Eos, Vol. 79, No. 39, September 29, 1998 EOS,
T R A N S A C T I O N S ,
A M E R I C A N
G E O P H Y S I C A L
U N I O N
VOLUME 79 NUMBER 39 SEPTEMBER 29,1998 PAGES 461-476
Satellite Radar Interferometry Measures Deformation at Okmok Volcano Zhong Lu, Dorte Mann, and Jeff Freymueller PAGES 4 6 1 , 4 6 7 - 4 6 8
A, eventually covering an area of about 15 k m . The red line in Figure 1 shows the extent of the lava flows. Satellite thermal imagery and inspection from the air indicated that eruptive activity had ended by late April 1997. 2
The c e n t e r of the Okmok c a l d e r a in Alaska subsided 140 c m as a result of its Feb ruary-April 1997 eruption, a c c o r d i n g to satel lite data from ERS-1 and ERS-2 synthetic aperture radar (SAR) interferometry. The in ferred deflationary s o u r c e was located 2.7 km b e n e a t h the approximate c e n t e r of the caldera using a point s o u r c e deflation model. Researchers believe this s o u r c e is a m a g m a c h a m b e r about 5 km from the erup tive s o u r c e vent. During the 3 years before the eruption, the c e n t e r of the caldera up lifted by about 23 c m , which researchers be lieve was a pre-emptive inflation of the magma c h a m b e r . Scientists say such meas urements demonstrate that radar inter ferometry is a promising s p a c e b o r n e t e c h n i q u e for monitoring remote v o l c a n o e s . Frequent, routine acquisition of images with SAR interferometry could m a k e near real time monitoring at such v o l c a n o e s the rule, aiding in eruption forecasting.
Satellite R a d a r Interferometry To measure the deformation associated with the 1997 eruption, interferograms were constructed from the phase difference of
pairs of SAR images recorded by ERS-1 and ERS-2 satellites. T h e two satellites have Cband microwave radars with a 5.66 c m wave length. T h e phase difference b e t w e e n two images taken on different satellite passes cor responds to the c h a n g e in the round-trip path length of radar waves to ground targets. Essentially, an interferogram is a contour map of the c h a n g e in distance to the ground surface along the look direction of the satel lite. In the c a s e of ERS-1 and ERS-2, the look direction is inclined 23° from the vertical. As a result, interferograms constructed from ERS data are more sensitive to vertical displace ments in the target area than to horizontal dis placements. Path length c h a n g e s are related to topography, the difference in the satellite positions between the two passes, ground de formation occurring b e t w e e n the times of the two passes, and noise. E a c h c y c l e of the interferometric phase, or fringe, is represented by a series of colors on the interferogram (Fig ures 2 a-c).
Fig. 1. SAR amplitude image of Okmok vol cano, 16x20 km in di mension. Floor of the caldera is filled by la vas and ash from erup tions in 1945, 1958, 1986, 1988, and 1997. The extent of lava from the latest 1997 erup tion is outlined by red line. The solid white rectangle indicates the position of Figures 3a and 3b. Original color image appears at the back of this volume.
The Okmok Volcano Okmok V o l c a n o (Figure 1) is a 10-kmwide caldera that o c c u p i e s most of the north eastern end of Umnak Island, in the Aleutian Islands. It erupted extrusively in 1 9 4 5 , 1 9 5 8 , and 1997, and explosively in 1986 and 1988. All historic eruptions of O k m o k originated from C o n e A, which is located on the south ern edge of the c a l d e r a floor, and produced a b u n d a n t ash emissions and mafic lava flows that crossed the c a l d e r a floor. Voluminous rhyodacitic caldera-forming eruptions oc curred 8 0 0 0 and 2400 years ago. The latest eruption of Okmok V o l c a n o be gan in mid-February 1997 and lasted for 2 months. Ash and lava were vented from Cone For more information, contact Zhong Lu, Raytheon STX Corporation, EROS Data Center, Sioux Falls, SD 5 7 1 9 8 USA; E-mail: l u @ e d c m a i l . cr.usgs.gov.
This page may b e freely c o p i e d .
