UBC Faculty Research and Publications

Modeled Downward Transport of a Passive Tracer over Western North America during an Asian Dust Event.. Hacker, Joshua P.; McKendry, Ian G.; Stull, Roland B. 2001-09-30

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


52383-Stull_AMS_2001_JAM1617.pdf [ 976.08kB ]
JSON: 52383-1.0041828.json
JSON-LD: 52383-1.0041828-ld.json
RDF/XML (Pretty): 52383-1.0041828-rdf.xml
RDF/JSON: 52383-1.0041828-rdf.json
Turtle: 52383-1.0041828-turtle.txt
N-Triples: 52383-1.0041828-rdf-ntriples.txt
Original Record: 52383-1.0041828-source.json
Full Text

Full Text

SEPTEMBER 2001 1617HACKER ET AL.q 2001 American Meteorological SocietyModeled Downward Transport of a Passive Tracer over Western North Americaduring an Asian Dust Event in April 1998JOSHUA P. H ACKER,IAN G. MCKENDRY, AND ROLAND B. STULLAtmospheric Science Program, Department of Geography and Department of Earth and Ocean Sciences,The University of British Columbia, Vancouver, British Columbia, Canada(Manuscript received 31 July 2000, in final form 9 December 2000)ABSTRACTAn intense Gobi Desert dust storm in April 1998 loaded the midtroposphere with dust that was transportedacross the Pacific to western North America. The Mesoscale Compressible Community (MC2) model was usedto investigate mechanisms causing downward transport of the midtropospheric dust and to explain the highconcentrations of particulate matter of less than 10-mm diameter measured in the coastal urban areas of Wash-ington and southern British Columbia. The MC2 was initialized with a thin, horizontally homogeneous layer ofpassive tracer centered at 650 hPa for a simulation from 0000 UTC 26 April to 0000 UTC 30 April 1998. Modelresults were in qualitative agreement with observed spatial and temporal patterns of particulate matter, indicatingthat it captured the important meteorological processes responsible for the horizontal and vertical transport overthe last few days of the dust event. A second simulation was performed without topography to isolate the effectsof topography on downward transport.Results show that the dust was advected well east of the North American coast in southwesterly midtroposphericflow, with negligible dust concentration reaching the surface initially. Vertically propagating mountain wavesformed during this stage, and differences between downward and upward velocities in these waves could accountfor a rapid descent of dust to terrain height, where the dust was entrained into the turbulent planetary boundarylayer. A deepening outflow (easterly) layer near the surface transported the tracer westward and created a zonal-shear layer that further controlled the tracer advection. Later, the shear layer lifted, leading to a downwardhydraulic acceleration along the western slopes, as waves generated in the easterly flow amplified below theshear layer that was just above mountain-crest height. Examination of 10 yr of National Centers for EnvironmentalPrediction–National Center for Atmospheric Research reanalyses suggests that such events are rare.1. IntroductionAir pollution in western North America has generallybeen considered to be of regional scope and local origin.However, recent observations show that 1) anthropo-genic pollutants from rapidly industrializing east Asiareach western North America (Jaffe et al. 1999; Bernt-sen et al. 1999) and 2) mineral dust from semiarid re-gions of Asia, long known to influence the north Pacificincluding Hawaii (Duce et al. 1980; Merrill et al. 1989),can also reach North America (Husar et al. 2001, here-inafter HUS; McKendry et al. 2001, hereinafter MCK).The potential impact of long-range transport of Asianpollutants to North America was highlighted during lateApril 1998, when mineral dust from a severe dust stormin the Gobi Desert of western China was rapidly trans-ported across the north Pacific to create high surfaceconcentrations of particulate matter of less than 10-mmdiameter (PM10) that led to reduced visibility acrossCorresponding author address: Joshua P. Hacker, University ofBritish Columbia, Department of Earth and Ocean Sciences, 6339Stores Road, Vancouver, BC V6T 1Z4, Canada.E-mail: jhack@eos.ubc.cawestern North America for several days (HUS). Jaffeet al. (1999) document anthropogenic pollutants fromAsia reaching the Pacific Northwest, but this was thefirst documented case of mineral dust reaching NorthAmerica from Asia (although there is ice-core evidenceof Asian dust reaching Greenland in the past; Biscayeet al. 1997). This spectacular and unusual event raisedquestions about mechanisms of free-tropospheric–plan-etary boundary layer (FT–PBL) exchange that wouldpermit mineral aerosol transported in the middle andlower troposphere to be subsequently mixed to the sur-face in high concentrations (MCK). If mineral aerosolcould be entrained into the PBL over western NorthAmerica, then burgeoning east Asian anthropogenicemissions could suffer the same fate.Although processes by which pollutants from the PBLfind their way into the FT are well known (McKendryand Lundgren 2000), processes by which pollutants aretransported downward to where they can be mixed backinto the PBL after long-range transport are less welldescribed. In this case study, the Asian dust event ofApril 1998 provides the motivation to examine mech-anisms of FT–PBL exchange over mountainous western1618 VOLUME 40JOURNAL OF APPLIED METEOROLOGYFIG. 