Vol49_1_2006def 235 ANNALS OF GEOPHYSICS, VOL. 49, N. 1, February 2006 Key words hyperspectral data – surface tempera- ture – spectral emissivity – Solfatara (Phlegraean Fields) – DAIS 1. Introduction Surface temperature and spectral emissivity of materials are valuable parameters in volcanol- ogy since they provide information on several as- pects of the current activity and of past events. Mapping surface temperature allows us to esti- mate the heath flux from the surface and there- fore to control activity inside vents and fractures. Spectral emissivity in the thermal infrared is a physical parameter necessary to calculate tem- perature, and moreover it enables recognition of surface materials on the basis of their spectral pattern (Salisbury and D’Aria, 1992); it is partic- ularly suitable to study volcanic materials be- cause the main absorption bands of silicates are located in the thermal infrared (Gillespie, 1985). Remote separation and identification of surface deposits in volcanic areas is useful to map lava flows on the basis of their composition, to char- acterize surface alteration due to hydrothermal phenomena and to identify newly formed de- posits related to fumarolic activity. Surface kinetic temperature and emissivity properties of materials can be estimated with high accuracy by remote sensing if images collected on several channels in the thermal infrared are avail- able. Thermal infrared multi-channel sensors have been operating for a long time at low spatial reso- lution (less than one km, mainly for meteorologi- Spectral emissivity and temperature maps of the Solfatara crater from DAIS hyperspectral images Luca Merucci (1), Maria Paola Bogliolo (2), Maria Fabrizia Buongiorno (1) and Sergio Teggi (3) (1) Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy (2) Istituto Superiore per la Prevenzione e la Sicurezza del Lavoro, Monteporzio Catone (RM), Italy (3) Università di Modena e Reggio Emilia, Modena, Italy Abstract Quantitative maps of surface temperature and spectral emissivity have been retrieved on the Solfatara crater at Pozzuoli (Naples) from remote sensing hyperspectral data. The present study relies on thermal infrared images collected on July 27, 1997 by the DAIS hyperspectral sensor owned by the German aerospace center (DLR). The Emissivity Spectrum Normalization method was used to make temperature and emissivity estimates. Raw data were previously transformed in radiance and corrected for the atmospheric contributions using the MODTRAN radiative transfer code and the sensor response functions. During the DAIS flight a radiosonde was launched to collect the atmospheric profiles of pressure, temperature and humidity used as input to the code. Retrieved tem- perature values are in good agreement with temperature measurements performed in situ during the campaign. The spectral emissivity map was used to classify the image in different geo-mineralogical units with the Spec- tral Angle Mapper method. Areas of geologic interest were previously selected using a mask obtained from an NDVI image calculated with two channels of the visible (red) and the near infrared respectively. Mailing address: Dr. Luca Merucci, Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Roma, Italy; e-mail: merucci@ingv.it 236 Luca Merucci, Maria Paola Bogliolo, Maria Fabrizia Buongiorno and Sergio Teggi Table I. DAIS 7915 main technical and spectral characteristics. Spectrometer Spectral region No. of channels Wavelength 1 0.4-1.0 nm 32 15-30 nm 2 1.5-1.8 nm 8 45 nm 3 2.0-2.5 nm 32 20 nm 3.0-5.0 nm 1 2.0 nm 4 8.0-12.6 nm 6 0.9 nm Dynamic range 15 bit Ground resolution 5-20 m (function of the flight altitude) Scan line dimension 512 pixels Fig. 1. Topographic map of the Solfatara crater. Gray squares indicate the surface temperature measurement sites; the black circle (RS) locates the radiosonde launch site. 237 Spectral emissivity and temperature maps of the Solfatara crater from DAIS hyperspectral images cal or oceanographic use), and only few instru- ments provide multi-channel images