Sources of Data

Satellite data

Crucial to the concept of this volcano compilation was the existence of a consistent set of Landsat Thematic Mapper images covering the entire Central Volcanic Zone: a Landsat image forms the key component of each volcano description. 29 TM scenes were obtained to cover the region (Figure 3). Several acquisitions of one quad covering Lascar volcano, north Chile (VCA no. 27), were obtained to study changes in the volcano related to the eruption of September 16, 1986 (Francis & Rothery, 1987). An additional acquistion was obtained of Sabancaya (VCA no. 2) to investigate possible changes during the 1988 activity. One night-time scene was acquired of north Chile to investigate possible thermal inertia effects. These TM scenes formed by far the most important data source used in this study. Fieldwork supported by conventional air photography was carried out in a number of selected areas to provide ground truth for the TM data. Supplementary data sources which have beenFigure 3 useful include the experimental Modular Optoelectronic Multispectral Scanner (MOMS-01) flown on Space Shuttle missions STS-7 (June 1983) and STS-11 (February 1984). The most useful attribute of the MOMS data proved to be the serendipitous low angle, westerly illumination (Rothery & Francis, 1987). SPOT images of two volcanoes (Socompa and Llullaillaco; VCA 33 and 34) and of the La Pacana Caldera (VCA section 4.2) have also been acquired. The 20 m resolution multispectral SPOT data proved to have little advantage over the TM data, but the 10 m resolution panchromatic stereo SPOT images represented a powerful additional means of making morphological and textural studies.

The Landsat Thematic Mapper - some statistics

The first Thematic Mapper instrument (Figure 4) was flown on the American remote sensing satellite, Landsat 4 launched on July 16, 1982, which is now moribund. A second instrument is still functioning (1990) aboard Landsat 5, launched March 1, 1984. Both satellites orbit at at altitude of 705 km, in sun-synchronous orbits whose ground tracks precess westwards at the same rate as the sunrise; thus they provide coverage of all areas at approximately the same local time of day, about 10 a.m. (Figure 5). Figure 4Any given point on the Earth's surface can be viewed by each satellite once every 16 days. Image data acquired from the satellite are down linked to ground stations directly, if the ground stations are near by, or else via the TDRS (Tracking and Data Relay Satellite systems) to ground stations at distant points, such as the principle receiving station at NASA's Goddard facility (Figure 6). Because of the huge data stream generated by the sensors, images are not recorded continuously, but are only acquired for specific areas according to the needs of individual investigators or agencies. Under the Reagan adminstration, the US civilian remote sensing program was commercialised, and the EOSAT corporation now markets and distributes image data.

Figure 5A full scene covers an area 185 x 185 km and is marketed in four quadrants. Each scene is imaged in 7 different wavelength bands (TM bands 1-7), from the visible to the near infra-red (Table I). The field of view is scanned mechanically using a moving mirror, with 16 detectors at each wavelength recieving the input radiance. Data at each wavelength is quantised over 256 DNs (digital numbers). Each scan line is divided into 6,320 samples, and there are 5,964 lines in each scene. Thus, a full TM scene consists of 2.7 x 107 pixels (picture elements) for each band. A pixel can be thought of as representing an area of 30 x 30 m on the ground, but in practice any one pixel recieves some radiance from adjacent pixels. For a full explanation of remote sensing techniques, terminology and analysis, refer to standard texts such as Sabins (1986).

Table I. Landsat Thematic Mapper spectral bands

Spectral features

In preparing Volcanoes of the Central Andes, color images created from combinations of TM bands 7, 4 and 2, covering much of the spectral range and displayed in red, green and blue (RGB) respectively, proved to be both easy to interpret and aesthetically pleasing. A few images were prepared using bands 5,4 and 2. Details of the TM data used are provided in Appendix I.

Figure 6The short wavelength infrared bands (TM Bands 5 and 7) are especially valuable in two ways:

First, snow and ice on the high volcanoes tend to saturate the shorter wavelength bands (1-4), limiting interpretability. Fortunately, at these wavelengths water absorbs rather than reflects radiation, and thus TM bands 5 and 7 are useful for subtle morphological studies on the snow covered parts of volcanoes. In a typical RGB color composite image created from TM bands 7, 4 and 2, snow appears bluish, because it is 'dark' in the red component, contributed by Band 7.

