SIR-C Radar Investigations of Volcanism and Tectonism in the Northern Andes: Final Report

CA Wood, Space Studies, Univ. of North Dakota, Grand Forks, ND 58202 ;

SN Williams, RL Wessels, S Schaefer and C Gorman, Geology, Arizona State Univ., Tempe, AZ;

AW England and RT Austin, Electrical Engineering & Computer Science, Univ. of Michigan, MI;

MN Hall, Instituto Geofisico, Escuela Politecnica Nacional, Quito Ecuador.

The Northern Andes volcanic province of Ecuador and Colombia consists of a 1000 km long belt of active, dormant and extinct volcanoes, few of which have been well studied. Unlike the high desert volcanics of the Central Andes, field work is difficult in the Northern Andes because of heavy vegetation and sometimes dangerous political conditions, and frequent cloud cover lessens the value of traditional remote sensing sensors. The flights of the SIR-C/X-SAR radars have now provided a stunning high resolution synoptic dataset that transforms the Northern Andes into one of the best imaged volcanic provinces in the world. Much of the improvement comes from the high resolution (25 m) radar views, unimpeded by clouds, in multiple look angles and wavelengths. But another, non-technical characteristic of the SIR-C/X-SAR data is revolutionary. For the first time, a homogeneous dataset is conveniently available to researchers, without the high cost and data gaps associated with satellite optical wavelength data, and without the political and logistical complexity needed to acquire images from various agencies within two separate nations. We intend to explore various means to make this immense, invaluable but unwieldy dataset more widely available.

The investigations associated with our team's research span more than half a decade, two continents, four research groups and multiple themes. Three students used SIR-C data as integral parts of their PhD research, and others used it for MS projects. The results summarized here, largely based on these student efforts, concern theoretical and field characterizations of of radar roughness, tectonic and volcanic studies and investigations of hazards associated with active volcanoes.

Radar Roughness Studies (Austin and England)

Based on field measurements we found that the primary debris flows at Mt. St. Helens have power-law roughnesses over two intervals of spatial frequency (corresponding to the large and small scale surface measurements), and that these spectra align to suggest that the surfaces behave as power-law (or scaling) surfaces over spatial frequencies from 1 mm-1 to 0.1 m-1. We also found that the small scale roughness spectrum is reduced for eroded surfaces or surfaces with water-deposited sediment. These findings were summarized in Austin and England (1993).

We found that power-law surfaces introduce unique difficulties in the process of spectral estimation. Use of an improper estimator will allow leakage of low frequency energy into higher parts of the energy spectrum. The result is a spectral estimate that is insensitive to the spectral slope (that is, insensitive to the exponent in the power law spectrum). We believe that several previous analyses of the roughness of natural surfaces that appear in the literature were in error because of this difficulty. We showed how Capon's spectral estimator has less variance than Fourier-based estimators and measures the spectral slope more accurately. We also showed how estimates of a 2-D roughness spectrum can be obtained from estimates of the 1-D spectrum of an isotropic power-law surface. These findings were reported in Austin, England and Wakefield (1994).

Based upon the roughness spectra measured at Mt. St. Helens, we designed and constructed artificial dielectric analog surfaces with power-law roughnesses. Three surfaces were manufactured using a single roughness amplitude and three values of the spectral slope. Scattering coefficients for co-polarized and cross-polarized backscatter were obtained using a 35 GHz scatterometer. These measurements were described in Austin's dissertation and in a TGARS paper that is under review.

The measured values of the co-polarized scattering coefficients were compared to those predicted by the common scattering theories - physical optics, geometric optics, and small perturbation theory, and by a recent theory - phased Wiener-Hermite theory. The small perturbation model underestimated the measured backscatter by about 5 dB at incidence angles between 15 and 45 degrees, but followed the trends with angle and surface roughness fairly well. Other models generally performed worse, failing to mimic the trend of the measured data with respect to incidence angle. These results were described in Austin's dissertation and in the TGARs paper that is under review.

