Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/unpaywall/aerosol-dynamics-in-ship-tracks-cgAbR40TYl
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Publisher
- Unpaywall
- ISSN
- 0148-0227
- DOI
- 10.1029/1999jd900985
- Publisher site
- See Article on Publisher Site
Abstract
Calhoun: The NPS Institutional Archive DSpace Repository NPS Scholarship Publications 1999-12-27 Russell, Lynn M.; Seinfeld, John H.; Flagan, Richard C.; Ferek, Ronald J.; Hegg, Dean A.; Hobbs, Peter V.; Wobrock, Wolfram; Flossmann, Andrea I.; O'Dowd, Colin D.; Nielsen, Kurt E.... Journal of Geophysical Research, Vol. 104, No. D24, pp. 31,077-31,095, December 27, 1999. https://hdl.handle.net/10945/45718 This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States. Downloaded from NPS Archive: Calhoun JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. D24, PAGES 31,077-31,095, DECEMBER 27, 1999 Lynn M. Russell/ John H. Seinfeld,2 Richard C. Flagan,2 Ronald J. Ferek,3 Dean A. Hegg,4 Peter V. Hobbs,4 Wolfram Wobrock,5 Andrea 1. Flossmann,5 Colin D. O'Dowd,6 Kurt E. Nielsen/ and Phillip A. Durkee Abstract. Ship tracks are a natural laboratory to isolate the effect of anthropogenic aerosol emissions on cloud properties. The Monterey Area Ship Tracks (MAST) experiment in the Pacific Ocean west of Monterey, California, in June 1994, provides an unprecedented data set for evaluating our understanding of the formation and persistence of the anomalous cloud features that characterize ship tracks. The data set includes conditions in which the marine boundary layer is both clean and continentally influenced. Two case studies during the MAST experiment are examined with a detailed aerosol microphysical model that considers an external mixture of independent particle populations. The model allows tracking individual particles through condensational and coagulational growth to identify the source of cloud condensation nuclei (CCN). In addition, a cloud microphysics model was employed to study specific effects of precipitation. Predictions and observations reveal important differences between clean (particle concentrations below 150 cm -3) and continentally influenced (particle concentrations above 400 cm- ) background conditions: in the continentally influenced conditions there is a smaller change in the cloud effective radius, drop number and liquid water content in the ship track relative to the background than in the clean marine case. Predictions of changes in cloud droplet number concentrations and effective radii are consistent with observations although there is significant uncertainty in the absolute concentrations due to a lack of measurements of the plume dilution. Gas-to-particle conversion of sulfur species produced by the combustion of ship fuel is predicted to be important in supplying soluble aerosol mass to combustion-generated particles, so as to render them available as CCN. Studies of the impact of these changes on the cloud's potential to precipitate concluded that more complex dynamical processes must be represented to allow sufficiently long drop activations for drizzle droplets to form. 1. Introduction nity to study cloud processing, including both marine boundary layer' chemistry and interactions between an Ship-generated anomalous lines in marine stratiform thropogenic aerosols and marine clouds. During the cloud structures, visible in satellite imagery and com June 1994 Monterey Area Ship Track (MAST) exper monly referred to as ship tracks, provide an opportu- iment, off the coast of Monterey, California, the effect of ship exhaust was observed under continentally influ 1 Department of Chemical Engineering, Princeton Univer enced and typical background marine conditions, with sity, Princeton, New Jersey. aerosol and cloud droplet measurements being made 2Department of Chemical Engineering, California Insti within the boundary layer, in -the cloud layer: and in tute of Technology, Pasadena, California. the free troposphere [Durkee et al., 2000a]. 3 Office of Naval Research, Arlington, Virginia. 4 Department of Atmospheric Sciences, University of The goal of the present study is a comprehensive ex Washington, Seattle, Washington. amination of the evolution of ship emissions in the ma 5Department of Atmospheric Physics, Universite Blaise rine boundary layer. We compare observations of the Pascale, Clermont-Ferrand, France. chemical and microphysical characteristics of ship emis 6University of Sunderland, School of the Environment, sions as functions of time since release with predictions Center for Marine and Atmopsheric Sciences, Durham, Eng of aerosol modification by condensation (in and below land. cloud), coagulation, and homogeneous/heterogeneous 7Department of Meteorology, Naval Postgraduate School, Monterey, California. nucleation. This analysis provides a theoretical basis for understanding the processes important in the formation Copyright 1999 by the American Geophysical Union. of ship tracks. We investigate the role of aerosols from several marine boundary layer sources, including sea Paper number 1999JD900985. spray, biogenic sulfur, and ship stack combustion prod- 0148-0227/99/1999JD900985$09.00 31.077 31.078 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS cloud, undergoing successive cooling in updrafts and ucts. The ability of different particle populations to heating in downdrafts [Russell et aI., 1994]. Ship tracks serve as sites for cloud droplet activation provides a ba frequently persist for multiple days allowing time for sis for estimating the impact of anthropogenic aerosols numerous cycles of a single air parcel through cloud on marine stratocumulus. [Durkee et al., 2000b]. Ship track formation results in changes to the air parcel particle size distribution, 2. Ship Track Evolution which will affect the fraction of particles activating in In the cloudy marine boundary layer, aerosols may subsequent updrafts. By predicting the relative rates of grow by gas- to- particle conversion both below and in number and size change due to gas-to-particle conver cloud, may activate to droplets in cloud, may coagulate sion, coagulation, and cloud-processing, one may study with each other below cloud and with droplets in cloud, how particles grow as the ship track evolves. Both the and may coalesce while activated in cloud. Sources of magnitude of the peak supersaturation in each succes additional particles are those entrained from the free sive cloud cycle and the presence of other particles that troposphere and from adjacent air masses. These pro compete for water vapor will control the CCN available cesses control aerosol evolution for both natural and for droplet formation and associated cloud processing anthropogenic particles in the atmosphere. growth [Kaufman and Tam ... 1994]. Since this varia Marine stratus clouds form when aerosol particles in tion in supersaturation cannot be measured with cur moist air are activated, and this air is supersaturated by rent instrumentation, LES predictions are required to cooling as it is lifted to the top of the boundary layer. determine the impact of this spatial and temporal vari The activation of particles (of, e.g., 0.2 I'm diameter) ability. to cloud droplets (exceeding 10 I'm diameter) increases particle surface area and collision cross section by sev 3. Aerosol Dynamics Model eral orders of magnitude. Anthropogenic aerosol particles can provide addi To understand aerosol growth and activation pro tional potential cloud condensation nuclei (CCN) be cesses and subsequent cloud droplet dynamics, we have yond those naturally present. If the anthropogenic par constructed a numerical model and used it to evaluate ticles are sufficiently large and hygroscopic to activate at data measured in the atmosphere [Russell and Sein the peak supersaturation reached in the updrafts form feld, 1998]. The impact of anthropogenic aerosol has ing the cloud, then the number of particles activated been studied using three distinct aerosol populations in in supersaturated conditions increases over that under an externally mixed, fixed-sectional model of the nucle conditions without anthropogenic influence. However, ation, condensation, coagulation, and deposition pro since the number of particles activated and the max cesses of the aerosol. Growth and evaporation of par imum supersaturation attained in cloud are intrinsi ticles during cloud formation are calculated by explicit cally dependent. both must be calculated simultane accounting of wet and dry particle size to retain the ously. The consequence of activating more particles separate compositions of activated particles. is that water is removed from the vapor phase more The aerosol dynamics model is based on a fixed quickly, such that given a constant rate of cooling in sectional approach to representing the size domain, with the air parcel, there is no longer as much "excess" wa internally mixed chemical components and externally ter in the vapor phase, leading to a lower maximum mixed types of particles [Russell and Seinfeld, 1998]. supersaturation. By allowing a fixed set of species to condense onto any The structure of updrafts and downdrafts within one of several predefined aerosol populations, the model clouds varies greatly with significant vertical and hori allows each population to be described by a different in zontal variations [Stevens et aI., 1996]. Predicting these ternal mixture of species at a different size. Coagulation would tend to define increasing numbers of types of new structures and the magnitude of the associated lapse rates is important for following the track evolution over "populations)' resulting from combinations of existing the timescales of hours to days but would require a large populations, but to maintain a manageable number of eddy