typically pose great risks to personnel on the ground, inhibiting the collection of necessary data. Remote-sensing techniques allow crucial data to be collected for eruption forecasting, while minimizing the hazard to scientific investigators. SAR-based deformation monitoring would complement information from seismic, geochemical, and other monitoring methods, where they are available. It is worth noting that deformation is likely to precede measurable seismic activity in many cases--e.g., a rock must be stressed before it fails --suggesting that deformation changes may be the first indicator of an impending eruption. Furthermore, many remote volcanoes have no local seismic, geochemical, or other monitoring systems in place.
Figure 4.3 indicates one aspect of the problem. If a relatively shallow magma chamber is the source of unrest, as is typical of many active volcanoes, then surface deformation will be concentrated relatively close to the top of the volcano--precisely the area that is most difficult to reach and most likely to be dangerous in the event of an eruption. Clearly, the ability to remotely sense deformation would be an enormous boon to volcano studies.
Systematic monitoring of ground deformation at potentially hazardous volcanoes around the globe should be our long-term goal. SAR interferometry has the potential to warn of impending eruptions, mitigating the impact on lives and property. For this reason, we strongly endorse a series of projects over the next few years to test the suitability of SAR interferometry for volcano monitoring and to determine requirements for future satellite missions, partly or wholly dedicated to volcano monitoring. We envision a global system in which SAR images will be acquired and interpreted on a time scale useful for forecasting eruptions, reducing loss of life and increasing our understanding of volcanic processes.
Fig. 4.3. A simple Mogi (point-source) model for volcanic deformation, showing displacement rate vs. horizontal distance from the center of deformation. For a shallow (5-km in this example) deformation source that is typical of many volcanoes, deformation decreases rapidly with increasing distance from the source.
While SAR interferometry holds tremendous potential for volcano monitoring, there are important questions that need to be addressed in the next few years before a worldwide moni-toring system can be designed and implemented or data from it properly interpreted. For example, can we reliably identify artifacts that might be misinterpreted as volcanic deformation? These artifacts could be due to atmospheric effects (e.g., troposphere or ionopshere; Figs. 2.3 and 2.4) or bad topographic data (Fig. 2.2), all of which may yield quasiregular areas of "pseudo-deformation" crudely analogous to expected volcanic deformation patterns. Early results from two flights of NASA's space shuttle in 1994 suggested rapid rates of volcanic deformation in Hawaii based on SAR interferometry. Later research failed to confirm the high rates, and suggested that the anomalies were most likely associated with water vapor variation in the troposphere. Several approaches are available to identify artifacts and, with more experience, we believe they will not pose a critical problem, at least when repetitive coverage is available. For example, multiple images of the same area may allow discrimination of true volcanic deformation from atmospheric effects, the latter being highly variable in both space and time. Data from short-term missions, such as the Shuttle, will continue to pose problems in interpretation.
A critical question is whether or not SAR images obtained at different times from heavily vegetated or snow-covered volcanoes can be correlated, so that interferograms that measure ground deformation can be constructed. Results to date suggest that coherent interferograms can be obtained in arid and semiarid areas over months to years. Much less is known about the time scales over which surfaces decorrelate in other areas. Many active volcanoes are covered by snow and ice, at least seasonally. Other volcanoes have significant vegetation cover that may inhibit or preclude correlation of SAR images. It is presently unknown whether multipass SAR interferometry can be done on these volcanoes, so a vigorous attempt to answer this crucial question is the highest priority for study.
There are a number of possible ways to overcome the surface decorrelation problem. First, some volcanic deformations are so large that useful information can be obtained by differencing DEMs obtained by interferometry. Space-based SAR can provide 10- to 30-m horizontal resolution and 1- to 3-m vertical accuracy. Before its 1980 eruption, parts of the north flank of Mt. St. Helens moved as much as 2.6 m horizontally and 0.07 m vertically each day. In these situations, useful data can be obtained even if there is total surface decorrelation between SAR images acquired at different times. For this method to succeed, it is beneficial if the SAR system is capable of determining a DEM with a single pass--either by having a dual antenna system or tandem satellites.
Unfortunately, all volcanoes cannot be counted on to deform as rapidly as Mt. St. Helens. Also, it would be nice to have more warning of an eruption, perhaps by studying the smaller precursor strain events that seemed to precede the meters-per-day deformation at Mt. St. Helens. To maximize opportunities to measure small amounts of deformation, it will be important to determine the rates of surface decorrelation on a variety of volcanic surfaces. Minimizing the time between passes will limit surface decorrelation. This goal implies a dedicated space mission optimized for change detection, including wide-swath, high-accuracy knowledge of orbit parameters, and perhaps two (or more) spacecraft. Other alternatives include placing radar corner reflectors or transponders on the ground or focusing on the bare flanks and tops that often characterize some parts of active volcanoes. The problem with the last two suggestions, however, is that we lose the broad spatial coverage.
