Using EO-1 Hyperion Data as HyspIRI Preparatory Data Sets for Volcanology Applied to Mt Etna, Italy

Transcription

1 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL Using EO-1 Hyperion Data as HyspIRI Preparatory Data Sets for Volcanology Applied to Mt Etna, Italy Michael Abrams, Dave Pieri, Vince Realmuto, and Robert Wright Abstract One of the main goals of the Hyperspectral and Infrared Imager (HyspIRI) mission is to provide global observations of surface attributes at local and landscape spatial scales (tens of meters to hundreds of kilometers) to map volcanic gases and surface temperatures, which are identified as indicators of impending volcanic hazards, as well as plume ejecta which pose risks to aircraft and people and property downwind. Our project has created precursor HyspIRI data sets for volcanological analyses, using existing data over Mt. Etna, Italy. We have identified 28 EO-1 Hyperion data acquisitions, and 12 near-coincident ASTER data acquisitions, covering six eruptive periods between 2001 and These data sets provide us with 30 m hyperspectral VSWIR data and 90 m multispectral TIR data (satellite). They allowed us to examine temporal sequences of several Etnaean eruptions. We addressed the following critical questions, directly related to understanding eruption hazards: 1) What do changes in SO emissions tell us about a volcano s activity? How well do these measurements compare with ground-based COSPEC measurements? 2) How do we use measurements of lava flow temperature and volume to predict advances of the flow front? 3) What do changes in lava lake temperatures and energy emissions tell us about possible eruptive behavior? Index Terms Hyperspectral imaging, infrared image sensors, volcanic activity. I. INTRODUCTION I N2004, the U.S. National Aeronautics and Space Administration commissioned the National Research Council (NRC) to conduct a Decadal Survey for Earth science and applications from space. The 2007 report is titled Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond [23]. The purpose of this study was to provide NASA with a blueprint for the next 10 years, prioritizing science missions, in response to community inputs and recommendations. The report identified key science measurements and recommended a small number of missions to acquire those measurements. The topical areas included Earth science applications and societal benefits, and solid Earth hazards. Included in the recommended missions was the Hyperspectral and Infrared Imager (HyspIRI) instruments that would provide global observations at local and landscape scales (tens of Manuscript received February 29, 2012; revised May 28, 2012 and July 18, 2012; accepted August 28, Date of publication March 12, 2013; date of current version May 13, Work by Abrams, Realmuto, and Pieri was done at the California Institute of Technology, Jet Propulsion Laboratory, under contract with the National Aeronautics and Space Administration. M. Abrams, D. Pieri, and V. Realmuto are with the Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA USA. R. Wright is with the University of Hawaii, Honolulu, HI USA. Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /JSTARS meters to hundreds of kilometers) [2]. The mission would include a hyperspectral visible-near infrared-shortwave infrared (VSWIR) imaging spectrometer, and a multispectral thermal infrared (TIR) scanner. Operating from low earth orbit, the instruments would provide global coverage, frequent repeat observations, and data at 60-m spatial resolution. One of the primary science goals of the mission is to map volcanic gases and surface temperatures, which are identified as indicators of impending volcanic hazards; as well as plume eject a which pose risks to aircraft and people and property downwind. The Decadal Survey laid out a timeline to implement the missions: HyspIRI was originally included in the second tier, with recommended launch dates between 2013 and 2016, but the expected launch date has been slipped to 2020 or later. NASA assigned preliminary study activities for HyspIRI to the Jet Propulsion Laboratory. The goal of our project was to create simulated HyspIRI volcanological data sets. Our primary focus was on Mount Etna, Sicily (Fig. 1), where we investigated active lava flows and volcanic plumes. Mount Etna, a stratovolcano rising over 3300 meters above the Mediterranean, is one of the world s most active volcanoes, and its eruptions have been historically documented for longer than any other volcano [8], [33]. Two styles of eruptive activity typically occur at Etna. Persistent explosive eruptions, sometimes with minor lava emissions, take place from one or more of the three prominent summit craters, the Central Crater, the Northeast (NE) Crater, and the Southeast (SE) Crater (formed in 1978). These are often accompanied by effusive plumes, containing water vapor, ash, and SO. Flank vents, typically with higher effusion rates, are active less frequently and originate from fissures that open progressively downward from near the summit, usually accompanied by strombolian eruptions at their upper ends (Fig. 2). Lavas and pyroclastics of alkali affinity make up the main bulk of Etna representing over 98% by volume of the volcano [8]. Cinder cones are commonly constructed over the vents of lower-flank lava flows. Lava flows extend to the foot of the volcano on all sides and have reached the sea over a broad area on the SE flank. Highly explosive events produce tall eruption columns (e.g., over 6 km high during the 2002 eruption [5], [6]), typically laden with both ash and SO. These are then transported horizontally for hundreds of kilometers. All of the eruption phenomena pose hazards to property and life. For instance, there are numerous towns located on the volcano s flanks, and the nearby city of Catania has been inundated by lava numerous times in the past (e.g., the famous 1669 eruption [33]). Etna s ash plumes also pose direct hazard to aircraft over a large area, including those based at the military airfield at Sigonella, near Catania, actively used by the U.S. Navy, NATO, and the Italian /$ IEEE