simulation (LES) approach. However, there is not aerosol types, we have used here a categorization based yet a LES algorithm available which describes external on the experimentally measurable quantity of particle mixtures of particles, forcing a choice of either accurate volatility [Jennings and O'Dowd, 1990], in which parti dynamics with simplified aerosol populations or detailed cles of different populations that coagulate are assigned aerosol descriptions with simplified boundary layer cir to the population with the least volatile components. culation. In this study we use the latter approach to fo Alternative schemes can be tailored to track specific cus on the role of aerosol particle concentration, size dis chemical components, but this scheme also provides tribution, and composition on cloud droplet activation an optically important characterization of the mixture since those particles with involatile black carbon cores and subsequent dynamics, using measured lapse rates and prescribed updraft velocities as fixed constraints in are the same particles that have insoluble, absorbing one-dimensional thermodynamic profiles. components. This algorithm provides a solution which In a uniform, stable stratus cloud layer, aerosol par retains some information about the degree of external ticles in parcels of air may be cycled repeatedly through mixing in a population although it necessarily lumps RUSSELL ET A L. : AEROSOL D YNAM ICS IN SHIP T RAC KS 3 1,079 Figure 1. Di agram of chenli cal a nd microphysical p ro cesses included in th e P rinceton model of a ship track. M echanisms in r ectangl es indicate processes causing aerosol gr owth. Mechanism s in rectan gles with flattened corn ers represent processes t h at change th e am b ien t size of parti cles by addition or rem oval of water . \1 echanisrns in rectan gles ~" i th rounded corn er s show processes for particle re m ova l. Fluxes fro m nat ural and an t hro poge ni c sources a re s how n in hexa gons . som e particles of d iss im ilar compositions (fo r example, th e app roach of Jacobs on et al. [1994], to a llow t he ac ,cpur e" salt particles wi t h salt particles o nt o which nou cur ate c alculation of evap orati on and cond ensa tion of sea-salt sulfate h as cond ensed) in t h e sam e category. wate r in conditions of var y in g humidity. However , we believe it is preferabl e to t he a pproach o f T he m odel uses a du al m o m ent m eth od based o n Ja cobson et al. [1 994] in which all coag ul a t ed p articles T zivio n et al. [1987] to a llow accurate acco un t ing o f a re placed into a sing le "mi xed " p opulat io n ret a ining no bo t h ae rosol numbe r and m ass . This algo ri t hm in corpo d istinction b etween pa r ticles of differ ent com positions , rat es indep endent calcula t io ns o f the change in pa.rt.id e since over time this ca t egory will repr esent an increasing numb er a nd mass for all p rocesses other th a n gr owth. fr action of the p ar t icle number, thus failin g t o preserve For particle number, xternal mixt u re p r operties of the origin a l scheme. t he e pik dN J~ucJ + J ~ux _ K ?epn N . + J !!row Aero sol microp h ys ics) gas-phase and heterogeneous ~ = ,k t k , k Ptk tk sulfur chemi str y, a nd bo und a ry laye r co ncen t ra tio n a nd ~ '\' '\' '\' f{~oag N . N t emper a tur e g radi ents a re d escribed e;q)licitl y (see Fig 2 L L L !I t'l P l t k , Pi ,k i l $i t'J$i k l '5.k ur e 1). The m o d el em ploys a fixed secti ona l represen tat ion of the si ze doma in with a du a l m o men t (number L L I{~lOtg NPi l kl Np ik' (1 ) a nd mass) alg orith m t o calculate growt h o f particles i l $ irn u k l $ k m ... " fr om one sedion t o th e next for non evapo rating species (namely, all comp o nents ot her than water ). Water is where N is the number of p articles in siz e section i of p i k treated in a m ov ing sect io n representat ion 1 similar to ar t icle populat ion k Here J !1ud J~ux and J ~ row a re P . I k , l k , t k 31,080 RUSSELL ET AL: AEROSOL DYNAMICS IN SHIP TRACKS the rates of particle production by nucleation, trans ammonia keeps the pH from becoming too acidic. As port from external sources, and growth to other size a resuit, aqueous oxidation of S02 was limited largely bins, respectively. The rate constants for deposition by the particulate surface area controlling the impinge and coagulation are K~epn for particles with the ambi.., ment rate of molecules. Also, the sulfuric acid activity ent size of section i in population k and K~l~~g between of the particle was assumed to be negligible, making particles with the ambient sizes of sections i and i • surface area the limiting factor for the H S0 conden 1 2 2 4 The change in particle mass, including condensational sation rate. growth, is then described by 3.2, Fluxes To identify the potential of particles to act as CCN, the chemical and microphysical properties of the vapor and aerosol phases must be described. Here we have used measurements from the MAST experiment to pro vide initial conditions for particle size distributions and for 80 concentrations. Ambient particle and vapor concentrations for the case studies here are described in section 4, DMS and H S0 were not measured and 2 4 have been initialized by marine boundary layer mixing ratios of 5 parts-per-trillion (ppt) and 0 ppt, respec tively [Pandis et 01,. 1994J, in accordance with mea surements in temperate marine conditions [Bates et at., 1990; Weber et aI" 1995; DeBruyn et 01" 1998J, A con 2 1 servative value for marine DMS flux of 2 pmol m- d- is used on the basis of low wind speeds in clouded con ditions in the midlatitude Pacific Ocean [Bates et 01" 1987], The average ion ratio of NHt to SO~- is based on a midlatitude eastern Pacific Ocean average value of 1.5 [Quinn et 01" 1990J, Filter measurements of submicron aerosol ionic com positions were collected [Hobbs et 01" 2000], but size resolved information for each externally mixed particle population was not measured. "Sea sale' particles are considered to be primarily sodium chloride, and DMS where M is the mass of species j in section i of parti pijk derived "marine sulfate" particles are dominated by am cle population k, and ffiik is the mass of a single particle of population k, in section i. In this expression, V j, Pjco , monium sulfate and bisulfate, Since the stack dilution rf conditions are not known, "plume" emissions are esti and P are the diffusivity, bulk partial pressure, and iJk mated to be mixtures of 50% organic carbon and 50% surface partial pressure of vapor-species j 1 Dp::;b and black carbon (consistent with the range of compositions Knik are the ambient diameter and associated Knudsen measured by Hiidemann et 01,(1991]), where all sulfur number of particles in section i of population k, and the dioxide produced in stack combustion and emitted with correction factors F(Knik) and A(Knik) account for free the particles is assumed to be in the vapor phase at molecular effects and mass transfer limitations, respec stack exit, although it condenses rapidly during the ini tively. tial period of the simulation. While this assumption neglects the sulfur vapors that condense prior to exit 3.1. Nucleation ing the stack, it is useful in providing an artificial dis The binary, homogeneous nucleation rate for sulfuric tinction between primary and secondary particle mass. acid and water is calculated on the basis of K ulmala "Continental" emissions are taken to be a mixture of and Laaksonen [1990J, We consider vapor-phase oxida several combustion sources (including industrial boilers tion of DMS to SO, with a yield of 80% and of S02 and automobiles), onto which coemitted sulfur species to H,S04 [Pandis et aI" 1994J, Aqueous oxidation of have already condensed, resulting in internal mixtures S02 is assumed to be limited by the rate of transfer of of 50% ammonium sulfate and bisulfate, 25% organic SO, from the vapor as long as an oxidant is available, carbon, and 25% black carbon [Hildemann et ai" 1991J, In the absence of size-segregated information about the distribution of the primary potential aqueous oxidants, 3,3. Deposition H202 and 0 , we have used a midrange value of 10 pM H 0 from the measured cloud concentrations of Particle deposition is calculated from the gravita 2 2 Richards et 01, [1983J and have taken 0 available for tional settling velocity based on the ambient size and oxidation to be in excess since the presence of sufficient density of particles of type k in size bin i [Seinfeld and RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS 31.081 changes in number concentrations of interstitial aerosol Pandis, 1995J. Since water is not included in the mass and cloud droplets. When discrete size sections are used to represent the fixed-section grid, the ambient used, artificial diffusion of particles among adjacent size diameter of particles does not correspond to the grid bins can occur [Wexler and Seinfeld, 1990; Dhaniyala (dry) diameters. The following identity illustrates this and Wexler) 1996]. Even accurate advection algorithms, important distinction: such as accurate space derivatives [Wexler and Sein feld, 1990] and the Bott method [Batt, 19S9] can fail (3) to conserve particle number by apportioning additional mass to a particle number in a fixed size bin. For these studies we have determined that numerical diffusion is where the ambient diameter of particles and droplets is the duration of the events studied, corre negligible for Dp;:;b for section i of population k. Cons.equently, at sponding Lo Limescales of up to 2 hours, by confirming each time step for each section of each partIcle type, an that the number and shape of the size distribution are ambient diameter is calculated on the basis of the total conserved by the model in the absence of both conden mass of all species present and their volume-weighted sation and coagulation. We have employed 40 size bins densities. The ambient diameter is then used in order in the dry diameter range of 0.005 I'm to 50.0 I'm. The to calculate the slip correction factor needed for depo time step was varied from 1 s to 10 s where the shorter sition as well as the surface area available for canden:'" time steps were needed to follow water condensation in satio~ and the collision cross section needed for coagu updrafts, and longer time steps were used when water lation. was partitioned at equilibrium below cloud. 