The use of Multispectral Satellite Data to retrieve temperatures from Volcanic Lakes


A. Bernard (



With the exception of the pioneering work of Oppenheimer (1997), satellite remote sensing has been rarely used for the monitoring of volcanic lakes. During the last few years, significant developments in spaceborne sensor technology have been achieved offering new tools for the study of volcanic lakes by remote sensing methods.

ASTER satellite launched in 1999, is probably one of the most powerful satellite for monitoring volcanic activity (Ramsey, Flynn and Wright, 2004). ASTER has 14 different spectral bands in the visible (VNIR : 0.52-0.86µm), near-infrared (SWIR : 1.60-2.43µm) and thermal infrared (TIR : 8.12-11.65µm) wavelengths. The TIR bands with a spatial resolution of 90 meters give the ability to detect small thermal anomaly (a few degrees C), perform thermal mapping and monitor temporal variations in the lake surface temperatures.

Many volcanic lakes are located in remote areas with difficult accessibility, these lakes are rarely visited or monitored. For these lakes remote sensing by satellite sensors can provide very useful information at relatively low cost.

Ruapehu lake in New Zealand and ASTER TIR image of April 11, 2001 showing hot spots at the lake surface.


Volcanic lakes act as calorimeters trapping most of the heat released by the magmatic-hydrothermal system (see CVL fundamentals).

Their temperatures are reflecting the balance between heat input from hydrothermal fluids and heat output by radiation and evaporation of the lake surface to the atmosphere.

So, lake surface temperature is a key parameter to detect any changes occurring in the activity of the volcano.


Heating episodes, sometimes cyclic, are relatively frequent in some volcanic lakes (like Ruapehu) and usually reflect changes in the flow rate or in the enthalpy of hot fluids entering the lake. These heating episodes always represent an alarming situation because an increasing lake temperature could be a precursory signal for the renewal of magmatic activity as was observed 3 months before the 1990 eruption of Kelud volcano (Vandemeulebrouck et al. 2000).

A sensor like ASTER with multispectral TIR is crucial for obtaining accurate surface temperatures. ASTER is clearly an improvment compared to Landsat TM or ETM+ which have only one TIR band and for which the correction of atmospheric effects is much more difficult or impossible.


Delmelle, P. and Bernard, A. 1999. Volcanic lakes. In: Encyclopedia of volcanoes. Ed. H. Sigurdsson. Academic Press, pp: 877-895.

Oppenheimer, C. 1993. Infrared surveillance of crater lakes using satellite data. Journal of Volcanology and Geothermal Research, 55:117-128.

Ramsey, M., Flynn, L. and Wright, R. 2004. Volcanic Observations from Space: New Results from the EOS Satellite Instruments. Journal of Volcanology and Geothermal Research, 135 1/2.

Vandemeulebrouck, J., Sabroux, J. C., Halbwachs, M., Surono, Poussielgue, N., Grangeon, J. & Tabbagh, J. 2000. Hydroacoustic noise precursors of the 1990 eruption of Kelut volcano, Indonesia. Journal of Volcanology and Geothermal Research,  97: 443-456.


1. Development of a new method:

The retrieval of temperature by remote sensing methods is based on the measurement of the spectral radiance emitted by the target (Planck’s law). Radiance is a function not only of temperature but also emissivity and the radiance (Ll) measured by a satellite sensor need to be corrected for atmospheric effects according to:

Ll = t [e B(T) +((1-e) Ld)]+Lu

where t is the transmissivity of the atmosphere, e is emissivity, B(T) blackbody radiance, Ld: downwelling atmospheric irradiance and Lu: upwelling path radiance.

A new technique for the retrieval of surface temperatures of volcanic lakes with ASTER data has been developped. This technique is based on the classical Split-Window method which has been successfully used with other satellites to obtain accurate sea surface temperatures (SST). The Split-Window method is based on establishing an empirical relationship between water temperatures measured at the surface and the brightness temperatures measured in at least two TIR channels. The use of two TIR channels enables a differential absorption measurement in order to remove the effects of atmospheric vapor and other absorbing constituents. This approach has the principal advantage to correct the atmospheric effects on a pixel by pixel basis i.e. for the local atmosphere and at the exact time the satellite is collecting TIR data. With ASTER data, the difference in brightness temperatures between bands 13 and 14 (BT13-BT14) was calculated to remove the atmospheric effects.

ASTER On-Demand L2 Brightness Temperature at the Sensor or the AST_04 product is used for brightness temperature data. This product is available from the NASA LP DAAC web site at a small cost (80 US$). More details on these data can be found here.

The crater lake of Taal volcano (Philippines) was used as a test site to develop the method. Taal volcano is a 15 x 22km prehistoric caldera occupied by the large freshwater Lake Taal (LT) and an active vent complex of the small Volcano Island where an acidic lake the Main Crater Lake (MCL) is present. The 1.2 km2 acidic lake in the Main Crater is only 2 m above the level of LT and its volume is estimated at 40 millions m3 (Delmelle et al., 1998). Since both lakes have almost the same elevation, the caldera lake can be used as a reference to remove the influence of atmospheric temperatures.

Delmelle, P., Kusakabe, M., Bernard, A., Fisher, T., DeBrouwer, S. and Del Mundo, E. 1998. Geochemical and isotopic evidence for sea water contamination of the hydrothermal system of Taal volcano, Luzon, the Philippines. Bulletin of Volcanology, 59:562-576.


