Measuring volcanic degassing of SO2 in the lower troposphere with ASTERband ratios

Robin Campion a,!, Giuseppe Giovanni Salerno b,d, Pierre-François Coheur e, Daniel Hurtmans e,Lieven Clarisse e, Kohei Kazahaya f, Michael Burton c, Tommaso Caltabiano b,Cathy Clerbaux e, Alain Bernard a

a Université Libre de Bruxelles, Département des Sciences de la Terre et de l'Environnement. 50 Av. Roosevelt, CP160/02, 1050 Bruxelles, Belgiumb Istituto Nazionale di Geo!sica e Vulcanologia—Sezione di Catania. Piazza Roma 2, 95125, Catania, Italyc Istituto Nazionale dei Geo!sica e Vulcanologia—Sezione di Pisa. Via della Faggiola 32, 56126, Pisa, Italyd University of Cambridge, Department of Geography. Downing Place, Cambridge, CB2 3EN, United Kingdome Université Libre de Bruxelles, Chimie Quantique et Photophysique, 50 Av. Roosevelt, CP160/09, 1050 Bruxelles, Belgiumf Geological Survey of Japan, Institute of Advanced Science and Technology, Tsukuba, Ibaraki 305-8567, Japan

a b s t r a c ta r t i c l e i n f o

Article history:Received 23 November 2009Accepted 21 April 2010

Keywords:remote sensingSO2ASTERDOASEtnaMiyakejima

We present a new method for measuring SO2 with the data from the ASTER (Advanced Spaceborne ThermalEmission and Re!ectance radiometer) orbital sensor. The method consists of adjusting the SO2 columnamount until the ratios of radiance simulated on several ASTER bands match the observations. We present asensitivity analysis for this method, and two case studies. The sensitivity analysis shows that the selectedband ratios depend much less on atmospheric humidity, sulfate aerosols, surface altitude and emissivity thanthe raw radiances. Measurements with b25% relative precision are achieved, but only when the thermalcontrast between the plume and the underlying surface is higher than 10 K. For the case studies we focusedon Miyakejima and Etna, two volcanoes where SO2 is measured regularly by COSPEC or scanning DOAS. TheSO2 !uxes computed from a series of ten images of Miyakejima over the period 2000–2002 is in agreementwith the long term trend of measurement for this volcano. On Etna, we compared SO2 column amountsmeasured by ASTER with those acquired simultaneously by ground-based automated scanning DOAS. Thecolumn amounts compare quite well, providing a more rigorous validation of the method. The SO2 mapsretrieved with ASTER can provide quantitative insights into the 2D structure of non-eruptive volcanicplumes, their dispersion and their progressive depletion in SO2.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Volcanism is the leading natural source of sulfur dioxide (SO2) to theatmosphere, with average annual emissions estimated at 10 to40.109 Gg/year, according to different authors (Le Guern, 1982; Stoiberet al., 1987; Andres and Kasgnoc, 1997; Graf et al., 1997). Theseemissions occur through eruptions and passive degassing. The lowatmospheric background of SO2, together with its strong absorptionfeatures in the ultra-violet (UV) and infrared (IR), makes this gas a goodtarget for spectrometric measurement from "eld and space. SO2 !uxmeasurements have been performed since the 1970s by UV correlationspectrometry (e.g.: Newcomb and Millán, 1970; Stoiber et al., 1983).They have proven to be valuable for our understanding of eruptiondynamics and eruption forecasting (see Williams-Jones et al., 2008, foran extensive review of theory, practice and case studies of SO2 "eld

measurements on volcanoes). The most commonly used instrumentsfor SO2 "eldmeasurements are now compact UV spectrometers with orwithout calibration cells (Galle et al., 2002, Elias et al., 2006). However,these sensors can only measure scans or transects of SO2 columnamounts, while volcanic plumes are actually dynamical 3D objects. Toexpand the investigations on volcanic plumes, "eld UV imaging deviceshave been developed recently (Bobrowski et al., 2006;Mori and Burton,2006; Bluth et al., 2007, Dalton et al., 2009) and these technologies arecurrently being rapidly developed as volcano monitoring tools. Inaddition, techniques using several spectrometers are starting to beapplied to evaluate vertical distribution of SO2 in the plume throughtomography algorithms (e.g. Kazahaya et al., 2008).

Spacebased sensors are also able to detect andmeasurevolcanic SO2.They provide a more complete view of large scale volcanic plumes andare the only tools allowing a quanti"cation of thematerial (gas, ash andaerosols) injected into the atmosphere by large eruptions. Currentspace-borne sensors sensitive to SO2 (see Annex table of Appendix A)are MODIS, ASTER, AIRS, SEVIRI and IASI operating in the thermalInfrared (TIR), and GOME-2, SCIAMCHY, and OMI in the UV. Thanks to

Journal of Volcanology and Geothermal Research 194 (2010) 42–54

! Corresponding author. Tel.: +32 26502239.E-mail address: Robin.Campion@ulb.ac.be (R. Campion).

0377-0273/$ – see front matter © 2010 Elsevier B.V. All rights reserved.doi:10.1016/j.jvolgeores.2010.04.010

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research

j ourna l homepage: www.e lsev ie r.com/ locate / jvo lgeores

their high sensitivity and daily (or nearly daily) global coverage, OMI(Carn et al., 2008), AIRS (Prata and Bernardo, 2007) and IASI (Clarisseet al., 2008) have already proved very useful for quantifying eruptiveemissions of SO2 as well as tracking SO2 transport up to several weeksafter the eruption. Thanks to its high acquisition frequency (1 imageevery 15 min), the geostationary sensor SEVIRI can also be very usefulfor near real-time monitoring of volcanoes located in its observationzone. However the low spatial resolution of these three sensors, and/ortheir weak sensitivity to tropospheric SO2 complicate the quanti"cationof SO2 !ux in passive degassing plumes.