Experiments Needed to Address Key Questions
The working group recommends that data be obtained from existing systems [i.e., ERS-1, the Japanese Earth Resources Satellite (JERS-1), Topographic SAR (TOPSAR)] to examine the question of surface decorrelation on volcanic surfaces. It is important that a number of different types of volcanoes be studied--basaltic shields, stratovolcanoes, silicic calderas, and active lava domes, and that a variety of climactic zones be investigated--arid regions, snow- and ice-covered terrain, and rain forest. Some form of deformation monitoring is already underway at a number of volcanoes that we are suggesting for study (below). This list is not intended to be complete in any sense. To study these areas, we recommend the use of both satellite- and aircraft-based SAR instruments and other techniques to assess the suitability of SAR interferometry for volcano monitoring.
The group recommends obtaining the necessary technological support to obtain data that will prove (or disprove) the utility of SAR interferometry for eruption forecasting. This support includes installation of an ERS-1 ground receiving station in Hawaii and selected other locations, improvements in real-time navigation and control for the NASA DC-8 to extend the capability of the TOPSAR instrument to perform repeat-pass interferometry, and support for the required improvements in image processing.
(1) Hawaii. The group strongly endorses the proposal to install a ground receiving station for ERS-1 in Hawaii (scheduled for operation in 1995). This would allow SAR images to be obtained from Hawaii 's two most active volcanoes, Kilauea and Mauna Loa, as well as the rest of the Big Island where surface conditions vary from arid desert to lush rain forest. Both Kilauea and Mauna Loa are known to be actively deforming. Current rates of horizontal deformation on the south flank of Kilauea are 10 cm/yr (Owen et al. 1995; Fig. 4.4). Since both volcanoes are currently being studied with a variety of ground-based geodetic methods (leveling, tiltmeters, and GPS), ground truth is available for comparison to SAR results. It is worth noting that this effort is much more concentrated on Kilauea than on Mauna Loa, partly for logistical reasons that would not be a factor in SAR studies of this largest of Earth's volcanoes.
The upper slopes of Mauna Loa and the southwest flank of Kilauea are extremely dry and devoid of vegetation, providing an ideal environment for SAR interferometry. Parts of Mauna Loa's lower slopes and Kilauea's east rift zone, however, are largely covered by tropical rain forest. Thus, two different environments, one ideal, the other difficult for SAR interferometry, can be examined on two nearby active volcanoes.
(3) Aniakchak. This 10-km caldera formed at the summit of an andesitic stratovolcano, part of the Aleutian island arc, during a very large eruption about 3400 years ago. Since then, the volcano has produced numerous eruptions, including a series of violent explosions and formation of a lava dome in 1931. Each of the postcaldera eruptions has produced magma that is chemically similar to that erupted during the caldera-forming event, which suggests that a large reservoir of silicic magma still exists in the upper crust beneath the caldera--as well as the potential for another catastrophic caldera-forming eruption. High geothermal heat flux keeps the caldera free of permanent ice and snow, so that essentially year-round observations by remote sensing may be possible. Currently, Aniakchak is not being monitored. The U.S. Geological Survey plans to establish a leveling traverse across the caldera during summer 1994.
(4) Long Valley. At this restless caldera, intensive monitoring programs already exist: subweekly geodimeter observations, continuous tiltmeters, continuous GPS observations (Fig. 4.1), static GPS networks, and leveling. In addition, both TOPSAR and ERS-1 data exist that could be used to test multiple-pass SAR for deformation detection. The Inyo volcanic system adjacent to the Long Valley caldera has erupted repeatedly during the last 40,000 years, most recently about 550 years ago. Future eruptions will threaten the Mammoth Lakes community and interstate travel though the area.
(5) Misti, Peru. The city of Arequipa (pop. ~800,000) is built on the southwestern flanks of this active volcano. A fuming dome in the summit crater of this volcano shows periodic variation in the intensity of fumarolic activity. Recent mapping of this volcano shows a past history of episodic explosive activity, as well as more recent lahar and mudflow hazard. This is a very dangerous volcano on which no current monitoring is being undertaken. The aridity of the region and its high altitude (base elevation ~3000 m) makes Misti a very good target for SAR interferometry.
(6) Sabancaya, Peru. Some 70 km to the north of Arequipa is Sabancaya, which has been in eruption since 1989. During this time, mudflows and ash have been a constant hazard to the 30,000 inhabitants that live 15 km away and 2 km below the volcano. In 1992, an earthquake (M 5.2) destroyed a village and commu-nications just north of the volcano. A rudimentary three-station seismic network has been in episodic operation since 1990, but there is a real need for an early warning system. Like Misti, the aridity of the region makes this a very good target for deformation studies using SAR interferometry.
Of course there are many other possible targets around the globe, including volcanoes that have the potential to erupt explosively putting nearby populations at risk, that can be studied with existing data.
Recommendations
(1) Perform research to assess the impact of temporal decorrelation on SAR-based monitoring of volcanic deformation in a variety of environments.
(2) Install a ground receiving station in Hawaii for ERS data, and use it to assess ground deformation and temporal decorrelation.
(3) Pursue a space mission for global monitoring and eruption forecasting of active volcanoes. Assuming moderate success in the first recommen-dation, SAR interferometry holds the greatest promise for providing the requisite data.
Previous Section
Next Section
References
Imaging Radar Home Page
Return to Table of Contents