Landsat 7 night imaging of the Nissyros Volcano, Greece

A. GANAS

National Observatory of Athens, P.O. Box 20048, 11810 Athens, Greece;e-mail: aganas@gein.noa.gr

E. LAGIOS

Space Applications Research Unit in Geosciences, University of Athens,Panepistimiopolis, 15784 Athens, Greece

(Received 6 July 2001; in final form 1 November 2002 )

Abstract. A night-time image of Nissyros from the ETMz (EnhancedThematic Mapperz) scanner onboard the Landsat 7 satellite was acquired inOctober 2000. The image was processed to compute surface temperature on thevolcano while contemporaneous field measurements on the island’s surface werecollected. We confirm that the thermal sensor of Landsat 7 can map (a) the‘orographic effect’ on land surface temperatures (temperature falling withincreasing elevations) and (b) the crater surface temperature within an accuracyof 0.4–2‡C. In addition, the low-temperature fumarolic activity of the volcanocould not be detected on the mid-infrared bands (5 and 7). However, there aresome high-frequency temporal variations of surface temperature inside the maincrater that cannot be mapped because of the revisit capability of the sensor(16 days).

1. Introduction

Many workers have used Landsat 5 Thematic Mapper (TM) data to estimate

the surface temperature of volcanic surfaces (e.g. Harris and Stevenson 1997). In

addition, the advent of ETMz (Enhanced Thematic Mapperz) sensor onboard

the Landsat 7 satellite offers the capability of detecting emitted energy from the

Earth’s surface at the spatial resolution of 60m every 16 days. This capability gave

us the opportunity to study small volcanoes with crater diameters ranging between

100 and 250m such as those of the Nissyros volcano, Aegean Sea, Greece (figure 1).

Nissyros is a collapsed stratovolcano at the eastern end of the South Aegean

volcanic arc (e.g. Lagios et al. 1998), and shows both seismic unrest and fumarolic

activity, accompanied by recent gas and hydrothermal explosions (Papadopoulos

et al. 1998). The crater region is relatively flat (figures 1 and 2) so there is a

horizontal datum for detecting temperature anomalies.The purpose of our work is to study the effectiveness of Landsat 7 as a volcano

monitoring tool when combined with simple image processing tools. This Letter

presents our preliminary results. We chose to process night-time imagery in order to

(a) remove the solar heating signal evident in the day scenes and (b) maximize the

temperature difference between the crater and the surroundings (figure 2).

International Journal of Remote SensingISSN 0143-1161 print/ISSN 1366-5901 online # 2003 Taylor & Francis Ltd

http://www.tandf.co.uk/journalsDOI: 10.1080/0143116031000066279

INT. J. REMOTE SENSING, 2003, VOL. 24, NO. 7, 1579–1586

2. Field work

Two field campaigns were conducted in September 2000 and October 2000,

respectively. The campaigns aimed at collecting ground temperature data, twice a

day, to calibrate the surface temperatures computed from band 6 of the ETMz

sensor. The land surface data were obtained inside the Stefanos crater, and at the

two geothermal wells nearby. We also measured sea surface temperatures before

and after the sampling of land surface temperatures. In addition, we collected local

Figure 1. Map showing ground localities during October 2000 campaign. Backgroundcontours are from the 1:50 000 scale map sheet ‘Nissyros’ of the Hellenic ArmyGeographical Service. The black circle indicates the caldera rim. The black box showsthe extent of figure 2. The inset at lower left shows the location of the area within theGreek territory.

1580 A. Ganas and E. Lagios

meteorological data (air temperature, humidity, atmospheric pressure, wind speed,

wind direction) from a meteorological station established within the framework of

the GEOWARN project (www.geowarn.org), located at the point BAR in figure 2

(27‡ 10’ 03.1@E, 36‡ 34’ 46.4@N, elevation 114m). The sampling points were

established in a grid during the first campaign by use of a hand-held GARMIN

(a)

(b)

Figure 2. (a) Enlarged portion of figure 1 showing the ground temperature samplinglocalities around the main crater. Points A, B and C are located inside the crater andspaced 30m along the north–south direction, whilst point D is 30m to the east. PointBAR shows the location of the local meteorological station. (b) Field photograph ofthe Nissyros caldera showing the Stefanos crater. View is to the west.