ASTER was launched in December 1999 on board the EOS Terrasatellite on a sun-synchronous orbit (Pieri andAbrams, 2004). Dependingon latitude, theminimal timedelaybetween two imagesvaries between1and7 days, somewhat limiting its use for routine volcanomonitoring. Thepresence of a tight cloud cover at the timeof over!ight can further extendthe time delay between two exploitable images. ASTER has "ve spectralbands in the TIR, with a 90 m ground resolution at nadir. The bands arelabelled from B10 to B14 with increasing wavelengths and have theircentroids located at 8.3, 8.6, 9.1, 10.6 and 11.3 µm. Following thepioneering works of Realmuto et al. (1994, 1997) for SO2 retrieval withTIR sensors, ASTERwas recognized, prior to launch, as a potential tool formeasuring volcanic SO2 from space (Corradini et al., 2003). Urai (2004)adapted their retrieval scheme to an ASTER image ofMiyakejima volcano(Japan). The method consists of applying a non-linear regression toMODTRAN-simulated radiance in twoASTER spectral bands (band11andband 14) for varying SO2 column amounts. More recently, Pugnaghi et al.(2006) developed several fast retrieval schemes, adapted forMount Etna,approximating the full radiative transfer calculations with functions ofmonthly averaged atmospheric parameters. Both studies computed SO2

!uxes in good agreement with ground-based measurements.

However a number of parameters other than SO2 can affectretrievals that use ASTER radiances. These parameters are plumetemperature, atmospheric humidity, surface altitude and surfaceemissivity. To reduce the effect of the last three parameters, wedeveloped a new method for measuring SO2, which uses radiativetransfer simulation and "tting of ASTER band ratios (B10+B12)/(2B11) and B14/B11. These ratios were selected because of their verylow sensitivity to atmospheric humidity, surface altitude andemissivity, which makes the retrievals more robust and straightfor-ward (see Section 3).

2. Methodology

2.1. Radiative transfer simulations

SO2 molecules in the plume absorb IR radiation emitted by theEarth's surface and lower atmosphere. The !1 absorption band of SO2,extending from 8.1 to 9.5 "m, mostly affects ASTER band 11, and, to aless extent, bands 10 and 12 (see Figs. 1 and 2). The radiance RAmeasured by the sensor is given by the general equation of the TIRradiative transfer in the atmosphere, where wavelength dependenceis omitted for sake of clarity

RA = #:B T0! " · $0 + !A

0

B Tz! " · "$z"z dz + 1##! "Rd · $0 !1"

B(T) is the blackbody Planck function for temperature T, ! is thesurface emissivity, z is the altitude, "Z is the transmittance betweenaltitudes z and A, and Rd is the downwelling radiance at the Earth's

Fig. 1. (a) Modelled spectra of Top of Atmosphere radiance for a “clean” atmosphere (black) and for an atmosphere containing 10 g/m! of SO2 at an altitude of 3 km (red dashed).(b) Normalised Spectral Response Function of ASTER TIR bands.

43R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

surface. The altitudes of the surface and the sensor are respectively 0and A. In this equation, the "rst term represents the Planck radiancethat is emitted by the Earth's surface at temperature T0, and partlytransmitted by the atmosphere. The second term is the radianceemitted by each layer of the atmosphere, including the plume (the so-called “path radiance”), and the third term is the part of downwellingradiance that is re!ected by the surface. In absence of sun glint, thislast term becomes negligible in the TIR.

Radiance is simulated with Atmosphit, a line-by-line radiativetransfer code developed at the Université Libre de Bruxelles (e.g.Coheur et al., 2005). Atmosphit simulates high-resolution spectra,considering a layered model atmosphere with spectral parametersfrom the HITRAN database (Rothman et al., 2008) and the MT_CKDmodel (Clough et al., 2005) for H2O and CO2 continua. Fixed

parameters for the simulations include the surface altitude andemissivity, and the atmospheric pro"les of temperature and concen-tration of 38 atmospheric molecules. In the spectral region of interestfor ASTER TIR, themolecules that absorb radiation are, in their order ofimportance, H2O, O3, CH4 and N2O. The concentration pro"les of thelatest three molecules are kept "xed and based on standardatmospheric models, because of their weaker absorption and lowervariability over short spatial and temporal scale. Temperature andH2Opro"les were de"ned using radiosonde data from the nearestlaunching site (available at the RAOB website: http://raob.fsl.noaa.gov/ ). For the simulations, the volcanic plume is considered as a layercontaining given SO2 column amounts in thermal equilibriumwith theatmosphere. Its thickness is assumed to be 400 m, but the in!uence ofthis parameter on the retrievals was found to be negligible. Surfacetemperature, plume altitude and SO2 amount in the plume, are variedincrementally.

2.2. Retrieval scheme

The simulated spectra are then integrated over the spectralresponse functions of ASTER (available at http://www.science.aster.ersdac.or.jp/en/about_aster/sensor/index.html), to get the simulatedASTER radiances on the "ve bands. The simulations are used togenerate "ve (one per band) 3D matrices containing the radiancevalues for each SO2 column amount, ground temperature and plumealtitude.

A region free of SO2, representing the variability of temperatures andsurface types present under the plume, is selected close to the plume.The pixels of this region are used as a blank to correct slightdiscrepancies between the simulated and measured radiances. Plumealtitude, together with wind speed, is measured on the Visible-Near-Infrared (VNIR) stereoscopic images with themethod described by Urai(2004). This method is based on the displacement of distinctive plumefeatures between the nadir and backward looking images, but has somelimitations (i.e. wind direction parallel to satellite trajectory and veryfaintplumes). A clouddetection routine, basedon their lower brightnesstemperature, is applied to exclude them from further analysis.