Eos, Vol. 79, No. 39, September 29, 1998 An interferogram is usable only in those areas where the radar returns in both images are coherent. T h e degree of c o h e r e n t interfer e n c e of SAR images d e p e n d s on the c h a n g e s of the backscattering characteristics of the ground surface b e t w e e n the two radar acqui sitions. Reduction of c o h e r e n c e c a n b e c a u s e d by surface alterations such as growth of vegetation, c h a n g e s in snow or i c e cover, or the covering of the old surface with new v o l c a n i c ash or lavas [Zebkeretai, 1996]. On Alaskan v o l c a n o e s , c o h e r e n c e c a n b e maintained over a time interval of at least 2 years if images are acquired during subarctic s u m m e r and fall, July to November [Lu and Freymueller, 1998]. T h e deformation associ ated with the 1997 eruption was analyzed us ing images acquired during the s u m m e r and fall of 1992, 1 9 9 3 , 1 9 9 5 , and 1997. No data were recorded b e t w e e n March and Novem ber of 1996. All of the images were acquired in the descending m o d e of the ERS-1 and ERS-2 orbits, where the satellites travel ap proximately from north to south, and look to the west, inclined 23° down from vertical. No Global Positioning System or other geodetic data are available for Okmok V o l c a n o . The c o m p o n e n t of the phase c h a n g e con tributed by ground surface deformation c a n b e obtained by removing the topographic c o m p o n e n t of an interferogram using a digi tal elevation model or another interferogram [e.g., Massonnet et al., 1995; Peltzer and Rosen, 1 9 9 5 ] . T h e effect of topography was re moved using a digital elevation model and the known satellite geometry, leaving only the contributions of ground deformation and noise (Figure 2 a ) . T h e existing elevation model for Okmok V o l c a n o was interpolated from 90 m spacing to 20 m (the dimensions of e a c h pixel), and was improved using an inter ferogram derived from ERS-1 and ERS-2 tan dem images acquired 1 day apart. The topography of the caldera floor, where most of the c o h e r e n t signal is observed, is rela tively flat, with less than 2 7 0 m relief. Figure 2. see next Deflation of O k m o k C a l d e r a Figure 2a shows an interferogram with to pography removed for a 16 x 20 km area of O k m o k V o l c a n o , showing ground deforma tion b e t w e e n O c t o b e r 9 , 1 9 9 5 , and Septem ber 9, 1997. Each fringe c a u s e d by ground deformation corresponds to a 2.83 c m c h a n g e in the distance from satellite to ground between two radar acquisitions and is represented by a c o m p l e t e c o l o r c y c l e in an interferogram. Phase signals are c o h e r e n t in the northern portions of the image, both within and outside of the caldera, as well as
in the western and southeastern parts of the image. T h e area of loss of signal c o h e r e n c e within the caldera corresponds mostly to re gions covered by loose materials or by new lava and ash of the 1997 eruption. An enlargement of the main area of coher e n c e is shown in Figure 2 b . Both in the main c o h e r e n t region and the small c o h e r e n t re gion east of the caldera, the elliptical fringes and the order of colors (repetitions of blueyellow-magenta from the caldera outward) show that the deformation field is consistent with a deflation source located in roughly the c e n t e r of the caldera. The three fringes in the
This page may b e freely c o p i e d .
page
southeastern part of the image are important b e c a u s e they show the s a m e sense of motion as the other fringes, down toward the c e n t e r of the caldera, and rule out the possibility that the fringes observed in the northern part of the image are erroneous "junk fringes" c a u s e d by poor orbital control. The altitude of ambiguity (the amount of topographic c h a n g e required to generate o n e interferometric fringe) of this interferogram is 9 5 0 m; s o 9 5 0 m of topography would contribute only o n e fringe to the image, and the ex p e c t e d topographic error amounts to no more than 0.04 fringes.
