Forecasting exposure to volcanic ash based on ash dispersion modeling

  • Published on

  • View

  • Download


  • ha,


    Received 5 January 2007; accepted 12 October 2007Available online 4 November 2007

    demonstrated the damaging consequences that can occur. Whilevarying degrees of engine, windshield and fuselage damage

    The location of ash clouds can be determined using satellitedata, although these data are not always available when needed,and only areas with relatively high concentrations are detected.The detection limit depends on several things involving the

    Available online at

    Journal of Volcanology and Geothermal Roccurred in each case, fortunately no catastrophic lose of life1. Introduction

    Volcanic clouds pose a significant hazard to aircraft (Millerand Casadevall, 2000). Volcanic ash clouds can be composed ofsub-millimeter sized particles of rock (tephra), water vapor, andother gases such as sulfur dioxide (SO2). Many instances ofaircraft flying into ash clouds over the past three decades have

    resulted. The cost of repairing the damaged aircraft can be large.Windscreens and fuselage may need to be replaced, and enginesdismantled, cleaned and rebuilt. The single August 2000 erup-tion of Miyake-jima was reported to cause over 12 million USdollars of damage to aircraft (Tupper et al., 2004). There isobviously both a public safety and financial incentive for air-craft to avoid volcanic clouds.A technique has been developed that uses Puff, a volcanic ash transport and dispersion (VATD) model, to forecast the relative exposure ofaircraft and ground facilities to ash from a volcanic eruption. VATD models couple numerical weather prediction (NWP) data with physicaldescriptions of the initial eruptive plume, atmospheric dispersion, and settling of ash particles. Three distinct examples of variations on thetechnique are given using ERA-40 archived reanalysis NWP data. The Feb. 2000 NASA DC-8 event involving an eruption of Hekla volcano,Iceland is first used for analyzing a single flight. Results corroborate previous analyses that conclude the aircraft did encounter a diffuse cloud ofvolcanic origin, and indicate exposure within a factor of 10 compared to measurements made on the flight. The sensitivity of the technique todispersion physics is demonstrated. The Feb. 2001 eruption of Mt. Cleveland, Alaska is used as a second example to demonstrate how thistechnique can be utilized to quickly assess the potential exposure of a multitude of aircraft during and soon after an event. Using flight trackingdata from over 40,000 routes over three days, several flights that may have encountered low concentrations of ash were identified, and theexposure calculated. Relative changes in the quantity of exposure when the eruption duration is varied are discussed, and no clear trend is evidentas the exposure increased for some flights and decreased for others. A third application of this technique is demonstrated by forecasting the near-surface airborne concentrations of ash that the cities of Yakima Washington, Boise Idaho, and Kelowna British Columbia might have experiencedfrom an eruption of Mt. St. Helens anytime during the year 2000. Results indicate that proximity to the source does not accurately determine thepotential hazard. Although an eruption did not occur during this time, the results serve as a demonstration of how existing cities or potentiallocations of research facilities or military bases can be assessed for susceptibility to hazardous and unhealthy concentrations of ash and othervolcanic gases. 2007 Elsevier B.V. All rights reserved.

    Keywords: exposure; ash; aircraft; model; healthAbstractForecasting exposure to volcanic as

    Rorik A. Petersona Mechanical Engineering, University of Alaska, Fb Geophysical Institute, University of Alaska, Fai Corresponding author.E-mail addresses: (R.A. Peterson),

    (K.G. Dean).

    0377-0273/$ - see front matter 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.jvolgeores.2007.10.003based on ash dispersion modeling

    , Ken G. Dean b

    anks, PO Box 755905, Fairbanks AK 99775 USAnks, PO Box 757320, Fairbanks AK 99775 USA

    esearch 170 (2008) and environmental conditions, and there is currently nodeterministic value. Despite these limitations, satellite detectionin conjunction with VATD models compose the common

  • ogymethod for identifying ash-contaminated air space (Dean et al.,2002, 2004).

    There is an obvious need for information about the potentialdamage volcanic ash can cause to jet engines. There have beenlimited experimental results published in the open literature.Experiments in the late 1980s measured the performance of aturbofan and a turbojet engine when subjected to variousconcentrations of sand, clay and ash collected from Mt. St.Helens (Batcho et al., 1987; Dunn et al., 1987). Theseexperiments concluded that the four major sources of enginedamage are compressor blade erosion, deposition of glassifiedmaterial, blockage of fuel nozzles, and blockage of coolingducts. Based on these results, more comprehensive experimentswere conducted in the 1990s using a hot section test system(HSTS) that comprised the hot sections from two differentrepresentative combustor engines (Kim et al., 1993; Dunn et al.,1996), reducing the cost of obtaining the complete engineassemblies. Again using a range of particulates that includedvolcanic ash, concentrations from 100 to 500 mg/m3 and expo-sure times from 7 to 14 min resulted in damage that includedabrasion and glassified deposition. Another notable conclusionis that higher turbine inlet temperatures (TIT) result in greaterdegrees of deposition. Accordingly, the International CivilAviation Organization (ICAO) recommends that when an air-craft does inadvertently encounter an ash cloud, the crew shouldimmediately reduce thrust to idle (ICAO, 2001), therebyreducing the turbine inlet temperature and any further deposi-tion of fused material.

    Aircraft engines continue to evolve with advances in hightemperature materials and changes in design. The changes inaircraft engines will probably always out pace the under-standing of how volcanic ash may affect them during operation.The current view of the aviation industry is to avoid any areawhere discernible ash can be detected either visually or bysatellite (Guffanti et al., 2005); an effective zero-tolerancepolicy. However, some recent events have indicated that somevolcanic ash and/or gases may be present in non-trivial con-centrations long after falling below the detection limit ofsatellites. A well-known example of this occurred after theFebruary 2000 eruption of Hekla, when a research aircraftequipped with scientific instruments made several measure-ments of particles and trace gases of volcanic origin (Rose et al.,2006). This encounter is used as the first example of theexposure calculation technique in this paper.

    2. VATD modeling

    VATD models are useful tools for forecasting the movementof volcanic ash clouds. As a complement to satellite data andvisual observations from both the ground and aircraft, VATDmodels provide another source of information about manyaspects of airborne volcanic ash clouds such as cloud height,relative concentration of ash and predicted cloud movement.VATD models have their own limitations which include uncer-

    R.A. Peterson, K.G. Dean / Journal of Volcanoltainty in the volcanic ash source and meteorology (Servranckxand Chen, 2004). Additionally, some complex physicalphenomena such as turbulent diffusion are modeled usingempirical parameters because solution of the turbulent fluiddynamics equations is quite formidable and time consuming.Despite this, VATD models have proven to be important forreanalyzing data after a volcanic eruption in an attempt to betterunderstand the event, and determine methods for improving thecapabilities of the model for future use.

    When VATD models are used to help reanalyze a past event,model data are combined with satellite observations and infor-mation from the air (pilot reports or PIREPS) and ground. It isduring these post-event analyses that techniques and methodsfor improving the forecast capability of VATD models areidentified and eventually implemented. Numerous comparisonsof VATD simulations to satellite observations have shown themto be reasonably accurate in most cases (e.g. Dean et al., 2004).However, VATD models will likely never forecast the exacttrajectory and composition details of an evolving ash cloud, andthe forecasts may range from exceptionally good to quiteinaccurate, which is why satellite observations are often used tovalidate models. When VATD models appear to agree well withthis observational data, some insight into the physical processesthat occurred during the event is achievable. It is possible toassume that these accurate model forecasts indicate that themodeled physical processes correspond to the actual physicalprocesses that occurred, although that is not guaranteed.Proceeding with this assumption, the model can then providefurther information about other processes that occurred duringthe event, which are more difficult to observe otherwise, such asatmospheric reaction or wet removal (particle scavenging).