Remote Sensing Letters 1581

Table 1. Temperature ground data. A, B, C and D are sampling points shown in figure 2. Sea surface temperature (SST) is measured at point FAROin figure 1. Shaded text indicates ground measurements during the satellite overpass. All temperatures are in ‡C. Time is local (GMTz3 h).

20 Oct 2000 13:56 13:35 14:36 13:25 14:55 12:12 15:30Depth A B C D Meteorol. Data Meteorol. data Well A Well B SST SST2 cm 31.2 33.6 35.0 33.6 Air temp.~19.7 Air temp.~20.1 23.4–23.5 23.34 cm 31.8 37.0 39.3 35.3 Hum.~48% Hum.~46% 28.6 28.57 cm 33.0 42.3 45.8 37.8 Pr~1000HPa Pr~1000Hpa10 cm 34.5 50.0 51.9 41.2 no clouds no clouds

41.2 kmh{1 47.4 kmh{1

20 Oct 2000 22:16 20:20 21:50 22:30Depth A B C D SST Meteorol. Data Meteorol. Data2 cm 22.2 25.2 28.3 24.2 21.6 Air temp.~18.1 Air temp.~17.64 cm 25.6 31.6 36.3 29.8 Hum.~50% Hum.~49%7 cm 28.0 36.3 41.8 35.5 Pr~1002HPa Pr~1003Hpa10 cm 31.9 43.9 48.6 35.1 no clouds no clouds

40 kmh{1 51.4 kmh{1 no wind

21 Oct 2000 12:20 11:50 11:35 11:20 10:40 14:45Depth A B C D Meteorol. Data Well A Well B SST SST2 cm 28.0 27.0 33.6 30.0 Air temp.~17.5 23.3 23.4 22.5 22.84 cm 28.1 27.0 37.1 30.6 Hum.~44%7 cm 29.1 27.2 42.8 30.7 Pr~1005Hpa10 cm 30.6 42.5 49.8 30.9 no clouds

36.8 kmh{1

21 Oct 2000 22:50 22:26 23:06 22:15 21:47 23:40Depth A B C D Meteorol. Data Meteorol. data Well A SST SST2 cm 19.8 20.9 25.8 20.0 Air temp.~16.2 Air temp.~15.9 20.6 22.4 21.84 cm 24.6 27.5 31.1 26.1 Hum.~44% Hum.~45%7 cm 28.1 40.0 39.0 30.9 Pr~1006HPa Pr~1005Hpa10 cm 30.0 47.6 47.6 34.9 no clouds no clouds

34.8 kmh{1

1582

A.GanasandE.Lagios

12-channel GPS, with a planimetric accuracy of 4–6m. At each point, ground

temperatures were recorded at specified depth intervals: 2, 4, 7 and 10 cm (table 1).

All points share the same elevation: 90m.The WGS84 positions of the sampling points inside the crater are as follows:

. point A: 27‡ 10’ 04.5@E, 36‡ 34’ 42.1@N;

. point B: 27‡ 10’ 04.4@E, 36‡ 34’ 41.0@N;

. point C: 27‡ 10’ 04.2@E, 36‡ 34’ 40.0@N;

. point D: 27‡ 10’ 05.6@E, 36‡ 34’ 40.9@N.

The field measurements were conducted by use of digital thermometers with

piercing probes because surface temperatures at the fumaroles do not exceed

103‡C. In October 2000, we used the FLUKE2 Type K thermocouple 80PK-5A.