Late in our study, we attempted to apply our new understanding of scattering by power-law surfaces to the SIR-C data. We tried to classify the surface at Galeras by its spectral slope using the difference in L- and C-band backscatter. If the surface had been a power-law surface, we should have been able to estimate the spectral slope by the way self-affine scattering surfaces behave at different scales. We were never able to get reliable solutions. The derived slopes were spread far outside (both smaller and larger) than the range of values that make physical sense. Our best guess about the failure of the classification is that either the surfaces were not power-law, or that other factors like vegetation were affecting the backscatter intensity.

In summary, we learned much about characterizing surface roughness in volcanic terrains and, to the extent that a surface is purely a power-law surface, we developed a quantitative means for classifying rough surfaces by their spectral slope using radar backscatter at more than one frequency. The classification did not work in the single SIR-C application that we tried. There was neither the time nor the resources to fully analyze the reason for the failure, but the most likely cause was that the surfaces were not power-law surfaces. The discrepancy could have been caused by significant vegetation in the scattering scene or by erosion of the original surface so that the higher spatial frequencies were suppressed.

Volcanic-Tectonic Studies (Wessels, Williams, Wood, Hall)

Image Analysis and Data Products

Studies of the tectonic controls of volcanism has been based on detailed mapping of volcanoes and tectonic features, especially in Colombia, but also in Ecuador. Wessels (Ph.D. candidate at Arizona State) began initial image processing and interpretation of local tectonics of southern Colombia and the regional faults of the Northern Andes using SRL-1 SIR-C survey data in July, 1994. He has worked on high resolution MLC products over specific areas since March 1995. Analysis of the massive data files has been a major effort with supurb results that constitute a lasting contribution to studies of the Northern Andes. All of the image proccessing and most of the interpretation has been done by the Arizona State team.

One of our early accomplishments was the assembly of a geo-referenced online mosaic of twenty survey images over the active volcanic arc (covering 1300 x 120 km). This giant mosaic provides the absolutely best data set for regional volcano-tectonic investigations of the Northern Andes (Figure 1). Inspired by the results from the early mosaic, we have now mosaiced and geo-referenced six of the survey data takes over the Nevado del Ruiz, Colombia region (500 x 200 km). The Ruiz data provides additional geometric control for documenting how the major strike-slip faults in the northern part of the volcanic arc curve away from the Central Cordillera. We are investigating how such mammoth photo images might be published. This is a case where looking at a compressed version or a series of individual sections on a computer monitor is vastly inferior to having the entire hard copy rolled out on a table!

We have also compiled mosaics over southern Colombia using several high resolution (MLC) subscenes from three data takes. The subscenes of MLC products have been used to document variations in active volcano morphology related to regional fault geometry. Both the large survey and the local high resolution mosaics provide a geo-referenced base layer for our geographic information system (GIS) work. The data SIR-C data was used to precisely measure orientations, areas, and vent locations. We also digitized and co-registered structural field data, seismicity and published faults and incorporate them as interactive layers of information to aid our interpretations. We plan to gradually expand the database to include geochemistry, geochronology, and eruptive activity.

X-SAR scenes over Galeras, Cumbal, and Ruiz volcanoes were coregistered and merged with the SIR-C datasets (MLC). The co-registration and merging was performed with commercial software and is fairly straightforward. We've found that the combination of the three different wavelengths (X, C, and L) enhance textural differences in the volcanic features. False color images combining SIR-C and X-SAR also provide a better definition of structural features in the ice covered areas of Ruiz volcano (Figure 2).

We've also have begun experiments merging SIR-C data with Landsat TM as a means to better distinguish different volcanic products. We anticipate that the merged radar/optical data will help map deposits by providing a combination of rock/ash alteration information and textural information. This work will be limited to the recent deposits of the most active volcanoes because vegetation is widespread over most of the volcanoes. Initial results over Galeras volcano (using commercial software) show that the layover effects of the SIR-C data create fairly large rms errors in georectification and the merged data have a several pixel mismatch. We have postponed this work until we can find a more robust package for georectification.