    There is a wide array of VATD models used throughout theworld. Use of a particular model is often associated with ageographical region or institution, although some of the morepopular models are used world-wide. An operational model,sometimes called a runtime model, must be capable ofproducing trajectory forecasts fast enough to be used for hazardmitigation. In the case of preventing encounters with aircraftthat can travel 500 mph, this correlates to minutes. In order toprovide ample time for ash avoidance, model results still mustbe available in less than an hour. Many runtime VATD modelsuse a Lagrangian framework to describe turbulent dispersion.Tracer particles that represent and behave as individual ashparticles are tracked as they advect, diffuse and fall due togravity. Diffusion is simulated by the random-directional walkof each particle, with a spatial step size determined by a localdispersion coefficient. This formulation is often much fastercomputationally than the conventional Eulerian diffusiontechnique, and the results are comparable when the number oftracer particles is large. Obtaining a relative value of con-centration is straight-forward by counting the number of tracerparticles in a given volume at any time, with each tracer particlerepresenting some small fraction of the total eruption mass.Because each particle can have a range of characteristics such assize, shape, density and chemical identity, spatial gradients ofeach characteristic within the cloud can also be determined.

    The Puff model uses a Lagrangian framework with an

    231and Geothermal Research 170 (2008) 230246adjustable number of tracer particles. The dispersion coeffi-cients can be calculated locally based on velocity deformation(Smagorinsky, 1963; Draxler and Hess, 1998), or a constant

  • and true aircraft trajectories that ascend, descend and changebearing with time can be utilized.

    The methodology for calculating ash-cloud exposure for amoving object (e.g. airplane) is illustrated in Fig. 1. The totalexposure, E, is calculated by integrating the concentration ofash, c, along the trajectory, s, of the object. The trajectory canbe decomposed into the product of the velocity, V, and the traveltime, t.

    E R c ds R cV dt 1

    The SI units of E are g/m2. Both cloud concentration andvelocity may be functions of time. Numerically, the integral is

    gy and Geothermal Research 170 (2008) 230246value can be specified for the entire domain. The latter optioneliminates the need to continually calculate the local dis-persion coefficient, and eliminates the tendency of large grid-size NWP data to yield unrealistically large local diffusioncoefficients.

    All VATD models rely on meteorological data from a NWPmodel for wind velocity, temperature, and pressure information.There are several different NWP models used throughout theworld covering either a global or regional domain. Regionalmodels typically have a higher resolution, which can potentiallyprovide more detailed information near the ground surfacewhere topography has a greater influence. The primary draw-back of using regional NWP data with a VATDmodel is in long-term forecasting when the ash cloud may leave the modeldomain. Another drawback is that data from many regionalmodels is only available over regions of high population such asthe contiguous United States, while many volcanoes are locatedin remote regions.

    A particular challenge can arise when using different NWPmodels covering the same domain because they can providedifferent forecasts. The variability stems from differences intheir underlying modeling techniques and also assimilation ofdifferent initialization data. Under those conditions, a VATDmodel will also produce different forecasts when using differentNWP data sets. Because this variability is beyond the control ofa VATD model, only the ERA-40 global NWP will be usedthroughout this paper. This choice was made to keep the focuson the utility and limitations of the exposure technique only.The importance of forecast variability due to different NWPdata sets should not be downplayed, however, and acomprehensive study of NWP effects on VATD forecasts willbe the subject of a future manuscript.

    The ERA-40 model (Uppala et al., 2005) is a product of theEuropean Centre for Medium-Range Weather Forecasts(ECMWF). The model assimilates a range of meteorologicaldata from satellite, radiosonde, aircraft, and ships among othersources. The project then produces a reanalysis of theatmospheric conditions using a global NWP model. Theresolution of the data used here is 2.52.5 with 23 verticallevels from 1000 to 1 mbar, or about ground level to 50 km.

    3. Exposure calculation

    The calculation of potential exposure to volcanic ash by anaircraft was first proposed by Armienti et al. (1988). They useda Eulerian-based VATDmodel with the Stokes Settling Law andconstant, parameterized diffusion coefficients. They createdhypothetical, linear aircraft trajectories during the May 1980Mt. St. Helens eruption event and calculated the total quantity ofairborne ash that each trajectory would encounter. The windfield was based on a linear interpolation of three points ofmeteorological soundings. Advances since then in computa-tional speed and methodology permit an improvement on thistechnique. High resolution wind fields from NWP models can

    232 R.A. Peterson, K.G. Dean / Journal of Volcanolobe used with VATD models that incorporate more detailedtransport and settling physics. Exposure to different sizes,shapes and identities of particulates can be rapidly quantified,approximated by summation at discrete time steps. The VATDmodel provides concentration as a function of time and space.The object trajectory is a list of three-dimensional locations as afunction of time. The numerical approximation of exposure istherefore

    E XN

    i0ci Vi ti ti1 2

    for the time steps from 0 to N.Total exposure has units of mass per area. The area in the

    denominator could correspond to the cross-sectional area of ajet's engines, or a correlated equivalent area of air intake.Therefore, multiplying the exposure by the area results in thetotal mass of ash that passed through the engines. Limitedexperiments using an HSTS have shown surface abrasion andinternal deposition when the total mass ranged from 100 to500 g (Kim et al., 1993; Dunn et al., 1996). These experimentswere also able to calculate a capture ratio: the ratio of deposits tothe total mass through the HSTS. Capture ratios were typicallyless than 5%, and showed evidence of increasing with furtherexposure. It is this captured mass that causes the most sig-nificant damage by clogging vent holes inside the engine andprecluding adequate ventilation and cooling. It is relativelystraight-forward to include increasing capture ratios in Eq. (2)by including the term Ei 1 on the right-hand side. However,because the exact quantitative manner that the capture ratioincreases with exposure is not entirely clear at this time, it is notincluded here.

    Fig. 1. Diagram illustrating a three-dimensional aircraft trajectory that sweepsout a volume of constant cross-sectional area, A. The velocity V is a function of

    aircraft position and time, and the cloud concentration is also a function ofposition and time. Exposure calculations result in the integrated cloudconcentration along this trajectory.

  • ogyThere is no requirement that the time step size of the VATDmodel match that of the trajectory. There is a maximum time stepfor the trajectory to prevent the object from flying completelythrough an entire grid volume between time steps

    Dtmax DxVmax 3

    wherex is the spatial dimension of the concentration grid fromthe VATDmodel. This grid size is adjustable with the Puff VATDmodel and specified in degrees horizontally and metersvertically. It should be noted that this grid size is not neces-sarily the same as that of the NWP, and is often up to an orderof magnitude smaller to help resolve more complex cloudmorphology.

    4. February 26, 2000 Hekla event

    The details of the February 2000 eruption of Hekla volcanoin Iceland have been presented by Rose et al. (2003) andLacasse et al. (2004). The eruption began on 26 Februaryaround 18:19 UTC. The major explosive activity ceased around2200. Weather radar data indicate that the initial height of theplume was at least 12 km, the maximum detection elevationfrom Keflavk international airport (Lacasse et al., 2004). Apilot report at 2000 UTC indicated the plume top at 1012 km.Satellite data analysis using the brightness temperaturedifference method indicated the plume top at 1013 km.Measurements made on February 29 and March 1 by the PolarOzone Aerosol Measurement (POAM-III) instrument indicatethe cloud top was between 11 and 12 km. POAM data has 1-kmvertical resolution (Rose et al., 2003). The minimum mass oftephra erupted during this time has been estimated as 0.1 Tg,and was likely more.

