ÐÓ°ÉÔ­´´

In an average year, about sixty volcanoes disgorge ash or lava over the Earth’s surface. Another three thousand or so, which lie dormant, could come to life. Volcanoes have been erupting for millennia, but the risk associated with these eruptions – the potential loss of lives and livelihoods – is increasing as populations mushroom near active craters from Italy to Indonesia. The past decade has seen the worst volcanic disasters since the start of the century, and 25 000 people were killed as a result of just three eruptions: El Chichon in Mexico (1982), Nevado del Ruiz in Colombia (1985) and Mount Pinatubo in the Philippines (1991). There is a pressing need to improve ways of forecasting eruptions. (see Graphic)

Although few, if any, volcanoes erupt without warning, most eruptions are not predicted because only a small fraction of active sites is under routine surveillance. The signs of unrest are often subtle, requiring sensitive instruments to detect them, or they are simply hidden from view within lofty summit craters. Also, volcanologists are hampered by their incomplete understanding of how physical factors, such as the movement of ground water, and the gas content, temperature and viscosity of the magma, trigger and influence the style of eruptions. While there are many theoretical models of volcanic activity, their link with observation remains weak. The size and inaccessibility of lava flows, and the vast plumes of ash, gases and aerosols propelled into the atmosphere by explosive eruptions, defy close measurement. But remote sensing satellites, in orbits hundreds, or even tens of thousands of kilometres above the Earth, can safely image any volcano that is not hidden beneath the sea, no matter how remote or active it is.FIG-mg18763901.jpg

Traditionally, volcanologists monitor volcanic activity by two principal methods. The first involves analysing patterns of earthquakes taking place within and beneath volcanoes; the second, tracking deformation of the ground surface. Seismic techniques rely on detecting the quivering of volcanoes caused by magma rising up from below. Ordinary seismometers can pick up the distinctive ‘harmonic tremors’ caused by magma flow. These movements can also make the volcano swell slightly, and this bulge can be measured by precise ground surveys. Unfortunately, only a relatively small number of volcanoes are subject to this kind of scrutiny. Nor is it practical to set up a monitoring team and its equipment at all potentially active sites.

But with modest computing facilities and access to satellite data, a team of volcanologists could regularly ‘visit’ numerous volcanoes, perhaps all those in a given country or region, without leaving the comfort of their observatory. They could provide information on the progress of erupting craters too dangerous to approach. Equally usefully, they could target sites whose signs of unrest warrant more intensive studies at ground level.

Precisely what remote sensing satellites can ‘see’ depends on their orbits, optics and, most of all, on the part of the electromagnetic spectrum to which their sensors are tuned. Images taken in the ultraviolet and microwave bands of the spectrum can reveal constituents of the atmosphere that the eye cannot see, such as water vapour and ozone. Infrared detectors are sensitive to the heat radiated by the Earth. And the transfer of heat from one place to another is a fundamental feature of volcanic activity.

The heat becomes most obvious when incandescent lavas ooze from vents or fissures to form flows, lakes or domes, or when roaring columns of ash with a temperature of around 1000°C shoot into the atmosphere. Even if magma remains confined to its subterranean chamber or conduit, it may vaporise ground water, as well as expel its own dissolved gases. The resulting mixture of hydrothermal and magmatic fluids carries heat up to sulphurous fumarole vents on crater floors. Sometimes they create hot, corrosive crater lakes which can be brightly coloured by the iron and sulphur compounds they contain. Because a volcanic eruption is often preceded by a gradual ascent of magma, an increase in thermal activity at a volcano may indicate an impending eruption – in other words, volcanoes sometimes get hotter before they erupt.