3.4. Growth . 'b d b h Jgmw . 3.5. Coagulation Growth of partIcles, descrI e y t e term ik ,IS evaluated as a net contribution to each section by eval Coagulation is treated by an algorithm that consid uating the remaining terms of equations (1) and (2) ers particles to have a comparatively involatile core sub subject to conservation of the zeroth and third moments stance, so coagulated particles are assigned to one of the between adjacent sections) namely for section i and sec original particle types based on the least volatile core tion i + 1, component [Russell and Seinfeld, 1998]. This approach assigns "mixed" particles to different bins and allows the original composition of particle types to change, (4) while retaining characteristics that could be compared to volatility-based measurements. 3.B. Condensation (5) The model includes a dynamic scheme for activat- ing particles to cloud droplets [Russell and Seinfeld, 1995]. In subsaturated conditions below cloud base ry aerosol particles can be considered to be in local equilib where D 1 is the mass mean diameter of dry compo rium with water vapor; for the conditions of the MAST nents of section i and is fixed during the simulation such experiment, the characteristic time to reach water va- that por equilibrium is always less than 1 s. The surface D dry P(i+l) area used in calculating condellsation rate accounts for = constant. (6) dey the water associated with each particle dry mass by Pi using the ambient particle diameter defined by equa In some cases the accuracy of this approach is limited by tiO!1 (3). Our predictions assume that these thermody numerical diffusion [Jacobson, 1997J, but for the num namic parameters are the ones on which aerosol parti ber concentrations and growth rates considered over the cles have a negligible feedback effect due to small asso time scales of 20 min to 75 min (corresponding to up ciated changes in the heat of condensation evolved and draft velocities of 0.5 m s-l to 0.2 m S-l) used here, in the density and viscosity of the nondrizzling particle these errors are negligible. laden air parcel. Tracking changes in particle size requires careful at tention in numerical models of aerosol populations. 4. Case Studies Since the particle number distribution determines the number of nuclei available for cloud droplet formation, A clean marine case and a continentally influenced conservation of particle number in the calculation of case were chosen to represent the different ambient con growth and coagulation is critical to modeling cloud ditions and ship emissions sampled during MAST: (1) processes such as those involved in ship tracks, espe Star Livorno, University of Washington (UW) C131- cially since ship tracks are characterized by significant A flight 164S, June 29 (JDT ISO) (clean marine case); 31,082 RUSSELL ET AL., AEROSOL DYNAMICS IN SHIP TRACKS Table 1, Measured and Estimated Track and Background Conditions for the Clean Marine Case (JDT 180) and the Continentally Influenced Case (JDT 178) Clean Marine Case Continentally Influenced Case Background Background Track Track Case Description Ship Star Livomo TaiHe June 27, 1994 Date June 29, 1994 Julian Date (JDT) 180.51 to 180.60 178.49 to 178.55 Latitude to 35.8°N 36.l N to 37.3°N 35.5°N Longitnde -125.3°W to -126.3°W -123.4°W to -123.9°W Thermodynamic Quantities Boundary layer height (m) 450 450 405 405 Cloud base (m) 230 230 173 173 Cloud base temperatore (K) 273.7 273.7 285.3 285.3 1 6.5 6.5 6.1 6.1 km- Lapse rate (K ) 11.4 11.4 3 10.3 10.3 Total water content (g m- ) 12 12 12 12 Wind speed (m s-l) Microphysical Characteristics 0.3±0.2 0.3±0.2 0.3±0.2 0.3±0.2 Updraft velocity (m sol) I 32±41 Cloud droplet number (cm") 49±1O 110±32 l62±37 0.28±0.05 0.31±0.11 0.35±0.05 3 0.35±0.05 Liquid water content (g m- ) Effective radius (iJIIl) 7.87 5.10 5.98 5.67 Chemical Species S02 (ppb) 0.34 3.7 1.0 6.5 0.3 NO (ppb) 0 0 0 0.44 0.05 0.29 0.18 N02 (ppb) 0.1 2 1 HCHO(Ppb) 2 26 21 25 27 03 (ppb) C02 (ppb) 360 360 360 360 OH (ppb) 6 7 7 7 1.43xlO- 3.77xlO- 6.42xlO- 4.4lxlO- 88 88 88 88 H202 (J.IM) H2S04 (ppt) 0 0 0 0 5 5 5 DMS (ppt) 5 2 2 2 2 2 DMS flux (iJIIlol m- dol) Aerosol Populations Total aerosol (cm") 18528 1113 2894 0.4 Primary mode peak (iJIIl) 0.08 0.05 0.4 NA 0.2 0.07 0.06 Secondary mode peak (iJIIl) Marine sulfate aerosol (cm") 94 94 81 81 (NH,),SO, mass (dry) 53% 53% 53% 53% 47% 47% 47% 47% (NH,)HSO, mass (dry) Sea Salt aerosol (cm-') 10 10 10 10 NaCI mass (dry!, 100% 100% 100% 100% Plume aerosol (cm- ) 1781 0 18424 0 NA 50% OC mass (dry) NA 50% EC mass (dry) NA 50% NA 50% Continental aerosol (cm-') 0 0 1022 1022 NA NA 25% 25% (NH,),SO, mass (dry) (NH,)HS04 mass (dry) NA NA 25% 25% OC mass (dry) NA NA 25% 25% NA NA 25% 25% EC mass (dry) "NA" indicates that this composition is not applicable since there were no particles present of this type for this case. and (2) Tai He, UW C131-A flight 1646, June 27 (JDT imagery [Noone et 01., 2000b; Hobbs et 01.,2000]. The 178) (continent.ally influenced case). The Tai He case Star Livorno was sampled in clean marine conditions illustrates ship tracks in continentally influenced ma~ with low background aerosol concentrations [Hobbs et rine air, in which the in-track aerosol signal was weak al.) 2000]. The Tai He and Star Livorno were studied and the ship track was only faintly visible in satellite primarily with instrumentation aboard the UW C131-A RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS 31,083 [Russell et a!., 1995,1996; Hobbs et 01., 2000], although apparent source strength of the particles emitted by the Toi He is significantly less than that for the Star the MRF C130 flew near similar tracks in adjacent re gions on the same days [Hobbs et aI., 2000]. Livorno, as can be seen from the significantly larger Case studies incorporating data of simultaneous mea number of particles measured on JDT 180 than on JDT surements of microphysical, chemical, and meteorologi 178. The resulting track is characterized by a relative cal parameters from the MAST experiment appear else change in particle number and mass that is smaller in where [Durkee et aI., 2000b; Noone et al., 2000a, b; the continentally influenced case than in the case with a Hobbs et 01., 2000]. Supporting information about the clean marine background. There are also minor differ MAST operations and the ships sampled during the ences in the boundary layer structure in the two cases, including a slightly higher but almost identical in thick study are described by Gasparovic [1995]. ness cloud on JDT 180. The case studies of clean marine (JDT 180) and con tinentally influenced (JDT 178) conditions illustrate the Particle populations in the initial size distribution range of conditions measured during the MAST exper are assigned by estimating the basic sea salt contribu iment. For each case, the measured and estimated con tion and assigning the remaining clean marine submi ditions used to initialize the prediction of the evolution cron particles to the marine sulfate population. Plume of particle size are shown in Table 1, with the initial particles are determined by the difference between the particle number and mass distributions described be total plume particle number size distribution for. each low. In both cases, comparing the evolution of particles case and the total background size distribution for that in background aerosol and in track conditions allows case (namely, sea salt particles plus marine sulfate par comparison of the influence of particle conditions under ticles). Continentally derived particles are assigned as essentially identical meteorological forcing. Here the those present in the (background) continentally influ background conditions provide an experimental control enced case which exceed the number distribution of par for comparison to the track conditions. ticles present in the (background) clean marine case. There are two main microphysical differences between In this section, we describe predictions of cloud for the two cases studied. The background cases differ in mation for the cases studied, with cloud characteristics that particle size distributions measured on JDT 178 for each case summarized in Table 2. All simulations show a clear influence of continentally derived anthro were run for one complete cycle through the boundary pogenic particles in background air. In addition, the layer, which was equivalent to about 50 min for the Table 2. Predicted Cloud and Aerosol Characteristics for Track and Background Conditions for the Clean Marine Case (JDT 180) and the Continentally Influenced Case (JDT 178) Clean Marine Case Continentally Influenced Case Background Track Background Track Updraft Cloud Updraft Cloud Updraft Cloud Updraft Cloud Onl~ Averase GnlX Averase Gnll: Averase Gnll: Averase Cloud Properties Liquid water content (g m") 0.34 0.19 0.36 0.20 0.39 0.22 0.39 0.22 Effective radius (filll) 10.8 10.8 3.4 3.4 5.7 5.7 4.1 4.1 Droplet number (em") 96 48 3880 2130 533 293 1790 985 Maximum supersaturation 0.68% NA 0.18% NA 0.26% NA 0.15% NA Droplet Populations (> 1 pm) Marine sulfate droplets (em") 86 47 48 26 55 30 52 29 Sea salt droplets (cm") 10 6 7 4 7 4 6 3 Plume droplets (em") 0 0 3820 2101 0 0 1270 699 Continental droplets (cm") 0 0 0 0 471 259 471 259 Aerosol Populations (total) Marine sulfate aerosol (cm") 94 94 93 93 81 81 81 81 Sea Salt aerosol (em") 10 10 10 10 10 10 10 10 Plume aerosol (em·') 0 0 18300 18300 0 0 1750 1750 Continental aerosol (cm·') 0 0 0 0 1020 1020 1010 1010 Aerosol Species « I pm) (NH')2S0, mass (I'g m") 0.88 0.88 10.4 10.4 3.8 3.8 17.9 17.9 (NH,)HSO, mass (I'g m") 0.78 0.78 9.3 9.3 3.3 3.3 15.9 15.9 OC mass (I'g m,3) 0 0 1.51 1.51 0.90 0.90 1.36 1.36 EC mass (1!lI m·') 0 0 1.52 1.52 0.90 0.90 1.36 1.36 Values are given both for the updraft predictions (in the column labeled "Updraft Only") and the estimated average assuming a cloud updraft area fraction of 55% (in the column labeled "Cloud Average"). 