(A) : simulated natural color of a VNIR ASTER image and (B): TIR image of 24 January 2003.
Hot spots from shallow hot springs are clearly visible in the North part of MCL in an area adjacent to a fumarolic field.


The method was originally developped by the combination of radiance data measured by two sensors ASTER and MODIS. Because these two instruments are on board the same satellite EOS Terra , they offer coincident Nadir observations of the same scene but with different spatial resolutions.


A good agreement exists between ASTER and MODIS radiances measured at sensor level attesting the capability of ASTER TIR to derive precise temperature data.


The MODIS SST data (MOD28L2 product) were used to provide lake surface temperatures and to derive empirically a new algorithm: AST_SW. MODIS SST is believed to be accurate within about 0.35°K for clear sky conditions.

The use of two TIR channels enables a differential absorption measurement in order to remove the effects of atmospheric vapor and other absorbing constituents. This approach has the principal advantage to correct the atmospheric effects on a pixel by pixel basis i.e. for the local atmosphere and at the exact time the satellite is collecting TIR data. A good agreement between MODIS SST and ASTER_SW is observed for both night and day images.


2. Application to Taal.

Evolution of MCL temperatures during 2000-2006:


With a record of 34 historical eruptions since 1572, Taal is considered as one of the most active volcanic centers in the Philippines. Recent eruptions were dominated by hydro-volcanic activity that produced devastating base surges and pyroclastic flows. These eruptions caused extensive damages to lake Taal shore and villages. Several million people are living within a 20 km radius of Taal’s caldera rim. The most recent unrest of Taal volcano occurred in 1991-95 when elevated seismicity, deformation and increasing lake temperatures were recorded.

More information on PHIVOLCS website.


During normal (quiet) activity, MCL temperatures are always a few degrees above those of the caldera lake (LT).

30 different ASTER images were processed with the new AST_SW algorithm. The results show relatively large fluctuations in the MCL surface temperatures for the period 2000-2006.
Most of these variations are non-volcanic in origin but are due to seasonal variations in the local air temperatures. LT lake can be used as a reference to remove the influence of the atmospheric component.

The thermal anomaly (Dtemp) calculated as the temperature differences between the two lakes remained almost constant during 2000-2006 and averaged: 2.5 ± 0.7°C.

The last heating episode at MCL occurred between 1994-1995 when temperatures increased as high as 39°C with a DT between MCL and LT of: 9 -10°C.


The satellite measures the skin temperature i.e., the temperature of the top millimeter (or little less) of the water surface which is generally a few tenths of a degree C cooler than the ‘bulk’ temperature, i.e., the temperature at ~1 meter deep. This skin temperature is partly influenced by the temperature of the air above it.



MODIS SST (monthly average) January 2005
There is a clear seasonal variation in the atmospheric temperatures with January as the coldest month (= dry season in the Philippines). The skin temperatures of Lake Taal closely follow this seasonal trend. Monthly average temperatures delivered by MODIS SST were used for Lake Taal (MO04MM product).

3. Validation of the method:

The method was developped on a low elevation volcano (Taal lake is at sea level) and for a tropical (humid) atmosphere. The validity of the AST_SW algorithm was thus checked on different lakes with different elevation and atmospheric conditions. Part of this validation was performed by Laura Trunk (University of washington) who applied independtly the algorithm developped here.


Here is an example of such validation made on Ruapehu volcanic lake in New Zealand (elevation: 2797 meters asl).

Here the satellite data (ASTER_SW) are compared to ground based measurements. Last image November 23, 2005. Data are accessible on the Hazard Watch web site.
Hazard Watch is a service of the Institute of Geological & Nuclear Sciences (GNS).

4. Other examples:


Ijen volcano (Indonesia)


The Ijen volcano is situated in East Java and is part of the easternmost volcanic complex in the island of Java.
Ijen lake is probably the largest natural accumulation of strongly acid waters in the world. The maximum depth of the lake is 170m and its volume = 35 millions cubic meters.
The chemistry of the lake waters is typical of acid sulfate-chloride fluids and its composition is extreme with a pH~0.2, SO4 = 80,000ppm, Cl = 25,000 ppm and Al = 6,000ppm. (Takano et al., 2004, Delmelle and Bernard, 1994).


Delmelle, P. and Bernard, A. 1994. Geochemistry, mineralogy, and chemical modeling of the acid crater lake of Kawah Ijen volcano, Indonesia. Geochimica Cosmochimica Acta, 58: 2445-2460.

Takano, B., Suzuki, K., Sugimori, K., Ohba, T., Fazlullin, S. M., Bernard, A., Sumarti, S., Sukhyar and Hirabayashi, M. 2004. Bathymetric and geochemical investigation of Kawah Ijen crater lake, East Java, Indonesia. Journal of Volcanology and Geothermal Research, 135 :299-329.


The caldera complex hosts a large number of volcanic edifices of which Ijen and Raung are the most active. Both volcanoes show clear thermal anomalies on TIR images.


Lake temperatures peaked at 48.1°C on July 13, 2004. For the same period, signs of increasing activity were reported by local observers. See Bulletin of the Global Volcanism Network, BGVN (11/2004)



More examples: Voui and others, soon...

and a useful link: ASTER image database for volcanoes...




All satellite images and data appearing on this page are free to copy or download if proper credit is given. These are original ASTER or MODIS data (courtesy of NASA, USA) processed at ULB Brussels.

Back to CVL home page

This site is hosted by the Université Libre de Bruxelles, Brussels, Belgium

Université Libre de Bruxelles. Last modification: November 18, 2005

Webmaster: A. Bernard