2 376 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL 2013 Fig. 1. Location map of Mt. Etna volcano, Sicily, Italy. TABLE I DESCRIPTION OF TWO SATELLITE SENSORS USED TO CREATE PRECURSOR HYSPIRI TIR DATA SETS Fig. 2. Aerial view of Mt. Etna during simultaneous summit and flank eruptions (plume on right). Image taken from the International Space Station on October 30, Air Force; commercial airline traffic from the Fontanarossa airfield that serves Catania; and the Comiso airport. Mount Etna has been, and continues to be, a frequent target for both satellite and aircraft optical instruments [1], [13], [14], [17], [18], [20], [27], [32], [37]. We assembled HyspIRI TIR data from two instruments to create our simulated HyspIRI data sets (Table I). Hyperion began operation in early 2001 as the VSWIR hyperspectral scanner on NASA s Earth Observing-1 (EO-1) satellite platform [38]. It continues to operate, acquiring 196 useable data channels. The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) has been acquiring data since 2000 on NASA s Terra satellite platform [50]. ASTER combines multispectral VSWIR instruments with a multispectral TIR instrument. The combined data sets provided us with 30-m hyperspectral VSWIR data and 90-m multispectral TIR. We analyzed currently available data sets and tools to address the following critical questions, directly related to understanding eruption hazards for configuring the HyspIRI TIR: 1) What do changes in SO emissions tell us about a volcano s activity? 2) How do we use measurements of lava flow

3 ABRAMS et al.: USING EO-1 HYPERION DATA AS HyspIRI PREPARATORY DATA SETS FOR VOLCANOLOGY APPLIED TO MT ETNA, ITALY 377 Fig. 3. Proposed HyspIRI TIR bandpass placements for eight channels. temperature and volume to predict advances of the flow front? 3) At what temperature should we set the saturation limits for HyspIRI s TIR bands (Fig. 3)? For some of these questions we were able to provide answers. For others, we find that current capabilities do not provide adequate data sets (due to temporal limitations) to fully answer the science questions. We have had to content ourselves with showing how the techniques and tools will one day allow us to advance our understanding more fully. II. DATA SETS The simulated HyspIRI TIR data sets over Mount Etna, Italy were assembled from EO-1 Hyperion data, and ASTER data. The five daytime data sets were assembled from the only five coincident daytime ASTER+EO-1 data collection obtained during the eleven overlapping years of operations of the two instruments. They occurred on 15 January 2003, 19 July 2003, 11 August 2003, 8 October 2008, and 30 June The single nighttime data set was acquired14september2004. Corresponding eruptive activity was reported by the Smithsonian Global Volcanism Program ( index.cfm), from which the following were extracted: On 15 January 2003, ash emission increased at the 2750-m-elevation pyroclastic cone on the volcano s upper south flank. There was also an associated increase in lava emission towards the south. On 19 July 2003, no activity was reported. On 11 August 2003, a weak, fluctuating glow was observed at the base of a dense gas column emitted from the NE Crater. On 10 September 2004, a 300-m-long lava flow poured out from a vent; on 13 September another 1-km-long lava flow occurred. On 8 October 2008 and 30 June 2011, no activity was reported. III. LAVA FLOWS The Hyperion data were scaled to radiance-at-the-sensor in units of W/m sr m, resampled to 60-m pixel size; bands 1 to 57 and 79 to 242 were combined to make 196-channel data sets, removing the overlap between Hyperion s two spectrometers. ASTER s five thermal bands were resampled to 60-m pixel size, scaled to radiance-at-the-sensor, and registered to the Hyperion coverage using standard tie-point identification, and nearestneighbor resampling. The Hyperion and ASTER data were combined to a 201-band image. The data sets are in BSQ format, with ENVI text header files. The data sets can be downloaded from the web site The nighttime Hyperion data sets from 12, 14, and 16 September 2004 were analyzed as part of a study to determine cooling rates of lava flows [43], [46], though only one of these had a matching ASTER data acquisition (Fig. 4). On the 12th, only a single flow (Flow I) existed with a flux of 441 MW. On the 14th, two flows were seen, I and II. Because of the high lava flow temperature, coincidently acquired ASTER TIR pixels were mostly saturated (Fig. 4(a)) for both flows. However, Hyperion data had most of the channels unsaturated (Fig. 4(b)), allowing calculation of radiant fluxes and convective heat flux (Fig. 4(c)). For September 14, total heat flux of lava flow I was 761 MW, and lava flow II 161 MW. Two days later these flows had cooled and were putting out energies of 384 MW and 57 MW respectively (flux measurements from [43], [46]). The general decrease in surface temperature was pronounced between the 14th and 16th. This is explained by field observations that the open lava channel had roofed over after the 14th. The combined Hyperion and ASTER daytime data set from 15 January 2003 was characterized by both ash emissions and lava flows. The flows, like those shown for 12 September 2004, were incandescent in visible wavelengths, and saturated some pixels at higher wavelengths in the Hyperion data (Fig. 5). This makes it difficult to accurately calculate energy fluxes from the lava flows; an instrument with much higher saturation levels would be of great value. The ASTER TIR data are also saturated formostofthelavaflow, and cannot be used to determine flow temperatures. However, the spectral color of the emission plume indicates that the major composition at this time was SO -rich (Fig. 6) (more about ASTER and SO later). IV. SATURATION STUDIES The surface temperature of an active lava body, be it a flow, a dome, or a lake, determines the rate at which the lava cools and is an important boundary condition for estimating, for example, how quickly an active lava flow cools, rheologically stiffens, and eventually ceases to flow [27]. Temporal variations in the spectral radiance emitted from erupting volcanoes has also been shown to act as a reliable proxy for eruption intensity (e.g., [19]), and has been shown to correlate with variations in geophysical variables such as seismic energy release and degassing [49]. At Lascar volcano in Chile, simply using satellites to record how the spectral radiance emitted from its summit crater varies over time has been demonstrated as a genuinely robust method for predicting when that volcano is likely to erupt explosively (see [22], [26], and [41]). Although sensors such as the Landsat Thematic Mapper (TM) and its successor the Enhanced Thematic Mapper (ETM) have been used to quantify the surface temperature distributions of active lava flows (see [9], [12], [25], [27], [34], [48]) all have failed to a certain degree. This is because these sensors were designed to observe spectral reflectance from targets such as vegetation, shallow water, rocks and soils, and because of instrument engineering constraints. As such, the upper measurement limit of these sensors (the maximum spectral radiance that a particular waveband can measure, ) is optimized to observe tar-