Existing topographic maps were compared with individual volcano DEM data from the 1993 AIRSAR/TOPSAR flights for Cumbal and Galeras volcanoes. The TOPSAR DEMs provide a very detailed representation of each volcano. We have used the data to construct three-dimensional representations to aid in interpretation of lava flow morphology and source. We are able to use the DEM visualizations to distinguish many flow lobes that appear to be the same flow on the radar images (Figure 3). Although the 3-D representations are quite useful, the no data zones in the DEM created by radar shadow made contouring the data quite difficult. TOPSAR/AIRSAR images were used for geologic field research by Gorman and Wessels.

Geologic Interpretation Results

We've been able to precisely map the structural detail of faults and fractures that have ripped apart the crust in the volcanic arc. For our regional volcano-tectonics study, we have combined remote sensing, seismic, and field data to document several volcano-tectonic features: Arc-parallel and subsidiary fault geometries and kinematics, volcano location and spacing along the arc, and volcano orientation and morphology. This information relates to the interaction of regional stress with volcanism in the arc.

We can estimate the orientation and direction of the movement of many of the faults connecting the major volcanoes. In addition to the possible large volcanic centers found in our preliminary work, the high resolution mosaics have revealed previously unknown lines of small cinder cones and possible maars not visible on the regional data (Figure 4). This information is critical for our regional tectonic models and may lend additional support to the segmentation model proposed by Hall and Wood (1985).

Our analysis defines three distinct volcano-tectonic domains in the Northern Andes arc. Each zone is characterized by a dominant fault/fracture orientation which controls volcanic vent alignments or edifice elongations (Figure 4). We have also found several examples where we can demonstrate that the proximity of a volcano to the major faults may affect the amount of elongation or alignment. In the linear volcanic arc of Colombia the fracture trend orientations appear to reflect the age of the crust in which they formed. In Ecuador, the volcanic arc is more dispersed with no dominant subsidiary fracture trends. The trends in Colombia may reflect the changes in subduction directions since the Jurassic, while the distribution in Ecuador may be affected by the subduction of the Carnegie Ridge. We are presently evaluating models of fault geometry that occur in transpressional zones to understand the overall distribution of volcanoes along the arc.

Hazardous Volcanoes (Schaefer, Gorman, Williams, Hall and Wood)

The Northern Andes hosts 34 active volcanoes, including Ruiz which had one of the most deadly eruptions in history (25,000 casualties in 1985) and Galeras (which had a fatal eruption in 1993 while three of our team were on the mountain). Many other volcanoes in this region can be expected to have future tragic eruptions, especially considering the deadly combination of glacier-capped summits and increasing populations along the flanks. Thus, one of our continuing goals is to use SIR-C/XSAR data to better understand the geologic histories of potentially dangerous volcanoes so as to better forecast their future activities. As part of the SIR-C contribution, Schaefer (Ph.D. 1995) conducted detailed field mapping of the recent pyroclastic deposits of Ruiz volcano, Colombia, prior to the first SIR-C flight. A future student will use Schaefer's work to aid interpretation of the quad-pole, multi-look angle SIR-C/X-SAR data of Ruiz. Another Arizona student, Caitlin Gorman, used both SIR-C and AIRSAR data in the field to map the structure and lava flows of Cumbal Volcano, Colombia for her masters thesis. She discovered that unlike nearby volcanoes, Cumbal is not composed of highly explosive, rhyolitic deposits, but rather is made of andesitic lava flows.

SIR-C data provided discoveries within the first few minutes of examining the browse images. Two volcanoes were discovered in Colombia: one is Volcan San Diego, the northern-most volcano in the Northern Andes, exactly along a postulated segment boundary (Wood et al., 1994). The youthful appearance of this caldera lake volcano implies that it has had relatively recent activity and thus may erupt again. The other newly recognized feature is less certainly a volcano, but appears to be a large caldera in southern Colombia. Unfortunately, dangerous travel circumstances have made it impossible to visit these structures on the ground. The cloud-free radar images are also being used to map geology and assess hazards for many of the active volcanoes in Ecuador. One example of a poorly known structure which SIR-C images were used to better understand is the large volcano Cayambe. One of Hall's students in Ecuador mapped the geology and hazards of this volcano using both SIR-C data and traditional aerial photos. Hall has added the SIR-C images to his arsenal of information to systematically evaluate the hazards posed by all volcanoes in Ecuador. Scientific and public access to processed and browse SIR-C/X-SAR images of the main volcanoes is provided through the South America section of the World Wide Web site VolcanoWorld.