    A detailed account of the engine damage that occurred to aNASA DC-8 aircraft that flew in the proximity of Hekla vol-cano about 1.5 days after the eruption began is given in thereport by Grindle and Burcham (2003) and discussed in Huntonet al. (2005). The aircraft crew, flying on a scientific researchmission from Edwards AFB, California to Kiruna, Sweden, wasmade aware of the Hekla eruption and associated ash cloud bythe London Volcanic Ash Advisory Center (VAAC) that wasmonitoring the event. AVAAC will commonly compile satellitedata, information from seismic measurements, pilot and groundreports, and VATD-model forecasts in order to help advise air-craft on the current location and projected trajectory of an ashcloud. The London VAAC issued an advisory around 2200UTC on the 26th which described the area believed to enclosethe current ash-cloud position and a trajectory forecast forthe next 12 h. The forecasts used by the London VAAC wereproduced by a different VATD model than discussed here.Updated advisories of the cloud position were issued through-out the next several days as the London VAAC continued tomonitor the situation. The northern extent of the bounding box

    R.A. Peterson, K.G. Dean / Journal of Volcanolbelieved to contain the ash cloud was never extended furthernorth of 65 through the 28th. The original route planned for theDC-8 did not enter this zone, but was modified about 200 mifurther north of the original plan to provide a margin of safety(Grindle and Burcham, 2003).

    Numerous satellite images of the Hekla eruption wereanalyzed to better understand the dynamics of the eruptioncloud (Rose et al., 2003). Forward trajectory model results usingtheHysplit VATDmodel have indicated that particles originatingat a 9000 meter elevation coincide with the position of the ashcloud observed in satellite data on the morning of the 28th (Roseet al., 2006). The purpose here is not a similar attempt toreconstruct an accurate VATD model simulation of the eruptionbased on information gained after the event. Instead, the Puffmodel is used in a defaultmode based only on the sparse initialinformation available, such as an operational center might doduring an event when initial information is limited. The defaultmode is designed to provide conservative forecasts, meaning itattempts to provide the maximum probable spatial extent of thecloud. This mode includes an even initial distribution of tracerparticles throughout the entire column height. This type ofdistribution (often called linear) gives a conservative trajectoryforecast because initial concentrations of ash at all elevations areequal, and the variations in wind speed and direction willdisperse ash in different directions. This type of columninitiation is in contrast to the distribution within moderate tohigh magnitude eruptions where ash is initially more concen-trated in a region just slightly below the top of the ash plume(Sparks et al., 1997) in an umbrella-type distribution. This Heklaeruption had a Volcanic Explosivity Index (VEI) of 3 (GlobalVolcanism Program, Smithsonian National Museum of NaturalHistory:

    The route of the DC-8 was obtained from the SOLVE in-ternet data site ( which hasthe longitude, latitude, and elevation at one-second intervals forthis mission. Figs. 2 and 3 show this route with the dotted purpleline along with the predicted ash-cloud position from Puff at0515 UTC on the 28th. The black X marks the exact aircraftlocation at 0515 UTC according to the SOLVE data. Fig. 2 is atracer particle plot where the color corresponds to the particleelevation. Only the higher elevation particles above 10 km areshown. The predicted location of the cloud closely resemblesthe AVHRR and MODIS data discussed in Rose et al. (2003),even without any custom tuning of model parameters (i.e.default mode). The same forecast data is shown in Fig. 3 as arelative concentration plot of the particles between 10 and15 km. The scale shown is not quantitative because the con-centration of tracer particles simply scales with the total numberof particles in the simulation. This type of qualitative scale iscommon with VATD models because the total eruption mass isoften not immediately known.

    It is straight-forward, however, to convert to absoluteconcentration if an estimate of the total eruption mass is used.Here a value of 0.1 Tg is used (108 kg), which Rose et al. (2003)estimate was the total ash mass in the early cloud. Using thisestimate and an initial plume height of 13.7 km, the totalexposure is 57.6 g/m2 with a maximum instantaneous con-


    233and Geothermal Research 170 (2008) 230246centration of 0.413 mg/m . Using the 0.1 Tg estimate, the scaleof the color bar is then 0.0 (green) to 0.64 mg/m3 (red) in fourequal increments of 0.16 mg/m3 each. Direct measurements of

  • gy234 R.A. Peterson, K.G. Dean / Journal of Volcanoloparticle concentration made during the flight peaked between0.3 and 0.5 mg/m3, however these particles were predominantlyice nucleated from a much smaller ash particle. The forecastexposure and maximum concentration are on the order of 10times greater than were actually encountered, which is reason-ably accurate considering the variety of default assumptionsabout the initial cloud morphology that have been used. At thispoint it would be possible to adjust the default parameters in theVATD model in an attempt to more closely match the experi-mental data by adjusting the particle size and distribution,diffusion coefficients, or the initial vertical distribution in theplume. However, in an operational situation this is not possiblebecause this information is not available, and the more pressingquestion is whether any exposure to volcanic ash might occur.In this case, the technique does successfully predict that expo-sure to ash is probable, and the factor of 10 error is of lesserconsequence.

    The minimum amount of ash required to cause damage to ajet engine has been the subject of intense discussion and debatewithin the volcanological and airline safety communities forquite some time. Currently, the threshold concentration belowwhich damage will not occur is not known, and may never becompletely quantified for all types of situations (Guffanti et al.,2005). Furthermore, the amount of published data based on

    Fig. 2. Puff forecast of the ash-cloud location on 28 February 2000 at 0515 UTC. Tracand the 'X' indicates the exact aircraft position at 0515. (For interpretation of the referarticle.)and Geothermal Research 170 (2008) 230246simulations, experiments, and actual encounters is in shortsupply. Therefore, the current advice is to adopt a zero-tolerancepolicy when it comes to the possibility of an aircraft encounterwith an ash cloud. When all four engines from the DC-8 werelater overhauled, fine ash was found inside passages, turbineblades were eroded, and cooling air holes were blocked (Grindleand Burcham, 2003). Subsequent analysis of recovered materialfrom air filters indicated a high probability of originating fromthe eruption of Hekla (Grindle and Burcham, 2003), althoughthe results are not definitive.

    Any ash that ended up in the DC-8 flight route likelyoriginated from the upper elevations of the early ash plume.Calculating a backwards trajectory in time is impossible whendiffusion is taken into account. Backwards trajectory simulationsthat ignore diffusion are useful for estimating the source locationof an existing cloud, such as was done using Hysplit in Rose et al.(2006). Similar information that includes diffusion can beobtained from performing a set of forward simulations usingvery small initial packets of ash. A set of 70 simulations usingvery thin (100-meter deep) initial plumes originating from 10 to17 km were performed using otherwise default parameters fromTable 1. The potential ash exposure as a function of this initialelevation is shown in Fig. 4. Based on these data, no exposure toash would have occurred if the initial cloud height was below

    er particles are color-coded by elevation. The DC-8 trajectory is shown in violet,ences to color in this figure legend, the reader is referred to the web version of this

  • 235ogy and Geothermal Research 170 (2008) 230246R.A. Peterson, K.G. Dean / Journal of Volcanolabout 10,000 m. The potential exposure quickly reaches a maxi-mum for ash originating from an initial plume that reached13 km, and then a slightly more gradual decrease for even higherelevations, becoming negligible above 17 km. A previoussensitivity analysis of Puff parameters has indicated that thehorizontal diffusion coefficient often has the most influence ofthe model forecasts. Therefore, the exposure calculations wereperformed using four different values of the diffusion coefficient,ranging from 0 to 5000m2/s. The results in Fig. 4 indicate that thediffusion coefficient has only a small effect on the total exposure,however. The fact that complete removal of the diffusionmechanism (K=0) results in nearly the same results indicates thatthe majority of ash that the aircraft encountered was directlyadvected into the flight route, and was not the result of turbulentdiffusion.