4 378 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL 2013 Fig September 2004 nighttime ASTER and Hyperion data for two lava flows. (a) Combined ASTER band 12 (red) and Hyperion (blue) data where most ASTER TIR pixels are saturated (bright red and white). (b) Color coded Hyperion 1.6 m data. (c) Radiant heat flux calculated from Hyperion data. gets that reflect less than 100% of albedo (and for most targets, much less). They were not designed to image active lavas for which temperature can be as high as 1150 C (1423K) and from which emission of spectral radiance can be two orders of magnitude greater than Landsat TM can reliably measure [43], [46]. As a result, widespread sensor saturation has been noted as the primary barrier to successfully resolving lava surface temperatures, and volcanogenic thermal fluxes, using these instruments. Although the Terra ASTER TIR sensor was designed with volcanism in mind, the increased dynamic range of the sensor when operating in volcano mode offers only marginal improvement over that afforded by the Landsat family [47]. The significance of this is that if the spectral radiance emitted from a lava surface is unknown, then the common approaches to inverting at-satellite spectral radiance to obtain surface temperature (e.g., the dual-band method of [34]; the more complex mixture models of [44]) cannot be applied without great uncertainty. Neither can temporal variations in the energy emitted from an active volcano be reliably quantified. An important advantage of ASTER data is to measure with high precision the background temperatures in volcanic areas, an important component in the thermal analysis of lave flows. Finally, neither ASTER nor Hyperion instruments are capable of systematic monitoring of any volcanoes (Hyperion) or more than just a few (like ASTER), as compared to Landsat. Temporal repeat coverage (like HyspIRI) is a critical capability. The baseline HyspIRI TIR design includes a middle infrared waveband at 4 m, the purpose of which is to accurately record the thermal radiance emitted from vegetation fires and active volcanoes. Heritage for this capability lies in the design of the Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) sensors, which include an equivalent spectral band referred to as the fire channel [15]. Throughout the 1980s and 1990s the wildfire community suffered from the propensity of its orbital workhorse, the NOAA Advanced Very High Resolution Radiometer (AVHRR) series, to easily saturate over high temperature fires. As a result, the MODIS fire channel (MODIS band 21) was designed to saturate at pixel integrated temperatures of 506K and 478K, on Terra and Aqua respectively, compared to temperatures of 320K for AVHRR. Therefore, the MODIS fire channels have demonstrated great resilience to saturation over vegetation fires. It has also been shown to represent a marked improvement over AVHRR for studying active volcanism (e.g., [44]). The 4- mregionisof special significance for both fire and volcano remote sensing. Over the temperature range K (covering the range of active lavas) emitted spectral radiance at 4- m increases approximately with the fourth power of temperature [42]. As the radiant exitance from a surface also increases with the fourth power of temperature (the Stefan Boltzmann law) a measurement of spectral radiance from an active lava at 4 m can be converted directly to an estimate of power loss (in W/m ; see [42] for a derivation) without the need to unravel the lava s detailed sub-pixel surface temperature distribution. Although the specifications of the MODIS fire channels have virtually eliminated the incidence of saturation over both fires and active lavas, in has been suggested that the for this

5 ABRAMS et al.: USING EO-1 HYPERION DATA AS HyspIRI PREPARATORY DATA SETS FOR VOLCANOLOGY APPLIED TO MT ETNA, ITALY 379 Fig January 2003 Hyperion daytime data. Bands 27 (0.620 m), 57 (0.925 m), and 84 (1.194 m) are in RGB. The plots show radiance at the sensor as a function of wavelength for lava pixels at three different temperatures. channel was set too high, resulting in a lack of precision when observing in the temperature range at which the majority of wildfires burn [10]. An objective of this study was to prevent a similar situation arising with the HyspIRI TIR. To this end we have simulated the response of the HyspIRI TIR to a range of volcanic targets to determine the most appropriate value of to ensure that it is high enough to provide unsaturated data over active lavas, without being too high, and in effect wasting bits. The first step in simulating the response of the HyspIRI 4 m TIR channel to active lavas is to determine the range of surface-leaving radiance that real lavas exhibit at the scale of a 60-m instantaneous field of view, at this wavelength. To do this, we used data acquired by the EO-1 Hyperion sensor to determine the surface temperature distributions for a suite of active lavas. The fact that Hyperion measures a contiguous spectrum between 0.4 and 2.2 m means that as the temperature of an active lava increases, the corollary of saturation (and increasing numbers of unusable wavebands) at longer wavelengths is increasing amounts of emitted thermal radiance (and useable data) at increasingly shorter wavelengths. In this way, Hyperion provides sufficient unsaturated spectral radiance measurements for complex sub-pixel radiance-to-temperature mixture models to be employed, even for the very hottest active lava surfaces. These more complex mixture models allow a more accurate determination of lava surface temperature distributions, and hence surface leaving radiance, than those attainable using the dual-band method (see [44]). This allows the surface temperature distribution of active lavas, and consequently the surface-leaving radiance at the 30-m scale, to be determined in a manner not possible with Landsat/ASTER-class data. Once the sub-pixel temperature distribution is resolved using data ac-

