Satellite analysis and PUFF simulation of the eruptive cloud generated by the Mount Etna paroxysm of 22 July 1998

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. B12, 2373, doi: /2001jb000630, 2002 Satellite analysis and PUFF simulation of the eruptive cloud generated by the Mount Etna paroxysm of 22 July 1998 Marco Aloisi, 1 Marcello D Agostino, 1 Kenneson Gene Dean, 2 Antonino Mostaccio, 1 and Giancarlo Neri 3,4 Received 23 May 2001; revised 7 May 2002; accepted 30 May 2002; published 27 December [1] We applied the PUFF algorithm [Searcy et al., 1998] to simulate the space-time evolution of the eruptive cloud generated by one of the main paroxysms to have occurred this past century at Etna volcano, namely the 22 July 1998 explosive event at the Voragine summit crater. The comparison of PUFF simulations to satellite images of the ash cloud at different times allowed us to estimate several parameters of the eruptive event, such as onset time, duration, ash cloud height and shape, and pyroclast size. Based on this analysis, we concluded that the paroxysm started around 1645 GMT, lasted between 20 and 40 min, and generated a Poissonian-shaped ash column nearly 13 km high and composed of particles with a mean diameter (MD) of 10 3 m and size logarithmic standard deviation (LSD) equal to 1.5. The PUFF simulations using horizontal and vertical diffusivity values of 5000 and 10 m 2 s 1, respectively, provided the best agreement to the eruption clouds observed on satellite images. The results were compared to the information available in the literature concerning the eruption, e.g., volcanological, video camera, and volcanic tremor data. The analysis showed that the method of simulating the eruptive clouds and comparing simulations to satellite images can give a contribution to the study of paroxysmal events generated by the Etna volcano. Also, the degree of accuracy of the cloud simulation leads us to be optimistic about the potential of using this method as a tool for hazard mitigation in the Etna region, with particular applications to air traffic. INDEX TERMS: 8419 Volcanology: Eruption monitoring (7280); 0933 Exploration Geophysics: Remote sensing; 9335 Information Related to Geographic Region: Europe Citation: Aloisi, M., M. D Agostino, K. G. Dean, A. Mostaccio, and G. Neri, Satellite analysis and PUFF simulation of the eruptive cloud generated by the Mount Etna paroxysm of 22 July 1998, J. Geophys. Res., 107(B12), 2373, doi: /2001jb000630, Premise and Goal of the Investigation [2] The volcanic activity of Mount Etna, Sicily (Figure 1), mainly consists of nearly continuous degassing from summit craters, strombolian phases of highly variable intensity, and frequent basaltic lava flows representing a primary source of volcanic hazard in the area. High magnitude explosive events at summit craters can also be included among typical eruptive manifestations of the volcano, as witnessed in the last decades by the occurrence of several paroxysms leading to the formation of eruption columns more than 10 km high. Recent stratigraphic investigations of eruption deposits revealed the occurrence of one Plinian and tens of sub-plinian eruptions over the last 100 kyr [Coltelli et al., 1998, 2000]. 1 Istituto Nazionale di Geofisica e Vulcanologia, Catania, Italy. 2 Geophysical Institute, University of Alaska, Fairbanks, Fairbanks, Alaska, USA. 3 Dipartimento di Scienze della Terra, Università di Messina, Italy. 4 Also at Istituto Nazionale di Geofisica e Vulcanologia, Catania, Italy. Copyright 2002 by the American Geophysical Union /02/2001JB000630$09.00 [3] Volcanic ejecta from the most intense explosive phases are a notable cause of damage to local cultivations ( and of risk to vehicles and aircraft in a wide area around the volcano (e.g., see the La Sicilia newspaper of 23 July 1998). One of the most recent examples of volcanic hazards is represented by the encounter between an aircraft and an eruption cloud from Etna on 26 April 2000 ( Shortly after the end of the main paroxysmal phase at 0739 GMT, an Air Europe jet, that had departed from the Fontanarossa International Airport of Catania in the direction of Milano, encountered the plume at an altitude of about 1000 m. The aircraft received windshield damage from the violent impact with scoriaceus lapilli and was forced to return immediately to the airport of Catania ( [4] Italian research institutions are studying the application of satellite technology for monitoring volcanic plumes and eruption clouds in the Mount Etna area (Figure 1). In the present study, satellite images of the eruption cloud produced by the 22 July 1998 paroxysm at Voragine crater have been used to study the movement of the cloud away from the volcano and as reference data for theoretical modeling of ash emission and dispersion. ECV 9-1