Surface temperature is measuredwith bands 13 and 14 with a splitwindow algorithm (Trunk and Bernard. 2008 for sea or, for landsurfaces, by inverting the radiative transfer with the look-up tables.For each pixel, knowing the measured surface temperature, plumealtitude and radiance ratios, the SO2 column amount is interpolated inthe 3D matrix of the simulated radiance ratio. Column amount is "rstretrievedwith (B10+B12)/(2B11). If the result is higher than 15 g/m!(very high value, rarely encountered in passive degassing plumes), itis remeasured with B14/B1, to avoid saturation effects (see Section 3).Once the SO2 column amounts are measured, over the entire image,traverses are de"ned automatically as lines of pixels perpendicular tothe transport axis. The SO2 !ux is "nally computed by integrating theSO2 column amount over all the pixels of each traverse, andmultiplying by the wind speed. The above retrieval scheme isintegrated in one program, where the user just de"nes the contoursof the plume and “background” regions as well as the transportdirection of the plume. The program also includes an error estimationroutine that is based on SO2 retrievals with biased altitude, windspeed, and takes into account radiometric noise.

3. Sensitivity analysis

In this section, we present a sensitivity analysis of radiance ratiosin response to SO2, thermal contrast, atmospheric humidity, surfacealtitude and emissivity, and sulfate aerosols. For comparing thesensitivity of radiance ratios and radiance to each of these parameters,we use the %SO2 equivalent to the bias that affects retrieved columnamounts if the parameter is varied.

Fig. 2. Simulations, as a function of SO2 column amount (increment=4 g/m!), of ASTERradiances, (B10+B12)/2B11 and B14/B11. The simulations reproduce a Miyakejimacase, with surface temperature equal to 289 K and plume altitude set to 1100 m(T=273 K). Error bars represent the typical 1# observed on homogenous portions ofthe sea.

44 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

3.1. Sulfur dioxide

Fig. 2 shows the simulation of ASTER radiances, (B10+B12)/2B11and B14/B11 as functions of SO2 Column amount. Absorption of SO2

causes a slight decrease of B10 and B12 and a sharp decrease of B11,while B13 and B14 are unaffected. Thus, both radiance ratios increasewith SO2 Column amount. As SO2 increases, the sensitivity of (B10+B12)/2B11 decreases until complete saturation around $50 g/m!.Such very high values are extremely rare in volcanic plumes. To avoidthe negative impact of saturation, retrievals of amounts higher than15 g/m! are made with the B14/B11 ratio, which presents a highersensitivity at these high SO2 burdens.

3.2. Thermal contrast

Previous studies (Realmuto, 1994, 1997) have demonstrated theimportance of thermal contrast between the gas plume and theEarth's surface, for both accuracy and precision of the retrievals. Twosimpli"cations can be brought to Eq. (1) to emphasize the criticalimportance of thermal contrast.

1°) Surface emissivity # is approximated equal to 1. This is justi"edas spectral emissivity of water smoothly varies between 0.98 and 0.99in the considered spectral region.

2°) Atmosphere can be approximated as transparent except for theSO2 plume. The total atmospheric transmittance "0 is thus equal to thetransmittance of the plume. This can be done because atmospheric

Fig. 4. Relative retrieval error caused by a 400 m bias in plume height estimation. It isshown as a function of plume altitude. The temperature pro"le used for the simulations(Hachiyo-jima, 1st April 2001) is also shown. Surface temperature is set to 290 K.

Fig. 5. Noise Equivalent %SO2 as a function of thermal contrast for retrievals using (B10+B12)/2B11, B11 and B14/B11. Plume altitude is set to 1100 m (T=273 K).

Fig. 3. Simulations of B11 radiance, (B10+B12)/2B11 and B14/B11 as functions of SO2

column for several surface temperatures. Plume altitude is 3200 m and its temperatureis 283 K (typical Etna case).

45R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

transmittance on band 11 varies between 0.7 and 0.9, depending onatmospheric humidity.

Applying these simpli"cations, Eq. (1) becomes thus

RA = B T0! " · $p + B Tp! "

· 1#$p! "

;

which can be rearranged as

RA = B T0! "#Ap B T0! "#B Tp! "! "

!2"

Ap, the plume absorptance, is, according to the Beer's law, equal to1#exp (#$%Cp), where Cp is the column amount of SO2 and & theabsorption coef"cient of SO2. Eq. (2) clearly shows that net absorption

Fig. 6. Simulation of B11 radiance, (B10+12)/2B11, and B14/B11 for increasing SO2 column amounts. (a) As a function of atmospheric humidity, varied in the model bymultiplying themixing ratiosofH2O in the lowest 4 kmof the referencepro"le. The referencepro"le is the oneused in the imageofApril 1st ofMiyakejima. Theplume temperature, 273 K, is determined byits altitude (1100 m). (b) As a function of surface elevation. Plume altitude is 3200 m (T=281 K), and the atmospheric pro"le is the one used for the image of Etna on 3 August 2001.

46 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

of the plume depends not only on the SO2 content of the plume, butalso on the thermal contrast between the plume and the surface. Inorder to analyse the sensitivity to thermal contrast, we varied surfacetemperature (Fig. 3) in the simulations, while keeping the plume atthe same altitude and temperature (1100 m and 275 K). In this "gure,the slope of the curves indicates the sensitivity to SO2, while thevertical gap between the curves indicates the dependence totemperature. As expected from Eq. (2), SO2 is undetectable whenthe ground has the same temperature as the plume, and sensitivityincreases with thermal contrast. Fig. 3 also shows that, at low columnamounts of SO2, (B10+B12)/2B11 is nearly insensitive to groundtemperature. Plume altitude is also critical for thermal contrast, astemperature in the troposphere decrease with altitude (except locallyin case of thermal inversion). Therefore, error on plume altitudeestimation will cause a systematic bias in the SO2 retrievals. Themagnitude of this error decreases when plume altitude (or thermalcontrast) increases, as shown in Fig. 4.