at high spa- tial resolution in this spectral range. The airborne hyperspectral sensors TIMS (NASA), MIVIS (CNR-LARA Project) and DAIS (German aero- space center DLR: http://www.op.dlr.de/dais/) and the spaceborne sensor ASTER onboard TER- RA and AQUA satellites (NASA) are examples of this kind of instruments. With the aim of understanding the capabilities of the DAIS (Digital Airborne Imaging Spec- trometer) hyperspectral sensor to provide infor- mation of volcanological interest, we analyzed the thermal infrared DAIS data of an active vol- canic area, according to three work phases: 1) re- trieval of the radiance emitted by the surface, through atmospheric modeling; 2) estimation of surface temperature and spectral emissivity; 3) application of spectral mapping techniques to produce a thematic map of the main surface units. 2. Data set Remote sensing data used in this study were provided by the DAIS 7915 spectrometer owned by the German aerospace center (DLR). The main spectral and technical characteristics of DAIS are summarized in table I. The image un- der investigation was collected on July 27, 1997 at 10:27 (local time) on the Solfatara crater at Pozzuoli (Naples, Italy) in the framework of the E.C. Large Scale Facilities Project, and under re- quest and supervision of the Remote Sensing Laboratory of the Italian National Institute of Geophysics and Volcanology (INGV) of Rome. DAIS was flown on a DO 228 aircraft at an alti- tude of 1600 m a.s.l., and the collected images ground resolution is about 5 m. The data spectral subset used here corresponds to the six DAIS in- frared thermal channels. Simultaneously with the DAIS flight, a ground measurement campaign was carried out to measure the parameters necessary to atmos- pheric correction and validation of the retrieved maps. Vertical atmospheric profiles of pressure, temperature and relative humidity at about 10 m vertical resolution were measured by means of a radiosonde launch performed near the Sol- fatara (fig. 1). Surface brightness temperature was meas- ured with an infrared thermometer (EVEREST Inters. Inc.) at different sites (fig. 1), selected on homogeneous and horizontal surfaces easy to lo- cate on the images. In order to compensate for the spatial scale difference between the ground measures and the remote sensing data, 10 meas- urements were acquired on different points at each site: the average value was considered to be representative of the site brightness temperature. Temperature inside fumaroles was also meas- ured with a thermocouple. 3. Geological setting Solfatara is a volcanic crater located in the central area of the Phlegraean Fields caldera complex, west of the city of Naples (Italy) (figs. 2, 3). It represents the most active zone of Phle- graean Fields, and sits inside the sprawling ur- ban area of Pozzuoli. Activity in the Phlegraean area has been dominated by two eruptions that produced wide- spread ash-flow deposits: the Campanian Ign- imbrite and the smaller Neapolitan Yellow Tuff (Lirer et al., 1987; Scandone et al., 1991). The eruption producing the Campanian Ignimbrite occurred 37 000 years BP and resulted in the col- lapse of a large area including Campi Flegrei and part of the gulf of Naples. The eruption of Neapolitan Yellow Tuff occurred 12 000 years BP, had a very complex history that led to the formation of a caldera of smaller dimensions in- side that of the Campanian Ignimbrite. In the last 12 000 years, the bottom of the Neapolitan Yel- low Tuff caldera has been the site of intense vol- canic activity and ground deformation (Di Vito et al., 1999). Volcanism was concentrated in three periods: 12 000 and 9500 years BP, 8600 and 8200 years BP, and 4800 and 3800 years BP (e.g., at Cigliano, Agnano-Monte Spina, Astroni, Averno, Solfatara), followed by the last eruption in 1538 that led to the formation of the Monte Nuovo (Di Vito et al., 1987). Since 1800, sea-level measurements made at ancient roman ruins have indicated a slow sink- ing of the area. This slow sinking of the ground continued until 