Monitoring the earth’s volcanoes from space poses a variety of unique problems and opportunities related to spatial and temporal scales, as well as spectral range (e.g., the necessity for thermal IR observations). Some of these are well posed with respect to the ASTER instrument and mission, and some challenge ASTER’s limitations.
Volcanoes represent one of the most active features of landscape generation. The frequency of discernable volcanic feature generation (as opposed to indiscernible fault or landslide/glacial creep) is exceeded only by aeolian-generated landforms and beach landforms (at times constantly changing at an observable spatial scale). Volcanic eruptions reported during human history range in frequency from daily (e.g., Mt. Etna, Italy; Mt. Sakurajima, Japan) to once every several centuries (e.g., Mt. Pinatubo, Philippines; Mt. St. Helens, USA). Currently there are of order 1000 volcanoes worldwide that are considered active, in the sense that they could enter a restless state or erupt more-or-less at any time (Simkin and Siebert, 1994).
The earth’s volcanic activity manifests itself at fairly well-defined spatial scales. Lava flows typically exhibit characteristic length scales of 1-10km, with exceptional flows reaching length scales of the order of 100km (e.g., ancient high temperature komatiite flows). Central vent volcanic features on the earth have characteristic base diameters of the order of 10-100km, with basaltic shields tending toward the high end and more compact stratovolcanoes tending toward the lower end of the range. Thus, ASTER’s 60x60km footprint and 15-30-90m/pixel (VNIR-SWIR-TIR) resolution scale allow most central vent volcanoes to be captured in just a few frames centered on the feature, and in many cases with just one image. At 15m/pixel (VNIR) and 30m/pixel (SWIR), ASTER images reveal the spatial distribution of lava flow and summit crater hotspots (Figures 1 and 2). Landsat ETM+ can provide comparable imaging and hotspot assessments; however, its dynamic range in the SWIR region is less than ASTER.
Figure 1. A false color composite of the three ASTER VNIR channels showing the 29 July 2001 summit eruption of Mt. Etna. Spatial resolution is 15 m/pixel. Areas appearing red are vegetated. Strong water vapor, ash, and SO2 emissions are evident.
Figure 2. A shortwave / longwave false color composite of the same scene as Figure 1. ASTER VNIR (Band 3), SWIR (Band 4), and TIR (Band 10) are reconstructed here in the blue, green, and red color planes. The active lava flows appear red and yellow, and are visible through much of the volcanogenic plume.
In addition, ASTER VNIR Bands 3N (nadir-looking) and 3B (backward-looking) provide parallax information that permits the construction of digital elevation models (DEMs) using classic stereo photogrammetry techniques. Such products are available on-demand from the USGS Eros Data Center in Sioux Falls, South Dakota. Such data are useful in assessing post-eruption flow volumes, and in helping with pre-eruption assessments of lava flow routing paths (Figure 3).
Figure 3. A three dimensional perspective view created from an ASTER digital elevation model with a simulated natural color ASTER image. El Misti volcano towers above the city of Arequipa, Peru, with a population of more than one million. Geologic studies indicate that a major eruption occurred in the 15th century. Despite the obvious hazard, civil defense authorities see it as a remote danger, and development continues on the volcano side of the city.
Finally, the detection and multispectral imaging of low altitude (<10km ASL) proximal volcanic plumes with weather satellite instruments (e.g., AVHRR, GOES) at nadir resolutions ranging from 1-4km/pixel is somewhat problematic (Hufford et al., 2000). This is because the plumes themselves are often small with cross-wind dimensions of order 100m and down-wind extensions of 1-10km. Nevertheless, the detection and imaging of such plumes is important not only in the context of basic science, but also in the arena of aviation safety. Understanding the magnitude and distribution of ash and sulfur dioxide in such emissions can only be accomplished utilizing TIR-based multiband techniques at relatively high spatial resolution. In this regard, ASTER’s capabilities are unique.