3.3. Radiometric noise

Radiometric noise affects ASTER images and SO2 retrievals under theform of cross track lines of pixels above or below their average value.ASTER bands have aNE%T of about 0.15 K,which is better than the 0.3 Kdesign requirement (Arai and Tonooka, 2005). At 293 K, this translatesinto a standard deviation of 0.025 W sr#1m#2µm#1, 0.005 and 0.004for B11, (B10+B12)/2B11 and B14/B11 respectively. The impact ofradiometric noise on the retrievals was evaluated by considering anadditive 1# noise error on radiance and radiance ratios. The result isexpressed in terms of an error (called NE%SO2) between the perturbedand reference situation as NE%SO2=(SO2)pert#(SO2)ref. It is a randomerror that varies as a function of thermal contrast between the plumeand the ground, as shown in Fig. 5. It increases dramatically whenthermal contrast nears zero. Due to the combination of noises on each ofthe 3 bands, the NE%SO2 on retrievals based on (B10+B12)/2B11 isslightly higher with than with raw radiance.

3.4. Atmospheric humidity and surface elevation

Water vapor is the principal and more variable atmospheric gasinterfering with SO2 around 8.5 "m. It is mostly abundant in the lowtroposphere, with 95% of its total column residing in the lowest 4 km ofthe atmosphere. Atmospheric humidity varies on a regional scale as aresult of topography and atmospheric dynamics. Fig. 6a shows thesimulated effect ofwater vapor on the raw radiances andon the radianceratios for a Miyakejima-like atmosphere. Flat slopes indicate thatretrievals with radiance ratios are nearly independent on atmospherichumidity. For comparison, in the same conditions, a 40% increase ofatmospheric humidity in the four lowest km induces a systematic %SO2

of about 1 g/m! in retrievals with raw radiance.The total H2O column amount also strongly decreases with surface

altitude, because of the pathlength reduction between the surface andthe satellite. Fig. 6b shows the effect of varying surface altitude on B11radiance for an Etna-like atmosphere. As for atmospheric humidity,the retrievals with radiance ratios show a remarkable stability, whilethose with radiance are signi"cantly affected. A 500 m elevationdifference produces a systematic %SO2 of 1 g/m! on raw radianceretrievals. The effect of surface altitude is expected to be stronger in amoist atmosphere.

3.5. Surface emissivity

Temperature Emissivity Separation algorithms do not workproperly for surfaces lying under volcanic gas plumes, because theirspectral properties differ from the atmosphere. Therefore, a prioride"nition of emissivity is needed before retrieving SO2, which maybias the retrievals. The effect of varying emissivity is shown in Fig. 7.

Radiance and radiance ratios were simulated with several emissivityspectra: basalt, sea water (references basalt.7f and seawater in theAster Spectral Library) and grey bodies with emissivity of 0.96, 0.97and 0.98. The B14/B11 ratio and B11 radiance show amaximal%SO2 ofabout 2 g/m!, while (B10+B12)/2B11 is much less sensitive, with aninterestingly low 0.5 g/m!.

3.6. Sulfate aerosols

Aerosols are always present in passive degassing plumes. Theyscatter and absorb TIR radiation, causing an attenuation of theradiance measured at the sensor. Direct sampling campaigns (e.g.:Mather et al., 2004) have shown that sulfuric acid (H2SO4) is thedominant component of the particles. In order to test the effect of

Fig. 7. Simulations of B11 radiance, (B10+B12)/2B11, and B14/B11 as functions of SO2

for different surface emissivities: grey bodies with emissivities from 0.96 to 0.98, waterand basalt. The parameters of simulations correspond to an Etna-like case, with asurface temperature of 303 K and a plume altitude of 3400 m (T=281 K).

47R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

aerosols on ASTER SO2 retrievals, the extinction coef"cient of theaerosol were computed with MALUT (Clarisse et al., in review) a fastand accurate Mie algorithm. These coef"cients were then used inAtmosphit to simulate the radiative transfer in an atmospherecontaining a 400 m thick layer of SO2 and particles at an altitude of1500 m and a temperature of 272 K. The effective radius of theparticles was set to 0.5 "m, their composition was assumed to 50%mass ratio of H2SO4, and the column amount of particles was variedfrom 0 to 108 cm#2, according to the values reported in severalstudies based on sun photometers (Watson and Oppenheimer, 2001;Mather et al., 2004). Simulated spectra (Fig. 8a) clearly show thewideband extinction of the aerosol. Fig. 8b shows the effect of aerosolextinction in term of %SO2 for retrievals using raw radiance, andradiance ratio. Sulfate aerosols extinction affects retrievals signi"-cantly, but retrievals using (B10+B12)/2B11 are "ve times lesssensitive than radiance, making this ratio particularly interesting.

4. Case studies

4.1. Miyakejima

Starting in September 2000, Myake-jima emitted huge amounts ofSO2, passively degassed by a convecting magma body, through anewly formed caldera (Kazahaya et al., 2004). Maximal SO2 !uxespeaked at N100 Gg/day in December 2000, before gradually decliningto $7 Gg/day at the end of 2002.

An ASTER image (Fig. 9a) was acquired during this high rateperiod, on 1 April 2001 at 01 h44 GMT (11 h44 local time).Temperature and humidity pro"les come from the radiosondelaunched at 00H00 GMT from Hachiyo-jima, an island located110 km south of the volcano. In Fig. 9b, the map of SO2 columnamount shows a plume extending eastward from the volcano to adistance of 30 km, with the presence of concentrated puffs. Flux

Fig. 8. (a) Simulated radiance spectra for a Miyakejima-like atmosphere containing 400 m thick layer rich in H2SO4 aerosol. The mass fraction of H2SO4 in the droplets is 50%. Thedroplets column density is 6 107 droplet/cm! and their effective radius is 0.5 "m. (b) Bias induced by the presence of H2SO4 aerosols on the SO2 retrievals with B11 radiance, (B10+B12)/2B11 and B14/B11.

48 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

Fig. 9. (a) ASTER VNIR natural colour image of Miyakejima on 1st April 2001. (b) SO2 column amount map produced after application of the algorithm (UTM coordinates, in meters).Cold or cloudy pixels appear as black spots. Horizontal stripping is a consequence of radiometric noise. (c) Evolution of SO2 !ux with plume age. Proximal pro"les were removedbecause of the abundance of opaque clouds.