Second, because oxidised iron compounds such as limonite are extremely bright in the short wavelength infra red (TM bands 5 and 7) volcanic phenomena such as oxidised lava or scoria are extremely conspicuous, and thus some individual units are very easy to discriminate and correlate (Townsend, 1987). This spectral capability has been utilised for some more specific tasks: discriminating lithological units within debris avalanche deposits, and in detecting evidence of hydothermal alteration in fumarolic deposits.

Socompa volcano (VCA no. 33) provides a good illustration of the spectral capabilities of the TM system. The debris avalanche deposit from the volcano is the best exposed large avalanche deposit in the world. Although emplaced 7,200 yr BP, the deposit is exceptionally well preserved in the arid Andean environment. From a remote sensing viewpoint, the most interesting aspect of the deposit is the wide range of spectrally varied lithologies within the deposit (Figure 7). TM data have been invaluable in mapping the areal extents of different lithologies and thus tracing the trajectories of individual avalanche components back to the source regions on the volcano.

Most of the deposit is composed of clasts of dacite lava of different ages and provenances, together with ignimbrite from the sub-volcanic basement. Most prominent on the image are trains of oxidised dacite boulders which appear red to the naked eye and are bright in the short wavelength infrared. More important volcanologically are narrow streaks that are dark in all wavelengths; field inspection shows that these trains correspond to boulder trains of fresh, glassy black dacite. Many of the boulders exhibit 'breadcrust' and prismatic jointing structures, and they are clear evidence that there was hot magmatic material present within the volcano at the time of collapse. Figure 7P>A lithology which was generally of high albedo but with a marked absorption in Band 7 presented more challenging problems of interpretation. The lithology was identified in small patches only in the north-eastern part of the volcano, and it was interpreted initially as being derived from an area of hydrothermal alteration within the volcano. Field studies showed that this interpretation was erroneous, and that the deposits were composed of a mixture of chert, diatomite and carbonate material, clearly formed in a fresh-water lake and derived from sediments from the sub-volcanic basement.

Socompa is the volcano which we have studied most closely in purely remote sensing terms. Our work there emphasizes that the TM is remarkably useful in discriminating between volcanic rock units, largely as a consequence of varying degrees of iron oxidation. It is much more difficult to identify different lithologies - in TM terms, a fresh, glassy, dacite lava has similar spectral characteristics to a fresh basalt.

Thermal infrared data

We have shown elsewhere that emitted thermal radiance from high temperature (> 400ºC) volcanic thermal features can be unambiguously detected in TM bands 7 and 5, and that the sizes and temperatures of the anomalies can be constrained (Francis & Rothery, 1987; Glaze et al., 1989).

It might seem logical that the TM's thermal infra red sensor (Band 6) would be of most value in detecting active volcanoes, and would therefore play a prominent role in this compilation. Unfortunately, however, the 120 m pixel size of the Band 6 sensor greatly limits its utility in this respect, since magmatic thermal anomalies are typically much smaller than this. Furthermore, Matson & Dozier (1981) have shown that while it is possible to detect thermal anomalies which are much smaller than the pixel size, this requires that the anomalies are sufficiently hot that the anomalous pixels can be distinguished from the background. On the day time data that are routinely available, only gross thermal anomalies can be distinguished from the background of solar warming. Several of the volcanoes in the study area are known to possess persistent, low temperature steam fumaroles (< 100 ºC), such as that on Ollague volcano, Chile (VCA no. 18). We have been unable to detect these low temperature anomalies even on night time data. This is most likely due to the combination of the relatively low temperatures of the thermal sources, their small size, and the large pixel size of the Band 6 sensor. Figure 8

The thermal band has proved useful for monitoring lake water temperatures as illustrated by a study of Cerro Bonete, Argentina (CVA section 4.4). Derivation of surface temperatures from remotely sensed thermal infra-red data in the absence of ground and atmospheric reference parameters can only be carried out approximately. The measured at-satellite radiance can be derived from the raw data using known sensor calibration data, but derivation of the at-surface radiance requires assumptions concerning atmospheric temperature and transmission coefficients (Markham and Barker, 1986).