6 380 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL 2013 Fig January 2003 ASTER TIR color composite of Etna emission plume. The yellow color streaming from the summit to the northwest indicates the presence of SO gas due to spectral absorption of SO in the ASTER bands. White pixels are saturated. Bands at 11.3 m, 10.6 m, 8.6 maredisplayedinrgb after a decorrelation stretch enhancement. quired in the m region, the emitted spectral radiance at other wavelengths can be predicted. The method for retrieving lava surface temperatures from Hyperion data is described and validated in [43], [46]. Briefly, the spectral radiance emitted from a lava surface can be written simply as where is the temperature of the th component, is the wavelength, is the number of components, and is the fraction of the sensor s instantaneous field of view that it occupies. Saturated wavebands were excluded, the remaining spectral measurements were corrected for the non-unitary, graybody emissivity of the lava (here a value of 0.95 was assumed for basalt) and absorption of the surface-leaving radiance by the atmosphere (estimated using the MODTRAN radiative transfer code, assuming an appropriate atmosphere, season, and altitude for Etna). Nighttime data were used to avoid contamination of the emitted radiance by reflected sunlight. The best-fit solution of equation (1) to each spectral radiance spectrum was then obtained, to yield a solution for each image pixel, using the Levenberg-Marquardt method. was allowed to float between limits of 100 C (the theoretical lower sensitivity limit of Hyperion) and 1200 C (beyond the basalt liquidus), while was allowed to float between values of 0 and 1. The sum of the values was constrained to not exceed one but could (and usually did) sum to less than one. The remainder of the pixel was assumed too cool to radiate significantly at these wavelengths, and was set to a background temperature of 8 C, appropriate for Mount Etna s flank region. (The background temperature cannot be computed using Hyperion data because the spectrometer is not sensitive to emission from surfaces with a whole pixel at 900 m,temperatureof 90 C.) [43], [46] found that most solutions converged when 2or3.Thesolution with the lowest root mean square error was retained. The set of and values that result allow the surface-leaving radiance, at 4 m and a spatial resolution of 30 m, to be predicted. Fig. 7 shows the results obtained for a sequence of images of lava flows erupted at Mount Etna in late On the left is the simulated 4- m brightness temperature at 30-m resolution calculated from Hyperion data using the methodology described; on the right, the simulated response of HyspIRI s 4 m channel at 60-m spatial resolution. The 60-m data were derived from the 30-m data. Spatial resampling assumed the point spread function of Landsat TM s longwave infrared channel (albeit at 60 m) which, like HyspIRI TIR s infrared subsystem is a whiskbroom scanning instrument. The spectral response function of MODIS band 21 was assumed to perform the spectral resampling. Although the images have been rotated so north is to the top, they have not been geo-registered. Channel-fed aa lava flows, which are much longer than they are wide and are characterized by lava flowing in a channel confined by lateral levees that are much cooler than the lava in the channel itself, are common at Mount Etna, much more so than the rather amorphous pahoehoe flow-fields characteristic of Kilauea volcano in Hawaii. For comparison, Fig. 8 shows an equivalent result obtained from analysis of a Hyperion scene for a much larger lava flow erupted at Nyamuragira, a volcano in the Democratic Republic of Congo. Comparison of the simulated HyspIRI 4- m TIR data with the raw 30-m data reveals that while the central lava channel was somewhat resolved in the 30-m Hyperion data, it becomes indistinct at 60 m. Thus, for this flow, while the HyspIRI TIR would allow for cooling from the lava flow as a whole to be determined the spatial resolution would be insufficient to allow the cooling from the channeled lava itself to be estimated. To place this in context, however, this flow is much smaller than those that are typically of concern at Mount Etna insofar as the hazards the volcano poses. Fig. 8 shows how a more substantial lava flow erupted at Nyamuragira volcano would appear in the HyspIRI TIR data. This flow is more comparable in size to those erupted at Etna in 1983, , 2001, 2006, , flows which in several cases have either destroyed infrastructure or threatened infrastructure to the point that artificial lava diversion measures were taken. The results presented here indicate that for a flow in this size range (i.e., 5 km long and greater) the 60-m spatial resolution of HyspIRI will be sufficient to estimate the cooling rate of the feeder channel itself. Obviously,asthedataareresampledfrom30to60m,so the maximum predicted 4- m brightness temperature decreases. For the purposes of this study, we are interested in how high (or expressed as an equivalent whole pixel brightness temperature) should be, to ensure that saturation of HyspIRI s 4- m TIR channel does not occur over the active lava flow. For the seven Etna data sets depicted in Fig. 7, a of 700K would ensure that no regions of the flow surface would be sufficiently radiant to saturate the HyspIRI 4- m TIR channel at 60 m. However, as a recommendation regarding the necessary to ensure that HyspIRI provides unsaturated data over a wide range of active lava bodies, the Etna example provides insufficient information, given that the flow itself was relatively small. Fig. 9 shows how the incidence of saturation decreases as increases for the much more substantial Nyamuragira