1. Observed distribution of 24-h average ground-level PM10concentrations for 29 Apr 1998 over western North America. Thesize of the boxes is proportional to the observed dust concentrationsat the box location.North America. A numerical model with atmospherictracer capability is utilized to identify important factorscontributing to this event. Distinct spatial and temporalpatterns of surface PM10concentrations give evidencethat the model is capturing the important dynamics as-sociated with the downward transport of the dust. Thenext section is a summary of the Asian dust event, andsection 3 describes the experimental approach. Resultsare presented in section 4, including a brief comparisonof modeled surface tracer distribution with the observeddust and an analysis of the transport mechanisms. Asummary is presented in section 5.2. The 1998 Asian dust eventA detailed examination of the 1998 dust event is pro-vided in HUS, MCK, and Tratt et al. (2001, hereinafterTFW). In summary, during April 1998, several unusu-ally intense dust storms were generated over the Gobi.These events are common during spring and produce amineral aerosol rich in Fe, Si, Ca, and Al (Braaten andCahill 1986). An unusually intense storm on 19 Aprilproduced a dust cloud that was rapidly transportedacross the Pacific. The dust cloud was tracked in satelliteimagery and observed by LIDAR on both sides of thePacific (TFW; Murayama et al. 2001, hereinafter MUR).Ground-level concentrations of PM10were observed topeak on 28–29 April from British Columbia to Cali-fornia, with 24-h average concentrations exceeding 100mgm23in parts of Washington on 29 April. The dustchemical composition over North America bore the sig-nature of Asian dust (McKendry et al. 2000), was uni-form, and had a volume mean diameter of 2–3 mm. Thecombination of an unusual dust storm in Asia and me-teorological conditions conducive to the rapid transportacross the Pacific is rare, as will be shown.Observations from sparsely distributed routine mon-itoring stations provide the only information on the spa-tial and temporal patterns associated with the arrival ofAsian dust in the PBL over western North America. InFig. 1, the distribution of daily PM10concentrations isshown for 29 April 1998, the peak of the event. Contoursin Fig. 1 were generated using an objective analysisscheme with weights determined by inverse squared dis-tance. A zone of high concentrations (50–100 mgm23)was observed in northern Oregon, central Washington,and southern British Columbia. It is important to rec-ognize that these PM10observations represent a com-bination of particulate matter of local origin and Asiandust. MCK estimate that up to 50% of PM10in the LowerFraser Valley of southwestern British Columbia duringthe event was of Asian origin, while HUS estimate thatAsian dust contributed about 40 mgm23greater thanthe local regional PM10average of 10–25 mgm23forApril–May 1998.Temporal variations during the event are not pre-sented here because most stations do not monitor PM10on a daily basis (approximately 30 of 230 western U.S.stations sample on a daily basis; the remainder sampleevery sixth day). However, in southern British Colum-bia, MCK note that observed concentrations peaked inthe southern central interior of the province first (28April) and later shifted westward. On Vancouver Island,concentrations were highest on 30 April, and in theLower Fraser Valley, they reached a maximum on 29April. A similar pattern was observed in Washington(HUS). A second feature of the event was a strong di-urnal variation in concentrations characterized by high-est surface concentrations during the afternoon and low-est concentrations at night. This feature was observedas far south as California (HUS).3. Modeling environmenta. MC2 characteristics and initial conditionsMeteorological transport processes associated withthe April 1998 event were examined with the CanadianMesoscale Compressible Community (MC2) atmo-spheric model (Benoit et al. 1997). In the MC2, advec-SEPTEMBER 2001 1619HACKER ET AL.FIG. 2. Map of region with model grids. The number of grid pointsin each domain: outer domain is 65 3 70, Dx 5 90 km; middledomain is 75 3 85, Dx 5 30 km; and inner domain is 151 3 101,Dx 5 10 km.tion is semi-Lagrangian with a semi-implicit time step,facilitating straightforward tracking of atmospheric trac-ers. The Recherche en Pre´vision Nume´rique full physicspackage is used, which is similar to that used by theCanadian Meteorological Centre (CMC) in their suiteof operational models. The force–restore method (Dear-dorff 1978) describes surface exchanges, and a 1.5-orderturbulence closure (retaining a turbulent kinetic energypredictive equation, Benoit et al. 1989) parameterizesvertical turbulent diffusion. Surface values are specifiedat shelter height (2 m) and anemometer height (10 m)using a surface-layer model based on similarity theory.The MC2 uses a one-way (cascade) nesting strategy, inwhich coarser grid simulations provide initial andboundary conditions for finer grid simulations, and thereis no upscale feedback. A Dx 5 90 km grid simulationcovering a large domain is initialized from 0000 UTCanalyses. With this coarse mesh output, a Dx 5 30 kmgrid initialized at 0600 UTC, which in turn drives a Dx5 10 km grid initialized at 1200 UTC for the smallestdomain. Model domains are shown in Fig. 2. The 6-hoffset between grid initialization is chosen to allow in-ertial gravity waves to disperse from the coarser gridbefore subsequent initialization of the finer grids.To examine meteorological mechanisms by whichmineral aerosol was transported to the PBL in the LowerFraser Valley (Vancouver area), the MC2 was modifiedto advect and diffuse a passive tracer. A Dx 5 90 kmgrid simulation was initialized 0000 UTC 26 April 1998,and 30-, and 10-km grids were run beginning 6 and 12h later, respectively. The 96-h simulation ending at 0000UTC 30 April used grids from the National Centers forEnvironmental Prediction (NCEP) Eta Model analysesfor initial and boundary conditions. Vertically, the tracerwas initially distributed as a 100-hPa deep (;1000 m)single layer, with a concentration maximum at 650 hPa.Specification of tracer depth was based on 1) serendip-itous aircraft observations near Seattle on 27 Aprilshowing a distinct dust layer at about 2–3 