31,084 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACK S collected near t he to p of the cloud layer are shown in Figure 4. As a result of t he significant variability along the length and wid t h of the ship track, with values of liquid wate r conten t vary ing between 0.2 and 0.5 g m- , th ere is no statistically sig nificant distinctio n between background and track values of liquid water. There is also no clear trend in liquid water with track age. In t h e clean marine case study, the background par ticle size distribution is composed entirely of sea salt and marine sulfate particles with a mean particle num ber concentration of 104 em-a Contrasting this dis tribuLion to the track initial size distribution in this case shows the track distributi on dominated by over 18 ,400 cm- plume particles. This large part icle num b er was measured close to (with in 100 m) the stack of t he Star Livomo. The resu lt ing factor of over 100 increase in part icle concentrat ion also over whelms t he Figure 2. AV HRR image from t he 3.7 pm channel at 1026 PDT on June 29 , 1994 , for t he clean m a rin e str a cloud droplet dist ributi o n during the ini t ia l cloud cycle , tocumu l us cloud case (JOT 180). T he s hi p position is even tho ugh more dilution will occur before the parcel s hown with whi te d ots o n t he satelli te image and with reaches cloud base than we have assumed in the initial gray dots and labcled position t imes on the magnified case study. inset. The inset also shows the flight track of the Uni 4,1.1. Pre dicted background aerosol. The versity of Washington C131-A aircraft as a black line particle size distribut ion below cloud initialized from with gray dots where the track was sampled. Times la beling dashed g r ay lin es indicate the approximat e age of th e track measured from the time of emission to th e salnp li ng point. Solid gray lines tlhow the observed track Observed location at the indicated satellite overpass tim es. Cloud Top 400 (450 m) base case updraft velocity of 0.3 m ,- I [Nicholls and Leighton, 1986J. 4.1. C lean Mal·ine Case: S ta." LivU1'fI,U , June 29 (JDT 180) Observed Figure 2 illustrates t he cloud r efl ectance for th e sh ip Cloud Base track m easured on Jun e 29, 1994. The particle num (230 m) b er concent ration measured was 104 cm - , indicative <I> :r5a6~!!!!.IMtt~ "'C of clean marine a ir [Hoppel et ai., 1990J. The aeroso l :l .... in the plume of the Star Livorn o was characterized by a E 200 peak at 0.05 pm (dry) diameter as measured directly af 1"" + ter emission from the stack below cloud. The associated + Lowest track m eas ured in cloud was characterized by a corre t Observation sponding peak of interstitial particles at 0.05 pm (dry) t -210 m diameter. Over the length of track sampled (from < 1 h our to 2.5 hours after emission from the ship), particles 100 from the mode between 0.030 and 0.100 pm diameter were d epleted. T he vertical temperature profile for t he cloud sam pled o n this day is shown in Figure 3. The below-cloud temperature was extrapolated from t he lowest flight al OL---~~LL--~--~ titude (- 50 m) to the surface as a constant temper 282 283 284 285 286 ature since th e surface conditions were not measured Temperature (K) directly. The temperature profile corresponds to an av erage lapse rate in the cloud layer of6.5 K km-I. Total Figur e 3. Measured temperature p rofi le for the clean water content of the boundary layer air was measured marin e cloud case (JOT 180). P luses indicate data mea to be approximately const ant below cloud at 10.25 g sured during a profile throu gh th e bo undary layer by the m- Measured particle and cloud droplet distributions University of Washington C131-A aircraft. RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS 31,085 ____ Track (measured) -B- Total -B- Background (measured) -0- Marine sulfate ····b··· Sea salt ~ 10' O~ ~ .Q '" :!1 :!1 10' 10' "C Diameter (~m) Diameter (~m) Figure 4. Measured in-cloud particle and droplet Figure 5. Predicted below-cloud particle size distribu size distribution for track and background clouds in the tion for background air in the clean marine case (JDT clean marine case (JDT 180). Solid circles represent 180). Circles show the .total aerosol n~mber distr~bu track cloud measurements and open circles represent tion, squares show manne sulfate partIcles, and tnan background cloud measurements. gles show sea-salt particles. -B- Total the measured aerosol concentration is shown in Figure -EJ- Marine sulfate .... & ... Sea salt 5. As particles are activated in the updraft region at a prescribed velocity of 0.3 ill 8- , the maximum super 4 saturation attained is 0.68%, which is shown in Figure 7. The resulting in-cloud distribution of particles and droplets is shown in Figure 6. Of the 96 em -3 predicted to be activated to cloud droplets, 86 cm- are marine sulfate particles and 10 cm- are sea-salt particles (Ta ble 2). The predicted droplet distribution of the cloud has a vertically averaged liquid water content of 0.34 g m- and effective radius of 10.8 f.<m. The predicted liquid water in the updraft is greater than the average mea sured liquid water content for this cloud of 0.28 g m- , but since the measured value reflects an average in the cloud rather than in just the updraft, we have also in cluded in Table 2 an estimate of the cloud average, as suming that updrafts represent 55% of the cloud area and downdrafts account for the remaining 45% [de Laat and Duynkerke, 1998J. The predicted average for the cloud with 55% updraft fraction is only 0.19 g m- , which is below the measured value. Both are within . Diameter (~m) the range of the values measured in cloud. It is inter Figure 6. Predicted in-cloud particle and droplet size esting to note that the predicted number of droplets of distribution for· background cloud in the clean marine 48 cm -3 for the cloud average is almost identical to the case (JDT 180). Symbol definitions are the same as in reported number of measured droplets (49 em -3). Figure 5. 31.086 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS Relative Humidity (%) simultaneous appearance of particles at larger sizes 0.2- 0.3,.,m (dry) diameter as the track ages. This change is 100.4 100.6 100.8 101.0 consistent with the predicted activation of particles at this size range as well as with the growth of the emitted plume particles by below-clond H S0 condensation. 2 4 4.2. Continentally Influenced Case: Tai He, June 27 (JDT 178) The signature of the ship track from the Ta; He in continentally influenced air measured on June 27 (JDT 178) is faintly visible in the AVHRR 3.7 ,.,m chan E 350 nel image shown in Figure 10. The stratus cloud on "0 this flight was characterized by high levels of back ." ground particles as indicated by the condensation nu clei and cloud droplet number measurements reported [Gasparovic, 1995J. The background aerosol is char acterized by two modes, one at O.4,.,m (dry diameter) and a smaller mode at 0.07 jJm, with concentration of 1110 cm- The plume of the Tai He exhibited a peak at 0.06 ,.,m. In cloud the track of the Ta; He was characterized by increased interstitial aerosol and a minor mode at 0.01 ,.,m (dry diameter). The cloud measured for this 0.8 1.0 0.0 0.2 0.4 0.6 case is very similar to the JDT 180 case in thickness Liquid Water Content (g/m ) but was approximately 50 m lower, extending from 175 m above sea level to 405 m. The measured temperature Figure 7. Predicted profile ofliquid water content and relative humidity (supersaturated) for the clean ma rine case (JDT 180) for background and track clouds. Thin solid line shows liquid water content in background cloud, and thick solid line shows liquid water content in ___ Total track. Thin dashed line shows relative humidity in back -. Marine sulfate ground cloud, thick dashed line shows relative humidity ......... Sea salt in track. "'I''' Plume 4.1.2. Predicted track aerosol. In track the plume aerosol population provides so many particles that can act as CCN that almost all of the resulting cloud droplets are predicted to have plume particles as nuclei. Particles available to act as CCN are shown in the below-cloud size distribution in Figure 8. The high number of CCN available results in more droplets activating sooner in cloud, consequently depleting the supply of water available to condense, so a maximum supersaturation of only 0.18% is reached (Figure 7) for the track case ~ 20 m below the maximum supersatu ration altitude in the background cloud. Average droplet distribution in cloud for the track is shown in Figure 9. The effective cloud droplet radius predicted in this case, 3.4 j.tm, is significantly smaller than the observed ambient cloud. This size range cor responds to the smallest channel of cloud probes, in Diameter (!