2 ECV 9-2 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 Figure 1. Satellite image of southern Italy (copyright EUMETSAT-Meteosat archive) showing cultural and geographic features in the Mount Etna area. The modeling approach has helped us to obtain information concerning the eruptive process and to make some progress toward the definition of new hazard mitigation strategies to be adopted in the Mount Etna volcanic region. [5] Satellite remote sensing and ash dispersion modeling are proven techniques to mitigate hazard associated with airborne ash [Brantley, 1990; Kienle et al., 1990; Tanaka, 1991; Dean et al., 1998]. Thermal infrared images from GOES (Geostationary Environmental Satellite) and AVHRR (Advanced Very High Resolution Radiometer) satellites have been particularly useful for detection of airborne ash clouds [Matson, 1984; Prata, 1989; Holasek and Rose, 1991; Dean et al., 1994a, 1994b, 1998, 2002; Holasek and Self, 1995; Schneider et al., 1995, 1999, 2000; Searcy et al., 1998]. Timesequential satellite data have been used to track the movement of eruption clouds generated by several volcanoes, such as, Spurr in August 1992 [Schneider et al., 1995] and Cleveland in February 2001 [Dean et al., 2002]. [6] In addition to satellite data, atmospheric dispersion models have been developed to predict the movement of ash clouds, such as MEDIA ( vaac), VAFTAD [Heffter and Stunder, 1993], PUFF [Searcy et al., 1998] and HYSPLIT ( ready/hysplit4.html). These models require gridded wind fields at the time of the eruption to predict the position and shape of the ash cloud. PUFF is designed to predict the extent and movement of airborne ash from intermediate size explosive eruptions in an operational setting, that is, it runs quickly, on a variety of relatively inexpensive platforms and is user friendly. Simulations begin by placing hypothetical particles above a selected volcano and using a modeled, gridded wind field to predict particle movement. The model is designed to track the movement of small (<0.1 mm) particles. The initial distribution of the ash particles in the eruption column is assumed using a column shape option. Simple column shapes (linear, exponential, or Poisson) have often shown to lead to satisfactory results in the modeling of the cloud movement [e.g., Searcy et al., 1998]. PUFF uses default parameters that represent best-guess initial eruption conditions but these can be modified to generate the best fit of an eruption cloud observed on satellite images. Satellite images have been used to validate the accuracy of PUFF which has shown to be satisfactory for intermediate level eruptions, such as Rabaul volcano in 1994, Klyuchevskoy in 1994, and Spurr in 1992 [Searcy et al., 1998]. Therefore, PUFF may provide valuable information for studying the eruption of 22 July 1998 at Mount Etna. The modeling of larger and more complex eruptive processes can be expected to require further work on the algorithm and

3 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 ECV 9-3 will probably be achieved in the near future by coupling PUFF with other models under development, such as ATHAM ( atham). Current efforts for improving PUFF s predictive capabilities include seam less reinitialization as new wind field forecasts become available during the tracking of ash, ashfallout that includes new settling parameters, and tracking of multiple eruptions ( alaska.edu). Efforts are also being made in order to develop a computational technique for estimating the range of uncertainty of the eruption model found by simulations. 2. The 22 July 1998 Mount Etna Paroxysm [7] During June and the first weeks of July 1998 the Etna volcano was characterized by nearly continuous strombolian activity at all summit craters [La Volpe et al., 1999]. In the last 10 days of July the eruptive activity underwent a clear style change to more sporadic but very energetic explosive events, the first and strongest of which took place on 22 July at Voragine crater between 1630 and 1730 GMT, approximately (Figures 2 and 3) ( [IIV, 1998; La Volpe et al., 1999]. This event was followed [IIV, 1998] by four additional notable explosive events over a period of 2 months, at Voragine (6 August) and SE Crater (15, 25, and 30 September). [8] The paroxysm of 22 July 1998 at Voragine was one of the strongest explosive events at Mount Etna in the last century. The pyroclastic deposit volume was estimated between 1 and 5 million m 3 ( boris). The event produced an eruptive column