km altitudeand virtually no dust below (HUS) and 2) a short sen-sitivity experiment showing that a realistic surface-layertracer distribution is modeled with a tracer initialized atthat altitude (section 3b). To ensure that surface con-centrations result from dominant downward-transportmechanisms, the layer used in this simulation is higherthan the aircraft observed. Lidar backscatter data takenover Pasadena on 27 April (TFW) showed additionallayers at higher altitudes. Multiple layers were excludedfrom the simulation because of the absence of completetropospheric aerosol profile data over Washington/Brit-ish Columbia with which to confirm the presence ofmidtropospheric dust and the sensitivity analysis dem-onstrating higher layers would contribute less to ob-served surface concentrations. Last, the simulated dustwas distributed homogeneously horizontally and wascontinuously supplied at the boundaries of the modeldomain. To isolate the impact of the terrain on down-ward transport of tracer, simulations with both realisticterrain and flat terrain were conducted.b. Meteorological and tracer distribution uncertaintyTo evaluate the mechanisms responsible for high trac-er concentration at ground level, a model simulationmust capture the important features of both the mete-orology and the surface tracer distribution. In a limited-area domain, the solution is strongly controlled by theboundary conditions (e.g., Warner et al. 1997). Giventhat the boundary conditions were analyses rather thanforecasts, this simulation never deviates substantiallyfrom the analysis. This is no doubt aided by the factthat advective speeds were large for this case, renderinginternal model physics relatively less important.The simulation is a tool to understand some phenom-ena responsible for the observed surface dust distribu-tion, and it is unnecessary to reproduce the event indetail. If modeled tracer resembles surface observations,the model is likely reproducing dynamics similar towhat actually occurred. Because of scant dust plumeobservations, some arbitrariness is unavoidable inchoosing the initial three-dimensional tracer distribu-tion. Observations show that the dust plume generallytracked over the Pacific Northwest at upper levels justbefore and while concentrations peaked at the surface(HUS). Earlier experiments (MCK) used an initial hor-izontal distribution mimicking satellite imagery thatshowed the plume over the Pacific to be located in themidtropospheric flow maximum. Simulated surface trac-er distributions were reasonable only in localized re-gions, and a tracer that was not perfectly collocated withthe proper meteorological conditions did not reach thesurface or was dispersed aloft. Because the goal here isnot to perfectly reproduce surface tracer concentrationsbut rather to examine what is responsible for its arrival1620 VOLUME 40JOURNAL OF APPLIED METEOROLOGYFIG. 3. Initial tracer profiles for the four runs of the sensitivitystudy.FIG. 4. Surface tracer concentration for the four runs corresponding to the initial profilesshown in Fig. 3, valid 1200 UTC 29 Apr 1998. Units are relative.at the surface, the attempt at simulating a realistic hor-izontal distribution aloft was abandoned. Initializingwith a horizontally homogeneous tracer distribution re-moves any assumptions about the horizontal distributionaloft and restricts our attention to vertical and low-levelhorizontal transport that occurs when the tracer is in thePBL. By maintaining the same profile at the boundariesthroughout the simulation, the restriction maintains itsintegrity.An extensive body of work examining haze layersover southern Africa (Garstang et al. 1996; Tyson et al.1997; Piketh et al. 1999) links the vertical distributionof atmospheric aerosols to the thermal structure of thelower troposphere. Namely, aerosols concentrate in ab-solutely stable layers, and extensive vertical mixing gen-erally occurs with deep convection that penetrates anddestroys the stable layers. This phenomenon could ac-count for the multipeak structure observed and reportedin HUS, even though this case is in a synoptically activeregime and the southern African work primarily ex-amines persistent anticyclonic conditions. If stable lay-ers develop in the MC2 simulation, the modeled tracershould also concentrate in these layers, and it is un-necessary to initialize with a tracer profile consistentwith the thermal structure. The lead time of 2 or 3 daysbetween model initialization and tracer arrival at thesurface is sufficient to allow the model physics to ver-tically distribute the tracer according to the stability.The initial tracer should be placed high enough sothat it will only reach the surface where the modeledvertical transport is large but low enough so that theamount of tracer reaching the surface can be qualita-tively compared with observations. Results from foursimulations, each initialized with different profiles (Fig.3), demonstrate how the surface tracer distributions varywith the initial height of the tracer layer (Fig. 4). Asmay be expected, a higher initialized layer results inlower simulated surface concentrations. The location ofSEPTEMBER 2001 1621HACKER ET AL.FIG. 5. Synoptic conditions during the dust event: (a) 1200 UTC 26 Apr 1998 and (b) 1200 UTC 29 Apr 1998. Solid linesare 500 hPa geopotential height every 6 decameter (dam) and dashed lines are mean sea level pressure every 4 hPa.the maxima over the Vancouver area and south of PugetSound does not change between runs 2 and 4. The quan-tities of tracer reaching the surface in run 4 were deemedsubstantial for evaluation in more detail. Besides sat-isfying the criteria, Run 4 is also initialized with a layerthat is higher than any of the lowest observed dust layers(HUS; TFW). Thus, surface concentrations are conser-vative in the sense that the height of origin is not aslow