    The maximum instantaneous concentration along this flightroute is shown in Fig. 5. The overall trend with height is similarexcept for a prominent shoulder between 14 and 15 km. Onceagain, the removal of the diffusion mechanism by setting K=0results in only small changes in the maximum concentrationencountered. Although the current policy for ash exposure iszero tolerance, it seems entirely possible that future refinementsto the policy may include both maximum allowable concentra-tion and net exposure. There is a financial incentive to developan improved policy because flight re-routing and cancellation

    Fig. 3. Puff forecast of the ash-cloud location on 28 February 2000 at 0515 UTC showis 00.64 mg/m3 based on assumptions discussed in the text. (For interpretation of theof this article.)are expensive, and this type of exposure forecasting will then bevery useful.

    There is another important aspect of the data shown in Figs. 4and 5 in addition to determining from where in the initial plumethe ash originated. The maximum height of the initial ash plumeis often difficult to know accurately, and there is even moreuncertainty during the initial stages of an event when detailedanalysis of satellite and meteorological data has not yet beenperformed. Initial estimates are often based on ground or airobservers, weather radar, and comparison between the thermalsignal of the cloud top with the atmospheric temperature pro-file. Furthermore, these initial estimates may be contradictory,

    Table 1Parameters used in the Puff simulations for the eruptions of Hekla, Clevelandand Mt. St. Helens Volcanoes

    Hekla Cleveland St. Helens

    Start time (UTC) 2000-02-26 18:19 2001-02-19 14:30 All 2000 4 dailyEruption duration (h) 4 1.53 3Plume top (km) 13.7 10.0 4.0Plume bottom (m) 1500 4000 2550Diffusion coeff. (m2/s) 05000 500 10,000Particle distribution Linear Poisson Linear

    Values with a range indicate that a sensitivity analysis of this parameter rangewas conducted.

    n as relative concentration. The color bar shows an absolute concentration rangereferences to color in this figure legend, the reader is referred to the web version

  • Fig. 4. Integrated potential exposure for the DC-8 flight route from the Hekla eruption as a function of the initial plume height for a 100-meter thick cloud. Symbolsindicate several different values of the dispersion coefficient from 05000 m2/s.

    236 R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246casting further uncertainty as to the plume top elevation. Datasuch as that shown in Fig. 4 can be used to assess the relativehazard to a potential flight route based on the uncertainty in theinitial plume height. In this case, relative certainty that the initialplume is below 10 km correlates to a very low hazard, whereasthe possibility that the plume top could be 23 km higher leads toa maximum of the potential hazard. This type of analysis hasparticular relevance to this event at Hekla volcano. The C-bandweather radar near Keflavk International airport in Iceland wasable to report real-time observations of the initial ash plume witha frequency of only a few minutes (Lacasse et al., 2004). How-ever, due to software limitations and radar settings, the upper

    detection limit was 12 km, and it is not possible to determine

    Fig. 5. Maximum concentration encountered by the DC-8 flight route as a function ovalues of the dispersion coefficient from 05000 m2/s.whether the initial plumewas higher based on the radar data. Ashadvisories placed the initial plume height at 30,000 ft (9143m) at1845 on the 26th, and that was quickly elevated to 45,000 ft(13,715 m) 15 min later. Speculation of the certainty of theseinitial estimates is beyond the purposes here, but illustrates theimportance of parameters such as plume height to this hazardassessment technique.

    5. February 19, 2001 Cleveland event

    Calculation of potential exposure to a single aircraft routeas described above can be performed rapidly and is easily

    automated. Another potential application is to unify the VATD-

    f the initial plume height for a 100-meter thick cloud. Symbols indicate several

  • model exposure calculations to the real-time tracking of aircraftroutes. An automated system that continually updates the ex-pected exposure due to a potential eruption could be used as areal-time hazard assessment tool for the airline industry. Thereare several commercially available software packages that pro-vide a real-time interface to the current location of most com-mercial aircraft world-wide. To illustrate how this can be doneefficiently, over 40,000 individual aircraft paths over NorthAmerica from 1922 February 2001 were obtained using theFlight Explorer software package. These data include thelatitude, longitude, elevation, and velocity of the aircraft at one-minute intervals. The precision in aircraft location is 1/60(i.e. 1 min) horizontally and 100 ft vertically. The entire listingof all 40,000 flights were input into the Puff model without any

    filtering based on location, time or identity. When intermediatedata points were missing, a linear interpolation in space andtime was performed. Because the purpose here is more todemonstrate the technique than glean the most possible in-formation from the available data, no further effort was made tofill in the gaps more accurately, such as with a spline in-terpolation or consulting another database of flight trajectoryinformation.

    The potential exposure of these 40,000 routes to the ashcloud from the 19 February 2001 eruption of Mt. Cleveland wasforecast using the exposure calculation technique. This rela-tively brief eruption produced a volcanic cloud that formed anarc over 1000 km long, and drifted to the NE across Alaska(Dean et al., 2002; Dean et al., 2004). The cloud was detected

    237R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246Fig. 6. Puff forecast of the ash-cloud position from the 19 February eruption of Clevframe is valid at 12-hour increments after the eruption commencement at 1430 UTCeland volcano showing tracer particles that are color coded by elevation. Eachon the 19th.

  • and its movement tracked using data from multiple satellitesensors, including GOES, AVHRR and MODIS for approxi-mately 50 h. The translucent cloud was detected in the GOESdata at half hour intervals after the start of the eruption,approximately 1430 UTC. The plume appeared to disconnectfrom the volcano between 2315 and 0242 UTC the next day,while a decreased seismic signal from 230 km away indicatesactivity may have ceased as early as 2230. The altitude of thecloud increased over time from 7.5 km a few hours after the startof the eruption to 12 km 8 h later (Dean et al., 2002). Althoughthe eruptive activity spanned nearly 8 h, the initial and mostpowerful portion of that was probably on the order of 1 h.Because the eruption duration can never be known in the shortterm when critical hazard assessments are made, the effectof eruption duration on the exposure technique will also beinvestigated. The sensitivity of the technique to the assumederuption duration will be shown.

    The Anchorage Volcanic Ash Advisory Center (VAAC),

    distribution agrees with the observed data as well as theVEI 3 classification (Global Volcanism Program, SmithsonianNational Museum of Natural History: Higher concentrations of lower-level ash would betracked using a linear distribution, but there is no direct evi-dence of low-level ash in the satellite data. In any case, thislower-level ash would not have entered into the higher elevationaircraft routes being investigated here, and neglecting it willhave no consequence on the results.

    For the most part, aircraft appear to have avoided the generalarea of the ash cloud during the first 16 h (1430 to 0630 thefollowing day), although there is a four-hour gap in the air-traffic data from 01000400 on Feb. 20. For the subsequent 30hour period, some aircraft flew through a vertical plane thatincluded the ash cloud. To determine the three-dimensionalextent of the ash cloud, the Puff model was used to simulate thisevent using the ERA-40 archive of NWP data. Fig. 6 shows theforecast ash-cloud position at 12-hour increments for the 48 h

    238 R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246working with the Alaska Volcano Observatory (AVO), issuedadvisories warning aircraft of the presence of the cloud and itsposition. The first advisory was issued 19 February at 1847UTC warning of a possible eruption at Cleveland volcano, ashort time later an 18-hour forecast was issued based on VATD-model simulations (not the same VATD model used here). Aportion of the cloud was reported at flight levels FL200FL400, or 2040 thousand feet. Several other pilots reportedobserving the cloud and gave height estimates that are sum-marized in Table 1 of Simpson et al. (2002).