that was more than 10 km high [La Delfa et al., 2000] ( a proximal deposit of huge bombs and spatter up to 15 m thick (Figure 3a) ( mtu.edu/boris) [IIV, 1998], and a heavy shower of lapilli and ash mainly on the SE quadrant of the volcanic area (Figure 3b) ( [IIV, 1998; La Volpe et al., 1999]. A several millimeter thick fall-out deposit covered the town of Catania at a distance of km from the volcano summit ( [9] A fairly detailed description of the eruptive episode is given by IIV [1998]. In the morning of 22 July 1998 fairly weak strombolian activity was present at all summit craters of Mount Etna. In the early afternoon, sporadic detonations from the Voragine crater started to be heard and, after 1 or 2 hours, strong steam emission episodes began to accompany the detonations. Around 1600 (all times described in the following discussion are GMT) steam emission episodes became vigorous and at 1626 the first ash emission was noted. At 1638 magma jets became recognizable on images from the video camera installed at the summit of the volcano (Figure 3b). The eruptive episode appears to reach the maximum intensity between 1637 and 1642, when magma jets of several hundred meters were observed. These jets were soon obscured by the formation of an eruptive column. The column development was nearly complete at At the same time the Bocca Nuova and SE craters (Figure 3) were obscured in the video camera images by the lapilli fall; the image obscuration was complete at The summit crater area was again observed in the video camera data at 1736, as the fall of pyroclastic material decreased at the summit. [10] Photographs of the eruption column ( geo.mtu.edu/boris) are shown in Figure 2, taken from west (Figure 2a), north (Figure 2b), and SE (Figure 2c), respectively. These photographs show variations in the structure of the column, and ash falling from the cloud SE of the volcano (Figures 2a and 2b). The photographs also show low-altitude material drifting to the south and east, and high-altitude material drifting to the north. No exact timing of photographs is available. [11] Additional information on the eruption is given by the seismic stations ESP and PDN located on the volcano edifice (Figure 3b) [Instituto Internazionale di Vulcanologia (IIV ), 1998]: the volcanic tremor amplitude recorded at these stations showed a progressive increase from 1545 and reached the maximum between 1648 and At 1716 the tremor amplitude quickly declined to nearly background values. 3. Satellite Images [12] Meteosat images of the eruptive cloud, kindly furnished to Istituto Nazionale di Geofisica e Vulcanologia by Eumetsat, have been used in the present study. The data are from a geostationary satellite that records images in three spectral bands at 30 min intervals. The three spectral bands are visible mm (VIS), thermal infrared mm (WV), and thermal infrared mm (IR). The spatial resolution is 2.5 km in the visible channel and 5 km in the thermal infrared channels (more details concerning Meteosat sensors and data can be found at the Web site The IR Meteosat data recorded between 1700 and 2000 of 22 July are shown in Figure 4. The plume was observed in all three bands but seen best in IR. The cloud on the satellite images of Figure 4 appears to be very well correlated in space and time to the Etna eruption. It can be noted that, although the eruption started prior to 1700, the 1700 image (Figure 4a) does not indicate volcanic activity due to the low resolution (5 km) relative to the small size of the cloud at that time. The cloud is evident above the volcano on the next image at 1730 (Figure 4b). There were no weather clouds in the region at that time. We also searched for AVHRR data in the NOAA Satellite Active Archive ( html), but discovered that no image of the Etna area was recorded by the orbiting systems in the hours following the paroxysm. 4. Physical Approach to the Eruptive Column and Cloud Modeling [13] Particle dispersion models have proven useful for predicting the location and movement of airborne volcanic ash [e.g., Heffter and Stunder, 1993; Searcy et al., 1998]. The PUFF model [Searcy et al., 1998] is currently used in the Pacific Region and operated at the University of Alaska Geophysical Institute, the National Weather Service in Anchorage, the Japan Weather Agency, the University of

4 ECV 9-4 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 Figure 2. Photographs of the 22 July 1998 paroxysmal event at Voragine crater (Figure 3) taken from west (a), north (b), and SE (c), respectively. No exact timing is available. The photographs were acquired from the Web site ( with the permission of Boris Behncke.