as the observations allow.4. Simulation resultsa. Synoptic backgroundSynoptic conditions, as simulated by the MC2, areshown in Fig. 5. A 500-hPa ridge and a surface highpressure center persisted throughout the period. Geo-potential heights (500 hPa) over the coast increased asthe ridge amplified, and the strongest flow aloft movednorthward. Simultaneously, the ridge axis moved slight-ly westward. During the period of the simulation, a sur-face anticyclone propagated eastward and strengthened.When surface isobars are broadly parallel to the coastover Washington and southern British Columbia (Fig.5b), continental air generally flows through gaps in themountain ranges and spills down western slopes. Suchevents are known as ‘‘outflow’’ in British Columbia andusually occur during winter months (Jackson 1996).They can cause strong gap-flow events (Jackson andSteyn 1994) and can be responsible for cold-air out-breaks along the coast and significant airborne transportof local crustal material out of the Fraser Valley in south-ern British Columbia (McKendry 2000). The outflowduring this April 1999 event was weak as comparedwith the much more intense winter events; nonetheless,it was important for dust transport, as will be shown.b. Surface tracer distributionSimulated surface-layer tracer concentrations and sur-face-layer winds at 30-h intervals are shown in Fig. 6.Initial maximum concentrations in the layer at 650 hPawere scaled to have a dimensionless value of 1000.Thus, the surface values in Fig. 6 are relative to thismaximum tracer value.Tracer at ground level was first apparent in the south-eastern portion of the simulation domain on the morningof 27 April (Fig. 6a). By midday on 28 April, the highestground-level concentrations were along the Washington/British Columbia interior border and to the south ofPuget Sound in southwestern Washington. These twozones correspond with the highest average elevations inthe Cascade Range. By this time, outflow had strength-ened considerably at the surface and continued to do sothroughout the simulation. By late afternoon on 29 April(0000 UTC 30 April; Figs. 6e,f) surface tracer concen-trations had increased in magnitude and spread hori-zontally across a broad zone including southern Van-couver Island. A second zone of high surface concen-trations also formed on the lee (east) side of the RockyMountains, bordering British Columbia and Alberta. Fortracer mixed downward in the vicinity of the Cascades,simulated outflow winds were responsible for advectionof the near-surface tracer westward and ultimately off-shore.Sudden arrival of the tracer at the surface in a zoneextending from immediately north of the Washington/British Columbia border southward through westerncentral Washington and near-surface westward transportof the tracer are roughly consistent with observationsdescribed by MCK and shown in Fig. 1. These results,combined with the fact that the simulated flow largelyreproduces the analyses, suggest that the simulation cap-tures the salient meteorological mechanisms responsiblefor bringing the dust downward and advecting it towardthe coastal urban areas.Comparing the modeled surface tracer concentrationwith the observations simply provides confidence thatthe simulation is capturing the gross features evident inthe observations and therefore some of the importantdynamics of the event. The modeled tracer concentra-1622 VOLUME 40JOURNAL OF APPLIED METEOROLOGYtions are instantaneous values, and the observations inFig. 1 are a 24-h average, making a direct, quantitativecomparison impossible. It is true that the simulated max-imum concentration, just west of southern VancouverIsland, appears to be farther west than the observed peakin Fig. 1. This may result from the tracer being advectedtoo quickly or a lack of any tracer deposition in theMC2. A maximum also could have actually occurredoffshore, but it was not observed. Even in the worstcase that the tracer advected too quickly, the analysisof the vertical and horizontal transport of the tracer isstill valid.c. Vertical transport processesIn Fig. 7, a west–east transect through the LowerFraser Valley illustrates processes contributing to thedownward transport of the tracer. In this cross section,the Pacific Ocean is west of 1268W, Vancouver Islandis at 1258W, and Vancouver is at approximately123.258W longitude. Georgia Strait lies in the flat areabetween 1238 and 1248W.Flow aloft during the dust event was broadly per-pendicular to the ridge lines of the Cascades, the BritishColumbia Coast Mountain Range, and the Rocky Moun-tains. At Dx 5 10 km, the MC2 simulated strong moun-tain wave activity aloft (Figs. 7b,d,f) and a shear layernear crest height between the developing easterly out-flow and predominately southwesterly flow aloft (Figs.7a,c,e). Aloft, the slow descent of an eastward-movingtracer (shaded in left panels) offshore throughout thesimulation is indicative of large-scale subsidence.Southwesterly winds remained constant in the upstreambranch of the ridge over British Columbia at 500 hPa,while wind speeds decreased slightly as the ridge am-plified and the jet core moved northward (Fig. 5). Nearthe surface, developing outflow conditions were evidentas a deepening layer with an easterly wind component.As the outflow deepened, the directional shear layerbetween it and the southwesterlies aloft rose from ap-proximately 850 to about 750 hPa at the end of thesimulation. By 1800 UTC 28 April (Fig. 7c), the tracerover the mountains was subject to much stronger sub-sidence than over the ocean. Near the surface, the tracerwas advected westward through the coastal valleys,reaching Vancouver a few hours later. A large proportionof the tracer was able to reach the surface.Stable stratification and weak shear in the southwes-terlies aloft allowed vertically propagating mountainwaves to amplify