    Table 1 shows the input parameters used for the Puff model.In contrast to the Hekla example earlier, the initial physicaldistribution of particles in the vertical direction was specified asa Poisson distribution with a variance of 3 km. This type ofinitial plume shape provides the umbrella-type concentrationdistribution with higher concentrations near the top of the initialplume. Satellite data at 1655 UTC indicate a shearing of theearly cloud around 6 km (Dean et al., 2004), and thisFig. 7. Integrated potential exposure for 12 select flights lettered Afollowing commencement of the eruption. The cloud forms agenerally northsouth arc that eventually spans nearly the entirestate of Alaska at 48 h while becoming more diffuse. Thisgeneral shape agrees well with satellite data. The front of theadvecting cloud reached the local airline hub city of Anchorageat about this time as well. Based on the large extent of thediffuse cloud at this point, a zero-tolerance policy would appearto make any traffic into or out of Anchorage impossible at thistime. However, air traffic continued and no incidents werereported regarding interaction with volcanic ash.

    Here the exposure calculation technique will be used todemonstrate that indeed most flights appear not have encoun-tered any ash, while a small number appear to have intersectedsome region of the cloud. If the total erupted tephra volume isestimated as 108 kg, the absolute concentration throughout theentire body of the ash can be calculated from the VATD-modeldata. Fig. 7 shows the potential exposure for 12 different flightsthat recorded the largest exposure for a 3-hour eruptionthrough M using four different values for the eruption duration.

  • Table 2Integrated exposure, E (g/m2), during the eruption of Mt. Cleveland for 12 select flights with the highest exposure values


    239R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246duration. The identity of each flight is indicated by a uniqueletter. The exposure of those same 12 flights when the eruptionduration is decreased to 2.5, 2.0 and 1.5 h is shown as well. Thesame data is given in tabular form in Table 2, where the flighttimes and general direction are also listed. Flight times in thetable are for 20 February except for the shaded rows, whichoccurred on the 21st. Fig. 8 shows the flight times and durationgraphically, and indicates that the most exposed flights occurredin two distinct temporal sets. Several flights had more than onesegment, and only their in-air times are shown with the darkbands.

    Fig. 9 shows the flight routes of the seven most exposedflights that occurred on 20 February along with the forecastcloud position at 1930 UTC. This forecast gives a generalindication of the cloud location when these flights occurred,although the exposure calculation is performed using themovingcloud at one-minute increments. All flights were traveling in thegeneral eastwest or reverse direction, and all but D were using

    Flight times are UTC on 20 February except for shaded rows that are on 21 Februof the trajectory. Bearing is the primary direction of the flight. Several flights hAnchorage airport. All except F were flying in between flightlevels 200 and 300, or approximately between 6 and 9 km. Ashin this elevation range is shown in the cyan to light green color.However, in this tracer particle plot, ash from higher elevation isplotted on top of ash below it, so not all lower-level ash is clearlyevident. Flights E, F, and G all traveled a similar route at different

    Fig. 8. Illustration of the in-air times for the 12 select flightstimes, and their trajectories overlap for the most part in thisfigure. Several flights traversed through the cloud twice duringthis time period as they returned or departed from Anchorage.

    The next five most exposed flights occurred on 21 Februaryand are shown in Fig. 10 along with the forecast cloud position at1230 UTC that day. Because the cloud is nearly coincident withAnchorage around this time, several northsouth routes havehigh exposure values as they flew nearly parallel to the cloudfront. The first leg of flight K is south to Anchorage, originatingnear the center of the state and overlapping with the northsouthrouteM. The second leg of route K is eastward out of Anchorage.There is an anomalous bend in this first segment of route K,although there is no indication that this deviation occurred inorder to avoid the cloud. Of the 12 most highly exposed flights,all on the 21st recorded lower values than those on the 20th,which agrees with the understanding that the cloud wasbecoming more diffuse. The cause of the temporal gap betweenthese two groups is not clear since several flights occurred

    Flight level is the approximate elevation (in hundreds of feet) during the majoritymultiple segments.during this time and resulted in small exposure values. Of themore than 40,000 flight routes analyzed over three days, about140 routes resulted in exposure values over 1 g/m2, and about 70resulted in exposure values over 10 g/m2.

    The sensitivity of these results to the eruption duration can beseen in Fig. 7, where the exposure is shown for four different

    shown in Fig. 7. Several flights had multiple segments.

  • gy240 R.A. Peterson, K.G. Dean / Journal of Volcanoloeruption durations from 1.5 to 3 h. Obviously the duration of aneruption is never known before hand, so the sensitivity to durationis one of the more important aspects to consider when evaluatingthe utility of these results. In general, there is only a moderatedeviation in the exposure values when the eruption duration isvaried by this factor of two. The largest percent deviationoccurred for routes H andMwhich show about a 50% increase inexposure when the duration is decreased from 3 to 1.5 h.According to these results, however, there is no trend in how theexposure values will vary with changes in eruption duration.Several routes demonstrate very little sensitivity to eruptionduration, such as A, C, and F. Other routes have increasedexposure with increased eruption duration, such asD and L,whileH, J andM decrease. The total eruption mass was held constant inall four scenarios, so the shorter duration corresponded with aninitially more concentrated cloud near the source. In the model,eruption rate is defined by the quotient of the eruption mass andduration. Apparently a higher initial concentration does notdirectly correlate with higher exposure. Those flights that havehigher exposure for longer eruption durations (i.e., more diffuseinitial cloud) are probably intersecting a portion of the cloud thatcorresponds with the end of the eruption.

    Fig. 9. Puff forecast of te ash cloud at 1930 on February 20 shown as tracer particles ctravel times are listed in Table 2. Some routes overlap and others have multiple segmenreferred to the web version of this article.)and Geothermal Research 170 (2008) 230246It would be beneficial to perform a similar sensitivityanalysis of the other eruption parameters that are also difficultto accurately assess during the early stages of an eruption.Some of these parameters include maximum plume height,grain size distribution, vertical particle distribution, total erup-tion mass and eruption rate. Initial investigations along theselines indicate that the results are also sensitive to the choice ofNWP model, such as the ERA-40 model used here. A com-plete analysis of all parameter combinations is a large taskthat is currently underway and will be the subject of a futuremanuscript.

    In conclusion, the meaning of these results deserves care-ful scrutiny because: (1) model simulations are not validated;(2) absolute ash concentration is based on several assumptions;and (3) the concentration of volcanic ash that severely damagesjet engines is still unknown. However, this type of investigationhas several merits. First, low-level exposure to volcanic ash maybe difficult to detect during routine aircraft inspection. If aircraftwith potential ash encounters are identified, a more thoroughinspection can be performed, possibly mitigating a future dan-gerous situation. Second, potential exposure can be calculatedahead of time for an existing air route during the early stages of

    olor coded by elevation. The routes of flights AG are shown in red and the exactts. (For interpretation of the references to color in this figure legend, the reader is

  • ogyR.A. Peterson, K.G. Dean / Journal of Volcanolan on-going volcanic event as a tool to determine the degree ofavoidance and caution necessary. This may be particularlybeneficial later when ash concentrations have dropped to levelsthat are difficult or impossible to detect with satellite techniques.Finally, the routine use of this technique is relativelyinexpensive, easy and fast. This example demonstrates that anautomated system that continually records the potential expo-sure of a large number of flights is quite feasible.