1968. In the periods 1970-1972 and 1982-1984 two important bradyseismic 238 Luca Merucci, Maria Paola Bogliolo, Maria Fabrizia Buongiorno and Sergio Teggi events (ground uplift) occurred in the Pozzuoli area (Corrado et al., 1976; Berrino et al., 1984), (maximum uplift of 1.7 and 1.8 m respectively) accompanied by shallow seismicity. More re- cently, two minor sudden ground uplifts and seismic swarms were recorded in 1989-1990 and in 1994. The Solfatara volcano is one of the most re- cent (about 4000 years BP) of the Phlegraean Fields caldera. It measures 0.5 × 0.6 km, with steep walls on the north, east and south sides. To the west the crater wall is missing. Its rectangu- lar shape is mostly due to the presence of faults at NW-SE and SW-NE. The volcanic cone is made of pyroclastic rocks, with the exception of Mt. Olibano that is a dome of trachytic lava. The flat-floored crater (Piano Sterile) is char- acterized by strong fumarolic activity which Fig. 2. Schematic geological map of the Phlegraean Fields (Scandone, 1997). 239 Spectral emissivity and temperature maps of the Solfatara crater from DAIS hyperspectral images causes both single vent emissions, with tempera- ture up to 160°C, and diffuse degassing. The gas- es emitted are mostly composed by H2O, CO2, H2S and smaller quantities of H2, CH4, He, HCl, Ar (Valentino et al., 1999). One of the most recent hydrothermal mod- els (Chiodini et al., 1997) describes a system divided into three parts: 1) a heat source which is made up of a relatively shallow magmatic chamber; 2) one or more aquifers located above the chamber; the degassing magma supplies fluids and heat to them; 3) an intensely frac- tured zone, sited above the uppermost aquifer and occupied by a pure vapor phase, which is produced through the boiling of the underlying aquifers. The intense fumarolic activity has given ori- gin to strong hydrothermal alteration of the original rocks (De Gennaro et al., 1980) and to secondary deposits. In fact, the trachytic rocks of the floor and of the flanks of the volcano are bleached and corroded by the effluent vapours, with formation of gypsum, alum, kaolin and alunite (Valentino et al., 1999). Moreover, sub- limation of the emitted gas causes deposition of sulphur, arsenic sulphide (realgar), ammonium chloride (sal ammoniac), mercury sulphide and antimony sulphide. 4. Atmospheric correction The spectral radiance LD received by a re- mote sensor in the thermal infrared can be ex- pressed as (4.1) where Lu is the upward atmospheric radiance, Ld is the downward atmospheric radiance, Ts is the surface kinetic temperature, f is the surface emissivity, x is the atmospheric transmittance between the surface and the sensor, and B(Ts) is the Plank function. In order to calculate the radiance emitted by the surface (L(m, Ts) =f(m) ⋅ B (m, Ts) ), remote sensing data must be corrected for atmospheric effects. To estimate the atmospheric radiative contributions (Lu, Ld and x), we used the MOD- TRAN 3.5 radiative transfer code (Berk et al., 1989), giving as input the atmospheric profiles measured during the field campaign. The ob- tained values were convolved with the sensor response functions of the six thermal infrared channels (fig. 4). The atmospheric terms were estimated for 20 different optical paths corresponding to view angle increments of ≈ 3° with respect to the nadir. Lower angle increments result in negligi- $=( ) ( ) ( , ) ( ( )) ( ) ( ) ( ) L B T L L 1D s d u $ $ $ + - + m f m m f m m x m m 6 @ Fig. 4. Response functions of the DAIS thermal in- frared channels. Fig. 3. Solfatara crater at Pozzuoli (Naples, Italy). Picture from «Pozzuoli dal cielo» by Aeromap Data. 240 Luca Merucci, Maria Paola Bogliolo, Maria Fabrizia Buongiorno and Sergio Teggi ble variations. The different sets of atmospher- ic terms were used to apply separate corrections on image stripes corresponding to the view an- gle increments and parallel to the image axis. 5. Evaluation of surface temperature and emissivity The spectral radiance emitted by a body having a