7 ABRAMS et al.: USING EO-1 HYPERION DATA AS HyspIRI PREPARATORY DATA SETS FOR VOLCANOLOGY APPLIED TO MT ETNA, ITALY 381 Fig. 7. The development of a lava flow at Mt. Etna, Sicily is shown in a time sequence of seven Hyperion images (i) (vii) using the 4- m band. These are shown on the let for native 30-m Hyperion resolution using the method described, paired (on the right) with simulated HyspIRI TIR 4- m images, both spatially (60 m) and spectrally. Hyperion images used to generate these results were acquired on (i) 12 September 2004, (ii) 14 September 2004, (iii) 16 September 2004, (iv)23 September 2004, (v) 7 October 2004, (vi) 9 October 2004, and (vii) 3 December Fig. 8. Simulated HyspIRI 4- m image of a lava flow at Nyamuragira volcano, Democratic Republic of Congo. Color bar gives the 4- m brightness temperature for each pixel (in Kelvin). On the left is the 4- m brightness temperature estimated from Hyperion data. On the right, the same data resampled spatially and spectrally to match the proposed specifications of HyspIRI. The Hyperion image used to generate this result was acquired on 20 May example. For this lava flow, a 4- m of 1000K would be required to ensure that the HyspIRI TIR provides unsaturated data at all points on the flow surface. Fig. 9. Histogram showing the number of lava flow pixels predicted to saturate HyspIRI s 4- mbandasafunctionof, for the Nyamuragira example depicted in Fig. 7. V. SO STUDIES The utility of ASTER and MODIS data in studies of volcanic SO plumes is evident from the quantity and variety of published investigations (e.g., [4], [16], [24], [29], [39], [40] or [51]. The use of ASTER data is an ideal preparatory activity

8 382 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL 2013 Fig. 10. Comparison of the TIR spectral response of ASTER, MODIS, and the HyspIRI TIR to the transmission spectra of SO and water vapor H O.(a)ASTER vs. SO,(b)MODISvs.SO, (c) Notional HyspIRI vs. SO2, and (d) Notional HyspIRI vs. H O. for HyspIRI-based SO mapping, due to the similarities in the spectral response (Fig. 10). In addition, the 60-m spatial resolution planned for HyspIRI is similar to the 90-m resolution of ASTER. High spatial resolution increases our sensitivity to SO as the fraction of a pixel occupied by a plume increases with increasing spatial resolution (i.e., decreasing IFOV). The heritage for the HyspIRI TIR channels is shown in Fig. 10. The transmission spectrum of SO is superimposed on the spectral response of ASTER (Fig. 10(a)), MODIS (Fig. 10(b)), and the notional spectral response of the HyspIRI TIR (Fig. 10(c)). In Fig. 10(d) we superimpose the transmission spectrum of water vapor H O on the HyspIRI TIR spectral response. The H O absorption, together with that of CO (not shown here), define the 8 12 m atmospheric window traditionally used in TIR remote sensing of the earth s surface. The transmission spectra shown in Fig. 10 represent a nadir view, or vertical optical path, from the top of the atmosphere (altitude 100 km) to sea level. The spectrum of SO between 7 and 13 m is distinguished by a strong absorption band centered near 7.3 m, a weaker band centered near 8.5 m, and transparency at wavelengths between 9.5 and 13 m. ASTER Channels 10, 11, and 12 cover the 8.5 m absorption band (Fig. 10(a)), and MODIS Channels 28 and 29 cover the 7.3 and 8.5 mso absorption bands, respectively (Fig. 10(b)). Accordingly, the HyspIRI TIR channels at wavelengths less than 10 marebasedonmodischannel28 and ASTER Channels 10, 11, and 12 (Fig. 10(c)). Regarding the 7.3 m channel, we note that strong H O absorption at wavelengths 8.0 m (Fig. 10(d)) will degrade the sensitivity of this channel to SO plumes at altitudes less than 5 km (cf. [28]). Figs. 11 and 12 are examples of SO abundance maps derived from ASTER data acquired on a daytime overpass on 29 July 2001 and a nighttime overpass on 6 June 2000, respectively. ThedatadepictedinFig.11wereacquiredduringaneruption that began on 17 July with the opening of new fissures on the northeastern and southern slopes of the edifice, between the elevations of meters, and the effusion of lava from these fissures (cf. [36]). The effusive activity continued until the end of the eruption on 9 August At the time of the ASTER overpass the eruption was characterized by Strombolian and Hawaiian explosions at the lower end (2740-m elevation) of the southern fissure. Fig. 11(a) is the false-color composite of the ASTER visible and near infrared (VNIR) data. The view of the active lava flow was obscured by steam clouds that formed over the southern fissure. Fig. 11(b) is a color-composite of ASTER TIR data from Channels 14, 13, and 11 (Fig. 10(a)) displayed in red, green, and blue, respectively. To enhance spectral contrast, the ASTER data were processed with the de-correlation stretch algorithm [11] prior to the compositing. The SO plume appears yellow in the TIR color composite (Fig. 11(b)), while the steam clouds, and associated veil of water droplets, appear blue [40]. The steam clouds obscured the view

9 ABRAMS et al.: USING EO-1 HYPERION DATA AS HyspIRI PREPARATORY DATA SETS FOR VOLCANOLOGY APPLIED TO MT ETNA, ITALY 383 Fig. 11. Analysis of ASTER data acquired during a daytime overpass of Mount Etna on 29 July (A) False-color composite of ASTER VNIR data aggregated from a native resolution of 15-m to the 90-m resolution of the ASTER TIR; (B) Color-composite of the data from TIR Channels 14, 13 and 11 (displayed RGB) where the SO plume appears yellow and droplets of liquid water (as opposed to water vapor) appear in blue; and (C) Map of SO column abundance derived from ASTER TIR. of SO was not affected by the lower land surface temperatures encountered during the nighttime overpass. We selected 184 ASTER scenes of Mount Etna acquired between 2000 and 2010 in an effort to construct a time-series of SO emissions from the summit craters. Our initial criteria for selection were clear views of the summit region during periods of surface activity. We applied the de-correlation stretch to each scene and selected 30 scenes for plume mapping based on viewing conditions and the presence of