The thermometer was calibrated against pots of hot water and was found to

measure the water boiling point with an accuracy of ¡0.3‡C. The October

measurements are shown in table 1. The temperature profiles of the crater points are

shown in figure 3. The September 2000 data were not used in further processing

because of a failure in acquiring the ETMz data by the ESA ground station in

Matera (Italy).

3. Image processing

The ETMz data were processed by the use of PCI software and ERDAS

Imagine 8.4. Fast-L7A imagery was acquired at the Level 1G (Systematic) Product.

The scene ID was L71045209_20920001020 (path 045, row 209 at the night world

reference system). The atmospheric correction software ATCOR2 (Richter 1996)

processed both thermal bands. ATCOR2 uses the theoretical approach by Singh

(1988). The calibration coefficients for the low gain band 6 (20 October 2000) were

as follows: offset~0, gain~0.00668 (mWcm{2 sr{1 mm{1). Then we geocoded the

Figure 3. Plot showing the temperature profile with depth at selected points inside theStefanos crater, on 19:28 GMT of 20 October 2000. The location of the points isshown in figure 2.

Remote Sensing Letters 1583

(a)

(b)

Figure 4. (a) Landsat 7 surface temperature map of the island of Nissyros, Aegean Sea. (b)Temperature map of the crater region. Black lines are elevation contours at 20mintervals. Note the thermal anomaly indicated by red pixels inside the Stefanos crater.

1584 A. Ganas and E. Lagios

product using a first-order polynomial and nearest-neighbour resampling. Ground

control points were selected from a 1:10 000 scale map coastline. We found that the

low-gain band mapped both sea and land surface temperatures accurately while the

high-gain band overestimated our ground measurements by 7–9‡C.Based on our local meteorological data and the relatively clear conditions

(visibility ranging between 15 and 25 km) during the night of the overpass we

selected the US 1976 Standard atmospheric model to perform the atmospheric

correction assuming constant atmospheric conditions over Nissyros. The US

Standard model was also used during the October, mid-latitude experiments of

Wukelic et al. (1989). The 20m elevation contours were also digitized and overlain

to show the temperature distribution with relief (figure 4(b)). Then, we checked if

our assumption for the ground spectral emissivity (0.98) was correct. In theory, for

surfaces in the emissivity range 0.97–0.99 (water, vegetation covered areas) the

ATCOR2 calculated ground temperature deviates less than 0.5‡C from the kinetic

temperature. This accuracy is comparable to the noise level of ETMz band 6. The

temperature result can be checked if the scene contains calibration targets (ideally

water surfaces of known temperature). Our results (figure 4(a)) confirmed that. The

sea surface temperature measured at Nissyros main port (Mandraki, figure 1) 1 h

before the time of the overpass was 21.6‡C, while the satellite sensor derived

temperature is 22‡C. Note that we have not filtered our results in order to remove

outliers and smooth out SSTs, because of the temperature differences reaching

more than 2‡C along the seashore. We think this temperature difference is real and

not an artefact of image processing. Moreover, filtering is more likely to affect

pixels at the open sea (figure 4) that are not of interest. However, emissivities for

rhyolitic volcanic surfaces such as Nissyros range between 0.95 and 0.97 (Harris

and Stevenson 1997). Therefore, we expect that ATCOR2 may underestimate land

surface kinetic temperatures up to 1.5‡C (Geosystems 2000).

4. Discussion and conclusions

A qualitative result in our temperature map is the ‘reproduction’ of the

orographic effect in land surface temperature (figure 4; compare with figures 1 and

2). It is reasonable to assign a high degree of confidence in this result as in day-time

imagery temperature also falls with increasing elevation (e.g. Warner and Chen

2001). A second result was the absence of any thermal anomaly in the short-

wavelength infrared bands (5, 7) of ETMz, indicating low temperature fumarole

activity (e.g. Rothery et al. 1988), as indeed was measured in the ground (table 1).

