UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mesometeorological modelling and trajectory studies during an air pollution episode in the lower Fraser… Miao, Yuelong 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1993_fall_miao_yuelong.pdf [ 4.61MB ]
Metadata
JSON: 831-1.0086265.json
JSON-LD: 831-1.0086265-ld.json
RDF/XML (Pretty): 831-1.0086265-rdf.xml
RDF/JSON: 831-1.0086265-rdf.json
Turtle: 831-1.0086265-turtle.txt
N-Triples: 831-1.0086265-rdf-ntriples.txt
Original Record: 831-1.0086265-source.json
Full Text
831-1.0086265-fulltext.txt
Citation
831-1.0086265.ris

Full Text

MESOMETEOROLOGICAL MODELLING AND TRAJECTORYSTUDIES DURING AN AIR POLLUTION EPISODE IN THELOWER FRASER VALLEY, BRITISH COLUMBIA, CANADAByYuelong MiaoB. Sc. (Atmospheric Physics) Nanjing University, P.R. China, 1984A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESATMOSPHERIC SCIENCE PROGRAMME, DEPARTMENT OF GEOGRAPHYWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAAugust 1993© Yuelong Miao, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of 1:..--0 I-R-A pHy The University of British ColumbiaVancouver, CanadaDate it 4/ IVDE-6 (2/88)Abst ractThe Lower Fraser Valley, straddling the USA-Canada border in western NorthAmerica, often experiences episodes of elevated tropospheric ozone in summertime. Themeteorology governing those episode days is characterized by a stagnant high pressuresystem, light wind, strong insolation and the occurrence of sea breezes.To shed some light on the mesometeorology of ozone episodes in the LFV, this studyemployed the Regional Atmospheric Modelling System from Colorado State University,a nonhydrostatic, three-dimensional mesoscale modelling system to simulate the detailedstructure of air flows over the valley for one specific episode day. Significant code modifi-cations have been made to enhance the model's ability to represent surface energy fluxesand predict surface temperatures in the surface of complex terrain and land-use patterns.Evaluation of the model performance was made against an extensive set of observationson the episode day. Pollutant transport on the modelling day was explored with a La-grangian particle dispersion model. Specifically, forward trajectories were calculated forparticles released at various locations in the Lower Fraser Valley and at different times.A systematic qualitative and quantitative model evaluation with the statistical methodof Willmott showed that the model could adequately simulate the observed sea breeze andother interactive terrain-induced flows such as slope wind and channel flow. Trajectorystudies indicated that pollutant recirculations occurred largely from sources originatingin the northwest part of the valley where most emission sources are located, and endedin a region where the highest ozone concentrations were observed. Recirculations weredue to pollutants traveling with the interactive flows of sea breeze and upslope winds,i iand later being captured and directed back to the valley by the downslope winds. Parti-cles released in other part of the valley all travelled outside the valley, having a minimalcontribution to the episode buildup.iiiTable of ContentsAbstract^iiTable of Contents^ ivList of Tables^ viiList of Figures viiiAcknowledgements^ xv1 Introduction^ 11.1 Introduction  ^11.1.1 Air Pollution and Mesoscale Meteorology  ^31.2 Objectives^ 91.2.1 Previous Air Pollution Studies in the LFV  ^91.2.2 Objectives  ^122 The RAMS Model and Its Application^ 162.1 Introduction  ^162.2 RAMS Configuration ^  162.2.1 Model Domain and Grid Structure ^  172.2.2 Initialization and Initial Data  ^212.2.3 Boundary Conditions ^  253 Results and Model Validation^ 31iv3.1 Quantitative Evaluation of Model Performance^  333.1.1 Surface Wind Evaluation ^  353.1.2 Surface Air Temperature Evaluation ^  363.2 Qualitative Evaluation of Model Performance  413.2.1 Surface Winds ^  413.2.2 Vertical Profiles  ^473.3 Conclusions ^  484 Trajectory Study 514.1 The RAMS Lagrangian Particle Dispersion Model ^ 514.2 General Pollutant Transport Investigation in the LFV 524.3 Intensive Investigation of Pollutant Transport in the LFV ^ 564.3.1^Dual Release-time Study of Pollutant Transport in Vancouver 564.3.2^Summary of the Dual Release-Time Study ^ 594.4 Discussions and Conclusions ^ 694.5 Significance and Speculations 72Appendix A Surface Energy Balance and Surface Temperature^73A.1 Introduction ^  74A.2 The Model  75A.2.1 The Net Radiation Flux ^  75A.2.2 The Storage Heat Flux  79A.2.3 The Latent Heat Flux ^  83A.2.4 The Sensible Heat Flux  85A.3 Numerical Procedures ^  86A.4 Energy Balance and Surface Temperature at Night ^ 87A.5 Model Evaluation ^  88vBibliography^ 94v iList of Tables2.1 RAMS meteorological model configuration used in the LFV simulations. 