49R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

!uctuations associated to these puffs range to one third of theaverage !ux value. Progressive dilution of the plume is observedwith distance, resulting in homogenization and decreasing Columnamounts. In Fig. 9c, the measurements of SO2 !uxes are plottedversus the plume travel time (calculated as the crater-traversedistance divided by the wind speed). A general decrease of the !uxis observed with plume ageing. A similar pattern is present onmany images of Miyakejima that we have analyzed (See Table 1).As the average value of SO2 retrievals outside of the plume isslightly positive (0.037 g/m!), this decrease is probably not anartefact due to progressive dilution of the plume below detectionlimit. The !ux decrease could be caused by SO2 oxidation duringplume transport and mixing with the atmosphere. The speed of SO2

depletion in volcanic plumes is thought to depend strongly on themechanism involved: slow gas phase oxidation by the OH radicals(McGonigle et al., 2004) or fast aqueous phase oxidation in aerosoldroplets (Oppenheimer et al., 1998). An exponential regressionapplied to the time series of !uxes on 1st April 2001 yields a "rstorder rate constant of about 8.7 10#5 s#1, which is in the sameorder of magnitude as recent measurements in Montserrat byRodriguez et al. (2008). Alternatively, column amount close to thevents might be overestimated due to sulfate aerosol absorption.Increased visible re!ectance in the proximal plume on Fig. 9asupports the presence of high concentrations of liquid aerosol,probably H2SO4 rich.

The image of 15 July (Fig. 10) illustrates a typical case where IRmeasurements are imprecise and inaccurate. Radiosonde pro"leshows a weak thermal contrast between the sea (T&297 K) and thewarm atmosphere at plume level (293 K at 1200 m). Furthermore,southerly wind disperses the plume in a direction sub parallel to thesatellite track, making plume height measurements imprecise.Radiometric noise on SO2 measurement is high ($4 g/m!) andrelative error due to imprecision in plume height becomes important(40%). Imprecision on the computed !ux in this very unfavourablecase amounts to (#50% +130%). However automatic errorestimation by the program is helpful to identify these unfavourablecases.

A total of eleven images of ASTER were analyzed for Miyakejima.The resulting SO2 !uxes were compared to "eld measurements thatwhere performed regularlywith a COSPEC (Kazahaya et al., 2004). Dueto the lack of simultaneity betweenASTER and COSPEC, and to the timevariability of the !ux, this does not constitutes a strict validation of themethod. Despite of this, the retrieved SO2 !uxes are generally in goodagreement with the COSPEC measurements, as shown in Table 1 andFig. 11. It thus indicates that SO2 measurements from ASTER can beused, with a reasonable con"dence, for long term evaluation ofvolcanic degassing.

4.2. Etna, 3 August 2006

Mt. Etna is the highest (3330 m) andmost active volcano in Europe,with four !ank eruptions and several tenths of summit eruptive eventssince the beginning of this century. Mt. Etna is also a permanent SO2

emitter, with !uxes varying from 500 to 30,000 T/day depending on theactivity. The !ux of SO2 emitted byMt. Etna ismonitored regularly since1987 using COSPEC (Caltabiano et al., 1994) and permanently by theFLAME network since 2004 (Salerno et al., 2009a, 2009b). The networkconsists of "ve "x scanning DOAS stations located on the Eastern andNorthern quadrants of the volcano.

On 3 August 2006, Mt. Etna was quiescently degassing between twoeruptive episodes at the South–East Summit Crater. The atmosphericpro"le from Trapani (220 km west of the volcano) indicates atemperature of 281 K and a relative humidity of 30% at 3200 m, theestimated plume height. The SO2map in Fig. 12a shows a homogeneousplume without signi"cant puffs. The time series of !uxes (Fig. 12b)show no signi"cant decrease with plume age, indicating that thedepletion of SO2 is much slower, as expected for a plume emitted in thedry free troposphere. The average !ux measured by ASTER above theEtna_Milo FLAME station (EMIL) station is 48 kg/s, which comparesquite well with the measurements of the station (51 kg/s). The windspeed and plume height used from !ux calculations with bothmethodsare respectively 9.5 m/s and 3200 m.

Stable atmospheric conditions and sustained constant degassingallowed the plume to be transported far from the volcano in acoherent way. Thanks to this, the column amounts above the EMILstation were still way higher the detection limit of ASTER, allowing acomparison of the two datasets. For doing this, FLAME measurementswere geolocated using the following relations

Xi # X0 $ Hp·sin!!"=tan!"i"

Yi # Y0 $ Hp·cos!!"=tan!"i"

Where X0 and Y0 are the UTM coordinates of the station, $ isthe azimuth of the scanning plan, %i is the scan angle and Hp is the plumeheight. As the plume was optically thin in the VNIR (hamperingstereoscopic measurements), this last parameter was determinedby"tting the position of the peak SO2 columnamount in the twodatasets.

The FLAME and ASTER measurements are in good agreement witheach other (Fig. 13). ASTER is, as expected, less precise than DOASwith standard deviation of 0.4 g/m!. A moving average applied onASTER retrievals to reduce radiometric noise reveals no systematicbias between the two measurements method, except for the right“wing” of elevated DOAS measurements. This is probably an artefact

Table 1Summary of ASTER measurements of SO2 !uxes (in tons per day). Source !uxes are calculated by applying an exponential regression to the time series of measurements, or byaveraging all data when no clear trend can be distinguished. Ground data "gured in italic are an average of the 2 closest measurements. (*) signi"es that SO2 depletion is insigni"cantor masked by plume puf"ng.