Publications

Austin, RT and England, AW, 1993, Multi-scale roughness spectra of Mount St. Helens debris flows. Geophys. Res. Lett. 20, 1603-1606.

Austin, RT, England, AW and Wakefield, GH, in press, Special problems in the estimation of power-law spectra as applied to topographical modeling. Trans. Geosci. Remote Sens., IEEE.

England, AW, 1992, The fractal dimension of diverse topographies and the effect of spatial windowing. in Ground Penetrating Radar (ed. by J Pilon) Geol. Surv Canada, Paper 90-4, 57-61.

Gorman, CE and Williams, SN, 1997, The eruptive history of Cumbal Volcano, southern Colombia. IAVCEI General Assembly, Puerto Vallarta, Mexico, p. 150.

Schaefer, SJ, 1995, Nevada Del Ruiz Volcano, Colombia: magmatic system and evolution. Arizona State University, Ph.D. Dissertation, 147 p.

Wessels, RL, Hall, MN, Wood, CA, and Williams, SN, 1997, Tectonic Controls of Volcano Morphology and Location in the Northern Andes (Colombia and Ecuador). GEODAZE '97, University of Arizona

Wessels, RL, Hall, MN, Wood, CA, and Williams, SN, 1997, Comparison of faults and volcanoes in the Northern Andes, Colombia and Ecuador using SIR-C, XSAR, and TOPSAR. IAVCEI General Assembly, Puerto Vallarta, Mexico, p. 109.

Wessels, RL, 1995, The interaction of transcurrent tectonics and continental arc volcanism in Colombia and Ecuador: Preliminary results from SIR-C radar analysis. EOS, Transactions, American Geophysical Union, v 46, n 7.

Wood, CA, Wessels, RL, Williams, SN and Calvache, ML, 1994, SIR-C images of little known Colombian volcanoes. Geol. Soc. Am. Ann. Mtg.

Wood, CA and VolcanoWorld Team, 1995-continuing, Sir-C images in VolcanoWorld World Wide Web site: http://volcano.und.nodak.edu/vwdocs/volc_images/south_america/south_america.html

Figures

Figure 1. NASA SIR-C survey image mosaic. This image offers a unique view over most of the active volcanic arc of the Northern Andes. The first looks at this data revealed several poorly known calderas (Wood and others, 1994). The image is a mosaic of 26 survey images over the center of the Northern Andes of Colombia and Ecuador. It is being used as a base map for several GIS overlays for various aspects of the Northern Andes

Figure 2. Nevado del Ruiz - Tolima Volcanic Complex, Colombia false color (Xvv, Lhh, Lhv). Nevado del Ruiz is notorious for its fatal November, 1985 eruption that created a large debris flow that wiped out the town of Armero and its 25,000 inhabitants. The volcanoes to the south of Ruiz appear to be strongly aligned along a N-S ridge. Several other lineaments are visible in spite of Recent volcanic deposits, suggesting a relatively young age. While most of the volcanic structures along the N-S lineament are elongate, Volcan Tolima, which lies to the east of the lineament, is conical.

Figure 3. Colored relief image of the Cumbal volcano TOPSAR DEM. Cumbal, in southern Colombia, is one of the best examples of an elongated volcanic edifice aligned with a regional fault and dome complexes. The AIRSAR/TOPSAR data was used as part of a field mapping project to establish the eruptive history of Cumbal.

Figure 4. SIR-C Lhh image of newly discovered cinder cones in the Laguna LaCocha area in southern Colombia. The vents (arrows) are located along the trace of a major strike-slip fault.