Data processing

Image data was routinely processed at the image processing laboratory of the Lunar and Planetary Institute, Houston using a Gould DeAnza system and processed images were written to film using an Optronics film writer. Linearly stretched false color composite images of TM bands 7,4 and 2 (in R, G and B respectively) further enhanced by spatial filtering, decorrelation stretches, and hue-intensity-saturation stretches were employed. Typical images presented in this compilation are composed of either 512 samples and 512 lines, covering an area about 15 km square on the ground, or else 1,024 samples and 1,024 lines, covering an area on the ground roughly 30 km square. For Socompa volcano, we have co-registered digitised 1:50,000 air photographs with TM data, and also used digital topography to construct a digital elevation model and then combined this with TM data (Figure 8). Using the TM image as a template, the pre-collapse topography of the volcano has been reconstructed, and the volume of the avalanche deposit calculated.


Overall, we found the most useful attributes of the TM data used in this study to be:

      (1) The 30 m pixel size, which permits possible subtle morphological interpretations not possible with 80 m pixel size MSS data. For example, we had noted the presence of an unusual deposit south of Tata Sabaya volcano, Bolivia, on early MSS images. When the higher resolution TM data were available , the hummocky topography of the deposit revealed unambiguously that it was a large volcanic debris avalanche deposit (Francis & Ramirez, 1985).

      (2) The synoptic view, so that a given structure can be studied in its entireity within the scope of a single TM quad. This attribute is particularly important for structural studies of large silicic calderas such as Pastos Grandes, Bolivia (Francis & de Silva, 1989; section 4.2).

      (3) The standardised coverage. All of the Central Andean region is covered by the same data set. Attempting to cover the same area with aerial photography would be fraught with difficulties, not least of which are the political problems of dealing with four seperate nations and their various regulations covering access to and dissemination of air photographs of frontier areas.

Space Shuttle photography

Collaboration with C.A. Wood at the NASA Johnson Spaceflight Center provided the source of the most useful supplementary data set: hand held photography from the Space Shuttle. Astronauts use Hassleblad 70mm cameras with 100 mm and 250 mm focal length lenses and Ektachrome film to photograph previously designated targets. Chief advantages of these photographs are:

       * synoptic coverage

       * high resolution (~25 meters at best)

       * variable viewing geometry

       * variable lighting geometry

       * stereo potential

       * low cost

Given the fixed, sun-synchronous nadir-looking views provided by the TM and its essentially one-time coverage, the variability afforded by the large numbers of Shuttle photographs provided a valuable means of making subtle morphological interpretations of features which were equivocal on TM scenes. The most important single use of the Shuttle photography was in subsidiary detailed studies of large volcanic debris avalanches. Following a special request, astronauts were able to identify the Socompa avalanche deposit from orbit, and obtain stereo photographic coverage which was valuable in interpretating the complex topography of the deposit. Since TM images are always obtained at mid-morning, with illumination from the east, Shuttle photographs obtained at different times of day, and particularly in the afternoon and evening, provided important additional perspectives of subtle topographic features. Fifteen previously unknown avalanche deposits were discovered using a combination of TM images and Shuttle photography (Francis et al., 1985, Francis & Wells, 1988). Others have been discovered during the preparation of Volcanoes of the Central Andes. In the Volcanoes of the Central Andes we have used several shuttle photographs to provide a regional perspective to the volcanoes we describe, as well as to highlight specific features which are particularly well illustrated on the image.

Aerial and ground photography

Where possible, we have included conventional ground and air photographs, since the different scales and viewpoints offered by such photographs often help considerably in the interpretation of the satellite images. We are grateful to our many friends and colleagues who helped in supplying these photographs, and they are credited in the appropriate places.

Topographic maps

Excellent topographic maps on scales of 1:50,000 exist for some parts of north Chile. Elsewhere in Chile, Bolivia and south Peru maps on scales of 1:250,000 are available. For northwest Argentina a few maps are available at 1;200,000, but for the most part 1: 1,000,000 Operational Navigation aeronautical charts had to be used. We have tried to include extracts from the best available topographic maps in each volcano entry.


Back to Volcanoes of the Central Andes