Date and time Source Max Min Estimated Rate COSPEC Detection

Miyakejima Flux Flux Flux Error Constant or DOAS Limit (3')

2000-11-08, 01:47 34,800 57,900 25,200 +18,000/#7000 $0* 29,600 4.5 g/m!2001-03-30, 12:58 23,800 26,400 13,600 Night 1.6 10#4 s#1 22,700 2.6 g/m!2001-04-01, 01:44 16,800 22,000 10,000 +4500/#4000 7.8 10#5 s#1 19,600 1.1 g/m!2001-04-15, 12:58 24,200 34,000 11,000 Night 1.3 10#4 s#1 20,600 4.8 g/m!2001-05-28, 01:37 13,700 21,300 5500 +3000/#2500 1.0 10#4 s#1 25,300 2.9 g/m!2001-07-15, 01:35 16,900 27,500 7300 +20,000/#8000 n.a.* 15,200 6.6 g/m!2001-08-23, 01:41 Unexploitable due to near-zero thermal contrast 8.7 g/m!2001-11-27, 01:38 13,800 22,700 5000 +3900/#2900 3.2 10#4 s#1 16,700 2.4 g/m!2002-03-19, 01:35 5800 9900 2800 +1700/#1500 $0* 8100 1.9 g/m!2003-04-07, 01:34 5800 7200 200 +3000/2000 2.3 10#4 s#1 5700 3.1 g/m!

50 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

Fig. 10. (a) ASTER VNIR image of Miyakejima on 15 July 2001. (b) Corresponding SO2 map. Due to low thermal contrast between the plume and the sea, NE%SO2 is so high ($4 g/m!)that it dominates the image. (c) Time series of !ux versus plume age.

51R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

Fig. 11. ASTER measurements of the SO2 !ux (in mg per day) fromMiyakejima compared to the evolution of !uxes as measured by helicopter based COSPEC (Kazahaya et al., 2004).

Fig. 12. Etna, 3 August 2006 (a) SO2 column amount map with the location of the summit craters, coastline and the “EMIL” UV scanner station. Noise lines are more prominent abovethe sea than above the land, due to difference in thermal contrast. (b) Time series of SO2 !uxes with plume age.

52 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

of DOAS due to a combination of geolocation inaccuracy and multiplescattering along a path close to the horizon.

5. Conclusion

We have developed a new method to retrieve SO2 !uxes withASTER TIR multispectral images. This method relies on band ratiosimulations and is directly applicable to any cloud free image of adegassing volcano, either island or continental. Retrievals arestraightforward and show no signi"cant dependence on atmospherichumidity, surface altitude and emissivity. The impact of sulfateaerosol on retrievals with (B10+B12)/2B11 is also much reducedcompared to retrievalswith raw radiance. For these reasons, the bandratio method represents an improvement compared with previousmethods relying on raw radiances. Our measurements of the SO2

!uxes onMiyakejima vary between 40,000 and 4000 Mg/day and arein agreement with "eld measurements. Validation with columnamounts measured simultaneously by automated scanning DOAS atMt. Etna was also successful. We emphasize that ASTER measure-ments achieve good accuracy and precision (25% or better) onlywhen the thermal contrast between the plume and the underlyingsurface is high enough (N10 °C). Hence, for low-altitude islandvolcanoes, precision and accuracy are increased in the winter, whenthe thermal inertia of the ocean maintains the surface waterssigni"cantly warmer than the atmosphere at the plume altitude.Conversely, over bare volcanic grounds, higher precision is achievedduring the summer.

Thanks to its highmultispectral ground resolution in the TIR, alongwith its ability to measure the plume's height and transport velocity,

ASTER is capable of measuring, with a reasonable accuracy, SO2 !uxesin small scale passive degassing plumes. Moreover, ASTER is animaging sensor, and the 2D distribution of the SO2 in the plumes canprovide insights into its dilution and depletion. The long return timeof ASTER presently limits its use for monitoring, but the retrievalscheme presented here is applicable to the images of the futureHyperspectral Infrared Imager (HyspIRI, JPL Publication 09-19, 2009).HyspIRI, planned for 2013–2016, will have an image frequency of 1–5 day thanks to its wider swath and continuous acquisition.

The total SO2 burdens measured with ASTER in the plumes ofMiyakejima are detectable by other sensors, such as OMI (notlaunched until July 2004) or MODIS (Watson et al., 2004). Cross-comparison of SO2 retrievals between these sensors and ASTER couldbe of great interest as the spatial resolution of ASTER imagesrepresents the missing link between these sensor and highly localDOAS measurements.

Acknowledgements

R.C. was supported by a grant from F.R.I.A (Fond pour la RechercheIndustrielle et Appliquée). GGS acknowledges a PhD grant funded bythe project “Sviluppo di sistemi di monitoraggio” funded byDipartimento di Protezione Civile della Regione Sicilia, INGV (IstitutoNazionale di Geo"sica e Vulcanologia, sezione di Catania—Italy) andNOVAC (Network for Observation of Volcanic and AtmosphericChange) EU-funded FP6 project no. 18354. P-F. C. is research associatewith FRS-FNRS and bene"ted from its "nancial support (F.4511.08).We also thank N. Bruno, V. Longo and F. Murè for technical assistance.

Fig. 13. Comparison between measurements of SO2 column amounts measured by ASTER and EMIL FLAME station in the plume of Etna on 3rd August 2006. (a): Pixel per pixelretrieval (b) Same data "ltered with a 9"9 weighted average to reduce noise.

53R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54

Appendix A

References

Andres, R., Kasgnoc, A.D., 1997. A time-averaged inventory of subaerial volcanic sulfuremissions. J. Geophys. Res. 103 (D19), 25251–25261.

Arai, K., Tonooka, H., 2005. Radiometric performance evaluation of ASTER VNIR, SWIRand TIR. IEEE Transactions on Geoscience and Remote Sensing 43 (12), 2725–2732.

Bluth, G.J.S., Shannon, J.M., Watson, I.M., Prata, A.J., Realmuto, V.J., 2007. Developmentof an ultra-violet digital camera for volcanic SO2 imaging. J. Volcanol. Geotherm.Res. 161, 47–56.

Bobrowski, N., Hönninger, G., Lohberger, F., Platt, U., 2006. IDOAS: a new monitoringtechnique to study the 2D distribution of volcanic gas emissions. Journ. Volcano.Geotherm. Res 150, 329–338.

Caltabiano, T., Romano, R., Budetta, G., 1994. SO2 !ux measurements at Mount Etna(Sicily). J. Geophys. Res. 99 (12), 12,809–12,819.