SO plumes. This low yield ( 16%) for candidate scenes limits the utility of ASTER data to monitor changes in SO emission rates on time scales from weeks to months. We will expand our search to include scenes with no evidence of surface activity. This new search will increase the number of candidate scenes for plume mapping, but we do not expect an increase in yield. The HyspIRI TIR will provide more frequent, consistent repeat coverage than either ASTER or Hyperion, so more good scenes will be available. HyspIRI s 60-m spatial resolution will be better than ASTER s 90 m, and improved spectral bands will allow better mapping of plume constituents. VI. CONCLUSIONS Fig. 12. Analysis of ASTER data acquired during a nighttime overpass of Mount Etna on 6 June (a) Color-composite of the data from TIR Channels 14, 13 and 11, (displayed as RGB) where the SO plume appears yellow and droplets of liquid water (as opposed to water vapor) appear in blue. (b) Map of SO column abundance derived from ASTER TIR. of the SO plume, indicating that the clouds were at a higher altitude than the plume. The wind field at this altitude transported the veil of water droplets to the south, while the wind field at the plume altitude transported the plume to the southeast. The SO plume was sheared to the south as it crossed the coast, presumably due to the presence of on-shore winds. Fig.11(c)isamapoftheSO column abundance (mass per unit area), which is the product of the SO concentration (mass per unit volume) and plume thickness. The fundamentals of the retrieval procedures were described in [30] and [31]. The maximum abundances were 6g/m, although there may have been higher concentrations of SO in the portion of the plume obscured by the steam clouds. The ASTER data depicted in Fig. 12 were acquired during a long-lived episode of activity characterized by lava fountains, lava extrusions, Strombolian eruptions, and high gas emissions. This episode began on 26 January and ended on 24 June 2000 (cf. [3]). The data were acquired immediately after the 61st (of 64) lava fountain event on 5 June 2000, which was accompanied by strong gas emissions. Fig. 12(a) is a color-composite of the ASTER TIR data, created with the same procedures described for Fig. 11(a). The SO abundance map, Fig. 12(b), captures the remnants of the high-emission event of 5 June The maximum abundances were 6g/m, with abundances in excess of 8 g/m recorded nearest to the summit craters. A comparison of the SO generated from day- and nighttime overpasses (Figs. 11(c) and 12(b), respectively) indicates that our sensitivity to low concentrations Our main objective in creating these simulated HyspIRI TIR data sets was to begin to illustrate the benefits to scientific analyses from the putative acquisition of multispectral TIR data by HyspIRI at over 2.7 times higher areal resolution as compared with ASTER TIR data (e.g., 3600 m pixel versus 8100 m pixel), and with an expanded TIR bandpass ( m) vs. ASTER (8 12 m). As a result of this study, we are able to draw several of conclusions that suggest HyspIRI TIR data will provide science advantages as compared to similar analyses possible with ASTER TIR data. 1) First, as a result of this effort, we were able to viably simulate HyspIRI TIR-like data using EO-1 Hyperion and ASTER TIR data, creating new HyspIRI-like hybrid data that contained a band at 4 m and seven additional bands ranging from 7.3 mto12 m. 2) As shown here, because of increased spectral resolution and spatial resolution, including a much higher thermal saturation temperature than was possible with ASTER, HyspIRI TIR will improve upon existing ASTER observations of volcanic gas and ash emissions, allowing greater sensitivity and lower detection limits for SO plumes, and better delineation of thermally emitting areas of active lava flows. 3) Improved spatial resolution (e.g., 60 m/pixel for HyspIRI versus 90 m/pixel for ASTER) will materially help in delineating the correspondence between flow morphology and radiantly emitting areas of lava flows. This is critically important to better understand rheo-thermal boundary conditions, thus improving flow dynamic models for lava flow hazard estimations. For instance, a HyspIRI TIR thermal saturation temperature set at 700K for the 4- m channel appears to be adequate to prevent sensor saturation. While for small open-channel aa lava flows (typical for Etna) thermally prominent central channels cannot typically be resolved at 60 m/pixel, for larger flows (e.g., Nymiragira and

10 384 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 6, NO. 2, APRIL 2013 possibly for future Mauna Loa eruptions), central channels are (or will be) visible, because they are significantly hotter than confining levees and flow aprons in the less thermally mixed (i.e., higher spatial resolution) HyspIRI TIR pixels. Such delineations are important in determining effusion rate and predicting ultimate flow run-out lengths, both parameters of crucial importance for mitigating lava flow hazards. 4) For volcanogenic SO studies, the higher spatial resolution of HyspIRI vs. ASTER and inclusion of the additional TIR band at 7.3 m (despite increased water vapor presence at altitudes below 5 km ASL vs. longer wavelengths) will provide increased sensitivity for detection and mapping of both passively and actively emitted cold volcanic plumes in the lower troposphere (e.g., PuuOo plume in Hawaii). Such detections are important in measuring increased or decreased passive degassing rates as potential eruption precursors. 5) HyspIRI TIR s 2.5-day repeat at the equator (includes both day and night observations), and greater frequency at higher latitudes is a major improvement over ASTER, EO-1 and Landsat. This will provide many more views of active volcanoes at better spatial resolution than ASTER. Overall, we feel that the increased spatial resolution and addition of 4- m and TIR bands that will be part of the HyspIRI TIR design offer substantial demonstrated benefits for analyses and hazard evaluations of volcanogenic thermal activity and of passive (and active) tropospheric sulfur dioxide emissions. REFERENCES [1] M.Abrams,R.Bianchi,andD.Pieri, Revisedmappingoflavaflows on Mount Etna, Sicily, Photogramm. Eng. Remote Sens., vol. 62, pp , [2] M. Abrams and S. Hook, HyspIRI: Hyperspectral and Infrared Imager, in Thermal Remote Sensing: Sensors, Methods, Applications,C. Kunzer and S. Dech, Eds. New York: Springer, [3] S. Alparone, D. Andronico, L. Lodato, and T. Sgroi, Relationship between tremor and volcanic activity during the southeast crater eruption on Mount Etna in early, J. Volcanol. Geotherm. Res., vol. 108, no. B5, p. 2241, 2003, /2002JB [4] R. Campion, G. Salerna, P.