up to about 500 hPa (Fig. 7b,c,f).Vertical velocities generated through a deep layer wereon the order of 0.1 m s21, with the downward velocitiesas much as 0.1 m s21greater than upward velocities.The zonal wavelength was approximately 45 km, andwind speeds at 500 hPa were about 10 m s21. Underthese conditions, an air parcel would travel one wave-length in 1.25 h, during which time it would descend450 m. Net downward transport by these waves rapidlyenabled the eastward-travelling tracer to be interceptedby the PBL over higher topography and the inland val-leys, despite the likelihood that the additional subsi-dence limited the depth of the PBL. Outflow in the PBLthen transported it coastward. Later in the simulation(Fig. 7f), mountain wave activity aloft abated. In theeasterly flow near the surface, isentropes deflect down-ward on the western slopes, a feature that grew morepronounced later in the simulation, suggesting strongerdownward motion here as the simulation progressed.A parallel simulation without terrain yields a first-order approximation of the effect of the western cor-dillera on downward transport (Fig. 7g,h). Though notshown, the long-wave pattern aloft was not greatly af-fected by the exclusion of terrain, and large-scale sub-sidence was similar for both simulations. However, by1800 UTC 28 April in the flat terrain case, the absenceof the tracer at low levels and isentropes that did notlower significantly (as in the real terrain case) indicatethat total subsidence was less. Variations in vertical ve-locities across the transect and perturbations to the is-entropes in the flat terrain case (Fig. 7g,h) can be at-tributed to differential heating over Vancouver Island(1258W) and the mainland coast (1238W).Figure 8 shows average vertical velocities over theeastern portion of the model domain and provides fur-ther evidence of the importance of terrain-induced waveactivity in transporting dust downward. With realisticterrain, vertical velocities at 600 hPa were negativethroughout the simulation and displayed a significantdiurnal fluctuation characterized by the strongest down-ward velocities during morning. With no terrain, verticalvelocities were of smaller magnitude and ranged be-tween net upward and downward motion. The same di-urnal fluctuation is evident in this case and suggests thatthermal forcing is responsible for the diurnal periodicityin the vertical motion field. MCK noted a diurnal var-iation in observed dust concentrations across the region,with maxima during the day. On the basis of these modelresults, early morning downward transport and then en-trainment and advection within the boundary layer like-ly contributed to observed daytime peak ground-levelconcentrations. Nocturnal deposition processes mayalso have contributed to the observed diurnal modula-tion of surface concentrations (HUS).d. Stability and shear effectsBulk Richardson number (RiB) calculations along thesame west–east transect as Fig. 7 illustrate the potentialfor turbulence in the PBL from shear and mountaineffects (Fig. 9). Although turbulence occurs when thegradient Richardson number (Ri) is less than a criticalvalue of 0.25, such is not the case when using RiB. Thewind and temperature differences across thick layers areinadequate to resolve thin turbulent layers. Thus, thereis a probability of turbulence even when RiB. 0.25(Stull 1988). Based on Stull’s criteria, the shear envi-SEPTEMBER 2001 1623HACKER ET AL.FIG. 6. Surface-layer concentrations (relative) of simulated tracer dust (left column) and topography (shaded) withsurface-layer wind vectors (m s21, right column). Times are shown above each plot. The middle (Dx 5 30 km) gridis shown here.ronment in this case would indicate a high probabilityof turbulence with RiB, 1.0, with decreasing but non-zero probability of turbulence up to RiB5 10.0.The turbulent environment during the morning andafternoon of 29 April suggests that turbulence generatedby the flow over the mountains was primarily respon-sible for the tracer entering the low-level outflow. TheRiBin the shear layer between the southwesterlies aloftand the outflow are between 0.25 and 5.0. Although theband of low RiBloses its coherence over the mountains,the PBL there is visible both in the early morning andin the afternoon. Absence of a significant diurnal pe-riodicity in RiBsuggests that wind shear over the com-plex terrain was primarily responsible for generatingturbulence rather than convection from surface heating.However, with the addition of daytime heating, RiBde-1624 VOLUME 40JOURNAL OF APPLIED METEOROLOGYFIG. 7. Zonal cross section (49.278N) for the Dx 5 10 km simulation valid 1200 UTC 29 Apr 1998. The left columnshows relative tracer concentration (shaded) and wind barbs (m s21), and the right column shows vertical velocity (ms21) (shaded) and potential temperature (K). The times correspond to Fig. 6. (g), (h) The same as (c) and (d) for thesimulation done without topography.SEPTEMBER 2001 1625HACKER ET AL.FIG. 8. Mean vertical velocity in the eastern half of the Dx 5 10km domain, at 650 hPa. Solid line shows results from simulation withtopography, and dashed line shows results from simulation withouttopography.FIG. 9. Bulk Richardson number along the same cross section shown in Fig. 7: (a) 1200 UTC 29 Apr and (b) 0000UTC 30 Apr 1998. The turbulent PBL between 1000 and 975 hPa is not adequately represented because of thecalculation method.creased in the afternoon, particularly over VancouverIsland (1258W) and the first ridge on the mainland (Fig.9b). Because RiBvalues were calculated using centeredfinite differences, the lowest level is not included. Con-sequently, the top of the PBL appears during the dayas a thin layer of values between 0.25 and 1.0 over theocean (Fig. 9b).Once mixed into the PBL, the tracer was advectedcoastward in the easterly flow. Along