Carn, S.A., Krueger, A.J., Arellano, S., Krotkov, N.A., Yang, K., 2008. Daily monitoring ofEcuadorian volcanic degassing from space. J. Volcanol. Geotherm. Res. 176, 141–150.

Clarisse, L., Coheur, P.F., Prata, A.J., Hurtmans, D., Razavi, A., Phulpin, T., Hadji-Lazaro, J.,Clerbaux, C., 2008. Tracking and quantifying volcanic SO2 with IASI, the September2007 eruption at Jebel at Tair. Atmos. Chem. Phys. 8, 7723–7734.

L. Clarisse, D. Hurtmans, A.J. Prata, F. Karagulian, C. Clerbaux, and P.F. Coheur. Retrievingaerosol properties from nadir observed high resolution spectra. Applied Optics, inreview, 2010

Clough, S.A., Shephard, M.W., Mlawer, E.J., Delamere, J.S., Iacono, M.J., Cady-Pereira, K.,Boukabara, S., Brown, P.D., 2005. Atmospheric radiative transfer modeling: asummary of the AER codes, short communication. J. Quant. Spectrosc. Radiat.Transfer 91, 233–244.

Coheur, P.-F., Barret, B., Turquety, S., Hurtmans, D., Hadji-Lazaro, J., Clerbaux, C., 2005.Retrieval and characterization of ozone vertical pro"les from a thermal infrarednadir sounder. J. Geophys. Res. 110, D24303.

Corradini, S., Pugnaghi, S., Teggi, S., Buongiorno, M.F., Bogliolo, M.P., 2003. Will ASTERsee the Etna SO2 plume? Int. J. Remote Sens. 24 (6), 1207–1218.

Dalton, M.P., Watson, I.M., Nadeau, P.A., Werner, C., Morrow, W., Shannon, J.M., 2009.Assessment of the UV camera sulfur dioxide retrieval for point source plumes.J. Volcanol. Geotherm. Res.. doi:10.1016/j.jvolgeores.2009.09.013

Elias, T., Sutton, A.J., Oppenheimer, C., Horton, K.A., Garbeil, H., Tsanev, V.,McGonigle, A.J.S.,Williams-Jones, G., 2006. Intercomparison of COSPEC and two miniature ultraviolet

spectrometer systems for SO2 measurements using scattered sunlight. Bull. Vol. 68,313–322.

Galle, B., Oppenheimer, C., Geyer, A., McGonigle, A.S.J., Edmonds, M., Horrocks, L.A.,2002. A miniaturised ultraviolet spectrometer for remote sensing of SO2 !uxes: anew tool for volcano surveillance. J. Volcanol. Geotherm. Res. 119, 241–254.

Graf, H.-F., Feichter, J., Langmann, B., 1997. Volcanic sulfur emissions: estimates ofsource strength and its contribution to the global sulfate distribution. J. Geophys.Res. 102, 10727–10738.

JPL HysPIRI group, 2009, NASA 2008 HyspIRI Whitepaper and Workshop Report, JPLPublication 09-19, downloadable at http://hyspiri.jpl.nasa.gov/downloads/public/2008%20HyspIRI%20Whitepaper%20and%20Science%20Workshop%20Report-r2.pdf.

Kazahaya, K., Shinohara, H., Uto, K., Odai, M., Nakahori, Y., Mori, H., Iino, H., Miyashita,M., Hirabayashi, J., 2004. Gigantic SO2 emission from Miyakejima volcano, Japan,caused by caldera collapse. Geology v.32 (no. 5), 425–428.

Kazahaya, R., Mori, T., Kazahaya, K., Hirabayashi, J., 2008. Computed tomographyreconstruction of SO2 concentration distribution in the volcanic plume ofMiyakejima, Japan, by airborne traverse technique using three UV spectrometers.Geophys. Res. Lett. 35, L13816.

Le Guern, F., 1982. Les débits de CO2 et SO2 volcaniques dans l'atmosphère. Bull. Volc.45, 197–202.

Mather, T.A., Tsanev, V.I., Pyle, D.M., McGonigle, A.J.S., Oppenheimer, C., Allen, A.G.,2004. Characterization and evolution of tropospheric plumes from Lascar andVillarrica volcanoes, Chile. J. Geophys. Res. 109, D21303.

McGonigle, A.J.S., Delmelle, P., Oppenheimer, C., Tsanev, V.I., Delfosse, T., Williams-Jones, G., Horton, K., Mather, T.A., 2004. SO2 depletion in tropospheric volcanicplumes. Geophys. Res. Lett. 31, L13201. doi:10.1029/2004GL019990.

Mori, T., Burton, M., 2006. The SO2 camera: a simple, fast and cheapmethod for ground-based imaging of SO2 in volcanic plumes. Geophys. Res. Lett. 33, L24804.

Newcomb, G., Millán, M.M., 1970. Theory, applications and results of the long-linecorrelation spectrometer. IEEE Trans. on Geoscience Electronics 8, 149–157.

Oppenheimer, C., Francis, P., Stix, J., 1998. Depletion rates of sulfur dioxide intropospheric volcanic plumes. Geophys. Res. Lett. 25 (14), 2671–2674.

Pieri, D., Abrams, M., 2004. ASTER watches the world's volcanoes: a new paradigm forvolcanological observations from orbit. J. Volcano. Geotherm. Res. 135, 13–28.

Prata, A.J., Bernardo, C., 2007. Retrieval of volcanic SO2 column abundance fromAtmospheric Infrared Sounder data. J. Geophys. Res. 112, D20204.

Prata, A.J., Kerkmann, J., 2007. Simultaneous retrieval of volcanic ash and SO2 usingMSG-SEVIRI measurements. Geophys. Res. Lett. 34, L05813.

Pugnaghi, S., Gangale, G., Corradini, S., Buongiorno, M.F., 74–90, 2006. Mt. Etnasulfur dioxide !uxmonitoring using ASTER-TIR data and atmospheric observations.J. Volcano. Geotherm. Res. 152, 74–90.