-1m) which significant uncertainties exist. Liquid water is Figure 8. Predicted below-cloud particle size distri increased to an average value of 0.36 g m- in track, an bution for track air in the clean marine case (JDT increase of 0.02 g m- from the background, as can be 180). Circles show the total aerosol number distribu seen in Figure 7. tion, squares show marine sulfate particles, triangles Measured submicron aerosol size distributions show show sea salt particles, and inverted triangles show plume particles. a depletion of particles at 0.1 ,.,m (dry) diameter and 31,087 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS _____ Total ++ .. ++ -. Marine sulfate ......... Sea salt ~o~7cl "'I ... " ... Plume Cloud Top (405 m) 10' '0 Observed ., '" .... . ., Cloud Base (173 m) +\+ ,... "',-' --r Lowest Observation + Diameter (~m) -80m I t Figure 9 .. Predicted in-cloud particle and droplet size distribution for track air in the clean marine case (JDT 285 286 282 283 284 180). Symbol definitions are the same as in Figure 8. Temperature (K) Figure 11. As for Figure 3 but for the continentally influenced cloud case (JDT 178). profile shows an average lapse rate of 6.1 K km -1 in the cloud layer (Figure 11). Distributions of particles and droplets in the cloud are shown in Figure 12. Below The presence of continentally influenced background cloud level, the total water content of the boundary air on JDT 178 resulted in increased background con layer air was measured to be 11.4 g m-a The liquid centrations of S02 and particles. Because many of water measured in this case was less variable in both the particles in this case are aged combustion particles the track and the background cloud) with an average from continental sources (and hence constitute the con value of 0.35 g m- . tinental aerosol population), they are represented in the model by a composition that is 50% carbonaceous and 50% condensed sulfur species (ammonium sulfate and ammonium bisulfate), as summarized in Table 1, These continentally derived particles are larger than the sea salt and more numerous than the marine sulfate parti cle populations. 4.2.1. Predicted background aerosol. The initial below-cloud' particle size distribution for back ground conditions on JDT 178 is shown in Figure 13. Particle number is dominated by continentally derived aerosol, although sulfate aerosol constitutes a fraction of particles large enough to act as CCN. When the air par cel reaches cloud base and becomes supersaturated, the predominant number of particles activated to droplets are continentally derived as shown in Figure 14. Throughout the vertical extent of this cloud, liquid water increases monotonically as shown in Figure 15, and supersaturation reaches a maximum of 0.26%. The Figure 10. Same as Figure 2 at 0810 PDT on June 27, average effective radius over the cloud depth is 5.7 pm 1994, for the continentally influenced case (JDT 178) with the Ta; He. with an associated liquid water of 0.39 g m- in up- 31.088 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS -+- Track (measured) --e- Total -B- Background (measured) -E}- Marine sulfate ... -6 .... Sea salt - (>- . Continental 10' ..,- 10' <? E 10' S- S- .Q '" '" 0 '!1 10' ." 10' '0 I l~" ,'\. 10' P i \. 10' 0. I \ ~, \ I1i IT Diameter (~m) 10 Diameter (~m) Figure 12. Measured in-cloud particle and droplet size distribution for track and background clouds in the con Figure 14. Predicted in-cloud particle and droplet size tinentally influenced CaBe (JDT 178). Symbol defini distribution for background air in the continentally in tions are the same as in Figure 4. fluenced case (JDT 178). Symbol definitions are the same as in Figure 13. -B- Total -E]- Marine sulfate drafts or 0.22 g m- averaged in the cloud. The mea ····fr··· Sea salt sured liquid water of 0.35 g m- falls between these - (>-. Continental two predicted values, but both predictions are within the range of liquid water content reported in cloud. 4.2.2. Predicted track aerosol. In the track in cloud, plume particles represent over half of the parti cle number distribution. Below-cloud particle number distribution and the contribution from different parti cle populations are shown in Figure 16. The activated droplet distribution is shown in Figure 17 including only a significant fraction of activated particles from plume particles, with almost all (1010 cm- in track compared to 1020 cm- in the background) of the continental par ticles that activated in the background case also forming droplets here. Predicted cloud water characteristics for the track on JDT 178 are only slightly different from the background case, with an average liquid water remaining at 0.39 g m- and a maximum supersaturation of 0.26% in the background decreasing slightly to 0.15% in track, aB il lustrated in Figure 15. Despite the small changes from Diameter (~m) the background to the track in liquid water and max Figure 13. Predicted below-cloud particle size distri imum supersaturation, the cloud effective radius de bution for background air in the continentally influ ereaBes from 5.7 I'm to 4.1 I'm and cloud droplet number enced case (JDT 178). Circles show the total aerosol 3 3 more than triples from 533 cm- to 1790 cm- . While number distribution, squares show marine sulfate par these changes do alter the albedo [Erlick et aI., 1999], ticles, triangles show sea-salt particles, and diamonds show continental particles. they are small compared to the changes predicted in the RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS 31.089 Relative Humidity (%) itation, we have modeled one case with the DESCAM (detailed scavenging and microphysics) cloud dynam 100.0 100.2 100.4 100.6 100.8 101.0 ics and microphysics model from the Clermont-Ferrand group [Flossmann ef al., 1985]. 5.1. Gas-to-Particle Conversion A microphysical mechanism that may be of impor tance in ship tracks is gas-to-particle conversion, en compassing processes from nucleation to homogeneous and heterogeneous oxidation [FeTek ef al., 1998]. Ferek ef al. [1998] studied measurements of ship tracks off the coast of Washington and found that gas-to-particle I 300 conversion, possibly in combination with increases in ." '" cloud peak supersaturation, may account for ship track .te .., persistence. These processes contribute to aerosol size distribution dynamics and influence the nature of their cloud interactions. The size and composition of CCN available will influence the maximum cloud supersatu ration, liquid water content, and droplet distribution, and hence the radiative features that characterize ship tracks. Here we consider separately the role of these processes in forming and maintaining tracks. In the formation of a ship track, one is interested in the roles of particles emitted directly from the stack and of the vapors emitted with them. To address this 0.0 0.2 0.4 0.6 0.8 1.0 Liquid Water Content (g/m ) Figure 15. As for Figure 7 but for the continentally influenced case (JDT 178). 10 r~~="~~="r~~~n,-~~,< __ Total -. Marine sulfate "-a-- Sea salt ..•. Plume clean marine case, where the effective radius dropped - +-. Continental from 10.8 I'm to 3.4 I'm in track. 5. Microphysical Processes Using case studies from section 4 as a basis, we can study the contrihutions from individual aerosol pro cesses in ship tracks by comparing predicted and ob served microphysical features. To study the role of gas to-particle conversion, we compare droplets activated with and without additional condensable vapor sources provided by SO, in the stack effluent. Enhanced parti cle growth and increased soluble fraction from the con densation of sulfate can change the predicted number of CCN and, consequently, the associated ship track fea 10 tures. Numerical modeling is used here to represent aerosol and cloud microphysical processes in the marine bound ary layer. With case study observations to provide a 10° measure of the quality of our representation of the at 0.01 0.1 mosphere, we compare the roles of several microphysical Diameter (~m) processes in aerosol evolution. The model described in Figure 16. Predicted below-cloud particle size distri section 3 includes aerosol dynamics and condensation bution for track air in the continentally- influenced case and coagUlation in a Lagrangian air parcel. Two sets of (JD'!' 178). Circles show the total aerosol number dis case studies described in section 4 provide field observa tribution, squares show marine sulfate particles, trian tions with a range of track and background aerosol con gles show sea salt particles, diamonds show continental centrations. To investigate the potential role of precip- particles, and inverted triangles show plume particles. 31,090 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS issue we need to know if the particles per se are CCN at typical marine supersaturations; some evidence from -+- Total observations [Hudson et aI., 2000] suggests that many -. Marine sulfate •••••• H Sea salt of the plume particles emitted do act as CCN. Hobbs et .. ", .. Plume al. [2000] inferred particle composition from the frac - ... Continental tion of particles activated to CCN. Here we take the alternative approach of using data from engine emis sions to prescribe compositions for stack-emitted par ticles. The model prediction that the particles in the track modify the cloud droplet distribution to include more smaller drops in the track is consistent with the prediction from the ob:.erved differences in the AVHRR image signatures [Durkee et 01., 2000b]. In the clean marine case we would like to determine how much of this signature results from gas-to-particle conversion. By considering the same conditions but restricting gas to-particle conversion by limiting the stack emissions to be only particles and no sulfur vapors, the resulting 10 ' distribution of activated droplets is predicted to be sig nificantly less than the predicted track conditions and , more similar to the background case. The results for predicted cloud droplet distributions with no S02 in the stack emissions (but with a back 0.1 ground S02 mixing ratio of 0.34 ppb) are shown in Ta ble 3. This result shows that some of the particles as Diameter (!