    6. Potential hazard of St. Helens ash

    Another application of calculating potential ash exposure isfor assessing the potential hazard of airborne ash to a stationarylocation such as a city, research installation, or military base. Asimilar technique was first performed by Papp et al. (2005) inthe generation of airborne ash probability distribution (AAPD)maps for aircraft-level altitudes in the North Pacific region.While taking into account the eruptive history of the volcanoesstudied, they concluded that the regions of highest probabilityhave seasonal variations due to the prevailing wind conditionsat higher altitudes. Their analysis did not compute the exposure

    Fig. 10. Puff forecast of the ash cloud at 1230 on February 21 shown as tracer particlexact travel times are listed in Table 2. Some routes overlap and others have multiplereader is referred to the web version of this article.)241and Geothermal Research 170 (2008) 230246of moving targets or provide a method for obtaining aquantitative estimate for the amount of ash in these regions.

    A modified and improved technique is employed here for theassessment of near-surface ash concentrations that would affectpeople and equipment or machinery on the ground. The purposeof this example is more to demonstrate the technique than tohighlight the product because there are a vast number of differentanalyses that could be performed depending on the particularlocation and nearby hazard. There are many cities throughoutthe world that are situated near active or potentially activevolcanoes. Some volcanoes pose an on-going hazard due toperiodic or sustained emissions of sulfur dioxide (e.g. Kilaeua inthe Hawaiian Islands) and low concentrations of ash (e.g. Etnain Sicily, Italy). Exposure to SO2 is believed to cause manyhealth problems including eye irritation, bronchitis, sore throatand headache (Delmelle et al., 2002). Low concentrations of ashcan be particularly hazardous when the particles are 10 micronsor less. When inhaled, these small particles adversely affect thelungs, causing respiratory problems such as asthma, bronchitis,and silicosis (Horwell and Baxter, 2006). The prevailing windconditions and current volcanic state of activity are the primary

    es color coded by elevation. The routes of flights HM are shown in red and thesegments. (For interpretation of the references to color in this figure legend, the

  • factors in how hazardous the situation is at any moment. Both ofthese conditions can vary widely, often, and with low pre-dictability. While the location of a population center may bedue to a wide range of historic reasons (e.g. cultural, economic,strategic, etc.), the location of new research and militaryinstallations is determined by a smaller set of strategic plansbased on purpose, cost, and logistics. Often, the optimumlocation must be evaluated in light of the potential natural andanthropic hazards, one of which may be a nearby volcano.

    The exposure calculation discussed earlier is modified bysetting the velocity of the target (e.g. city) to zero and the altitudenear the ground surface. Airborne ash at this lowest elevation isused to calculate exposure. Although the Puff VATDmodel doesinclude sedimentation physics and calculates ash deposition, thisash fallout is not used to calculate exposure. It is straight-forwardto calculate the potential exposure of this stationary locationfrom a single eruptive event, as the previous examples demon-strate. However, to assess the long-term hazard to a locationfrom a potential future eruption, an ensemble of hypotheticaleruptions is performed over a long time period such as months,years or even decades. Current archives of NWP data areavailable to do this for time periods of 50 yr or more. All aspectsof the potential eruption can also be varied and included in theensemble, such as eruption duration, average particle size orshape, and plume height. The periodicity of simulating

    with how often many NWP models are updated. It should benoted, however, that the frequency of NWPmodel runs is mainlydictated by computational resources and not necessarily thevariability in the weather conditions.

    As an illustrative example of this technique, the potentialexposure of three cities from an eruption of Mt. St. Helensvolcano in the state of Washington, USA, is assessed. Two ofthe cities, Boise Idaho, and Kelowna British Columbia, are asimilar distance from the volcano at 550 and 470 km, re-spectively, while the third, YakimaWashington, is closer at only150 km. As a first approximation based only on proximity to thevolcano, both Kelowna and Boise would appear equally lesshazardous compared to Yakima. However, Papp et al. (2005)have demonstrated that for high altitudes relevant to aircraft inflight, prevailing wind directions tend to skew the potentialhazardous regions from simple circles around the source. Here,it is demonstrated that a similar effect occurs at near-surfaceelevations, making some cities more vulnerable although theyare equidistant from the source. Furthermore, some quantifica-tion of the potential hazard is provided in terms of maximumconcentration and magnitude of exposure, which can also beuseful for assessing the potential hazard. The exact numericalvalues are dependent on knowledge of some eruption sourceparameters as discussed previously.

    The reference state of plume initiation parameters for a 3-hour

    242 R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246hypothetical eruptions should be determined by the variabilityin weather conditions, which is strongly dependent on where thesite being considered is located. A manageable and applicableperiodicity of model runs may be every 3 or 6 h, which correlatesFig. 11. Net potential exposure for three cities in the moderate proximity of St. HelenTable 1. Eruption simulations were performed four times daily for the entire year.eruption is shown in Table 1. Although there were no events ofthis magnitude during this year, it will serve as a representativedata set that can be used with this technique. Simulations wereperformed once every 6 h for the entire 2000 year using ERA-40s volcano for the year 2000 from a hypothetical eruption using the parameters in

  • data. Each simulation was run for 72 h, which providedsufficient time for the ash cloud to reach and eventually pass overall three cities when they were in the path of the moving ashcloud. Because the exposure calculation shown in Eq. (1) is anintegrated value over time, it is important that the simulated ashcloud be given sufficient time to completely advect over thetarget. In cases where very lowwinds prevailed, diffusion causedthe ash to sufficiently disperse that eventual exposure isminimal, and longer simulations did not result in substantialchanges in the calculated exposure.

    In this application, the exposure of a stationary location iscalculated slightly differently than a moving target such as anaircraft. The important parameters here are again the ash con-centration and the duration of exposure. The net potentialexposure is therefore

    E R c dt 4

    Comparison with Eq. (1) shows that the velocity of the targetis not used since it is zero. The SI units of potential exposure arenow the product of concentration and time [gs/m3]. The peakconcentration at any instant is also an important value for hazardassessment, and so the maximum concentration that occursfrom each eruption is determined in units of g/m3.

    The potential exposure of Yakima, Kelowna and Boise for allof year 2000 are shown in Fig. 11. Each plot shows the total

    such that the closest location (Yakima) is on top and the furtheston bottom, although the proximity of Kelowna and Boise arequite similar at 470 and 550 km, respectively. Clearly, Yakima islocated in a relatively more hazardous location based on boththe larger number of events that lead to exposure, and the higherexposure values shown in Fig. 11. The highest value for Yakimais 265 g s/m3 on June 16, and there are 39 occurrences withgreater than 100 g s/m3. The maximum exposure at Kelowna is186 gs/m3, with 13 occurrences above 100 gs/m3. AlthoughBoise is situated a similar distance from the source as Kelowna,the maximum exposure for 2000 is forecast to be only 83 gs/m3. Note that the vertical scale for Boise is one half of that forKelowna. It appears that Kelowna is in a relatively more haz-ardous location than Boise, although they are a similar distancefrom the source. The peaks in the Boise plot are all narrow,indicating the hazards are all short-lived relative to Kelowna.Many of the peaks, particularly those in June and July, havesignificantly longer durations on the order of one week.

    Fig. 12 shows the maximum instantaneous concentration thatoccurred during the time following each hypothetical eruptiveevent. Again note the variation in vertical scale in each plot.Similar to the net exposure values, Yakima appears more haz-ardous because of the larger number of events where theconcentration exceeded 10 mg/m3. There also appears to besome correlation between net exposure and maximum concen-tration for all three locations. It is interesting to note that peak

    243R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246exposure of that city that would occur if an eruption began at aparticular date and time. The vertical ordering of each figure isFig. 12. Maximum instantaneous concentration for three cities in the moderate proximparameters in Table 1. Eruption simulations were performed four times daily for theconcentrations do not always correspond with large netexposures, however. For example, the largest net exposure fority of St. Helens volcano for the year 2000 from a hypothetical eruption using theentire year.