DISTANT GLOW

Just as the fierce radiance from a red-hot lava flow can scorch flesh a few metres away, so it can reach, albeit much attenuated, an infrared sensor on a weather satellite 36 000 kilometres away. Once corrected for absorption and other atmospheric effects, a satellite infrared measurement at a given wavelength can be converted into a ‘model’, or apparent, surface temperature using a formula Max Planck put forward in 1900. This rather complicated looking equation relates a quantity called spectral radiance (which the satellite sensor records) to the wavelength of observation and the temperature of the surface viewed. As well as expressing what ‘red hot’ means in physical units, the equation explains why a red-hot surface glows red-orange, orange and then yellow as it becomes hotter: surfaces at a higher temperature produce more radiation at shorter wavelengths. So sensors for measuring ordinary Earth temperatures are set to respond to long wavelength ‘thermal infrared’ waves, and sensors for detecting red-hot or orange-hot lavas are tuned into the short-wavelength infrared waves that lie just beyond the visible spectrum.

Volcanic hot spots were detected from space as long ago as 1966, when one of the early meteorological satellites, Nimbus II, sensed thermal radiation from lava flows creating the new island of Surtsey off Iceland. But the 8 x 8 kilometre ‘footprint’ of the Nimbus instrument (the size of each pixel, or picture element, of the image, roughly equivalent to the resolution) meant that detailed thermal analysis could not be carried out, as the island itself measured only 1.5 kilometres across. More detailed observations of other volcanoes were made subsequently from aircraft, but these proved expensive and never provided a routine source of infrared data. Meanwhile, volcanologists became interested in the potential of images recorded by improved satellite-borne scanners, notably those carried by the Landsat series, the first of which was launched in 1972.

In 1986, Peter Francis and David Rothery from the Open University in Milton Keynes were examining a set of images of the central Andes, taken from an altitude of 700 kilometres by the Landsat Thematic Mapper . While inspecting the infrared bands, they noticed an intense ‘hot spot’ in the summit crater of Lascar, a remote, snow-dusted volcano perched high above the Altiplano plateau of Chile. Though eruptions of Lascar in the 1960s had been documented, this was the first ‘sighting’ of ongoing thermal activity there. While the scientists were awaiting publication of their discovery, Lascar erupted – on 16 September 1986. Ash drifted 300 kilometres downwind and rained gently over the town of Salta in Argentina. It was an almost imperceptible sprinkling compared with the tumultuous outburst of pumice from Mount Pinatubo in 1991, but clear evidence of the potential of satellites to pinpoint impending eruptions.

Remote sensing is capable of doing much more than detecting new thermal features or changes in existing ones. It can provide a quantitative insight into the thermal physics of volcanic processes. A better understanding of how volcanoes work should take the prediction of eruptions beyond an empirical science of pattern recognition. Recently, Lori Glaze and David Pieri at NASA’s Jet Propulsion Laboratory in California, and Francis, Rothery and myself at the Open University, have tried to refine methods for extracting thermal information from satellite data, asking whether it is possible to do more than simply say ‘yes, we can see an anomaly here’. We began, surprisingly, at sea.

Arguably the most rigorously established and widely practised application of thermal remote sensing is the estimation of sea surface temperatures. Comparisons with measurements taken using thermometers lowered into the sea from ships have shown that satellite measurements are capable of margins of accuracy better than 1°C. Many volcanoes have crater lakes, eruptions through which can be particularly devastating if the expelled water mixes with ash and other volcanic debris to form violent flows of mud and boulders. Several eruptions through crater lakes have been preceded by the temperature of the lake water rising by more than 10°C. In theory, the techniques for measuring sea surface temperatures could be adapted for monitoring crater lakes – a change of 10°C combined with a sensitivity of 1°C sounds favourable. But crater lakes are much smaller than oceans, and currently only the Landsat Thematic Mapper provides thermal infrared data with a spatial resolution high enough to be useful. Even then, its 120 x 120 metre footprint allows observation of only the larger volcanic lakes.

In February 1986, Landsat switched on its sensors above Poas, an active volcano in Costa Rica’s central cordillera. At the time, its crater contained a lake of warm, acid brine which was 300 metres in diameter. Applying Planck’s formula to the data recorded by the satellite suggested a water temperature of 36°C. Three days before the image was recorded, Jorge Barquero from Costa Rica’s Volcano Observatory had clambered down the vertiginous path to the water’s edge, where he measured a temperature of 38°C with an ordinary thermometer.