kinetic temperature Ts can be ex- pressed as a function of the spectral emissivity f of its surface and of the Plank function B (5.1) If the radiance L is measured at N wavelengths m, the relation is expressed by a system of N equations with N + 1 unknowns, i.e. the N values of spectral emissivity and the temperature. This makes it difficult to estimate surface temperature and emissivity from remote sensing data because the equation system is not closed. In the literature, different «non-exact» solutions are given to this problem, called «temperature and emissivity separation» (Gillespie et al., 1998; Li et al., 1999). In our work we applied the well established method of the Emissivity Spectrum Normaliza- tion (ESN) (Gillespie, 1985; Realmuto, 1990). According to this method, emissivity is normal- ized to a value fmax corresponding to the maxi- mum value expected in the scene. This method demonstrated good capability to maintain the spectrum shape, that is the most important fea- ture in applications of spectral analysis and mapping. Assuming f=fmax for each value of m, the in- version of eq. (5.1) gives N values of temperature; the highest of them is assumed to be the estimate of the kinetic temperature Ts. With this value, and using the Plank equation, the black body spectral radiance of the pixel is calculated. Finally, the emissivity is obtained dividing the surface radi- ance measured by the sensor by the calculated black body radiance, at each wavelength. Repeat- ing this procedure for every pixel, temperature and spectral emissivity maps are obtained. In this study we assumed a value fmax = 0.97 on the ba- sis of the spectral libraries data (Salisbury et al., ( , ) ( ) ( , ) .L T B Ts s$=m f m m 1991; Salisbury and D’Aria, 1992) available for the typical minerals of the Solfatara crater. 6. Selection of the areas of interest The atmospheric correction and the temper- ature and emissivity estimation were performed on an image window chosen in the Solfatara crater. Urban areas adjacent to the Solfatara were ex- cluded applying a suitable mask cropped on the original image. On these areas reliable estimates of surface temperature and spectral emissivity cannot be obtained by the ESN method due to the urban covers extreme heterogeneity that makes it impossible to define a value of fmax valid for the whole image. Moreover, the estimated surface temperature can be considered validated only in- side the crater, where the in situ temperature was measured during the ground campaign. In the images showing the elaboration re- sults (figs. 5, 6 and 8), non-processed areas were masked with an image of a thermal chan- nel to preserve topographic reference. A further selection was applied to identify the areas of interest for geo-mineralogical inter- pretation: the vegetation cover was excluded with a mask based on the Normalized Differ- ence Vegetation Index (NDVI). The data of two channels of visible (red) and near infrared are transformed by NDVI in a new image related to the green biomass using a spectral property of green vegetation: the sharp increase in re- flectance between 0.65 and 0.80 nm (red edge). The NDVI image was obtained using the DAIS channels 10 (0.659 nm) and 18 (0.802 nm); it was then converted in a mask by select- ing a threshold value between vegetated and non vegetated pixels. 7. Results Figure 5 shows the quantitative map of sur- face kinetic temperature. Values retrieved in the ground measurement sites (fig. 1) were ex- tracted from this map and compared with ground temperatures: the comparison revealed a good agreement, as shown in table II. Table II. Comparison between ground measured temperature and image retrieved temperature at the sites shown in fig. 1. Site Measurement Ground temperature Image retrieved temperature start/end (local time) (°C) (°C) Fumarole; dark grey soil (1FF) 11.10/11.15 45.18 ± 4.31 44.2 Central area (1CA) 11.20/11.25 44.91 ± 2.60 42.7 White tuffs (1WT) 11.30/11.31 37.18 ± 0.24 36.9 Campground (2CG) 10.40/10.43 44.54 ± 1.87 42.7 Dry grass (2DG) 