-F. Coheur, D. Hurtmans, L. Clarisse, K. Kazahaya, M. Burton, T. Caltabiano, C. Clerbaux, and A. Bernard, Measuring volcanic degassing of SO in the lower troposphere with ASTER band ratios, J. Volcanol., Geotherm. Res., vol. 194, pp , [5] S. Carn, L. L. Strow, de Souza-Machado, Y. Edmonds, and S. Hannon, Quantifying tropospheric volcanic emissions with AIRS: The 2002 eruption of Mt. Etna (Italy), Geophys. Res. Lett., vol. 32, p. L02301, 2005, /2004GL [6] S. Carn, L. L. Strow, de Souza-Machado, Y. Edmonds, and S. Hannon, Quantifying tropospheric volcanic emissions with AIRS: The 2002 eruption of Mt. Etna (Italy), Geophys. Res. Lett., vol. 32, p. L02301, 2005, /2004GL [7] J.Cipar,G.P.Anderson,andT.W.Cooley, Temperatureandpower output of the lava lake in Halema uma u crater, Hawaii, using a spacebased hyperspectral imager, IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. (JSTARS), vol. 5, pp , [8] D.Chester,A.Duncan,J.Guest,andK.Kilburn, Mount Etna: The Anatomy of a Volcano. Stanford, CA: Stanford Univ. Press, [9] S. Donegan and L. P. Flynn, Comparison of the response of the Landsat 7 enhanced thematic mapper plus and the Earth Observing-1 advanced land imager over active lava flows, J. Volcanol. Geotherm. Res., vol. 135, pp , [10] B. C. Gao, X. X. Xiong, R. R. Li, and D. Y. Wang, Evaluation of the moderate resolution imaging spectrometer special m fire channel and implications on fire channel selections for future satellite instruments, J. Appl. Remote Sens., vol. 1, 2007, / [11] A. Gillespie, A. Kahle, and R. Walker, Color enhancement of highly correlated images, I, decorrelation and HSI contrast stretches, Remote Sens. Environ., vol. 20, pp , [12] A.Harris,L.P.Flynn,L.Keszthelyi,P.J.Mouginis-Mark,S.K.Rowland, and J. A. Resing, Calculation of lava effusion rates from Landsat TM data, Bull. Volcanol., vol. 60, pp , [13] A. Harris, S. Blake, D. Rothery, and N. Stevens, A chronology of the 1991 to 1993 Mount Etna eruption using advanced very high resolution radiometer data: Implications for real-time thermal volcano monitoring, J. Geophys. Res., vol. 102, pp , [14] A. Harris, F. Massimiliano, and R. Wright, Hazard assessment of Mount Etna using a hybrid lava flow inundation model and satellite-based land classification, Natural Hazards, vol. 58, pp , [15] Y. Kaufman, C. O. Justice, L. P. Flynn, J. D. Kendall, E. M. Prins, L. Giglio, D. Ward, W. P. Menzel, and A. W. Setzer, Potential global fire monitoring from EOS-MODIS, J. Geophys. Res., vol. 103, pp , [16] C. Kearney, K. Dean, V. Realmuto, I. Watson, J. Dehn, and F. Prata, Observations of SO production and transport from Bezymianny volcano, Kamchatka using the moderate resolution infrared spectroradiometer (MODIS), Int. J. Remote Sens., vol. 29, pp , [17] V. Lombardo and M. Buongiorno, Lava flow thermal analysis using three infrared bands of remote sensing imagery: A study case from Mount Etna 2001 eruption, Remote Sens. Environ., vol. 101, pp , [18] V. Lombardo, F. Buongiorno, L. Merucci, and D. Pieri, Differences in Landsat TM derived lava flow thermal structure during summit and flank eruptions at Mount Etna, J. Volcanol. Geotherm. Res., vol. 134, pp , [19] J. Lyons, G. P. Waiter, W. I. Rose, and G. Chinga, Patterns in open vent, Strombolian behavior at Fuego volcano, Guatemala, , Bull. Volcanol., vol. 72, pp. 1 15, [20] A. Krueger, N. A. Krotkov, K. Yang, S. Carn, G. Vicente, and W. Schroeder, Applications of satellite-based sulfur dioxide monitoring, IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. (JSTARS), vol. 2, pp , [21] F. Marchese, N. Pergola, and L. Telesca, Investigating the temporal fluctuations in satellite advanced very high resolution radiometer thermal signals measured in the volcanic area of Etna (Italy), Fluctuation and Noise Lett., vol. 6, pp. L305 L316, [22] S. Matthews, M. C. SGardeweg, and R. S. J. Sparks, The 1984 to 1996 cyclic activity of Lascar volcano, northern Chile: Cycles of dome growth, dome subsidence, degassing, and explosive eruptions, Bull. Volcanol., vol. 59, pp , [23] National Research Council, 2007, Earth Science and Applications From Space: National Imperatives for the Next Decade and Beyond, Committee on Earth Science and Applications from Space: A Community Assessment and Strategy for the Future, National Research Council, , 456 pages. [24] M. Novak, I. Watson, H. Delgado-Granados, W. Rose, L. Cardenas- Gonzalez, and V. Realmuto, Volcanic emissions from Popocatepetl volcano, Mexico, quantified using moderate resolution imaging spectroradiometer (MODIS) infrared data: A case study of the December 2000 January 2001 emissions, J. Volcanol. Geotherm. Res., vol. 170, pp , [25] C. Oppenheimer, Lava flow cooling estimated from Landsat thematic mapper infrared data: The Lonquimay eruption (Chile, 1989), J. Geophys. Res., vol. 96, pp. 21,865 21,878, [26] C. Oppenheimer, P. W. Francis, D. A. Rothery, R. W. T. Carlton, and L. S. Glaze, Infrared image analysis of volcanic thermal features: Lascar Volcano, Chile, , J. Geophys. Res., vol. 98, pp , [27] D. Pieri, L. Glaze, and M. Abrams, Thermal radiance observation of an active lava flow during the June 1984 eruption of Mount Etna, Geology, vol. 18, pp , [28] A. Prata, W. Rose, S. Self, and D. O Brien, Global, long-term sulphur dioxide measurements from TOVS data: A new tool for studying explosive volcanism and climate, Geophys. Monograph, vol. 139, pp , [29] S. Pugnaghi, G. Gangale, S. Corradini, and M. F. Buongiorno, Mt. Etna sulfur dioxide flux monitoring using ASTER-TIR data and atmospheric observations, J. Volcanol. Geotherm. Res., vol. 152, pp , [30] V. Realmuto, A. Sutton, and T. Elias, Multispectral thermal infrared mapping of sulfur dioxide plumes: A case study from the East Rift Zone of Kilauea Volcano, Hawaii, J. Geophys. Res., vol.102,pp. 15,057 15,072, 1997.