5 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 ECV 9-5 Figure 3. Proximal (a) and distal (b) pyroclastic deposits of the 22 July 1998 eruptive event (modified from IIV [1998]). Tsukuba Japan, and the Air Force Weather Agency. In the present study, PUFF is used for tracking the space-time history of the eruptive cloud generated by the 22 July 1998 Mount Etna volcanic eruption. Results from PUFF were compared to the satellite data to assess the accuracy of the model. [14] Initiation and duration of the eruptive event, initial shape and height of the eruptive column, size distribution of the erupted particles, wind field and transport properties (i.e., convective diffusivity) around the volcano are input parameters for the PUFF model. The cloud situation at different times T i after the start of the eruption is given in

6 ECV 9-6 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 Figure 4. IR satellite images (copyright EUMETSAT-Meteosat archive) recorded between 1700 and 2000 of 22 July The temperature scale is given on the right bottom. the PUFF output files: each file contains the age, size, and location of all particles emitted until T i. The algorithm is based on the physical relationship R i ðt þ tþ ¼ R i ðþþwt t ðþt þ Zt ðþtþs i t ð1þ where: R i (t) = position vector of the ith particle at time t; W(t) = local wind velocity; Z(t) = turbulent dispersion vector; S i (t) = gravitational fallout vector relative to the ith ash particle.

7 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 ECV 9-7 Figure 5. model. Three possible vertical ash distributions that can be used in the initialization of the PUFF Z(t) is a vector containing three-component Gaussian random numbers with zero mean and standard deviation c h,c h,c n (h = horizontal; n = vertical) related to the diffusion rate. S i ðþis t ð0; 0; s i Þ ¼ 0; 0; 2rgd2 i 9h where r is the density of the particles, d i is the size of the ith particle, g is gravitational acceleration, and h is the dynamic viscosity coefficient. This use of Stoke s law is applicable for fine grain particles (<0.1 mm) [e.g., Sparks et al., 1997]. Shape factors can be introduced to account for nonspherical particles.in the present study PUFF was used as follows: 1. a grid was defined in the PUFF parameter space; the column shape option linear/poisson/exponential (Figure 5) was one of the parameters; 2. for each grid node (a set of parameter values) a run of PUFF was performed and the PUFF output (predicted location of the eruptive cloud at a given time) was compared to the satellite image recorded at the same time; 3. nodes leading to an acceptable agreement between the PUFF theoretical cloud and the cloud satellite image were identified. The most reasonable values for the eruptive process parameters and related uncertainties were therefore estimated. [15] Four-dimensional wind fields of 22 July 1998 useful to PUFF have been interpolated from meteorological data given on the following grid: 1. time: 22 July 1998, 1200, and 23 July 1998, 0000; 2. space (vertical): 925, 850, 700, 500, 400, 300, 200, and 100 hpa; 3. space (horizontal): E and N with step 0.5. The data on this grid have been kindly furnished by Capt. Lucio Torrisi, Italian National Meteorological Center, Rome. Figure 6 shows the wind field at 1200 of 22 July, at the 500 and 100 hpa altitude levels (nearly 6000 and ð2þ m), respectively. Note the clear change of wind direction with altitude. 