  • Kelowna occurs in October, but the maximum concentration atthe corresponding event is only one of several moderate-to-highvalues that occurred throughout the year.

    The peak concentration at any time is important becauseeven healthy individuals may experience adverse effects whencontaminant concentrations are sufficiently high. Sulfur di-oxide levels over 1000 ppbv can provoke asthma problems inhealthy people, for example (Delmelle et al., 2002). In Fig. 12for Yakima, the peak concentration exceeded 7 mg/m3 severaltimes; a concentration 500 times higher than the National

    Ambient Air Quality Standards for the United States (EPA,2006). In fact, the peak concentration in Yakima was higherthan the Standards values of 0.035 mg/m3 40% of the time. Thisis in marked contrast with Boise where the same standard isexceeded about 9% of the time. Kelowna falls in between thesetwo, exceeding the standard 21% of the time.

    Another representation of this data is shown in Fig. 13, whichis a composite plot of the relative near-surface concentration forall 1464 hypothetical occurrences in 2000. The values shown arethe product of concentration and number of occurrences, and can

    244 R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246Fig. 13. Integrated hazard map of forecast low-level ash for all of 2000. The relativeoccurrences at each point.hazard is determined by multiplying the forecast concentration by the number of

  • ogybe interpreted as a measure of the potential hazard. The highervalues shown in red occur predominantly to the southeast of thevolcano in the same direction as Boise. There is a secondaryregion of high potential hazard to the north in the direction ofKelowna. Yakima is right on the edge of the primary region.Figs. 1113 can be used in combination to assess the potentialhazard from a volcanic eruption.

    In general, the data for Boise in Fig. 11 indicate relativelyfew short periods of exposure to ash. Each period is usually 13 days long and separated by several weeks. For people withacute respiratory conditions, Yakima is a potentially more haz-ardous location in the event of a volcanic eruption from MSH.The fact that Kelowna is noticeably more hazardous than Boise,although the two cities are approximately the same distancefrom the source, highlights the major utility of using this tech-nique for hazard forecasting.

    An additional concern is that airborne particulates can bedamaging to machinery as well as people and animals. Land-based turbines such as used in power generation facilities aresusceptible to damage from dust and ash (Jensen et al., 2005).Obviously, the potential for damage due to airborne particulatesfrom a volcanic eruption is much higher in Yakima, and apreferred machinery location would be Boise. It should be notedthat the eruptive history of Mt. St. Helens is that of relativelyfew, large magnitude eruptions interspaced with long periods ofquiescence (Simkin and Siebert, 1994). However, othervolcanoes exhibit considerably different behavior characterizedby frequent, low magnitude effusive events. Near those vol-canoes, accelerated wear due to particulates would be a greaterconcern. In any case, this example demonstrates how quantify-ing potential exposure is a fairly direct and inexpensive tech-nique that can be part of a broader overall hazard assessment.

    The results shown in Figs. 1113 were obtained used oneyear of ERA-40 data and required about 60 h of computingresources on a 1.8-GHz desktop computer. Each individualforecast is an average of 50 Lagrangian trajectory simulations.A longer time series of data could be used to establish a longer-term average, or even begin to look for changing trends duepossibly to climate change. This same technique can also beused to assess the immediate danger of an imminent eruption byusing forecast NWP data. This is particularly relevant whenvolcano monitoring indicates a heightened probability that aneruption will occur within the next few hours or days. ForecastNWP data currently has 714 days of predicted meteorologicalconditions, and these predictions are updated about every 6 h.Therefore, forecasts about potential exposure to ash can bemade for several days into the future. This would help providetime for preparations such as installing air intake filters, shuttingdown machinery, evacuating aircraft and relocating particularlysusceptible people.

    7. Conclusions

    A technique has been developed that uses a volcanic ash

    R.A. Peterson, K.G. Dean / Journal of Volcanoltransport and dispersion model to quantify the potential expo-sure of aircraft and population centers to volcanic ash andhazardous eruption gases such as SO2. The velocity andlocation of aircraft relative to the developing ash cloud isaccounted for, and the total exposure is quantified as mass perarea of the moving target, e.g. the engine cross-sectional area.Exposure of population centers is quantified as the product ofconcentration and time, and can be directly related to air qualitystandards. Three separate examples were used to demonstratedifferent applications of this technique.

    Analysis of the potential DC-8 aircraft encounter with an ashcloud showed that archived NWP data can be used to assesspotential ash exposure of a single event and that low con-centrations of ash at the distal ends of clouds can be a problem.Using the information publicly available about the DC-8encounter during the 2000 eruption of Hekla volcano, theimportance of accurate initial plume height information was alsodemonstrated. The potential exposure of this flight was predictedto vary considerably for initial plume heights in the relativelynarrow range of 10 to 13 km. Several different sources ofinformation about the initial plume height indicate there wasprobably some uncertainty about the initial plume height, andthe analysis here indicates how this technique can be used toassess with some quantification the potential hazard of flightpaths near existing ash clouds such as that of the DC-8.

    The Mt. Cleveland event shows that ash exposure calcula-tions can be evaluated in an operational mode prior to aneruption for many aircraft routes rapidly and efficiently usingforecast data. The computational resources required for quan-tifying potential exposure is relatively small, and the analysis ofover 40,000 flights over 3 days during the February 2001 Mt.Cleveland eruption event was easily performed on a desktopcomputer. The results of this analysis indicate that severalflights may have encountered low concentrations of ash in thehours after the initial eruption began. The exact quantity ofexposure is dependent on the eruption duration, but there is noclear trend such as increasing exposure with longer durations.The relative change in exposure can be used to help assess thehazard when details about the eruption are not yet fully known.No mechanical problems were made public during or after thisevent. It is not unreasonable to believe that some accelerateddamage may have occurred, and this type of analysis can beused to identify those aircraft that may require premature oradditional maintenance after an eruptive event.

    The same exposure technique can be used for stationarylocations such as cities, research installations, and militarybases. The relative susceptibility of different locations can bereadily quantified using an ensemble of simulations over a longtime period. Also, the more immediate hazard for a particularsite can be quickly quantified using forecast NWP data, which iscurrently available 714 days in the future. Particle size isparticularly important for health hazard assessment, and therelative exposure to different sizes is easily determined usingthe Puff VATD model discussed here by changing theparameters describing the initial eruption.

    The sensitivity of results obtained using this technique to themany VATD simulation parameters must still be determined

    245and Geothermal Research 170 (2008) 230246before the overall utility can conclusively be determined. Someof the more obvious simulation parameters that likely affect theresults are eruption duration, maximum plume height,

  • dispersion coefficients, and particle size and distributions. Ourinitial investigations have also indicated that results can besensitive to the particular NWP model used. There is a limitednumber of different NWP models, and many areas of the globeare only covered by an even smaller number. A completeanalysis of all these parameters is a large task that is currently

    Kim, J., Dunn, M.G., Baran, A.J., Wade, D.P., Tremba, E.L., 1993. Depositionof volcanic materials in the hot sections of two gas turbine engines.J. Engineering for Gas Turbines and PowerTrans. ASME. 115 (3), 641651.

    Lacasse, C., Karlsdttir, S., Larsen, G., Soosalu, H., Rose, W.I., Ernst, G.C.J.,2004. Weather radar observations of the Hekla 2000 eruption cloud, Iceland.Bull. Volcanol. 66, 457473. doi:10.1007/s00445-003-0329-3.

    Miller, T.P., Casadevall, T.J., 2000. In: Sigurdsson, H. (Ed.), Volcanic AshHazards to Aviation. Encyclopedia of Volcanoes. Academic Press, San Diego,CA, pp. 915930.