11 ABRAMS et al.: USING EO-1 HYPERION DATA AS HyspIRI PREPARATORY DATA SETS FOR VOLCANOLOGY APPLIED TO MT ETNA, ITALY 385 [31] V. Realmuto and H. Worden, Impact of atmospheric water vapor on the thermal infrared remote sensing of volcanic sulfur dioxide emissions: A case study from the Pu u O o vent of Kilauea Volcano, Hawaii, J. Geophys. Res., vol. 105, pp. 21,497 21,508, [32] M. Rix, P. Valks, N. Hao, J. van Geffen, C. Clerbaux, L. Clarisse, P.-F Coheur, R. D. G. Loyola, T. Erbertseder, W. Zimmer, and S. Emmadi, Satellite monitoring of volcanic sulfur dioxide emissions for early warning of volcanic hazards, IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens. (JSTARS), vol. 2, pp , [33] G. Rodwell, Etna, A History of the Mountain and Its Eruptions. London, U.K.: C. Kegan Paul & Co., 1878, 218 pp. [34] D. Rothery, P. W. Francis, and C. A. Wood, Volcano monitoring using short wavelength infrared satellite data, J. Geophys. Res., vol. 93, pp , [35] A. Rybin, M. Chibisova, P. Webley, T. Steensen, P. Izbekova, C. Neal, and V. Realmuto, Satellite and ground observations of the June 2009 eruption of Sarychev peak volcano, Matua Island, Central Kuriles, Bull. Volcanol., vol. 73, pp , [36] S. Scollo, P. Del Carlo, and M. Coltelli, Tephra fallout of the 2001 Etna flank eruption: Analysis of the deposit and plume dispersion, J. Volcanol. Geotherm. Res., vol. 160, pp , [37] S. Teggi, M. Bogliolo, and M. Buongiorno, Evaluation of SO emission from Mount Etna using diurnal and nocturnal multispectral IR and visible imaging spectrometer thermal IR remote sensing images and radiative transfer models, J. Geophys. Res., vol. 104, pp , [38] S. Ungar, J. Pearlman, and J. Mendenghall, Overview of the Earth Observing One (EO-1) mission, IEEE Trans. Geosci. Remote Sens., vol. 41, pp , [39] M. Urai, Sulfur dioxide flux estimation from volcanoes using advanced spaceborne thermal emission and reflection radiometer A case study of Miyakejima volcano, Japan, J. Volcanol. Geotherm. Res., vol. 132, pp. 1 13, [40] M. Watson, V. Realmuto, W. Rose, A. Prata, G. Bluth, Y. Gu, C. Bader, and T. Yu, Thermal infrared remote sensing of volcanic emissions using the moderate resolution imaging spectroradiometer, J. Volcanol. Geotherm. Res., vol. 135, pp , [41] M. Wooster and D. A. Rothery, Thermal monitoring of Lascar volcano, Northern Chile, using infrared data from the along track scanning radiometer, Bull. Volcanol., vol. 58, pp , [42] M. Wooster, B. Zhukov, and D. Oertel, Fire radiative energy for quantitative study of biomass burning: Derivation from the BIRD experimental satellite and comparison to MODIS fire products, Remote Sens. Environ., vol. 86, pp , [43] R. Wright, H. Garbeil, and A. Davies, Cooling rate of some active lava flows determined using an orbital imaging spectrometer, J. Geophys. Res., vol. 115, pp. B06205 B06218, [44] R. Wright and L. Flynn, On the retrieval of lava-flow surface temperatures from infrared satellite data, Geology, vol. 31, pp , [45] R. Wright and L. P. Flynn, Space-based estimate of the volcanic heat flux into the atmosphere during 2001 and 2002, Geology, vol. 32, pp , [46] R. Wright, H. Garbeil, and A. G. Davies, Cooling rate of some active lavas determined using an orbital imaging spectrometer, J. Geophys. Res., vol. 115, p. B06205, 2010, /2009JB [47] R.Wright,D.A.Rothery,S.Blake,A.J.L.Harris,andD.C.Pieri, Simulating the response of the EOS Terra ASTER sensor to high-temperature volcanic targets, Geophys. Res. Lett., vol. 26, pp , [48] R.Wright,D.ARothery,S.Blake,andD.C.Pieri, Improvedremote sensing estimates of lava flow cooling: A case study of the 1991 to 1993 Mount Etna eruption, J. Geophys. Res., vol. 105, pp , [49] R.Wright,S.DeLaCruz-Reyna,A.J.L.Harris,L.P.Flynn,andJ. J. Gomez-Palacios, Infrared satellite monitoring at Popocatepetl: Explosions, exhalations, and cycles of dome growth, J. Geophys. Res., vol. 107, 2002, /2000JB [50] Y. Yamaguchi, H. Fujisada, and M. Kudoh, ASTER instrument characterization and operation scenario, in: Calibration and characterization of satellite sensors, Adv. Space Res., vol. 23, pp , [51] A. Rybin, M. Chibisova, P. Webley, T. Steesen, P. Izbekov, C. Neal, and V. Realmuto, Satellite and ground observations of the June 2009 eruption of Sarychev Peak volcano, Matua Island, Central Kuriles, Bull. Volcanol., vol. 73, no. 9, pp , 2011, doi: /s Michael Abrams received degrees in biology and geology from Caltech. Since 1973 he has worked at NASA s Jet Propulsion Laboratory, in geologic remote sensing. He has been on the science team for many instruments, including Skylab, HCMM, Landsat, and EO-1. Areas of specialization are mineral exploration, natural hazards, volcanology, and instrument validation. He has been on the US/Japan ASTER Science Team since 1988, and became the ASTER Science Team Leader in Dave Pieri received the B.S. degree in physics from Villanova University, and the Ph.D. degree in geology from Cornell University. He joined the JPL staff in His early interest in planetary geology led him to the position of Project Scientist on Viking Mars mission. His ore recent research interests are remote sensing of volcanic ash and gas plumes in the context of aviation hazards, and in-situ measurements related to the calibration and validation of their remotely sensed properties; applications of robotic vehicles for in-situ observations of volcanic processes; nature, character, and environmental history of valley networks on Mars and Titan; and the geomorphology of drainage networks in general. Vince Realmuto received the M.S. and Ph.D. degrees in geology from the University of Arizona. Since joining the JPL staff in 1988 he has followed his interests in application of quantitative remote sensing techniques to the study of volcanic and geothermal phenomena; application of 3D computer graphics and animation techniques to visualize earth science data sets; and conceptual design and evaluation of remote sensing instrumentation. Robert Wright is an Associate Researcher at the University of Hawaii. After receiving the Ph.D. from the Open University, U.K., he pursued his research interest in remote sensing of volcanoes. His current projects include hyperspectral temperature retrievals for active lava flows, domes and lakes; low temperature thermal precursors to large basaltic eruptions; operational thermal volcano monitoring using MODIS ( statistical analysis of low spatial resolution satellite data for volcano surveillance; field-based temperature measurements of active lava flows; HawaiiView, Satellite Remote Sensing data and images.