10.45/10.50 54.71 ± 3.31 50.7 Tuffs (2T) 10.55/11.00 42.32 ± 2.32 40.0 Fig. 5. Surface temperature map of the Solfatara crater retrieved by DAIS data. 241 Spectral emissivity and temperature maps of the Solfatara crater from DAIS hyperspectral images The six images of spectral emissivity al- lowed us to discriminate different surface cov- ers on the basis of their spectral characteristics. An example is reported in fig. 6, where an RGB composition of spectral emissivity calculated in channels 78, 76 and 74 is reported. By analyzing the emissivity images and the corresponding spectra, we could discriminate 6 main units, whose spectra differ strongly from each other. Fumarolic fields are well separable since they have a very distinctive spectral pat- tern, and materials with different emissive prop- 242 Luca Merucci, Maria Paola Bogliolo, Maria Fabrizia Buongiorno and Sergio Teggi Fig. 6. Spectral emissivity image in DAIS channels 78-76-74 (RGB). Fig. 7. Emissivity spectra retrieved from the DAIS data. Average values were calculated on regions of interest selected on the main surface units. 243 Spectral emissivity and temperature maps of the Solfatara crater from DAIS hyperspectral images erties are recognizable, probably corresponding to different degrees of hydrothermal alteration of the surface and to depositions from fumarolic gases. An example of the emissivity spectra of the separated units is reported in fig. 7. Spectral information retrieved from the emissivity images was used to produce a map of the main geo-mineralogical units outcrop- ping at the Solfatara crater, applying the Spec- tral Angle Mapper (SAM) technique (Kruse et al., 1993) (fig. 8). SAM is a physically-based spectral classification that uses an n-dimension- al angle to match pixels to reference spectra (end-members). The algorithm determines the spectral similarity between two spectra by cal- culating the angle between the spectra, treating them as vectors in a space with dimensions equal to the number of bands (RSI, 2002). Non- processed areas and zones covered by vegeta- tion were previously masked as already de- scribed. Pure spectra required by this technique were obtained averaging spectra of small groups of pixels, selected on the image on the basis of information drawn from field survey and spectral analysis (fig. 7). The six units detected were only separated but not yet identified at this stage of the work: to do this, representative emissivity measurements are needed to be compared with image spectra. Emissivity spectra of surface materials are strongly influenced by various factors (Salisbury and D’Aria, 1992) such as granulometry, tem- perature, compaction level, water content. In a remote sensing scene, pixel spectra are also in- fluenced by local un-homogeneities of the de- posits. For these reasons, a direct comparison be- tween image spectra and reference spectra of minerals from public spectral libraries is not fea- sible; field measurements of spectral emissivity are required to perform reliable comparisons. Fig. 8. Map of the main geo-mineralogical units outcropping at the Solfatara crater. 244 Luca Merucci, Maria Paola Bogliolo, Maria Fabrizia Buongiorno and Sergio Teggi 8. Conclusions and future work DAIS thermal infrared data were used to re- trieve surface temperature ad spectral emissivity maps of the Solfatara crater. Our results confirm DAIS capability to provide valuable volcanolog- ical information based on the thermal infrared properties of materials. The quantitative temper- ature map of the area we obtained allows us to locate thermal anomalies corresponding to active fumaroles and to characterize their thermal prop- erties. The spectral emissivity image allows the mapping of the main geo-mineralogical units on the basis of their spectral properties. This the- matic map points out that six thermal infrared channels are sufficient to detect and separate not only different surface covers but also similar ma- terials whose differences are mainly related to their alteration degree. In order to