western slopes ofVancouver Island, the Coast Mountain Range, and theCascade Range, the shear and stability environmentcaused further downward acceleration. At 1200 UTC 29April (Fig. 9a), the modeled atmosphere was stablystratified through most of the troposphere, with a slightdecrease in stability above 700 hPa. Stability was great-est in the lower troposphere, allowing gravity wave gen-eration in the easterly outflow. The wind reversal (crit-ical level) between the outflow and the southwesterliesaloft could be expected to absorb wave energy. Thiswould lead to a hydraulic (Bernoulli) acceleration sim-ilar to that modeled in idealized studies (Clark and Pel-tier 1984) and observed along the mountains of thenorthwestern United States (Colle and Mass 1998). TheRiBwas lowest on the western slopes of Vancouver Is-land and the coastal mountains in the easterly flow (Figs.7a,b) and was below 0.25 locally. This is consistent withgravity wave breaking in the hydraulic flow down theslopes (Clark and Peltier 1984). Persistent low RiBonthe eastern slopes of Vancouver Island at 1248W waslikely the result of the overwater fetch of the outflow.Cool air flowing over the Strait of Georgia between theBritish Columbian mainland coast and Vancouver Islandis subject to heating from the warmer ocean below, lead-ing to convection and low RiBvalues. Similarly, at theend of the simulation, outflow had reached well off-shore, and RiBsuggests deeper convection near the west-ern edge of the domain (Fig. 9b).The tracer caught in the easterly outflow was forcedrapidly toward the coastal valley floors. Time series ofzonal wind (not shown) are consistent with an increasingnocturnal cross-mountain pressure gradient, strength-ening surface high pressure, and with downslope windspeeds reaching a maximum in the early morning hours.Throughout the simulation, low-level stratification andsubsidence were sufficient to limit PBL heights to lessthan 1000 m but would not prevent entrainment of thenear-surface tracer because PBL convection was strong.Consequently, simulation results indicate that down-slope transport loaded the stable air within coastal val-leys with the tracer prior to any convective PBL de-veloping during daylight hours. This tracer was thenfurther mixed to the surface by entrainment and down-mixing associated with growth of the convective mixedlayer.A change from southwesterly to southerly winds at700 hPa and deepening of the outflow layer combinedto decrease gravity wave activity aloft over the last dayof the simulation (Fig. 7). A southerly shift reduced thecross-range wind component on the last day of the sim-ulation, but the northwest–southeast orientation of thecoast mountains still allowed some wave generation inthe stable environment. As described above, deepeningoutflow raised the critical shear level and captured thewave energy at low altitude.A discrete Fourier analysis of the vertical velocitiesat 600 hPa over the eastern portion of the cross section1626 VOLUME 40JOURNAL OF APPLIED METEOROLOGYFIG. 10. Discrete Fourier spectrum of vertical velocity at 600 hPaacross the cross section shown in Fig. 7. The peaks correspond tomountain wave activity, decreasing with time.FIG. 11. (a) Divergence (s21) and relative tracer concentration (shaded) at 600 hPa and (b) a south–north crosssection of tracer concentration and winds along 1208W, valid 1800 UTC 28 Apr for the Dx 5 30 km grid. Contourinterval for divergence is 2.5 3 1025.in Fig. 7 clearly shows this trend of reduced wave ac-tivity later in the period (Fig. 10). Peaks in the curvesbetween wavelengths of l 5 40 and 50 km correspondto the mountain waves displayed in Fig. 7. The dual-peak structure is an artifact of the Fourier analysis, anda single peak would be evident if the analysis was en-tirely robust. The small sample, including only 64 gridpoints in the eastern part of the domain, leaves the anal-ysis prone to errors related to discretization, such asaliasing and phase cancellation. Smaller peaks near 75km correspond to land–sea differences, because wavesof this length show up in the simulation that did notinclude topography (Figs. 7g,h).e. Horizontal tracer distributionFlow aloft exerted substantial control over the hori-zontal distribution of tracer in the simulation. A per-sistent, elongated region of divergence at 600 hPa, as-sociated with the wind maximum aloft (Fig. 7a) led toa minimum in the horizontal tracer distribution in aswath cutting diagonally across British Columbia im-mediately to the north of Vancouver Island (Fig. 11a).Mountain waves generated in the Dx 5 30 km gridappear in Fig. 11a as alternating centers of divergenceand convergence through central British Columbia. Asouth–north transect through 1208W (Fig. 11b) showsthe same split in the tracer distribution, with an absencein tracer aloft at 538–568N, where divergence in theupper level jet was strongest.That the tracer reached the surface to the south of theswath of divergence, but not to the north, can be ex-plained by the evolution of the mountain wave activityand outflow conditions. As the wind maximum in theupstream branch of the upper-level ridge moved north-ward during the simulation, so did the primary axis ofvertically propagating mountain waves. Prior to 1800UTC 28 April, vertical velocities were much strongerover Washington, and later in the simulation, the verticalvelocities were strongest over central British Columbia.In addition, the outflow layer deepened and strengthenednorthward (Fig. 6b,d,f). Time evolution of surface-layertracer concentrations reflects the northward propagationof processes that brought the tracer to the surface andlater advected it westward (Fig. 6a,c,e).f. Frequency of meteorological conditionsTo estimate the probability of a similar flow patternoccurring and thereby transporting pollutants to popu-lated coastal regions after