Despite the overall agreement of the image processing results with our ground data

there are some notable differences. This is evident inside the Stefanos Crater area

where the difference exceeds 2‡C in several localities. Table 1 shows that only the

2 cm temperature of point A (the most northern point) agrees with our ATCOR

calculations. The next closest point is D (24‡C ground versus 22‡C ETMz). Points

B and C differ by 3 and 6‡C, respectively, from the satellite sensor derived

temperature. We attribute this temperature difference to a combination of three

effects: (a) the smoothing effect of the ETMz pixel, (b) the lower spectral

emissivity of the crater’s surface than the one used during data processing (0.98)

and (c) the high energy flux in the south side of the crater (Brombach et al. 2001).

The latter effect may be the most influential as it is related to an east–west fracture

zone that crosses the crater and creates the vertical gradients seen in the profiles of

Remote Sensing Letters 1585

figure 3 (points B and C). This interpretation is enhanced by the groundmeasurements of the next day (table 1, 21 October 2000). Points B and C showtemperatures 4.3 and 2.5‡C less than the previous day, however, still higher thanpoints A and D. We also note that these high-frequency temporal variations ofsurface temperature cannot be mapped because of the revisit capability of thesensor (16 days). More research is needed to determine the role of other factors thathave contributed to the high-frequency effect, such as (a) local variations in airtemperature during the day, (b) local variations in wind speed, (c) fluctuations ofthe water table, and (d) variations in heat flux.

Acknowledgments

This research was funded by GEOWARN (IST 1999-12310, DGXIII). WummeDietrich, Carlo Cardellini, Irene Nikolaou, Vassilis Sakkas and Yannis Bakopoulosare thanked for useful suggestions. We also thank four anonymous reviewers forcomments.

ReferencesGEOSYSTEMS, 2000, ATCOR2 for ERDAS Imagine, User Manual (version 1.7), a Spatially-

adaptive Fast Atmospheric Correction Algorithm (Germany: Geosystems GmbH),93 pp.

BROMBACH, T., HUNZIKER, J. C., CHIODINI, G., CARDELLINI, C., and MARINI, L., 2001,Soil diffuse degassing and thermal energy fluxes from the southern Lakki plain,Nisyros (Greece). Geophysical Research Letters, 28, 69–72.

HARRIS, A. J. L., and STEVENSON, D. S., 1997, Thermal observations of degassing openconduits and fumaroles at Stromboli and Vulcano using remotely sensed data.Journal of Volcanology and Geothermal Research, 76, 175–198.

LAGIOS, E., CHAILAS, S., GIANNOPOULOS, J., and SOTIROPOULOS, P., 1998, Surveillance ofNissyros Volcano: establishment and remeasurement of Radon and GPS networks.Proceedings of the 8th International Congress of The Geological Society of Greece,Patras, 27–29 May, (Athens: ERE) vol. 32, pp. 215–227.

PAPADOPOULOS, G. A., SACHPAZI, M., PANOPOULOU, G., and STAVRAKAKIS, G., 1998, Thevolcanoseismic crisis of 1996–97 in Nissyros, SE Aegean Sea, Greece. Terra Nova, 10,151–154.

RICHTER, R., 1996, A spatially adaptive fast atmospheric correction algorithm. InternationalJournal of Remote Sensing, 17, 1201–1214.

ROTHERY, D. A., FRANCIS, P. W., and WOOD, C. A., 1988, Volcano monitoring using shortwavelength infrared data from satellites. Journal of Geophysical Research, 93,7993–8008.

SINGH, S. M., 1988, Brightness temperature algorithms for Landsat Thematic Mapper data.Remote Sensing of Environment, 24, 509–512.

WARNER, T. A., and CHEN, X., 2001, Normalization of Landsat thermal imagery for theeffects of solar heating and topography. International Journal of Remote Sensing, 22,773–788.

WUKELIC, G. E., GIBBONS, D. E., MARTUCCI, L. M., and FOOTE, H. P., 1989, Radiometriccalibration of Landsat Thematic Mapper Thermal Band. Remote Sensing ofEnvironment, 28, 339–347.

1586 Remote Sensing Letters