ASTER, given its 14 bands between 0.5µm and 12.6µm, can directly address a number of volcanological issues that require a multispectral approach. Much of what ASTER can do in this area can also be done with Landsat ETM+ or with experimental instruments on board the EO-1 spacecraft, though at somewhat coarser spatial resolution (e.g, identification of hydrothermal alteration and mapping of spectrally contrasting weathered units with VNIR, mapping of thermal radiance from active lava flows with SWIR, monitoring summit crater activity with SWIR). ASTER, however, has the unique capability of acquiring multispectral TIR data (5 channels) at less than 100m/pixel. This scale potentially allows the deconvolution of eruption precursor hot spot data with relatively low thermal contrast (theoretical limit given by NEΔT~0.3K; practical ΔT limit~5K) with relatively modest areas (≈7x104m2), as well as geologic mapping utilizing restrahlen signature principal-component contrasts, as was pioneered with the TIMS airborne instrument (Kahle et al., 1987).
As an example of low temperature thermal capability, warm spots associated with the currently erupting Chikurachki Volcano in the Kurile Islands were detected by ASTER as early as January 2003 before the April 2003 eruption (Pieri and Abrams, 2003). Enhanced heat flow, probably related to the subsequent explosive eruption, was detected in about a half dozen summit crater pixels, generating a consistent temperature contrast of between 5K and 10K above ambient, with temperatures hovering around the melting point of water. Between January and February 2003, the average temperature of these warm spots increased a statistically significant 1-2K. Though this analysis was carried out retrospectively, it shows the potential of ASTER as a very sensitive geo-thermometer for the detection of thermal precursors of volcanic eruptions.
SO2 is a fellow traveler with ash in explosive eruptions, and is often a key diagnostic that such an eruption has occurred. The TIR bands 10, 11, and 12 on ASTER are sensitive to an SO2 absorption band between 8 and 9 µm (Realmuto et al., 1994). Mt. Etna in Sicily is one of the world’s largest natural sources of SO2, erupting between 2500-5000MT/day during non-eruptive periods and 10,000-25,000 MT/day during paroxysmal eruptions. The NASA TOMS instrument, sensitive to UV absorption of SO2, has difficulty detecting the relatively low altitude (~3000-4000mASL) SO2 plume from Etna because of the general absorption of UV energy in the troposphere and TOMS band selection. ASTER, however, with higher spatial resolution and working on the TIR SO2 absorption feature at 8.5µm, has little difficulty in picking up the Etna SO2 plume (Figure 4).
Figure 4. One of the largest recent eruptions of Mt. Etna started on July 17, 2001 and continued until late August of that year. This ASTER image was acquired on Sunday, July 29, 2001 and shows the sulfur dioxide plume (in purple) originating form the summit, drifting over the city of Catania, and continuing over the Ionian Sea. The SO2 plume is distinguished by its absorption in ASTER TIR bands 10, 11, 12.
Even more problematic in terms of detection is the nearly constant very low altitude (<500-1000mASL) plume that has been emanating from the Pu’u O’o vent at Kilauea Volcano in Hawaii since 1983. While the emission rate is lower than Etna’s, the humid tropical air and SO2 conspire to produce coastal “vog,” or volcanic smog. Again, more weather-directed instruments like TOMS cannot see deeply enough into the troposphere to detect such a small plume. However, ASTER can clearly delineate the SO2 emission (Figure 5).
Figure 5. Kilauea Volcano, on the Island of Hawaii, has been in a constant state of eruption since January 3, 1983. The Pu'u O'o vent, formed soon after the onset of this eruption, is a persistent source of sulfur dioxide (SO2) gas emissions, as shown in this map of the Pu'u O'o SO2 plume derived from ASTER data. The map, produced from ASTER's thermal infrared channels, depicts color-coded concentrations superposed onto a false-color composite of ASTER's visible and near infrared channels. High SO2 concentrations (>4 gm/m2) are colored white, lower concentrations are red, orange, yellow, green and blue (<0.5 gm/ m2). ASTER is the only instrument in orbit that can detect passive venting of SO2 in plumes as small as the Pu'u O'o plume (typically less than 1.5 km wide over land). The image was acquired on October 30, 2001 and covers an area of 42 x 44 km. (Courtesy of Vince Realmuto, JPL).