initial interception of pollut-ants by mountain wave activity, daily average NCEP–NCAR (National Center for Atmospheric Research) re-analyses from 1990 to 1999, inclusive, were examined.In so doing, it is assumed that similar zonal wind shearhas the potential to generate similar wave dynamics.Because outflow conditions often persist for severaldays, daily mean averages were deemed sufficient toestimate the probability of occurrence. During the April1998 outflow event, the zonal wind aloft (500 hPa) wasgreater than 10 m s21while the zonal wind near thesurface (925 hPa) was less than 25ms21. This com-bination was used as the criterion for comparison (i.e.,zonal wind shear of 15 m s21).Table 1 shows frequencies of occurrence of zonalSEPTEMBER 2001 1627HACKER ET AL.TABLE 1. Number of occurrences of strong zonal wind shear witha critical level in daily average NCEP reanalyses from 1990 to 1999,inclusive. Criteria are exclusive, and both require U925, 0 and U500. 0.MonthNo. of occurrencesU925,25,U500. 10 m s21U5002 U925. 15 m s21JanFebMarAprMayJunJulAugSepOctNovDec10540100002713837170412516241637878778TotalProbability421.2%63517.4%wind shear when both the westerly component aloft andthe easterly component near the surface are in the rangeof this case study. Occurrences in the table’s secondcolumn, which is the strictest criterion, are stronglyskewed toward winter months, when outflow conditionsare more common (Jackson 1996). The third columnevaluates a relaxed criterion, in which the magnitude ofthe wind shear is similar to this case, but either thewesterly component aloft or the near-surface easterlycomponent has a smaller magnitude. Probabilities arenot cumulative; that is, occurrences in column threeexclude those in column two. The total probability thatthe zonal wind shear is greater than 15 m s21was 1.21 17.4 5 18.6%. In both cases here, U925, 0 and U500. 0, denoting a zonal wind reversal between these lev-els. In the 10 yr of reanalyses, no days were found thatcontained a zonal wind reversal with shear less than 15ms21, suggesting that daily averaged analyses do notcapture such events. Those are likely not synopticallydriven, as was this case, but are products of local to-pographically driven circulation with a diurnal evolu-tion.5. Summary and conclusionsA mesoscale model was used to simulate the arrivalof mineral aerosol from an intense dust storm in theGobi over western North America during late April1998. In the absence of detailed meteorological mea-surements during the event, the model was used to elu-cidate FT–PBL exchange mechanisms by which dustwas mixed downward and subsequently affected coastalregions. A 4-day simulation, nesting grids to Dx 5 10km, provided a realistic representation of the surfacedust distribution when it was initialized with a thin,horizontally homogeneous layer of the tracer with amaximum at 650 hPa. Modeled spatial and temporalpatterns of tracer concentrations were qualitatively con-sistent with observations during the event. A parallelsimulation was completed without topography, showingthat topography was important in generating downwardvertical velocities. Last, 10 yr of NCEP–NCAR re-analysis grids were examined to gain an estimate of thefrequency of similar meteorological conditions.Model results indicate that several processes com-bined to bring particles from aloft (650 hPa) to the sur-face. First, large-scale subsidence associated with astrengthening high pressure center inland and verticallypropagating mountain waves combined to force the trac-er to near the height of the topography over BritishColumbia and Washington. The tracer was incorporatedinto the low-level outflow by entrainment processes atthe top of the PBL in the mountains and inland valleysand through turbulence generated in the shear layer re-siding near crest height. The tracer was then advectedwestward and forced into the coastal valleys by hy-draulic (Bernoulli) acceleration along the westernslopes, which reached a maximum in the early morninghours. A shallow PBL could then mix it down to thesurface during the day. Outflow dynamics through themountain gaps may also have played a role in the trans-port of the tracer from the east of the mountain peakstoward the coast.As the ridge strengthened during the time period ofthe simulation, the region most affected by down-mix-ing processes and coastward advection moved north-ward. This included the vertically propagating mountainwaves aloft, the eastward advection near the surface,and the hydraulic downslope acceleration over westernslopes. The northern extent of the ground-level tracerwas constrained by a broadly zonal swath of upper-leveldivergence that reduced the midtropospheric source ofthe tracer available to downward-transport processes.Zonal wind shear was responsible for much of thefree-atmosphere dynamics that brought the tracer down-ward in the simulation, including vertically propagatingmountain waves in the westerly flow, potential mixingin the shear layer, westward transport near the surface,and a hydraulic acceleration with possible wave break-ing in the low-level easterlies. Daily average NCEP–NCAR reanalyses from 1990 to 1999 suggest that theprobability of similar meteorological conditions andcomparable zonal wind shear with comparable zonalvelocities is 1.2%. This may be interpreted as an upper-bound on the probability because other factors, such asstability and cross-mountain pressure gradient, were notconsidered and because of the coarse vertical resolutionof the analysis. The probability of a similar dust eventwould be difficult to estimate, but it is bound to be evensmaller because the frequency of occurrence was largelyskewed toward the winter months whereas Asian duststorms are skewed toward spring (Braaten and Cahill1986). Dust must also be transported across the Pacificwithout dispersing, being