-1m) emitted are not sufficiently efficient CCN to activate at Figure 17, Predicted in-cloud particle and droplet the predicted maximum supersaturation of 0.68%, so size distribution for track in the continentally influenced case (JDT 17S). Symbol definitions are the same as in that they never effectively compete with the previously Figure 16. existing background particles in taking up liquid water. Table 3. Predicted Variation of Cloud Characteristics for Track and Background Conditions for the Clean Marine Case (JDT ISO) and the Continentally Influenced Case (JDT 17S) Liquid Water Effective Radius Droplet Number Maxnnum 3 3 (cm- Content (g m- ) ([lm) ) Supersaturation Track Backll.round Track Backsround 8acksround Track Backll.round Track Updraft Velocity (w=O.3 m Sl) w=0.5 m S·l (JDT 180) +0.04 +0.06 +0.1 -0.6 +3 +1060 +0.23% +0.04% w=0.5 m S·l (JDT 178) +0.07 +0.06 -0.1 -0.4 +91 -40 +0.10% +0.05% w=0.2 m sol (JDT ISO) -0.02 -0.02 -0.2 +0.3 +1 +S70 -0.16% -0.02% w=O.2 m S·l (JDT 178) -0.04 -0.05 -0.2 +0.4 +58 -150 -0.05% -0.04% Updraft Area Fraction (55%) 65% (JDT ISO) +0.03 +0.04 0 0 +10 +390 0 65% (JDT 17S) +0.04 +0.04 0 0 +55 +179 0 45% (JDT ISO) -0.03 -0.04 0 0 -10 -390 0 0 45% (JDT 17S) -0.04 -0.04 0 0 -55 -179 0 Dilution ( 1 00% Emissions) 50% emissions (JDT ISO) NA 0 NA -0.3 NA +120 I'~A +0.04% 50% emissions (JDT 178) NA -0.01 NA +1.0 NA -710 NA +0.04% 10% emissions (JDT ISO) NA 0 NA +2.3 NA -3270 NA +0.15% 10% emissions (JDT 178) NA 0 NA +1.6 NA -1142 NA +0.12% Gas-Io-Particle Conversion (S02 and Particles Emitted) No emitted S02 (JDT ISO) NA 0 NA +0.2 NA -2080 NA +0.22% No emitted particles (JDT ISO) NA -0.01 NA +7.1 NA -3784 NA +0.49% All values refer to the difference from the "Updraft Only" base case value (specified in parentheses for each section) to the value noted for the case given. ("NA" indicates that there is no sensitivity calculated for this set of conditions.) RUSSELL ET AL.: AEROSOL DYNAMI CS IN SHIP TRACKS 31 .09 1 T he predicted cloud has a lower droplet concentration acti vated initially but , in t he abse nce of changes in t he t han in the track case, with a droplet concentra tion of supersaturation in later cloud cycles, does not suggest a only 1960 cm- . Wi th a faster updraft of 0.5 m 8- the mechanism for growing additi onal CCN that might be same no-S02 case yields a droplet concentration simi available to replace those particles lost to coalesce nce and scavenging in cloud. lar to the background cloud. In this case there is even less time for particle growt h fr om the small amount of 5.2. Precipitation background SO, (i.e. , 3 min rat her than 6 min ) a nd no plume particles grow large enough to be CCN at The C le rmont-Ferrand DESCAM m odel [Flossmann the maximum supersaturation . This difference from the et 01. , 1985) calculates cloud dynamics on the basis of actual case, which included stack-emitted SO, vapors, an aircraft sounding of the boundary layer structure. resu lts from the fact that the plume particl es at their The D ESCAM model differs from the Princeton model measured size on emission are both too small and too in t hat it has the ability t o consider a nonadi abatic air insoluble (without the addition of sulfate by conden parcel which explicitly a llows entrainment of air. The sation in and immediately after the stack) to be CCN features and limi tations of these models are su mmarized at lower supersaturations. The size on emission for the in Table 4. In addition , th e growth of precipitation Sta r Livornocase is derived fro m m easurements approx sized drops is calculated more accur ately by D ESCAM, imate ly 100 m away from the stack and so is likely to b e since it contains 69 droplet size classes and the fixed an overest imate of the actual size on leaving the stack. dry-size grid of the Princeton model has limited resolu Their composition of 50% organic and 50% black car tion for drops above 10 ",m diameter. The predictions bon is based on power plant engines burning heavy fuel for th e clean marine case confirm t hat sim ply by chang [Hi/demann et al. , 1991], providing a value that is well ing the input aerosol distribution from the clean back withi n the ran ge that can be expected for a ship engin e ground marine air to a pl ume aerosol from a ship stack , (excludin g the sulfate which we account for separately the resU lting cloud droplet distribu t ion was shifted to here). We note , however , that there is a lar ge range smaller sizes, as is shown in t he case studies in section 4 of possible particle compositions depending not only on for clean marine conditions. The track simulation also exact engine type but also on maintenance history, op showed much higher droplet concentrations than in the erating procedures, stack conditions, fu el source , and background. ship speed [Gosparovic, 1995]. Conversely, emitted SO, This marine air case involved an anomal ously clean is not suffici ent to nucleate and grow new CCN before background aerosol concent ration and so was chosen to the first cloud cycle. The results of t his sensit ivity st ud y study t he hypothesis that ship tracks could modify the for t he case of no p lume particles is shown in Table 3. cloud droplet distribution sufficiently to inhibit precipi Gas-to- p article fo rm ation may a lso be a m echanism tation locally. We studied the growth of droplets to de influential in the persistence of ship tracks for multiday termine if they become sufficiently large to form drizzle periods [Ferek et aI., 1998]. To addr ess this ques tion, we (approximately 20 ",m diameter) during their estimated firs t discuss th e general role of this pro cess illustrated time in the 220 m thi ck cloud at an updraft velocity of ill four case studies analyzed with the aerosol dynamics between 0.5 and 0.3 m S-I. In this case, the time re mode l. quired for growth to droplets by cond ensation and co Ship track formation is defined as the first cycle of alesce nce mechanisms was longe r t han t he t ime in t he t he plume particles through cloud in which they are ac cloud updraft region (approximately 7 to 12 min) in tivated and , consequently, change the predicted droplet both track and b ackground clouds, suggesting that in number and effective radius in track relative to the same order for drizzle to form, longer times within the cloud cycle in the background cloud. For gas-to-particle con layer are needed. This result implies that the simple ve rsion to be important in these secondary stages of single-parcel dynamics model used here is not sufficient evolution of the ship track, there has to be a source of to address drizzle formation and that a LES, which can condensable vapor present in t he air parcel after t hat exp licitly a llow mixing of parcels a nd can predict varia first cloud cycle to cause t he part icles to grow fur t her . t ions in supersaturation in multip le cloud cycles, would In t his work we h ave not employed models capable of be required in order to predict t he d etailed cloud r esi simulating large eddies that would allow us to predict dence times required for this questio n. the degree to which stack vapors may be mixed in from adjacent parcels that might not have undergone the 6. Measurement and Model same cloud cycle. Given this limit ation, our predic Uncertainties t ion s show that SO, is transferred rapidl y in the first cloud cycle to activated droplets , in agreement wit h t he Important uncertainti es underlie both measur~ments obser vations that concentrations of SO, in cloud near and m odel predictions in t he atmospheric processes the beginning of the trac k are comparable to t hose in compared here. The effects of several key un certai nties background cloud . As a result of this, gas-ta-par ticle are shown in Table 3. In t he MAST experiment, mea conversion of SO , to sulfate in droplets by heteroge surements of organic com position and of size-resolved neous oxidation contributes to growth of those CCN inorganic components were lacking. While th ese mea- 31,092 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS Table 4, Comparison of Features and Limitations of Princeton and Clermont-Ferrand (DESCAM) Models Clennont-Ferrand (DESCAM) Model Features Princeton General Mechanism studied gas-ta-particle conversion precipitation Reference Russell and Seinfeld [1998] Flossmann et al. [1989] JDT 180 background JDT 180 background Cases studied JDT 180 track JDT 180 track JDT 178 background JDT 178 track Thermodynamic parcel in I-D grid parcel Structure Aerosol Size description 40 dry fixed sections, moving section water 81 aerosol sections, 69 droplet sections external and internal internal Mixtures multicomponent nonideal equilibrium ideal equilibrium Vapor-liquid equilibrium kinetic rate of activation Droplet activation instantaneous activation of CCN Cloud Lapse rate prescribed measured profile nonadiabatic with entrainment Updraft velocity prescribed updraft velocity prescribed updraft velocity explicit coagulation and coalescence explicit coagulation, coalescence, and Microphysics scavenging Chemistry Homogeneous sulfur oxidation none Heterogeneous sulfur oxidation none surements are feasible with existing technology, they aggregate of parcels. Perhaps of greater importance is are limited by inlet losses and by long sampling times the limitation that both modeling and sampling in suffi required to colled sufficient material for off-line analy cient detail to instantaneously characterize a chemically detailed three-dimensional cloud in time remain tech sis [Huebert et al., 1998]. On-line single-particle com position measurements have been