The discrepancy is negligible. Moreover, the value derived from Landsat is arguably more representative of the heat losses from the lake’s surface, as it was averaged over most of the lake area. Certainly, satellite surveillance is an excellent alternative to perilous descents by observatory staff into active craters, and even to self-recording thermometers, which tend not to last very long when left in the highly acidic and sometimes explosive environments of crater lakes.

Lava flows can be just as dramatic as volcanoes, but they seldom threaten lives, as lavas tend to move slowly. For example, the toe of the Mount Etna lava flow, which menaced the town of Zafferana in March and April last year advanced only a few metres per day. If necessary, the inhabitants could have been safely evacuated. However, houses, factories and roads cannot be moved out of the way, and the economic motives for predicting the paths and speeds of lavas are compelling.

Broadly speaking, theorists who work on the dynamics of lava flows are divided as to whether topographic slope or temperature is the more significant factor in how flows evolve. The first group claims that heat losses from flows are more or less incidental, and that the rate of eruption and the angle of the slope are the decisive influences on flow dimensions. According to the second group, lava temperature, and particularly the rate at which it drops further on in the flow, controls the course of an eruption. If a flow cools rapidly, for example, its viscosity increases, and it may stall over a comparatively short distance. ÐÓ°ÉÔ­´´s have been unable to test these competing hypotheses adequately because collecting real thermal data for entire flow fields by conventional field methods is impractical. Satellites, on the other hand, can provide thermal images of volcanoes in fractions of a second.

On Christmas day 1989, after a century of repose, Lonquimay volcano in the southern Andes of Chile reawakened. For over a year, black rubbly lava poured from a vent on the flank of the partly ice-covered volcano, travelling 10 kilometres down a valley where it singed the native forest of monkey puzzle trees. About halfway through the eruption, Landsat’s Thematic Mapper activated its sensors while passing overhead. As well as its thermal infrared band, the satellite has two short-wavelength infrared channels that can detect hot lavas.

Temperature measurements of active lavas are not as straightforward as those for water. Water temperatures are typically uniform over a satellite sensor’s footprint. But an active lava can juxtapose brightly incandescent lava, at perhaps 1000°C, within centimetres of black crust at only a few hundred degrees, so that a simple satellite measurement of such a surface would give a temperature somewhere in between. However, observing at two wavelengths simultaneously gives a much more accurate idea of what is happening on the ground.

To see how this works, imagine a wide lava flow whose surface is divided into regions at two different temperatures: 1000°C and 250°C, say. Its radiative properties can be described in terms of three independent parameters: the two temperatures and the area that each region occupies relative to the other. For each of two short-wavelength infrared measurements of this surface, it is possible to write an equation incorporating the three factors. If a value for one of the three is assumed, the values of the other two can be determined by solving a set of simultaneous equations.

In an analysis of the Lonquimay image, I set the temperature for incandescent cracks on small areas of the surface at 1040°C, based on field measurements. Solving the two equations for all the pixels of the flow suggested that only tiny areas of incandescent lava were exposed – most of the flow was covered by an insulating crust that doubtless contributed to the long distance the flow covered.

The temperature of this crust also decreased steadily with distance, from about 250°C near the vent to 170°C at a distance of 1.5 kilometres. Although these temperatures were estimated using sensors 700 kilometres above the Earth, their precision is of the order of 10°C. Such a cooling trend is reasonable; lavas are good thermal insulators, so heat cannot be transferred rapidly enough by conduction from the interior of a flow to balance thermal losses at the surface. The crust of a given parcel of lava therefore cools as it travels away from the vent.

Superimposed on the cooling profile were some other interesting effects, including a 90°C drop in crust temperature where the flow made a right-angle bend, and a sudden jump in surface temperature precisely where the lava dropped over the cliff-like front of an old flow. Such topographic steps are important factors in accelerating the cooling of flows by exposing hotter rock from their interiors. And as the flows cool their rate of advance slows.