ASTER, with its pointable platform in low-earth orbit, occupies a kind of temporal niche between the high spatial resolution, multispectral nadir-looking infrequent Landsat ETM+-style observations, and the more frequent—but lower spatial resolution—weather satellite data (e.g., AVHRR and GOES). ASTER’s VNIR and SWIR scan platforms are capable of nominal off-nadir pointing up to 8.55º, and occasional off-nadir pointing up to 24º. As mentioned above, this results in a revisit interval of as short as 5 days at the equator, and much shorter revisit intervals at higher latitudes. Repeated acquisitions of ASTER data, then, could result in useful time-series observations of eruptions, if the eruptions have characteristic timescales of order 10 days or more.
A graphic example of monitoring of summit craters utilizing ASTER data is shown in Figure 6, a view of the Popocatepetl Volcano, Mexico summit crater during 2000 and 2001. Here, the summit crater exhibits thermally active pixels in both SWIR and TIR ASTER channels from September 2000 through the beginning of the following year.
Figure 6. ASTER view of the summit of Popocateptl Volcano, Mexico. Summit crater radiances in both the SWIR and TIR channels are shown on the right. Red pixels are hot (>100ºC). Surrounding dark pixels are at background (~25ºC).
An example of a longer interval time-series was the ASTER discovery that Chiliques Volcano in Chile was active. On January 6, 2002 an ASTER nighttime thermal infrared image of the Chiliques volcano showed a hot spot in the summit crater and several others along the upper flanks of the edifice, indicating new volcanic activity (Figure 7). Examination of an earlier nighttime thermal infrared image from May 24, 2000 showed no thermal anomaly.
Figure 7. Chiliques Volcano, Chile. ASTER nighttime thermal data discovered a thermal anomaly in January 2002, continuing in April 2002 (two left hand images). A visible image (two right hand images—lower one is a blow-up) from 14 March 2002 reveals two crater lakes at summit (dark areas against white snow), that have become hot. Chiliques has shown no historic activity, but is re-awakening.
Chiliques volcano was previously thought to be dormant. Rising to an elevation of 5778 m, Chiliques is a simple stratovolcano with a 500-m-diameter circular summit crater. Officials at the Chilean Geologic Survey reported that the summit hot spot indicated that the crater lake was heating up. During an aircraft overflight fumaroles were observed on the volcano’s flank. Such ASTER data point up the utility of time-series observations at high spatial resolution and the utility of simultaneous infrared observations.
The ASTER mission represents a fundamentally new type of observational capability with respect to studying the earth’s volcanoes. The new features of ASTER are:
- (a) an unprecedented number of thermal infrared channels (five);
- (b) the ability to point up to 24º off-nadir, resulting in a five day revisit interval;
- (c) the ability to carry out simultaneous along track stereo observations;
- (d) high spatial resolution: 15 m/pixel in the VNIR instrument, 30m/pixel in the SWIR instrument, and 90m/pixel with the TIR instrument.
Hufford, G., J.J. Simpson, L. Salinas, E. Barske, and D.C. Pieri, 2000, Operational considerations of volcanic ash for airlines, Bulletin of the American Meteorological. Society, 8 (4), 745-755.
Kahle A.B., 1987, Surface Emittance, Temperature, and Thermal Inertia Derived from Thermal Infrared Multispectral Scanner (TIMS) Data For Death-Valley, California, Geophysics 52 (7): 858-874.
Pieri D. and M. Abrams, 2003, ASTER pre-eruption thermal analysis of Chikurachki Volcano, in preparation for Geophysical Research Letters.