5. Modeling of the Eruptive Column and Cloud [16] Hundreds of simulations of the eruption cloud space-time history have been performed by changing values of the input parameters with the procedure described in the previous section. We used the following grid in the PUFF parameter space: (1) column height from 7 to 14 km with 0.5 km step; (2) eruption start between 1615 and 1730 with 5 min intervals; (3) eruption duration from 0.20 to 2 hours with step varying between a minimum of 3 min and a maximum of 10 min; (4) particle size mean value between 10 1 and 10 7 m with 10 1 steps; (5) logarithmic standard deviation (LSD) of the particle size distribution between 0.5 and 2.5 with 0.5 steps; (6) column shape: linear, Poissonian and exponential; (7) horizontal diffusivity between 10 and 20,000 m 2 s 1 with steps varying in the range m 2 s 1 ; and (8) vertical diffusivity between 0 and 1000 m 2 s 1 with steps in the range m 2 s 1. Since we compare the simulated cloud and the satellite image of the eruption cloud tens of kilometers from the vent, i.e., we analyze distal portions of the plume, only small particles (<0.1 mm approximately) are effectively used in the modeling process. Stokes settling law is an appropriate model for particles in this size range. It is evident that our estimate of the size Gaussian distribution of the erupted pyroclasts is based on the study of the motion of particles corresponding to a limited part of the Gaussian, only (<0.1 mm). [17] Simulations were performed using 5000 particles. Several tests showed that results do not change significantly using a larger number of particles. The visual comparison between the simulated cloud and the cloud signature appearing in the satellite image guided our evaluation of how the eruption model agreed with satellite data. PUFF input parameters that best fit volcanic clouds on 22 July 1998 satellite data (Figure 7) were: column height 13 km, eruption start at 1645, eruption duration 0.50 hours, particle size mean

8 ECV 9-8 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 Figure 6. Wind field at 1200 of 22 July 1998 in south Italy at the 500 hpa (a) and 100 hpa (b) altitude levels (these data have been kindly furnished by Capt. Lucio Torrisi, Italian National Meteorological Center, Rome). Velocity values can be deduced from the related legend (c); the wind direction is towards the point of application of the bar (standard convention).

9 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 ECV 9-9 Figure 7. Shows the PUFF simulation that best fits the time-sequential IR Meteosat images of the eruption cloud. The PUFF cloud was simulated under the following conditions: eruption start 1645, eruption duration 0.5 hour, column height 13 km, Poissonian column shape, particle diameter mean value 10 3 m, diameter log sta ndard deviation 1.5. The different shadings of the particles indicate different altitude (see legend).

10 ECV 9-10 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 Figure 8. Shows the effects of changing one input parameter in the PUFF model simulations, with the other parameters remaining fixed to the best model values (caption of Figure 7). Simulation time on the right bottom. (a and b) Changes in the eruption start time (ES). (c and d) Changes in the eruption duration (ED). (e and f ) Changes in the eruption column height (CH). (g and h) Changes in the eruption column shape (CS).