scoured out, or settling beforeit reaches the coast of North America. The probability1628 VOLUME 40JOURNAL OF APPLIED METEOROLOGYthat similar wind shear magnitude occurred while al-lowing one or both of the zonal wind velocities to beless than in this case was much higher, at 18.6%, butagain this must be interpreted as an upper limit and doesnot guarantee similar dynamics.Results described herein suggest that the meteoro-logical conditions that permit interception of pollutantsby the mountain ranges of western North America andthen subsequent transport of pollutants coastward arerare. However, further research is required to investigatethe role of the western cordillera in intercepting the fullrange of pollutants emanating from Eurasia. It is con-ceivable that pollutants (crustal or anthropogenic) maybe intercepted by other processes (e.g., precipitation) ormixed downward but not coastward. Clearly, this hasimportant implications for regional environmental qual-ity, particularly in light of predictions of growing emis-sions from Eurasia and emerging evidence that trans-Pacific transport of anthropogenic pollutants is not un-usual.Acknowledgments. The Canadian Natural Sciencesand Engineering Council and Environment Canada aregratefully acknowledged for their grant support for por-tions of this research. The authors thank Henryk Mod-zelwski for maintaining the computer hardware and sys-tems, Rudy Husar for preparing dust observation im-ages, RPN for development and support of the MC2model, and the NCEP–NCAR 40-Year Reanalysis pro-ject for data access (http://wesley.wwb.noaa.gov/reanalysis.html).REFERENCESBenoit, R., J. Cote´, and J. Mailhot, 1989: Inclusion of a TKE boundarylayer parameterization in the Canadian regional finite-elementmodel. Mon. Wea. Rev., 117, 1726–1750.——, M. Desgagne´, P. Pellerin, S. Pellerin, Y. Chartier, and S. Des-jardins, 1997: The Canadian MC2: A semi-Lagrangian, semi-implicit wideband atmospheric model suited for finescale processstudies and simulations. Mon. Wea. Rev., 125, 2382–2415.Berntsen, T., S. Karlsdottir, and D. A. Jaffe, 1999: Influence of Asianemissions on the composition of air reaching the northwesternUnited States. Geophys. Res. Lett., 26, 2171–2175.Biscaye, P. E., F. E. Grousset, M. Revel, S. VanderGaast, G. A. Zie-linski, A. Vaars, and G. Kukla, 1997: Asian provenance of glacialdust (stage2) in the Greenland Ice Sheet Project 2 Ice Core,Summit, Greenland. J. Geophys. Res., 102, 26 765–26 781.Braaten, D. A., and T. Cahill, 1986: Size and composition of Asiandust transported to Hawaii. Atmos. Environ., 20, 1105–1109.Clark, T. L., and W. R. Peltier, 1984: Critical level reflection and theresonant growth of nonlinear mountain waves. J. Atmos. Sci.,41, 3122–3134.Colle, B. A., and C. F. Mass, 1998: Windstorms along the westernside of the Washington Cascade Mountains. Part II: Character-istics of past events and three-dimensional idealized simulations.Mon. Wea. Rev., 126, 53–71.Deardorff, J. W., 1978: Efficient prediction of ground surface tem-perature and moisture with inclusion of a layer of vegetation. J.Geophys. Res., 83, 1889–1903.Duce, R. A., C. K. Unni, B. J. Ray, J. M. Prospero, and J. T. Merrill,1980: Long-range atmospheric transport of soil dust from Asiato the tropical North Pacific: Temporal variability. Science, 209,1522–1524.Garstang, M., P. D. Tyson, R. Swap, M. Edwards, P. Kallberg, andJ. A. Lindesay, 1996: Horizontal and vertical transport of airover southern Africa. J. Geophys. Res., 101, 23 721–23 736.Husar, R. B., and Coauthors, 2001: The Asian dust events of April1998. J. Geophys. Res., in press.Jackson, P. L., 1996: Surface winds during an intense outbreak ofarctic air in southwestern B.C. Atmos.–Ocean, 34, 285–311.——, and D. G. Steyn, 1994: Gap winds in a fjord. Part I: Obser-vations and numerical simulation. Mon. Wea. Rev., 122, 2645–2665.Jaffe D. A., and Coauthors, 1999: Transport of Asian air pollutionto North America. Geophys. Res. Lett., 26, 711–714.McKendry, I. G., 2000: PM10levels in the Lower Fraser Valley, B.C.,Canada: An overview of spatio-temporal variations and mete-orological controls. J. Air Waste Manage. Assoc., 50, 174–185.——, and J. Lundgren, 2000: Tropospheric layering of ozone in re-gions of urbanised complex and/or coastal terrain—a review.Prog. Phys. Geog., 24, 359–384.——, J. P. Hacker, R. Stull, S. Sakiyama, D. Mignacca, and K. Reid,2001: Long range transport of Asian dust to the Lower FraserValley, British Columbia, Canada. J. Geophys. Res., in press.Merrill, J. T., M. Uematsu, and R. Bleck, 1989: Meteorological anal-ysis of long range transport of mineral aerosols over the NorthPacific. J. Geophys. Res., 94, 8584–8598.Murayama, T., and Coauthors, 2001: Ground-based network obser-vation of Asian dust events of April 1998 in east Asia. Part 1:Transport mechanism. J. Geophys. Res., in press.Piketh, S. J., H. J. Annegarn, and P. D. Tyson, 1999: Lower tropo-spheric aerosol loadings over South Africa—the relative con-tribution of aeolian dust, industrial emissions, and biomass burn-ing. J. Geophys. Res., 104, 1597–1607.Stull, R. B., 1988: An Introduction to the Boundary Layer Meteo-rology. Kluwer Academic, 666 pp.Tratt, D. M., R. J. Frouin, and D. Westphal, 2001: The April 1998extreme Asian dust event: Southern California perspective. J.Geophys. Res., in press.Tyson, P. D., M. Garstang, A. M. Thompson, P. D’Abreton, R. D.Diab, and E. V. Browell, 1997: Atmospheric transport and pho-tochemistry over central southern Africa during the SouthernAfrica Fire–Atmosphere research Initiative. J. Geophys. Res.,102, 10 623–10 635.Warner, T. T., R. A. Peterson, and R. E. Treadon, 1997: A tutorial onlateral boundary conditions as a basic and potentially seriouslimitation to regional numerical weather prediction. Bull. Amer.Meteor. Soc., 78, 2599–2617.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items