Eos, Vol. 79, No. 39, September 29, 1998 eled using a spherical point deflationary pres sure s o u r c e in an elastic half-space [Mogi, 1958]. The deflationary s o u r c e is inferred to b e a shallow m a g m a c h a m b e r — t h e s o u r c e of the 1997 lavas. Synthetic interferometric fringes were generated from a variety of models, varying the location and depth of a single point s o u r c e deflation. T h e approximate hori zontal coordinates of the model deflation s o u r c e were c h o s e n to match the c o n c e n t r i c a p p e a r a n c e of the fringes. T h e s o u r c e depth and s c a l e of the pressure s o u r c e were varied to match the observed fringe spacing and to tal n u m b e r of fringes. The best-fitting model m a t c h e s the fringe spacing and shape within the caldera, the to tal observed deformation, and profile B-C. Based on this model, the 2.7 km s o u r c e depth (Figure 2 c ) and a volume c h a n g e in the magma c h a m b e r of 0.048 k m ( s e e Anderson [ 1 9 3 6 ] ) were inferred. T h e m a x i m u m subsi d e n c e in this model is 140 c m , directly a b o v e the deflationary s o u r c e . A m o d e l e d interfero gram is shown in Figure 2 c , which covers the s a m e region as Figure 2 b . T h e volume c h a n g e estimate assumes that no inflation or deflation o c c u r r e d b e t w e e n O c t o b e r 1995 and the beginning of the eruption. T h e lava c o v e r e d an area of 15 k m , so assuming the volume of lava is the s a m e as the volume re moved from the m a g m a c h a m b e r , the aver age flow thickness would b e 3.2 m. 3
E o C
o o E observed ( 9 5 - 9 7 ) —modeled ( 9 5 - 9 7 )
0
Fig. 2. a) Deformation interferogram span ning October 1995 to September 1997, which brackets the February J997 erup tion of Okmok Volcano. Each fringe repre sents a 2.83-cm change in distance along the satellite look direction. The image cov ers the same area as Figure 1; b) Portion of the interferogram of Figure 2a, outlined by the white box in Figure 2a. There are 44 fringes along a profile from B to C of 5.5 km; c) A modeled interferogram using the best-fit point deflation model and covering the same region as Figure 2a; d) The ob served (Figure 2b) and the modeled (Fig ure 2c) deformation along the satellite look direction on the profile from B to C. Original color image appears at the back of this volume.
1 2 3 4 5 distance (km)
The m a x i m u m topographic relief over the entire image is about 8 0 0 m. Atmospheric de lays c a n c a u s e up to two or three fringes of systematic, spatially correlated error in an in terferogram (see, e.g., Massonnet andFeigl [ 1 9 9 5 ] ) . Therefore systematic biases c a n ac count for no more than a few percent of the observed v o l c a n i c inflation. There are 44 fringes along a distance of 5.5 km from B to C (Figure 2 b ) and two or three more outside
the caldera, e a c h fringe representing 2.83 c m of ground surface motion in the satellite's look direction. Point B moved away from the satellite more than point C, corresponding to s u b s i d e n c e at the c e n t e r of Okmok caldera (Figure 2 d ) . T h e c e n t e r of s u b s i d e n c e is off set by about 5 km from the eruptive vent, which implies that significant lateral m a g m a transport may have o c c u r r e d during erup tion. T h e observed deformation was mod
This page may b e freely c o p i e d .
2
The spherical point s o u r c e model used suffers from two potential drawbacks: the s o u r c e could b e aspherical and the point s o u r c e approximation may b e invalid ( s e e , e.g.,McTigue [ 1 9 8 7 ] ) . One or both of these possibilities is suggested by the poor fit of the model fringes to those observed in the vicin ity of point B and the middle b e t w e e n B and C (Figures 2 b , 2 c , and 2 d ) . T h e location of the deflation s o u r c e in the approximate cen ter of the caldera and the inference that the s o u r c e is shallow are c o n c l u s i o n s that do not require investigation of more c o m p l e x mod els, but the estimated eruptive volume could d e p e n d on the s o u r c e model c h o s e n . Inflation of O k m o k C a l d e r a Interferograms were formed for time peri ods spanning O c t o b e r 3 1 , 1 9 9 2 to November 2 0 , 1 9 9 3 (Figure 3 a ) , and November 1,1993 to O c t o b e r 2 5 , 1 9 9 5 (Figure 3 b ) to measure how m u c h inflation may have p r e c e d e d the eruption. In general, pre-eruptive inflation is e x p e c t e d and the inferred pre-eruptive infla tion s o u r c e should b e l o c a t e d in the s a m e
Eos, Vol. 79, No. 39, September 29, 1998 b e e n significantly underestimated. Observa tions at several v o l c a n o e s suggest that both c a s e s are possible ( s e e , e.g., Newhall and Dzurisin [ 1 9 8 8 ] ) .
b.
a.
Acknowledgments
Fig. 3. a) Deformation interferogram spanning 1992-1993. The location of the image is out lined by the solid white rectangle in Figure 1. A total of 17 cm of uplift was observed along the profile from B to C; b) Deformation spanning 1993 to 1995. A total of 6 cm of uplift occurred along the profile from BtoC during 19931995; c) Deformation along the satellite look direction on the profile from BtoC for 19921993 and 1993-1995. Original color image ap pears at the back of this volume.
