Comment on ''Particle aggregation in volcanic eruption columns'' by ...

More documents

Recommendations

Info

XXXXXXTEXTOR AND ERNST: COMMENTARYXXXXXX303 (neglecting electrostatic forces for the moment). In addition,304 the efficiency of gravitational aggregation can be strongly305 enhanced in flows with strong turbulence [e.g., Pinsky and306 Khain, 1997; Pruppacher and Klett, 1997]. This super-307 position of purely turbulent and gravitational aggregation308 mechanism is considered by introducing a collision309 efficiency E in the equation for the gravitational kernel in310 cloud microphysics, as highlighted below.311 [16] The above discussion illustrates that it cannot be312 known a priori which aggregation mechanism(s) will dom-313 inate. VW01, however, do not describe or discuss which314 aggregation process they consider the dominant one in their315 model. They give for the aggregation kernel K VW a propor-316 tionality to the quadratic sum of the particle radii and to a317 collision velocity w p :K VW / r i þ r j 2wp:ð11Þ319 The collision velocity has a single value at a given height320 level and is depends only on the mean intensity of turbulent321 fluctuations, which in turn are controlled by the mean322 upward speed at that height. We agree that the collision323 velocity can be related to the intensity of turbulent324 fluctuations, especially in such strongly turbulent volcanic325 plumes. However, the effect of having contrasting masses,326 inertia, and fall velocities cannot be neglected, since it327 dominates the aggregation mechanism as demonstrated328 above. Hence the equation for the aggregation kernel in329 VW01 (their equation (4)) seems not to reflect any330 aggregation mechanism that would be relevant for particle331 growth in volcanic eruption columns. The solution of the332 aggregation equation (equation (6) of VW01) is performed333 using the well-known method of Berry and Reinhard [1974]334 by discretizing the equation on a logarithmic mass grid. It335 appears that VW01 use a constant collision velocity because336 it facilitates the solution of the aggregation equation.337 However, convenient this may be for mathematical treat-338 ment, our analysis suggests that this approach to aggrega-339 tion modeling has no meaningful physical basis.340 5. Collision efficiency341 [17] The collision efficiency is defined as the ratio of the342 actual collision cross section and the geometrical cross343 section. Hydrodynamic, electrostatic, and turbulent forces344 can modulate it. There still exist many uncertainties in the345 theoretical models for collision efficiencies. For charged346 particles in an electric field, for large particles of equal size347 in calm air, or for small particles in turbulent flow the348 collision efficiency can be enhanced. In turbulent flow the349 collision efficiency increases due to concentration inhomo-350 geneities resulting from the differential response of drops of351 contrasting size to fluctuations in the turbulent velocity352 field, and due to inertia-induced relative velocities in353 turbulent air. Reviews of the recent theory are given by354 Pruppacher and Klett [1997], Pinsky and Khain [1997],355 Pinsky et al., [2000], Vaillancourt and Yau [2000], and356 Shaw [2003].357 [18] VW01 use the approach of Mason [1971], Klett and358 Davis [1973], and Beard and Ochs [1984] tabulated by Hall359 [1980] and the approach of Rogers and Yau [1989] based on360 Oseen flow for small particles and superposition of potentialflow for larger ones. These collection efficiencies areobtained from theories, which are valid in calm air only.Hence they are in principle not consistent with the approachof VW01 where the relative particle velocities are determinedby turbulence only. In addition, the collision efficienciesemployed by VW01 are unlikely to be correct in avolcanic eruption column, where turbulence and electrostaticforces are likely to enhance particle collisions leadingto collision efficiency values larger than one.6. Sticking Efficiency[19] The sticking (or coalescence) efficiency is a complicatedfunction of particle size, shape and surface conditions,relative velocity and other factors [e.g., Pruppacher andKlett, 1997]. VW01 assume that two ash particles can sticktogether with an efficiency dependent on the particle sizeonly, as soon as liquid water is present in the eruptioncolumn. However, in reality, the relative mass mixing ratiosof water and ash in volcanic clouds is often smaller thanone, so that the amount of condensed water is not sufficientto completely moisten the ash particles. This is especiallytrue for porous volcanic ash, where capillary forces mayinitially absorb the water inside the particle and not at itssurface: such a particle may have a sticking efficiency like acompletely dry particle. VW01 do not consider the influenceof water availability on the particle surface, i.e., of thewater mass mixing ratio.