Realmuto VJ, Abrams MJ, Buongiorno MF, Pieri DC, 1994, The Use of Multispectral Thermal Infrared Image Data To Estimate the Sulfur-Dioxide Flux from Volcanoes - A Case-Study from Mount Etna, Sicily, July 29, 1986, Journal of Geophysical Research-Solid Earth 99 (B1): 481-488.
Simkin, T, and L Siebert, 1994, Volcanoes of the World (2nd Edition), Geoscience Press, Inc., Tucson, and the Smithsonian Institution, Washington, D.C., 349pp.
|8,313 hits since 09/17/99.|
Updated:9/10/2002 2:02:41 PM
Hazard assessment for flank eruptions
The historical records of flank eruptions at Etna date back over 2000 years, but are continuous and reliable only from the 17th century22. The eruptive behavior of Etna in the last 400 years is undoubtedly irregular, with important fluctuations in frequency, type of eruptions and lava discharge rates, both in the long (centuries) and short (decades) terms6,23,24,25. During the past century, major flank eruptions causing significant damage to crops and disruption in towns and villages have occurred in 1910, 1923, 1928, 1971, 1979, 1981, 1983, 1991–1993, 2001, and 2002–20035,8,22. An evident increase in the lava output and frequency of eruptions began in 1971. Since then, 17 eruptions have occurred mostly along the main rift zones (Fig. 1), showing that the volcano has been in a strongly active period during the last 400 years8.
In flank eruptions at Mt. Etna, lavas erupt from newly formed fissures and vents, hence, the potential spatiotemporal distribution of new vent openings must be estimated as part of the analysis. In order to identify the most probable emission zones of future lava flows, we analyzed the spatial location of past eruptive vents, as well as the eruption frequency within a time window. Recently, we conducted a detailed study22 of the main structural features of flank eruptions at Etna, including the outcropping and buried eruptive fissures active in the past 2000 years, the dykes in the Valle del Bove depression7,26 (Fig. 1a), and the main faults that can potentially be used as pathways for intruding magma27 and/or influence the surface stress field of the volcano. It is noteworthy that shallow feeder dykes guided by faults are not very rare on Etna, such as in the 1928 eruption27, while in other volcanoes worldwide the use of faults as channels is uncommon28. All these geological and structural features were used to construct the spatial probability map of vent opening at Etna27. Thanks to the completeness and accuracy of historical data over the last four centuries, we demonstrate spatial non-homogeneity and temporal non-stationarity of flank eruptions on Etna, showing that effusive events follow a non-homogenous Poisson process with space-time varying intensities8. Here, we assess the distribution of future vent locations using a probabilistic modeling (Methods) based on the locations of the pre-1600 eruptive fissures, dykes and faults (for which an exact date cannot be established), and on the spatial positioning and temporal sequencing of the post-1600 flank eruptions (that are accurately documented and well-dated). We calculate the recurrence rates (events expected per unit area per unit time) and produce a spatiotemporal probability map of new vent opening for the next 50 years (Fig. 2a). The distribution of probabilities is strongly non-homogeneous. The highest probability of new eruptions occurs in the areas closer to the summit of the volcano, at elevation higher than 2500 m. Diversity in the probability estimates also occurs in lower elevation areas, between 2500 and 1800 m. Below 1000 m, the expectation of new vent opening is very low, except in the South Rift, where probabilities slowly decrease to an elevation of ∼600 m.
(a) Spatiotemporal probability map of vent opening at Mt. Etna for the next 50 years. The probabilistic modeling is based on the locations of the pre-1600 eruptive fissures, dykes and faults, and on the spatial positioning and temporal occurrence of post-1600...