0
1 2 3 4 5 distance (km)
p l a c e as the coeruptive deflation s o u r c e [Dvorak andDzurisin, 1997]. More than four fringes were observed, corresponding to about 17 c m uplift of the c e n t e r of the cal dera from 1992 to 1993 (Figures 3a, 3 c ) and about two fringes corresponding to about 6 c m uplift from 1993 to 1995 (Figures 3 b - c ) . Within these interferograms, the magnitudes of possible systematic biases from atmos pheric delays or other s o u r c e s are a larger fraction of the inferred signal. However, be c a u s e the inferred uplift is generally consis tent in time, it is unlikely that the apparent 23-cm uplift a b o v e the m a g m a c h a m b e r from 1992 to 1995 is entirely the result of such er rors.
It is less c l e a r that the apparent reduction in the inflation rate from 1992-1993 to 19931995 is significant. The c e n t e r of 1993-1995 uplift (Figure 3 b ) correlates well with the cen ter of 1995-1997 s u b s i d e n c e in the model (Fig ure 2 c ) . T h e observation of pre-eruptive inflation of Okmok caldera could mean that a minimum of 0.05 k m of magma was emplaced in a shallow m a g m a c h a m b e r in the 3 years prior to the eruption, or that pressure in creased without additional m a g m a input by development of a vapor phase, perhaps by s e c o n d boiling during crystallization. B e c a u s e no images were recorded in the summer of 1996, and no usable images were recorded immediately prior to the eruption in 1997, it is unknown how much inflation oc curred between O c t o b e r 1995 and the begin ning of the eruption. If pre-eruptive inflation continued at a rate similar to that for 19921995, the 1995-1997 inflation would b e fairly small and the model would give a r e a s o n a b l e estimate of the coeruptive deflation. How ever, if Okmok caldera inflated rapidly in the year, months, or days prior to the eruption, then the coeruptive deflation would have 3
This page may b e freely c o p i e d .
W e thank C. Nye, S. McNutt, and J. Eichelberger for helpful discussions, and D. Meyer, D. Carneggie, D. Gesch, and T. Albright for in ternal U.S. Geological Survey reviews. W e also thank the c o m m e n t s of o n e reviewer. W e are grateful to C. Werner for helpful sug gestions during interferogram processing. This work was d o n e at Raytheon STX Corpo ration under contract 1434-CR-97-CN-40274 with the U.S. Geological Survey. Authors D. Mann and J. Freymueller were supported by NASA grant NAG5-4369. ERS-1 and ERS-2 SAR images are copyrighted 1 9 9 2 , 1 9 9 3 , 1 9 9 5 , 1997 by the European S p a c e Agency.
References Anderson, E., T h e d y n a m i c s of the formation of cone-sheets, ring-dykes, and caldron-subsi d e n c e , Proc. R. Soc. Edinburgh, 56, 128157, 1936. Dvorak, J . , and D. Dzurisin, V o l c a n o geodesy: the s e a r c h for m a g m a reservoirs and the for mation of eruptive vents, Rev. Geophys., 35, 343-384, 1997. Lu, Z., and J . Freymueller, Synthetic aperture radar (SAR) interferometry c o h e r e n c e analy sis over Katmai v o l c a n o group, Alaska, J . Geophys. Res., in press, 1998. Massonnet, D., P. Briole, a n d A. Arnaud, De flation of Mount Etna monitored by s p a c e b o r n e radar interferometry, Nature, 375, 567-570, 1995. Massonnet, D., and K. L. Feigl, Discrimination of geophysical p h e n o m e n a in satellite radar interferograms, Geophys. Res. Lett., 22, 1537-1540, 1995. McTigue, D. F., Elastic stress and deformation n e a r a finite spherical m a g m a body: resolu tion of the point s o u r c e paradox, J . Geo phys. Res., 92, 12931-12940, 1987. Mogi, K., Relations b e t w e e n the eruptions of various v o l c a n o e s and the deformations of the ground surface around them, Bull. Earth quake Res. Inst., Univ. Tokyo, 36, 99-134, 1958. Newhall, C. G., and D. Dzurisin, Historical un rest at large c a l d e r a s of the world, U.S. Geol. Surv. Bull. 1855, 1108pp., 1988. Peltzer, G., and P. Rosen, Surface displace ments of the 17 May 1993 Eureka Valley, California, e a r t h q u a k e observed by SAR inter ferometry, Science, 268, 1333-1336, 1995. Zebker, H., P. Rosen, S. Hensley, and P. J . Mouginis-Mark, Analysis of active lava flows on Kilauea v o l c a n o , Hawaii, using SIR-C ra dar correlation measurements, Geology, 24, 495-498, 1996.