    Papp, K.R., Dean, K.G., Dehn, J., 2005. Predicting regions susceptible tohigh concentrations of airborne volcanic ash in the North Pacific region.J. Volcanol. Geotherm. Res. 148, 295314.

    246 R.A. Peterson, K.G. Dean / Journal of Volcanology and Geothermal Research 170 (2008) 230246References

    Armienti, P., Macedonio, G., Pareschi, M.T., 1988. Air traffic risk evaluation involcanic ash clouds. Int. J. Model. Simul. 8 (1), 2932.

    Batcho, P.F., Moller, J.C., Padova, C., Dunn, M.G., 1987. Interpretation of gasturbine response due to dust ingestion. Trans. ASME 109 (3), 344352.

    Dean, K.G., Dehn, J., McNutt, S., Neal, C., Moore, R., Schneider, D., 2002.Satellite imagery proves essential for monitoring erupting Aleutian volcano.EOS, Trans. Am. Geophys. Union 83 (241), 246247.

    Dean, K.G., Dehn, J., Papp, K.R., Smith, S., Izbekov, P., Peterson, R., Kearney,C., Steffke, A., 2004. Integrated satellite observations of the 2001 eruptionof Mt. Cleveland, Alaska. J. Volcanol. Geotherm. Res. 135. doi:10.1016/j.jvolgeores.2003.12.013.

    Delmelle, P., Stix, J., Baxter, P., Garcia-Alvarez, J., Barquero, J., 2002.Atmospheric dispersion, environmental effects and potential health hazardassociated with the low-altitude gas plume of Masaya volcano, Nicaragua.Bull. Volcanol. 64 (6), 423434.

    Draxler, R.R., Hess, G.D., 1998. Description of the Hysplit_4 modeling system.NOAA Tech. Memo ERL ARL-224 25 pp,

    Dunn, M.G., Padova, C., Moller, J.E., Adams, R.M., 1987. Performance of aturbofan and a turbojet engine upon exposure to a dust environment. Trans.ASME 109 (3), 336343.

    Dunn, M.G., Baran, A.J., Miatech, J., 1996. Operation of gas turbine engines involcanic ash clouds. Trans. ASME 118 (4), 724731.

    EPA, 2006. National Ambient Air Quality Standards (NAAQS). ( accessed Oct. 1, 2006).

    Grindle, T.J., Burcham, F.W., 2003. Engine damage to a NASA DC-8-72airplane from a high-altitude encounter with a diffuse volcanic ash cloud.NASA Tech. Memo. NASA/TM-2003-212030.

    Guffanti, M., Ewert, J.W., Gallina, G.M., Bluth, G.J.S., Swanson, G.L., 2005.Volcanic-ash hazard to aviation during the 20032004 eruptive activityof Anatahan volcano, Commonwealth of the Northern Mariana Islands.J. Volcanol. Geotherm. Res. 146 (13), 241255.

    Horwell, C.J., Baxter, P.J., 2006. The respiratory health hazards of volcanic ash:a review for volcanic risk mitigation. Bull. Volcanol. 69 (1), 124.

    Hunton,D.E., Viggiano, A.A.,Miller, T.M., Ballenthin, J.O., Reeves, J.M.,Wilson,J.C., Lee, S.H., Anderson, B.E., Brune, W.H., Harder, H., Simpas, J.B.,Oskarsson, N., 2005. In-situ aircraft observations of the 2000 Mt. Heklavolcanic cloud: composition and chemical evolution in the Arctic lowerstratosphere. J. Volcanol. Geotherm. Res. 145 (12), 2334.

    ICAO, 2001. Manual on volcanic ash, radioactive material and toxic chemicalclouds. Doc 9691,

    Jensen, J.W., Squire, S.W., Bons, J.P., Fletcher, T.H., 2005. Simulated land-based turbine deposits generated in an accelerated deposition facility. Trans.ASME 127 (3), 462470.Rose, W.I., Gu, Y., Watson, I.M., Yu, T., Bluth, G.J.S., Prata, A.J., Krueger, A.J.,Krotkov, N., Carn, S., Fromm, M.D., Hunton, D.E., Ernst, G.G.J., Viggiano,A.A., Miller, T.M., Ballenthin, J.O., Reeves, J.M., Wilson, J.C., Anderson,B.E., Flittner, D.E., 2003. The FebruaryMarch 2000 eruption of Hekla,Iceland from a satellite perspective. Volcanism and the Earth's Atmosphere,Geophysical Monograph 139. American Geophysical Union, pp. 107131.

    Rose, W.I., Millard, G.A., Mather, A., Hunton, D.E., Anderson, B.,Oppenheimer, C., Thornton, B.F., Gerlach, T.M., Viggiano, A.A., Kondo,Y., Miller, T.M., Ballenthin, J.O., 2006. Atmospheric chemistry of a 3334hour old volcanic cloud from Hekla volcano (Iceland): insights from directsampling and the application of chemical box modeling. J. Geophys. Res.111, D20206. doi:10.1029/2005JD006872.

    Servranckx, R., Chen, P., 2004. Modeling volcanic ash transport and dispersion:expectations and reality. Proceedings of the 2nd International Conference onVolcanic Ash and Aviation Safety, Alexandria, VA.

    Simkin, T., Siebert, L., 1994. Volcanoes of the World. Smithsonian Institution,Global Volcanism Program, Washington D.C. and Geoscience Press Inc.,Tucson.

    Simpson, J.J., Hufford, G.L., Pieri, D., Servranckx, R., Berg, J.S., Bauer, C.,2002. The February 2001 eruption of Mount Cleveland, Alaska: case studyof an aviation hazard. Am. Met. Soc. V. 17, 691704.

    Smagorinsky, J., 1963. General circulation experiments with the primitiveequations: 1 the basic experiment. Mon. Weather Rev. 91, 99164.

    Sparks, R.S.J., Bursik, M.I., Carey, S.N., Gilbert, J.S., Glaze, L.S., Sigurdsson,H., Woods, A.W., 1997. Volcanic Plumes. John Wiley and Sons, WestSussex. 574 pp.

    Tupper, A., Kamada, Y., Todo, N., Miller, E., 2004. Aircraft encounters from the18 August 2000 eruption at Miyakejima, Japan. Proceedings of the 2ndInternational Conference on Volcanic Ash and Aviation Safety, Alexandria,VA 2124 June 2004. National Oceanic and Atmospheric Administration,Silver Springs, MD USA, pp. 59.

    Uppala, S.M., Kllberg, P.W., Simmons, A.J., Andrae, U., da Costa Bechtold,V., Fiorino, M., Gibson, J.K., Haseler, J., Hernandez, A., Kelly, G.A., Li, X.,Onogi, K., Saarinen, S., Sokka, N., Allan, R.P., Andersson, E., Arpe, K.,Balmaseda, M.A., Beljaars, A.C.M., van de Berg, L., Bidlot, J., Bormann,N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M., Fisher, M.,Fuentes, M., Hagemann, S., Hlm, E., Hoskins, B.J., Isaksen, L., Janssen,P.A.E.M., Jenne, R., McNally, A.P., Mahfouf, J.-F., Morcrette, J.-J., Rayner,N.A., Saunders, R.W., Simon, P., Sterl, A., Trenberth, K.E., Untch, A.,Vasiljevic, D., Viterbo, P., Woollen, J., 2005. The ERA-40 Re-analysis.Q. J. R. Meteorol. Soc. 131, 29613012.underway and will be the subject of a future paper.

    This link is 10.1007/s00445-0329-,",Forecasting exposure to volcanic ash based on ash dispersion modelingIntroductionVATD modelingExposure calculationFebruary 26, 2000 Hekla eventFebruary 19, 2001 Cleveland eventPotential hazard of St. Helens ashConclusionsReferences


View more >