[20] The sticking between wet ash particles with Stokesnumber larger than one is parameterized by VW01 usingexperimental data from Gilbert and Lane [1994]. However,these data were obtained under atmospheric conditions,which do not scale to relevant environmental conditions.It is worth emphasizing that the Gilbert and Lane experimentsfailed to form spherical ash aggregates analogous toash aggregates observed in the geological record. As a lastresort, VW01 were forced to introduce a large polystyrenesphere to simulate how growth may proceed on it, i.e., in asituation where a large aggregate has already formed.[21] Aggregation of ice particles or of ash particlesencased by ice is also neglected by VW01. However, it isclear from meteorological clouds that ice particles do growto form larger snow or graupel particles [e.g., Pruppacherand Klett, 1997]. The coalescence of ice particles is mostefficient close to the freezing temperature. Although itbecomes rapidly less efficient with decreasing temperature,it cannot simply be ignored in volcanic eruption columns,where the temperature is below the freezing level in largeregions and where ice frequently occurs [e.g., Rose et al.,1995; Mayberry et al., 2002; Lacasse et al., 2003].7. Comparison With Mount St. Helens Data[22] VW01 compare their model predictions to a wellknownash fall data set for the 18 May 1980 eruption ofMount St. Helens (MSH) [Carey and Sigurdsson, 1982],and after adjusting several free parameters they obtain agood fit and use this in support of their modeling approach.However, the MSH fall event did not lead to deposition ofaggregates, which could be associated with a substantialdegree of wet aggregation. On the contrary, the MSHaggregates observed were described as archetypal examples3613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184of6
XXXXXXTEXTOR AND ERNST: COMMENTARYXXXXXX419 of dry aggregates [Sorem, 1982], which are held together420 primarily by electrostatic attraction and possibly also by421 geometrical interlocking as in snowflakes, leading to ex-422 treme porosities and a loosely bound, open framework423 structure [see also Schumacher, 1994; James et al.,424 2002]. The good fit between the VW01 model predictions425 and the MSH data appears coincidental and related to the426 adjustment of free parameters (e.g., the collision Stokes427 number). The validation of the wet aggregation model428 simulated in VW01 is of little relevance, because wet429 aggregation was not the dominant mechanism during the430 MSH eruption.431 8. Conclusions432 [23] The series of sensitivity experiments in the VW01433 paper is misleading in giving the impression that the model434 assumptions have been validated against appropriate field435 data. Variations of the initial mass flux, temperature, and gas436 content, which is unrealistically low, do not mean much. In437 our view, the assumption of top hat geometry in the eruption438 column model is not suitable for the simulation of the wet439 aggregation process. The influence of varying the particle440 size distribution on the aggregation efficiency is of little441 interest: particle aggregation is not described according to442 sound first principles or parameterized in a way that would443 reflect any of the recent developments in cloud micro-444 physics. The comparison with the Mount St. Helens erup-445 tion is inappropriate to validate the simulations. The446 experiments do not increase our understanding of ash447 particle aggregation in volcanic eruption columns. The448 model leads to erroneous results and should not be used449 for important applications such as aircraft safety or predic-450 tions of hazards on the ground.451 [24] In general, numerical models can be extremely452 helpful for the investigation of processes, which are not453 accessible to direct observations, or which are not yet fully454 understood. Murray [2002, 2003] and Harry [2003] discuss455 the usefulness of numerical modeling as a geophysical456 research tool. Harry [2003] addresses the essential condi-457 tions, which have to be met by these models: ‘‘Simulations458 are based on quantitative descriptions of the relevant laws of459 physics as they are currently understood, empirical mea-460 surements of the physical properties involved, and the461 available field data.’’ This condition is not fulfilled in the462 VW01 model concept as highlighted above.463 [25] Clearly, on the one hand, the problem of particle464 aggregation in volcanic or meteorological clouds is complex465 and much work is needed. There are very scarce constraints466 on several of the key parameters, even in the much simpler467 and more thoroughly examined case of strongly convective468 clouds. For example, information about the following469 parameters would be desirable: (1) in situ temperature data470 from eruption columns, (2) characteristics of turbulence471 (intensity and spatial distribution), (3) particle concentra-472 tion, (4) chemical characterization and wettability of ash473 particles, (5) erupted and in situ size distributions of ash474 particles and aggregates, (6) coalescence efficiency of475 volcanic particles, and dry, wet, and icy ash aggregates,476 (7) collision efficiency of volcanic particles and aggregates477 in turbulent flow, (8) microphysical properties of hydro-478 meteors and aggregates (e.g., shape, physical condition,hydrodynamic stability), and (9) in situ electrostatic fieldmeasurements in volcanic clouds and effects on particleaggregation. The results of simulations of a poorly constrainedsystem, and any possible interpretation of them, arein their part, poorly constrained.[26] On the other hand, however, Harry [2003] indicatesthat ‘‘...numerical simulations may...be a very importanttool for determining which additional types of data areneeded to better constrain the system. Alternatively, itmay indicate that the system is sufficiently insensitive tocertain parameters that they cannot be determined by eventhe most thorough observations of the system behavior.’’Hence specific aspects of complex phenomena can beexamined with numerical models even for a system withmany unknown parameters in order to confine their sensitivityto certain parameters. These parameters can then befurther investigated in the laboratory or possibly in the field.We believe that the process of ash aggregation in volcaniceruption columns is one such problem.[27] Yet the adaptability of a certain model configurationdepends on the specific problem under consideration, andone must be careful to avoid developing models, which donot describe, in a physically sound way, at least one aspectof the complex natural phenomenon under study. The modelassumptions must be discussed and their limitations establishedthrough comparisons with relevant field or laboratorydata. The credibility of numerical simulations as a scientificmethod is in danger if model results are inadequatelyvalidated against such data. The exploration of factorsleading to discrepancies between model predictions anddata quite often results in the most important findings.[28] Our analyses indicate that the VW01 model of ashaggregation is not based on sound first principles. Afavorable comparison of the model predictions with irrelevantfield data completes to illustrate that it is of little valuein specifically and accurately describing the problem athand and that it fails to provide new insights.[29] In sum, no theoretical model of (either dry or wet)ash aggregation has been presented to date. The solutionrequired here, must be an unsteady eruption column modelincluding simplified but physically sound and fully justifiedprinciples of cloud microphysics. In the case of wet aggregation,comparison of model predictions must be againstany of the following: Laboratory data, constraints fromdirect and from remote sensing observations of water-richeruption columns [Rose et al., 1995, 2000; Lacasse et al.,2003] and from phreatomagmatic tuff deposits (i.e., more orless vesiculated layers containing accretionary lapilli (e.g.,Durant and Ernst, submitted manuscript, 2003)). We proposeto submit such a first numerical model of wet ashaggregation shortly.[30] Acknowledgments. The work of C.T. was supported by theBundesministerium für Bildung, Wissenschaft, Forschung und Technologie(BMBF) program AFO2000, grant 06-056-0180, and by the VolkswagenFoundation under the project EVA. G.G.J.E. acknowledges support fromthe Nuffield Foundation (NAL Award) and from the Fondation Belge de laVocation through the Golden Clover Prize; he also acknowledges supportfrom NSF through Bill Rose (Processes in volcanic ashclouds NSF grant),which allowed him to develop this work with C.T. while visiting MichiganTech in summer of 2002. G.G.J.E. is now supported by the Belgian NSF(FWO-Vlaanderen) at University of Ghent. We thank Francis Albarede andthree anonymous reviewers for constructive and helpful suggestions thathelped improve the manuscript.4794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415of6
Page 1 and 2: JOURNAL OF GEOPHYSICAL RESEARCH, VO
Page 3: XXXXXXTEXTOR AND ERNST: COMMENTARYX

Comment on ''Particle aggregation in volcanic eruption columns'' by ...

Create successful ePaper yourself

Delete template?

Save as template?