MEASUREMENTS AND MODELS

To estimate the temperatures of volcanic hot spots from Landsat Thematic Mapper images, some simplifying assumptions have to be made. The main one is that the surfaces of interest can be represented by just two temperature components. Rothery and I have carried out temperature surveys at several active fumarole fields, such as at the island of Vulcano in the Lipari Islands north of Sicily, and the volcano Momotombo in Nicaragua, some of them venting gases at almost 900°C. Our thermal maps supported the two-component model. In similar studies, Luke Flynn from the University of Hawaii, made a range of close-up infrared measurements of lava lakes and flows on the island of Hawaii. These also suggested that the distribution of surface temperatures could be approximated by a model using two thermal components, one corresponding to lava with a chilled black crust and the other to lava glowing red hot in small cracks and fissures.

Two-component models may not always be appropriate. For lava bodies – flows, lakes or domes – that are narrower than a satellite sensor’s footprint, a third component, representing the much cooler surrounding ground, is probably necessary for an adequate measure of temperatures. However, solving three-component thermal models – which have five independent variables – is impossible with the Landsat Thematic Mapper because it can only provide two short-wavelength infrared observations. Volcanologists are now looking towards NASA’s Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), an instrument with 224 spectral bands, and its successors, to give them more scope.

AVIRIS made its first tour of active volcanoes in the summer of 1991. Carried on an adapted US spy plane, it flew over Mount Etna just five months before the eruption began. Rothery, Francis and I were beneath the flight path at the time, making close-range thermal measurements of fumarole vents to compare with the AVIRIS observations.

As we had hoped, AVIRIS easily detected the small ponds of lava that occupied each of Etna’s four summit craters, and by fitting Planck’s formula to the resulting spectra, it was possible to apply and solve three-component thermal models for the hot pixels. If only Mount Etna’s current eruption had begun a few months earlier, we might have had a tremendous snapshot of the infrared spectrum radiated by a major lava flow. A future AVIRIS trip to the island of Hawaii will provide another chance to image active flows.

In the long term, volcanologists dream of a satellite dedicated to regular surveillance of volcanoes worldwide. Thanks to the efforts of Pieri, and Mike Abrahms and Lonne Lane, also at the Jet Propulsion Laboratory, Sergio Vetrella from the University of Naples and Remo Bianchi at the National Research Council of Italy, the wish may come true. Engineers are already redesigning an existing scanning system to carry several hundred sensors in the ultraviolet, visible and short-wavelength infrared regions. The Volcano Experimental Ultraviolet and Infrared Observatory (VEXUVIO) should be launched in 1996 on a European Space Agency platform to be deployed from the Space Shuttle. The instrument could be tested, retrieved within six to nine months, then refurbished and launched into a polar orbit. A free-flying orbiting volcano observatory could make its debut before the end of the millennium.

Clive Oppenheimer is in the Department of Earth Sciences at the Open University, Milton Keynes.

* * *

Landsat’s Thematic Mapper

The Thematic Mapper (TM) is a remote sensing scanner fitted to the Landsat series of satellites. The TM’s sensors record in seven spectral channels (see diagram) within the visible and infrared regions. Bands 1, 2 and 3 correspond roughly to blue, green and red wavelengths respectively. Band 4 is just beyond the red end of the visible spectrum. Bands 5 and 7, in the short wavelength infrared region, are particularly sensitive to thermal emission from very hot surfaces such as active lavas. Conventional thermal mapping is also carried out at the longer wavelengths of the ‘thermal infrared’ band 6.FIG-mg18763902.GIF

Each square pixel (picture element) of a TM image represents an area on the ground of 30 x 30 metres, except for those of band 6 which cover 120 x 120 metres. A standard image, available as a photograph or as digital data for computer display, represents an area of about 185 x 185 kilometres on the Earth’s surface. The repeat cycle of the Landsat orbit allows any given locality to be imaged twice (once by day and once by night) every 16 days. Since 1972, virtually the whole Earth has been imaged by Landsat scanners. Landsat 6 (see below) is due for launch on 1 July, with an improved TM.