complete the image interpretation and perform the automatic identification of the classes, field acquisition of emissivity spectra is planned. Spectral emissivity will be measured with a portable FTIR (Fourier Transform Infra- Red) spectrometer (Design & Prototypes LTD) working in the 2-16 nm spectral range with a spectral resolution between 2 and 16 cm−1. Sam- pled materials will be analyzed in the laboratory to retrieve chemical and mineralogical features corresponding to the field collected spectra. REFERENCES BERK, A., L.S. BERNSTEIN and D.C. ROBERTSON (1989): MODTRAN: a moderate resolution model for LOW- TRAN-7, AF-GL Technical Report GL-TR-89-0122 (Air Force, Hanscom, MA, U.S.A.). BERRINO, G., G. CORRADO, G. LUONGO and B. TORO (1984): Ground deformation and gravity changes accompany- ing the 1982 Pozzuoli uplift, Bull. Volcanol., 47 (2), 187-200. CHIODINI, G., R. CIONI, G. MAGRO, I. MARINI, C. PANICHI, B. RACO and M. RUSSO (1997): Chemical and isotopic vari- ations of Bocca Grande Fumarole (Solfatara volcano, Phlegraean Fields), Acta Vulcanol., 8 (2), 228-232. CORRADO, G., I. GUERRA, A. LO BASCIO, G. LUONGO and R. RAMPOLDI (1976): Inflation and microearthquake acti- vity of Phlegraean Fields, Italy, Bull. Volcanol., 40 (3), 169-188. DE GENNARO, M., E. FRANCO and D. STANZIONE (1980): Le alterazioni ad opera di fluidi termali alla solfatara di Pozzuoli (Napoli), Mineral. Geochim., 49, 5-22. DI VITO, M., L. LIRER, G. MASTROLORENZO, G. ROLANDI and R. SCANDONE (1985): Volcanological Map of Campi Flegrei (Min. Prot. Civ., Università degli Studi di Napoli). DI VITO, M., L. LIRER, G. MASTROLORENZO and G. ROLAN- DI (1987): The 1538 Monte Nuovo eruption (Campi Flegrei, Italy), Bull. Volcanol., 49, 608-615. DI VITO, M.A., R. ISAIA, G. ORSI, J. SOUTHON, S. DE VITA, M. D’ANTONIO, L. PAPPALARDO and M. PIOCHI (1999): Volcanism and deformation in the past 12 ka at the Campi Flegrei caldera (Italy), J. Volcanol. Geotherm. Res., 91, 221-246. GILLESPIE, A.R. (1985): Lithologic mapping of silicate rocks using TIMS, in The TIMS Data User’s Work- shop, Jet Propulsion Laboratory Publ. 86-38, 29-44. GILLESPIE, A., S. ROKUGAWA, T. MATSUNAGA, J.S. COTHERN, S. HOOK and A. KAHLE (1998): A temperature and emissivity separation algorithm for Advanced Space- borne Thermal Emission and Reflection Radiometer (ASTER) images, IEEE Trans. Geosci. Remote Sens- ing, 36 (4),1113-1126. KRUSE, F. A., A.B. LEFKOFF, J.B. BOARDMAN, K.B. HEIDE- BRECHT, A.T. SHAPIRO, P.J. BARLOON and A.F.H. GOETZ (1993): The Spectral Image Processing System (SIPS) – Interactive visualization and analysis of imaging spec- trometer data, Remote Sensing Environ., 44, 145-163. LI, Z., F. BECKER, M. STOLL and Z. WAN (1999): Evaluation of six methods for extracting relative emissivity spec- tra from thermal infrared images, Remote Sensing En- viron., 69, 197-214. LIRER, L., G. LUONGO and R. SCANDONE (1987): On the vol- canological evolution of Campi Flegrei, EOS, Trans. Am. Geophys. Un., 68, 226-234. REALMUTO, V.J. (1990): Separating the effects of tempera- ture and emissivity: emissivity spectrum normaliza- tion, in Proceedings of the 2nd TIMS Workshop, JPL Publ. 90-55. RSI (2002): ENVI 3.6 User’s guide (Research Systems, Inc.). SALISBURY, J.W. and D.M. D’ARIA (1992): Emissivity of terrestrial materials in the 8-14 mm atmospheric win- dow, Remote Sensing Environ., 65, 83-106. SALISBURY, J.W., L.S. WALTER, N. VERGO and D.M. D’ARIA (1991): Infrared (2.1-25 µm) Spectra of Minerals (Johns Hopkins University Press), pp. 294. SCANDONE, R. (1997): Campi Flegrei (on line: http://vul- can.fis.uniroma3.it/campi_flegrei/geol.gif/, modified after Di Vito et al., 1985). SCANDONE, R., F. BELLUCCI, L. LIRER and G. ROLANDI (1991): The structure of the Campanian Plain and the activity of Neapolitan Volcanoes, J. Volcanol. Geotherm. Res., 48 (1-2), 1-31. VALENTINO, G.M., G. CORTECCI, E. FRANCO and D. STAN- ZIONE (1999): Chemical and isotopic compositions of minerals and waters from the Campi Flegrei volcanic system, Naples, Italy, J. Volcanol. Geotherm. Res., 91, 329-344.