To estimate the probability of occurrence of eruptions, a complete revision of location and opening dynamics of all flank eruptions occurring in the past 400 years was carried out. For each lava flow field, we collected the main quantitative volcanological data, concerning the eruption start and end, the eruptive style, the area covered by flows, the lava volume emitted, and the fissure producing the lava flow. From an accurate analysis of the distribution of flow durations and lava volumes, we selected a time barrier at 30 days and two divisions for the total volume emitted, at 30 and 100 × 106 m3. Thus we obtained six eruptive classes, five of which are populated (Fig. 2b). This classification of expected eruptions permitted calculation of local occurrence probabilities for each class, and the deduction of essential information for the lava flow simulations (Methods).
The eruptive scenarios were simulated using MAGFLOW29,30, which is based on a physical model for the thermal and rheological evolution of the flowing lava. To determine lava flow emplacement, MAGFLOW requires several input parameters: a digital representation of the topography, the chemical composition of the lava, an estimate of the effusion rate, and the location of the eruptive vent. As digital representation of the topography, a 10-m resolution DEM of Etna updated to 2007 was utilized. For the chemical composition of the lava, the typical properties of Etna's basaltic rocks31,32 were used (Methods). The effusion rate, e.g. the discharge rate controlling how a lava body grows, was derived from the six eruptive classes obtained from the characterization of expected eruptions. Since the numerical simulations require flow duration and lava volume to be defined, short and long time periods were set at 30 and 90 days respectively, while 30, 100 and 200 × 106 m3 were fixed as values for total volume of lava emitted. These values of duration/volume produced six possible bell-shaped curves16 representing the flux rate as a function of time (Fig. 2c). Locations of eruptive vents are the nodes of a regular grid with boundaries shown in Fig. 1. For every potential vent, six simulations were executed, each one representative of a particular eruptive class.
Finally, assessment of the long-term hazard related to lava flows at Mt. Etna rests on combining the spatiotemporal probability of future opening of new eruptive vents, occurrence probabilities for each class of expected eruption, and the overlapping of a large number (28,908) of numerical simulations (Methods). The hazard map for the next 50 years is shown in Fig. 3. The highest hazard level is reached in the well-delimited zone inside the Valle del Bove, a 5 × 7 km morphological depression on the eastern flank of the volcano able to capture lava flows emitted from the eruptive vents below the summit craters, and along the upper portions of the South and North-East Rifts. Only eruptions exceptional in duration and lava output could form lava flows capable of travelling across the entire valley and seriously threatening the towns of Zafferana Etnea and Santa Venerina in the eastern sector. Other areas more likely to be threatened by lava flows are those on the southern flank, including densely populated areas near the towns of Nicolosi, Pedara, Trecastagni, and Ragalna. Towns located at lower elevations could be equally threatened by lava flow inundation, though the hazard progressively diminishes at greater distances from volcano summit. On the northern and western sectors of the volcano, the main population centers exposed to lava inundation are Linguaglossa, downslope of the North-East Rift, and Bronte, located along the likely trajectory of lava flows emitted from the West Rift.
Hazard map by lava flow inundation at Mt. Etna, based on 28,908 simulations of lava flow paths starting from 4,818 different potential vents.
Validation of the lava flow hazard map is an additional task in our four-stage methodology, giving a quantitative evaluation of its reliability. As historical eruptions provide the only available information, we propose a retrospective validation using more recent past eruptions to indicate potential future events33. In particular, we calculate a fitting score to measure the overlap of the probability distribution of lava flow hazards at Etna up to the year 1981 against the historical lava flow paths that occurred after 1981 (Fig. 1). The good agreement between the observed and expected inundation areas obtained for the hazard map up to 1981 confirms the consistency of our results33.