achieved for higher nologically challenging, so we still need to infer prop erties from various constant altitude averages to com altitude measurements but do not provide quantita pare them to the modeled updraft regions of our parcel tive mass composition [Murphy et 01., 1998]. Detailed model. Feingold et 01. [1998] modeled detailed inter knowledge of the distribution of species among external mixtures of aerosols would nonetheless be valuable in nally mixed size distributions with 500 air parcel tra future studies. jectories, providing a spatially resolved (but not fixed) An important overprediction by the model compared grid in which aerosol evolution could be tracked but to the measurements in the track for the clean marine without the additional computational burden of exter (JDT 180) and continentally influenced (JDT 178) cases nally mixed aerosol. Incorporating an externally mixed occurs in cloud droplet number concentrations and to aerosol in the Feingold et al. [1998] trajectory ensemble a lesser extent in liquid water content. This difference model (TEM) would allow one to study the range and results from the simplified parcel scheme in the model standard deviation of the results predicted here as well as well as instrument and sampling limitations in the as longer time evolution questions. The LES approach measurements. Predicted cloud droplet concentrations of Kogan et al. [1995] would be required to quantify the exceeding 1000 cm- cannot persist beyond short up mixing of parcels of background air with track aerosol, drafts and so comparing such predictions to ambient if a computationally efficient aerosol description can be averages is not possible. devised to incorporate track aerosol complexity. 6.1. Limitations on Sampling 6,2, Effect of Updraft Velocity Several fundamental parameters in cloud formation Uncertainty in the estimated updraft velocity sug cannot be measured with currently available techniques, gests a range of possible velocities between 0.2 and 0.5 including updraft velocity and rate of change of super m S-1 [Nicholls and Leighton, 1986]. The effect of this saturation for the time history of a single parcel or of an variation on the predicted cloud formation is shown in RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS 31,093 Relative Humidity (%) water contents than the downdraft regions with evapo rating and subsaturated droplets [Slevens el al., 1996]. 100.0 100.2 100.4 100.6 100.8 101.0 We have estimated thi s effect using an estimated up 450~--~--~.-----,----v.r---, • draft area fraction of 55% in the results in Table 2. The resulting predicted value is expected to represent t he measurement better since t he aircraft sampled both · • updraft and downdraft regions in unknown amounts . 400 Incorpor ating an explicit microphysics model in a LES · • h as been used by Kogan et al. [1995] to study the fine-sca le spatial structure of cloud properties using an internally mixed size-reso lved aerosol, but in th e ab sence of t his detail, we have estimated the sensitivity of cloud characteristics to variations in the area fraction that covers updrafts. While the maximum supersat \ urat ion a nd effective radius of droplets are essentially . unchanged by this parameter, the liquid water content \\~ and droplet number vary almost linearly. For instance, ", with an updraft fraction of 65% the cloud average liq '. ............ uid water content of both the background and the track " '.' ." ........ ~ conditions for the clean marin e case (JDT 180) increase \ . 3 3 to 0.23 g m- and 0.24 g m- , respectively, which are both within the standard deviations measured for the backgmund (0.28 g m - with a standard deviation of 3 3 0.05 g m- ) and the track (0.31 g m- with a standard deviation of 0.11 g m -3). 0.0 0.2 0.4 0.6 0.8 1.0 Liquid Water Content ( 3) 91m 6.4. Effect of Plum e Dilution Figure 18. As for Figure 7, but for three different updraft velocities. Thin gray solid line shows liquid Another consequence of t he one-dimensional ap water content , and thin g ray dashed line s hows .relative proach t hat we have employed here is that the rate of humidity in the background for the clean marme case dilution was not modeled ex plicit.ly and relied on est i (JDT 180) for the base case updraft velocity of w = 0.3 mates t.hat were insufficie ntly constrained by the obser m s-1. The thick solid line shows liquid water content, vat ions. The absence of explicit dilution of t he mea and the thick dash ed line shows relative humidity for sured initial size distributions may account for much the updraft velocity of w = 0.5 m s- l . The thin black of the overprediction observed in cloud droplet number solid line shows liquid water content, and th e t hm black dashed line shows relative humidity for the updraft ve concentrations in the track conditions. Given the uncer locity of w = 0 .2 m S - I. tainty o f both the peak supersaturation and t he degree of dilution , however , th ere were insufficient m easure m ents to independently constrain these two parameters. To quantify the potential effect of this uncerta in ty, we Figure 18 and is summarized in Table 3. The max have studied the impact of dilution of the plume to 50% imum supersaturation reached with an updraft veloc and 10% of its emitted particle and SO, concentrations. ity w = 0.5 m s-1 is increased to 0.91%, whereas for In th e clean marine case with only 10% of emissions the slower value of w = 0.2 m s-1 only 0.42% is pre (corresponding to an emitted particle concent ration of dicted. For low concentrat ions there is a significant im 1830 cm - ) the maximum su persaturation is 0.33%, pact on the maximum supersaturation reached in cloud} which is an increase of 0 .15% over the undiluted track although the small number ofvery efficient CCN present but still a significant drop from the background value in this case results in almost no change in duud droplet (0.68%). The updraft cloud droplet concentrat ion also number or liquid water content. The track prediction 3 decreases significantly to only 650 cm- , correspondi ng in the clean marine case shows a significa nt change in to an average cloud droplet concentration of 410 em-a, droplet number concentration and effective radius at an but the liquid water content is unchanged. updraft velocity of w = 0.5 m s-l but also shows an in Since t he in-track m easurements of cloud droplet con crease in cloud droplet number for the updraft velocity centrations s hown in Figures 4 and 12 were sampl ed at of w = 0.3 m s-1 since there is a longer time available p oints in t he tracks between 1 hour and 2.5 hours af to grow particles by condensation. ter emission, the plume concentrations are estimated to have been dispersed to between 50% and 10% of their 6.3. Effect of Updraft Area Fraction original values [Durkee el al., 2000c]. The magnitUde of Updraft regions with growing droplets will produce the dilution was not well-defined by the measurements, larger droplet distribut ions and high er average liquid but these calculations show that the uncertainty in the 31,094 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS degree of dilution is sufficient to account for much of sion suggests that the sulfur composition of fuels used the discrepancy between measured and predicted cloud in combustion processes has a direct impact on the indi characteristics. rect effect of these emitted particles on clouds. Further work is needed to determine if nitrogen oxides or or ganic species would replace the role of sulfur oxides in 7. Aerosol Processes III Ship Tracks the event that fuel sulfur were reduced. Aerosol particles provide additional CCN that acti vate in ship tracks resulting in enhanced droplet con Acknowledgments. This analysis was supported by centrations and decreased mean drop size [Durkee et NSF grant ATM-9732949 and ONR grant N00014-97-1- ai., 2000b]. Multiple processes contribute to the char 0673. The aerosol measurements on which this work was based were supported by ONR grant N00014-93-1-0872. acteristics of these CCN, from the point of their forma Measurements by several coauthors on this work were sup tion from vapors in the combustion process and subse ported under the ONR Accelerated Research Initiative enti quent emission from the ship stack to the atmosphere. tled Surface Ship Cloud Effects. This work would not have Condensation and coagulation contribute to the growth been possible without the cooperation and support of the of some particles and the removal of others. Aerosol crew of the University of Washington C-131A aircraft and of the RAF crew of the MRF C-130 aircralt. The authors growth by either condensation or coagulation has the express their gratitude to all members of the MAST Science effect of adding water-soluble mass to emitted aerosol Team. In addition the authors appreciate the comments of that makes them efficient as CCN at the effec nuclei Doug Johnson, Bjorn Stevens, and two anonymous review tive 5upersaturations expected for marine stratocumu ers who provided helpful suggestions that have improved this work. Ius. Condensation of H S0 from gas-phase oxidation 2 4 of S02 is primarily responsible for adding soluble sul fate mass to the newly emitted black and organic carbon References particles. Once these particles are activated) additional Bates, T.S., R.J. Charlson, and R.H. Gammon, Evidence for contributions from the aqueous oxidation of 80 on ac the climatic role of marine biogenic sulfur Nature, 329, tivated droplets also grow particles. Some activated and 319-321, 1987. some interstitial aerosol particles are lost by coagula Bates, T.S., J.E. Johnson, P.K. Quinn, P.D. Goldan, W.C. tional processes in cloud resulting in a shift in the size Kuster, D.C. Covert, and C.J. Hahn , The biogeochemical sulfur cycle