11 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 ECV 9-11 value 10 3 m, logarithmic standard deviation of the particle size distribution 1.5, Poissonian column shape, horizontal and vertical diffusivity equal to 5000 and 10 m 2 s 1, respectively. [18] Comparing other simulations to satellite data allowed us to evaluate the resolution power of the analysis, e.g., how changes in input parameter values around the best model affect the accuracy of the fit. For sake of brevity, only a few selected simulations are displayed in Figure 8, and these have been chosen to show the degree of uncertainty of the model parameters. In other words, Figures 8a 8h intend to show what are the minimum and maximum values of each model parameter (eruption start and duration, column height, etc.) marking the transition to unsatisfactory fits. [19] Figures 8a and 8b show the simulated eruptive cloud at 1730 by assuming the eruption start at 1630 and 1700, respectively. All other model parameters are fixed on the best model values (caption of Figure 7). Figure 8a shows that by assuming an eruption start time of 1630 (e.g., 15 min earlier than in the best-model case) the simulated cloud advances beyond the cloud seen in the satellite image. Overlying the simulated cloud and the satellite signature of the cloud shows a slightly different shape. Figure 8b shows that if the eruption starts at 1700 (e.g., 15 min later than in the best-model case) the simulated cloud is clearly behind the plume recorded by the satellite. [20] Figures 8c and 8d shows the cloud simulation at 1730 and 1800 in the respective cases of eruption duration T = 0.25 and 0.75 hour. When a duration of 0.25 hour is assumed (Figure 8c) a slightly greater particle dispersion can be observed compared to the best fit of Figure 7a (T = 0.50 hour): this may be easily explained by considering that the eruption in Figure 8c ends 15 min earlier than in Figure 7a. The upper bound of the eruption duration was investigated by using simulations at a slightly later time (1800, Figure 8d) than in the previous analysis of the duration lower bound (1730, Figure 8c). It can be seen from Figure 8d that an eruption duration of 0.75 hour (e.g., 15 min longer than in the best fit case of Figure 7b) produces a simulated cloud less dispersed than in the best fit case and in the satellite image. This is because in the simulation of Figure 8d the eruption ends 15 min later than in the case of Figure 7b. [21] Figures 8e and 8f show the cloud at 1900 corresponding to an eruption column height equal to 12 and 14 km, respectively. Figure 8e displays a simulated plume lagging behind the cloud in the satellite image. Conversely, Figure 8f shows numerous particles of the simulated plume clearly ahead of the cloud in the satellite image. The discrepancy is a consequence of variations in the wind direction and velocity as a function of altitude in the volcano region the day of the eruption (Figure 6). The wind variation with altitude was therefore instrumental to our estimation of the column height at 13 km. The column height estimated here is significantly greater than the previous estimate of 10 km based on visual observations. We requested the atmospheric temperature profile in the study area at the time of the eruption from the National Center of Meteorology and Climatology of Italy. Unfortunately, this information does not exist for the eruption day in the volcano area and surroundings. The closest atmospheric sounding to the volcano for 22 July 1998 was more than 250 km distant. If data were available in closer proximity, the column height could have also been estimated by comparing the cloud temperature derived from satellite data (Figure 4b) to the vertical atmospheric temperature profile from the sounding. This approach was successfully used by Dean et al. [1994a, 1994b] in the study of the eruptive column generated by the 8 January 1990 eruption at Redoubt volcano. [22] Figures 8g and 8h show PUFF model simulations at 1900 for linear- and exponential-shaped columns, respectively. Simulations using the linear-shaped column during initialization (Figure 8g) do not result in a cloud that matches the one observed on satellite images, while assuming a exponential-shaped column leads to a clearly better fit (Figure 8h) similar to that obtained assuming a Poissonshaped column (Figure 7d). Some minor differences in the shape of the simulated clouds derived from the exponentialand Poissonian-shaped columns (Figures 8h and 7d) can, however, be observed NE of the volcano edifice. The simulated cloud generated by the exponential-shaped column (Figure 8h) seems to undergo a slightly greater dispersion compared to the Poissonian simulation and the satellite image (Figure 7d). [23] Concerning the statistical distribution of the particle size, a mean diameter (MD) of 10 3 m with a logarithmic standard deviation of 1.5 was used for the best fit simulation of Figure 7. Other simulations (not shown) indicate that MD values between 10 2 and 10 4 lead to reasonable fit when compared to the cloud on the satellite data. If MD values greater than 10 2 are used, the particles tend to dissipate prematurely when compared to the cloud in the satellite image. MD values less than 10 4 lead to a suspended material distribution that seems unreasonable when compared to the satellite image of the cloud, i.e., too high of a concentration of simulated particles at the leading edge of the plume and too low towards the volcano edifice. [24] Finally, the simulations gave us the opportunity to constrain the atmosphere diffusivity coefficient used in PUFF from 2000 to 7000 m 2 s 1 (best model 5000 m 2 s 1 ) and 0 50 m 2 s 1 (best model 10 m 2 s 1 ) for the horizontal and vertical components, respectively. The above values are compatible with those found by previous investigators in other regions [see Searcy et al., 1998]. 6. Conclusions [25] Time-sequential Meteosat images clearly detected the shape and position of the ash cloud from the 22 July 1998 Mount Etna paroxysm and were good reference data for our theoretical modeling of the cloud space-time history using the PUFF ash tracking model. PUFF accurately modeled the movement and shape of the eruption cloud. Using the satellite image of the eruptive cloud as reference, the simulations made at the nodes of a wide and detailed grid in the PUFF parameter space allowed us to estimate several eruption parameters and to evaluate the respective uncertainties. Based on this analysis we concluded that the paroxysm started around 1645 GMT, lasted between 20 and 40 min, and generated a Poissonian-shaped column nearly 13 km high composed of particles with a MD of 10 3±1 m and size logarithmic standard deviation equal to 1.5. Mod-

12 ECV 9-12 ALOISI ET AL.: MOUNT ETNA PAROXYSM OF 22 JULY 1998 eling was based on the study of movement of the particles of diameter less than 0.1 mm, i.e., those that potentially travel to distances of tens of kilometers or more from the vent. Horizontal and vertical diffusivity values equal to 5000 and 10 m 2 s 1, respectively, provided the best match between the simulated cloud and the one observed on the satellite image. [26] Our estimates of the paroxysm beginning time and duration show a clear agreement with values inferred from the application of other techniques (video camera and volcanic tremor). On the contrary, the PUFF estimate of the column height (13 km) is significantly greater than reported from visual observations (10 km), the only previously available estimate. Variation of wind velocity and direction as a function of altitude in the volcano region at the eruption time was instrumental to the PUFF column height estimate and furnished conditions for constraining it to within 1 km. [27] The level of accuracy of the cloud simulation shows that the method used in the present study is appropriate for studying Mount Etna paroxysms of the type and energy observed on 22 July In this regard, the results of the present study mark a significant step toward eruptive cloud tracking in the Etna region, and offer a contribution to current projects for the mitigation of volcanic risk related to air traffic in the same area. [28] Acknowledgments. We are very grateful to Kevin Engle, Geophysical Institute of University of Alaska at Fairbanks, for his valuable contribution to the planning of the satellite data analysis useful to this study and for the helpful discussion of results. This research was supported by Istituto Nazionale di Geofisica e Vulcanologia in the framework of a scientific agreement existing between the same institution and the Earth Science Department of Messina University. References Brantley, S. R., The eruption