Hazard assessment for the summit eruptive activity
The summit area of Mt. Etna has frequently undergone major morphological changes due to its persistent eruptive activity, both effusive and explosive (Fig. 1a). A single, large Central Crater (CC) existed until the beginning of the 20th century, but soon afterwards the summit area was affected by repeated subsidence and collapse phenomena alternating with intracrater volcanic activity and the birth of new summit vents34. The North-East Crater (NEC; formed in 1911) opened a few hundred meters north of the CC rim. The Voragine (VOR; 1945) and the Bocca Nuova (BN; 1968) opened inside the CC. Finally, the South-East Crater (SEC; 1971) is the youngest and currently the most active summit vent, as also testified by the growth of a large new cone on its southeast flank, identified as the New South-East Crater1 (NSEC; 2007). The progressive increase in summit eruptive activity during the past 110 years is confirmed by the results of the statistical analysis8, indicating that the hazard from eruptive events is not constant with time and differs for each summit crater of Mt. Etna. A clear increase in the eruptive frequency is evident from the middle of the last century and particularly from 1971, when the SEC was formed.
The probability that a lava flow will inundate a certain area was estimated assuming an effusive eruption originated from a vent within the summit area of Mt. Etna. This was done using the methodology adopted for flank eruptions, including the development of the spatiotemporal probability map of future vent opening, the occurrence probabilities associated with the classes of expected eruptions, and numerical simulation of a number of eruptive scenarios.
Firstly, we calculated a spatiotemporal probability map of future vent opening for the next 10 years using the spatial location of the existing four summit craters (CC, NEC, SEC and NSEC) and the frequency of eruptions since 1900 (Methods). This produced the spatiotemporal probability map for future vent opening shown in Fig. 4a.
(a) Spatiotemporal probability map of vent opening at Mt. Etna for the next 10 years within the summit area. The grid of potential vents, defined by a circular region (dashed area), has a regular spacing of 100 m. The probability values are calculated...
Subsequently, the classes of expected summit eruptions were identified. We used flow duration/lava volume distributions for paroxysmal events and Strombolian activity at the summit craters1,2, to distinguish ten classes. For the paroxysmal events, lava flows and lava fountains occurring at VOR and SEC since 1998 were considered. We established three different duration intervals (≤4, 4–8 and >8 hours), and emitted lava volumes (≤1, 1–2 and >2 × 106 m3), resulting in nine possible eruptive classes1 (Classes I–IX; Fig. 4b). The last class of expected eruptions (Class X) was identified on the basis of Strombolian activity from 1955 to 1996. Classification of expected eruptions enables probabilities of occurrence to be calculated (Methods), as well as deriving the ten effusion rate trends associated with each class for the MAGFLOW simulations. For Classes I–IX of paroxysmal events, we used the bell-shaped curves shown in Fig. 4c. Since the Strombolian episodes are often characterized by a continuous mild effusive activity, we assigned to the Class X a constant rate equal to the Mean Output Rate (MOR ≈ 0.5 m3 s−1; see Fig. 4b) for the entire duration (150 days) of the eruption.
To evaluate areas more likely to be affected by lava flow paths we computed ten simulations, corresponding to the expected classes of eruptions (Classes I–X), for each of the potential vents of the grid defined around Etna's summit craters (Fig. 4a). The topographic base was the 10-m resolution DEM of Etna updated to 200734, while the chemical composition of the lava was defined using the properties of products erupted from Etna's paroxysmal events35.
The hazard map for summit eruptions was obtained by combining the spatiotemporal probability of future vent opening, the probability of occurrence for each eruptive class and a large number of simulations (5,950) computed with the MAGFLOW model. The resulting hazard map (Fig. 5) estimates probable areas of inundation by lava flows issuing from vents within the summit area over the next 10 years. The inundated area measures about 51 km2, while the maximum distance of the simulated lava flow paths is reached at the end of Valle del Bove (∼700 m elevation), and in the direction of Linguaglossa, ∼1200 m elevation. Close to the summit area, the northern tourist facilities (Piano Provenzana) located above 1800 m are exposed to lava inundation, although no simulated flows erupted from the summit vents inundated the southern tourist facilities (Rifugio Sapienza) at ∼1900 m. New pyroclastic cones formed on the southern flank during the 2001 and 2002–2003 eruptions, created an effective topographic barrier to these lava flows, diverting lava west or east of the cones.
Hazard map for lava flow inundation at Etna's summit area.