in the marine boundary layer over the North spectrum to larger particles. Whether there is enough east Pacific Ocean, J. Atmas. Chern., 10,59-81, 1990. residence time for particles in this type of cloud cycle Bott , A., A positive definite advection scheme obtained by for particles to grow sufficiently large by both condensa nonlinear renormalization of the advective fluxes, Mon. tion and coagulation to precipitate determines whether Weather Rev" 117, 1006-1015 1989. ship tracks will have feedback effects on both rain and de Bruyn. W.J., T.S. Bates, J.M. Cainey, and E.S. Saltz man, Shipboard measurements of dimethyl sulfide and boundary layer structure. S02 southwest of Tasmania during the First Aerosol With the exception of the limited information on Characterization Experiment (ACE 1), J. Geaphys. Res., chemical composition the above uncertainties primar 103, 16703-16711, 1998. ily interfere with our ability to predict the evolution of Dhaniyala, S., and A.S. Wexler, Numerical schemes to model ship tracks rather than their initial formation. While condensation and evaporation of aerosols, Atmas. Envi ron., 30 919-928 1996. we have initial evidence that gas-to-particle conversion , , Durkee, P.A., K.J. Noone, and R.T. Bluth, The Monterey is not likely to control the evolution of ship tracks af Area Ship Track (MAST) experiment J. Atmas. Sci., in ter formation due to removal of the bulk of condens press, 2000a. able species in the first cloud cycle, uncertainties in ox Durkee, P.A., et al., The impact of ship produced aerosols on idant concentrations as well as in thermodynamic pro the microphysical characteristics of warm stratocumulus files of subsequent cloud cycles make this conclusion clouds: A test of hypotheses 1.1a and 1.1b 1. Atmas. Sci., in press, 2000b. tentative until additional measurements and modeling Durkee, P.A" R.E. Chartier , A. Brown, E.J. Trehubenko is performed. S.D. Rogerson, C. Skupniewicz, K.E. Nielsen, S. Platnick, Gas-to-:particle conversion of S02 is necessary for ship and M.D. King, Composite ship track characteristics , J. track formation in the clean marine case studied here. Atmas. Sci., in press, 2000c. It is not clear how much of this conversion occurs be Erlick, C., L.M. Russell, and V. Ramaswamy, A microphysics based investigation of the radiative effects of aerosol-cloud fore emission from the stack and how much occurs after interactions for two MAST Experiment case studies J. emission due to the sampling constraints on particle Geophys. Res., in press, 2000. composition and the lack of detailed, direct stack mea Feingold, G" S.M. Kreidenweis, and Y. Zhang, Stratocumu surements of emissions. However 1 without any contri Ius processing of gases and cloud condensation nuclei J. bution from sulfate, the insoluble nature of both black Geophys. Res., 103, 19,527-19,542, 1998. Ferek, R.J., D.A. Hegg, P.V. Hobbs, P.A. Durkee, and KE. carbon and of the slightly soluble organic carbon frac Nielsen, Measurements of ship-induced tracks in clouds tion dictates that the plume particles are too small and off the Washington coast , 1. Geophys. Res., 103, 23,199- too nonhygroscopic to serve as CCN for the maximum 23,206, 1998. supersaturations of from 0.15% to 0.68% predicted for Flossmann, A.I. , W.D. Hall, and H.R. Pruppacher, A theo clouds sampled in the MAST experiment. This conclu- retical study of the wet removal of atmospheric pollutants, 31,095 RUSSELL ET AL.: AEROSOL DYNAMICS IN SHIP TRACKS part I, The redistribution of aerosol particles captured Richards, L.W., J.A. Anderson~ D.L. Blumenthal, J,A, Mc through nucleation and impaction scavenging by growing Donald, G.L. Kok, and A,L. Lazrus, Hydrogen peroxide cloud drops, J. Atmos. Sci., 42, 582-606, 1985. and sulfur (IV) in Los Angeles cloud water, Atmos. Env Gasparovic, R.F., MAST Experiment Operations Summary, iron.~ 11,911-914, 1983. Johns Hopkins Univ., Maryland, 1995. Russell, L.M., and J.R. Seinfeld, Size- and composition Hildemanu, L.M., G.R. Markowski and G.R, Cass, Chemical resolved externally- mixed aerosol model, Aerosol Sci. Composition of Emissions from Urban sources of fine or Technol.} 28} 403-416, 1998. ganic aerosol, Environ. Sci. Technol., 25, 744-759,1991. Russell, L.M., S.N. Pandis, and J ,H. Seinfeld, Aerosol pro Hobbs, P.V., et al., Emissions from ships with respect to duction and growth in the marine boundary layer, J. Geo their effects on clouds, J. Atmos. Sci., in press, 2000. phys, Res., 99} 20,989-21,003, 1994. Hoppel, W.A., J.W. Fitzgerald, G.M. Frick, and R.E. Lar Russell, L.M., R.C. Flagan, and J .B, Seinfeld, Asymmetric SOll, Aerosol size distributions and optical properties instrument response due to mixing effects in DMA-CPC fOlUld in the marine boundary layer over the Atlantic measurements, Aerosol Sci. Technol.} 23, 491-509, 1995. Ocean, J. Geophys. Res" 95, 3559-3886, 1990. Russell, L,M., M,R. Stolzenburg, S.H. Zhang, R. Caldow, R,C, Flagan, and J ,R. Seinfeld, Radially-classified aerosol Hudson, J.G., T.J. Garrett, P.V. Hobbs, S.R. Strader, Y. Xie, and S.S. Yum, Cloud condensation nuclei and ship detector for aircraft-based submicron aerosol measure track clouds, 1. Atmos. Sci., in press, 2000. ments. J, Atmos. Oceanic Technol., 13, 598-609, 1996. Huebert, B.J., S.G. Howell, L. Zhuang, J.A. Heath, M.R. Seinfeld, J,R" and S,N, Pandis, Atmospheric Chemistry and Litchy, D.J. Wylie, J .L. Kreidler-Moss, S. Coppicus, and Physics, John Wiley, New York, 1998. J.E. Pfeiffer, Filter and impactor measurements of anions Stevens, B., G, Feingold, Vi,T.R. Cotton, and R.L. Walko, El and cations during the First Aerosol Characterization Ex ements of the microphysical structure of numerically sim periment (ACE 1), 1. Geophys. Res.~ 103,16,493-16,509, ulated nonprecipitating stratocumulus, J, Atmos. Sci" 1998. 53, 980-1006, 1996. Jacobson, M.Z., Numerical techniques to solve condensa Tzivion, S" G. Feingold, and Z, Levin, An efficient numerical tional and dissolutional growth equations when growth is solution to the stochastic collection equation, 1. Atmos. coupled to reversible reactions, Aerosol Sci. Technol., 4, Sci., 44, 3139-3149, 1987. 491-498, 1997. Weber, R,J" P.B. McMurry, F.L. Eisele, and D.J. Tanner, Jacobson, M.Z., R.P. Turco, E.J. Jensen, and O.B. Toon, Measurement. of expected nucleation precursor species Modeling coagulation among particles of different compo and 3-500-nm diameter particles at Mauna Loa Obser sition and size, Atmos. Environ., 28, 1327-1328, 1994, vatory, Hawaii, ], Atmos. Sci., 52, 2242-2257, 1995. Jennings, S.G., and C,D, O'Dowd, Volatility of aerosol at Wexler, A.S., and J.B. Seinfeld, The distribution ofamm Mace Head, on the west-coast of Ireland, !. Geophys, nium salts among a size and composition dispersed Res., 95, 13,937-13,948, 1990. aerosol, Atmos. Environ., 24(A), 1231- 1246, 1990. Kaufman, Y.J" and D, Tanre, Effect of variations in super saturation on the formation of cloud condensation nuclei, Nature, 369, 45-48, 1994, L.M. Russell, Department of Chemical Engineering, A317 Engineering Quadrangle, Princeton' University, Princeton, Kogan, Y.L., M.P. Khairoutdinov, D.K. Lilly, Z.N. Kogan, and Q. Liu, Modeling of stratocumulus cloud layers in a NJ 08544. ([email protected]) large eddy simulation model with explicit microphysics, R,C, Flagan and J.H, Seinfeld, Department of Chem 1. Atmos. Sci., 52, 2923-2940, 1995. ical Engineering, California Institute of Technology, Pasadena, CA 91125. ([email protected], sein Kulmala, M., and A. Laaksonen, Binary nucleation of water sulfuric acid system: Comparison of classical theories [email protected]) with different H S0 saturation vapor pressures, J, Chern. R.J. Ferek, Office of Naval Research, 800 N, Quincy St., 2 4 Arlington, VA 22217. ([email protected],mil) Phys., 93,696-701, 1990. D.A. Hegg and P.V. Hobbs, Department of At de Laat, A.T.J., and P.G. Duynkerke, Analysis of ASTEX mospheric Sciences, University of Washington, Seat stratocumulus observational data using a mass-flux ap tle, WA 981950, (deanhegg@atmos, washington.eduj proach, Boundary Layer Meteorol., 86, 63-87, 1998. [email protected]) Murphy, D.M., D.S. Thomson, A.M. Middlebrook, and M.E. P,A. Durkee and K.E. Nielsen, Department of Meteorol Schein, In situ single-particle characterization at Cape ogy, Code MR/De,Naval Postgraduate School, 589 Dyer Grim, 1. Geophys. Res., 105, 16,485-16,491, 1998. Rd., Rm. 254, Monterey, CA 93943, ([email protected], Nicholls, S., and J. Leighton, An observational study of the kenielsen@ nps.navy.mil) structure of stratiform cloud sheets, part I, Structure, Q. A,I, Flossmann and W, Wobrock, Universite Clermont J. R. Mdeorol. Soc,} 112} 431-460, 1986. Ferrand, Laboratoire Meteorologie Physique, CNRS, 24 Noone, K,J., et al., A case study of ship track formation in a Ave. Landais, Clermont..,Ferrand F-63177 Aubiere, France. polluted boundary layer, J. Atmos. Sci., in press, 2000a. ([email protected]) Noone, K.J" et aL, A case study of ships forming and not C,D. O'Dowd, University of Sunderland, School of the forming tracks in a moderately polluted boundary layer, Environment, Center for Marine and Atmopsheric Sciences, J. Atmos. Sci., in press, 2000b. Benedict Bldg., St. Georges Way, Sunderland SR2 7BW, Panclis, S.N., L.M. Russell, and J .R, Seinfeld, The relation Durham, England. (colin.odowd@cmas,demon.co.llk) ship between DMS flux and CCN concentration in remote marine regions, J. Geophys, Res.} 99~ 16945-16958, 1994. Quinn, P.K., T.S. Bates, J.E. Johnson, D.S. Covert, and R.J, Charlson, Interactions between the sulfur and re duced nitrogen cycles over the Central Pacific Ocean, J. (Received May 13, 1999; revised September 7, 1999; Geophys. Res., 95, 16,405-16,416~ 1990. accepted September 15, 1999.)
Journal
Journal of Geophysical Research Atmospheres – Unpaywall
Published: Dec 1, 1999