Realmuto, V.J., Sutton, A.J., Elias, T., 1997. Multispectral thermal infrared mapping ofsulfur dioxide plumes: a case study from the East Rift Zone of Kilauea Volcano,Hawaii. J. Geophys. Res. 102 (B7), 15057–15072.

Realmuto, V.J., Abrams, M., Buongiorno, M.F., Pieri, D., July 29, 1986. The use of thermalinfrared image data to estimate the sulfur dioxide !ux from volcanoes: a case studyfrom Mount Etna, Sicily. J. Geophys. Res 99 (B1), 481–488.

Rodriguez, L.A., Watson, I.M., Edmonds, M., Ryan, G., Hards, V.L., Oppenheimer, C., Bluth,G.J.S., 2008. SO2 loss rates in the plume emitted by Soufrière Hills volcano,Montserrat. Journ. Volcanol. Geotherm. Res. 173 (1–2), 135–147.

Rothman, L.S., Gordon, I.E., Barbe, A., Benner, D.C., Bernath, P.F., Birk,M., Boudon, V., Brown,L.R., Campargue, A., Champion, J.-P., Chance, K., Coudert, L.H., Dana, V., Devi, V.M., Fally,S., Flaud, J.-M., Gamache, R.R., Goldman, A., Jacquemart, D., Kleiner, I., Lacome, N.,Lafferty, W.J., Mandin, J.-Y., Massie, S.T., Mikhailenko, S.N., Miller, C.E., Moazzen-Ahmadi, N., Naumenko, O., Nikitin, A.V., Orphal, J., Perevalov, V.I., Perrin, A., Predoi-Cross, A., Rinsland, C.P., Rotger,M., Simecková,M., Smith,M.A.H., Sung, K., Tashkun, S.A.,Tennyson, J., Toth, R.A., Vandaele, A.C., Vander Auwera, J., 2008. The HITRAN 2008molecular spectroscopic database. J. Quant. Spectrosc. Radiat. Transfer 82, 5–44.

Salerno, G.G., Burton, M.R., Oppenheimer, C., Caltabiano, T., Tsanev, V.I., Bruno, N.,2009a. Novel retrieval of volcanic SO2 abundance from ultraviolet spectra. Journ.Volcanol. Geotherm. Res. 181, 141–153.

Salerno, G.G., Burton, M.R., Oppenheimer, C., Caltabiano, T., Randazzo, D., Bruno, N.,Longo, V., 2009b. Three-years of SO2 !ux measurements of Mt. Etna using anautomated UV scanner array: comparison with conventional traverses anduncertainties in !ux retrieval. J. Volcanol. Geotherm. Res. 183, 76–83.

Stoiber, R.E., Maliconico, L.L., Williams, S.N., 1983. Use of the Correlation Spectrometerat Volcanoes. In: Tazieff, H., Sabroux, J.C. (Eds.), Forecasting volcanic events.Elsevier, Amsterdam, pp. 425–444.

Stoiber, R.E., Williams, S.N., Huebert, B., 1987. Annual contribution of sulfur dioxide tothe atmosphere by volcanoes. J. Volcanol. and Geotherm. Res. 33, 1–8.

Trunk, L, Bernard, A, 2008. Investigating crater lake warming using ASTER thermalimagery: Case studies at Ruapehu, Poás, Kawah Ijen, and Copahué Volcanoes. Journ.Volcanol. and Geotherm. Res. 178, 259–270.

Urai, M., 2004. Sulfur dioxide !ux estimation from volcanoes using advancedspaceborne thermal emission and re!ection radiometer: a case study ofMiyakejima volcano, Japan. J. Volcanol. Geotherm. Res. 134, 1–13.

Watson, I.M., Oppenheimer, C., 3561–3572, 2001. Photometric observations of Mt.Etna's different aerosol plumes. Atmosph. Environm. 35, 3561–3572.

Watson, I.M., Realmuto, V.J., Rose, W.I., Prata, A.J., Bluth, G.J., Gu, Y., Bader, C.E., Yu, T.,2004. Thermal infrared remote sensing of volcanic emissions using the moderateresolution imaging spectroradiometer. J. Volcanol. Geotherm. Res. 135, 75–89.

Williams-Jones, G., Stix, J., Hickson, C., 2008. The COSPEC Cookbook: making SO2-

Measurements at Active Volcanoes. IAVCEI, Methods in Volcanology 1 233 pp.

Annex tableCurrent satellite sensors sensitive to SO2. (*) For AIRS and IASI, the value refers to thediameter of circular footprint of the pixel.

Acronym Instrument Resolution Comments

OMI Ozone MonitoringInstrument

13"24 km UV imaging spectrometer,onboard Aura, GlobalCoverage (Carn et al.,2008)

SCIAMACHY SCanning ImagingAbsorptionspectroMeter forAtmosphericCartograpHY

15"26 km UV–Visible-Near IRimaging spectrometer, onboard ENVISAT

GOME-2 Global OzoneMonitoringExperiment

80"40 km UV imaging spectrometer,onboard METOP

ASTER Advanced SpaceborneThermal Emission andRe!exion radiometer

0.09"0.09 km Multispectral Sensor,programmable acquisition

MODIS MODerate resolutionImagingSpectroradiometer

1"1 km 2 multispectral sensorsonboard Terra and Aqua,global coverage (Watsonet al., 2004)

SEVIRI Spinning EnhancedVisible and InfraredImage

3"3 km Geostationary on boardMSG, centered on Europeand Africa (Prata andKerkmann, 2007)

AIRS Advanced InfraredSounder

13.5 km (*) IR imaging spectrometeronboard Aqua, onlysensitive to high level SO2

(Prata and Bernardo,2007)

IASI Infrared AtmosphericSoundingInterferometer

12.5 km (*) Imaging Fourier TransformSpectrometer onboardMETOP, with globalcoverage (Clarisse et al.,2008)

54 R. Campion et al. / Journal of Volcanology and Geothermal Research 194 (2010) 42–54