of Redoubt volcano, Alaska, December 14, 1989 August 31, 1990, U.S. Geol. Surv. Circ., 1061, 33 pp., Coltelli, M., P. Del Carlo, and L. Vezzoli, Discovery of a Plinian basaltic eruption of Roman age at Etna volcano, Ital. Geol., 26, , Coltelli, M., P. Del Carlo, and L. Vezzoli, Stratigraphic constraints for explosive activity in the past 100 ka at Etna Volcano, Italy, Int. J. Earth Sci., 89, , Dean, K. G., S. Bowling, G. Shaw, and H. Tanaka, Satellite analysis of movement and characteristics of the Redoubt Volcano plume, J. Volcanol. Geotherm. Res., 62, , 1994a. Dean, K. G., L. Whiting, and H. Jiao, An aircraft encounter with a Redoubt ash cloud, in Proceedings of the First International Symposium on Volcanic Ash and Aviation Safety, edited by T. F. Casadevall, U.S. Geol. Surv. Bull., 2047, , 1994b. Dean, K. G., M. Servilla, A. Roach, B. Foster, and K. Engle, Satellite monitoring of remote sensing volcanoes improves study efforts in Alaska, Eos trans. AGU, 79(47), 413, , Dean, K. G., J. Dehn, S. McNutt, T. Neal, R. Moore, and D. Schneider, Satellite imagery proves essential for monitoring erupting Aleutian volcano, Eos Trans. AGU, 83(22), 243, , Heffter, J. L., and B. J. Stunder, Volcanic ash forecast and dispersion (vaftad) model, Weather Forecast., 8, , Holasek, R. E., and W. I. Rose, Anatomy of 1986 Augustin Volcano eruptions as recorded by multispectral image processing of digital AVHRR weather satellite data, Bull. Volcanol., 53, , Holasek, R. E., and S. Self, GOES weather satellite observations and measurements of the May 18, 1980, Mount St. Helens eruption, J. Geophys. Res., 100(B5), , Istituto Internazionale di Vulcanologia (IIV), The current state of Mt. Etna volcano, July September 1998, Open-File Rep., IIV-CNR, Catania, Italy, Kienle, J., K. G. Dean, H. Garbeil, and W. I. Rose, Satellite surveillance of volcanic ash plumes: Application to aircraft safety, Eos Trans. AGU, 71, 266, La Delfa, S., G. Patanè, and J. C. Tanguy, Kilometer-scale heterogeneities inside volcanoes revealed by using a set of geophysical methods: Variable stress field at Mount Etna, Sicily, Phys. Earth Planet. Inter., 121, , La Volpe, L., P. Manetti, R. Trigila, and L. Villari, Volcanology and chemistry of the Earth s interior, in Italian Research Activity ( ) Report to IAVCEI, Boll. Geofis. Teor. Appl., 40, , Matson, M., El Chichòn eruptions: A satellite perspective, J. Volcanol. Geotherm. Res., 23, 1 10, Prata, A. J., Observations of volcanic ash clouds in the mm window using AVHRR/2 data, Int. J. Remote Sens., 10, , Schneider, D. J., W. I. Rose, and L. Kelly, Tracking of 1992 eruption clouds from Crater Peak Vent of Mt. Spurr Volcano, Alaska, using AVHRR data, in The 1992 Eruptions of Crater Peak Vent, Mount Spurr Volcano, Alaska, edited by T. Keith, U.S. Geol. Surv. Bull., 2139, 27 36, Schneider, D. J., W. I. Rose, R. L. Coke, and G. Bluth, Early evolution of a stratospheric volcanic eruption cloud as observed with TOMS and AVHRR, J. Geophys. Res., 104(D4), , Schneider, D. J., K. G. Dean, J. Dehn, T. P. Miller, and V. Y. Kirianov, Monitoring and analysis of volcanic activity using remote sensing data at the Alaska Volcano Observatory: Case study for Kamchatka, Russia, December 1997, in Remote Sensing of Active Volcanism, Geophys. Monogr. Ser., vol. 116, edited by P. Mouginis, J. Crisp, and J. Fink, pp , AGU, Washington, D.C., Searcy, C., K. Dean, and W. Stringer, PUFF: A high resolution volcanic ash tracking model, J. Volcanol. Geotherm. Res., 80, 1 16, Sparks, R. S. J., M. I. Busik, S. N. Carey, R. S. Gilbert, L. S. Glaze, H. Sigurdsson, and A. W. Woods, Remote sensing of volcanic plumes (Ch 12), in Volcanic Plumes, pp , John Wiley, New York, Tanaka, H. L., Development of a prediction scheme for the volcanic ash fall from Redoubt Volcano, in First International Symposium on Volcanic Ash and Aviation Safety, Seattle, WA, U.S. Geol. Surv. Circ., 1065, 58, M. Aloisi, M. D Agostino, and A. Mostaccio, Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Piazza Roma, Catania, Italy. (aloisi@ct.ingv.it; dagostino@ct.ingv.it; mostaccio@ct.ingv.it) K. G. Dean, Geophysical Institute, University of Alaska Fairbanks, 903 Koyukuk Drive, PO Box , Fairbanks, AK , USA. (ken. dean@gi.alaska.edu) G. Neri, Dipartimento di Scienze della Terra, Università di Messina, Salita Sperone, Messina, Italy. (geoforum@imeuniv.unime.it)