182.2 RAMS vertical discretization and initial conditions. Z denotes scalar levelsand ZZ denotes vertical velocity levels in the RAMS Arakawa C stagger inmeters. p, 0, q, u and v are initial pressure (in Pa), potential temperature(in K), water vapour mixing ratio (in gkg'), east-west and north-southwind component (in ms') respectively, corresponding to scalar levels Z. 222.3 Roughness Length and Albedo for Each Land-use Type. ^ 29A.1 Summary of Coefficients for Some Surface Materials. ^ 81A.2 Summary of Coefficients for Various Land-use Types. 82A.3 Summary of P&T Coefficient for Each Land-use Type. ^ 85viiList of Figures1.1 Location and land-use maps of the Lower Fraser Valley and the surround-ing area. Inner labels are the corresponding latitude or longitude valuesfor the boundaries.  1.2 The essentially closed atmospheric circulation system characteristic ofVancouver while under the influence of a stagnant high-pressure system.Taken from Hay and Oke (1976) . ^1.3 Examples of coastal meteorology and resultant air pollution. a) plumetrapping. b) fumigation. c) sea breeze. d) plume behaviour within the seabreeze regime. Taken from Lyons (1975) .1.4 Maximum ozone concentration isopleths on August 23, 1985 in the LFV.Thick solid line refers to ozone contour, and thick dashed line refers touncertain ozone contour. Thinner soild line is the coastline, and thinnerdashed line is the 100m terrain contour indicating the edge of the valleywalls. Values beside dots are the ozone concentrations at respective sta-tions. Inner labels refers to the corresponding latitude or longitude of theboundary. Outer labels are in units of UTM coordinates (x 10 3 ). . . .^15viii2.1 Smoothed terrain contour map for the model domain that includes themodel domain of interest (inside the dashed line) and the buffer area (out-side the dashed line). Contour interval is 100 metres. Outer labels arein units of km. Symbol refers to Vancouver;^to Tsawwassen;^toBellingham; to Abbotsford;^Chilliwack; to Mission City and^toPitt Lake.  ^202.2 a) 500 mb chart for August 23, 1985 1200Z(0400 PST on August 23) ^Units of contours are in dm. b) Corresponding sea level pressure chart ^Units of contours are in mb. ^  232.3 a) 500 mb chart for August 24, 1985 OZ(1600 PST on August 23). Unitsof contours are in dm. b) Corresponding sea level pressure chart. Units ofcontours are in mb.   243.1 Observation network for August 23 1985 in the LFV. The solid line is thecoastline. The dashed line is the 100 metre terrain contour, showing theedge of the valley wall. Outer labels are in units of UTM coordinates(x103). 323.2 The evolution of modelled and observed: a) average wind direction, b)average wind speed for August 23, 1985. ^  373.3 Statistic comparison of modelled and observed wind as time series of a)standard deviation of wind speed, b) total (RMSD), systematic (RMSDs)and unsystematic (RMSDu) root mean square deviations and c) index ofagreement for August 23, 1985.   383.4 The evolution of the modelled and observed: a) average temperature, b)standard deviation of temperature for August 23, 1985.^ 39ix3.5 The evolution of the modelled and observed: a) total (RMSD), system-atic (RMSDs) and unsystematic (RMSDu) root mean square deviations oftemperature, b) index of agreement for August 23, 1985. ^ 403.6 (a-h) The observed and modelled surface winds at a) 0900 PST and b)1100 PST of August 23, 1985. Bold arrows are the observed winds. Thedashed line is the 100 metre terrain contour, refering to the edge of theLFV wall. The solid line is the coastline. Outer labels are in units ofUTM coordinates (x103 ). Inner labels are latitude or longitude values forthe boundaries. The spacing between vector tails represents 7.5 ms -1 . . 433.6 c) 1300 PST and d) 1500 PST of August 23, 1985. ^ 443.6 e) 1700 PST and f) 1900 PST of August 23, 1985. 453.6 g) 2100 PST and h) 2300 PST of August 23, 1985. ^ 463.7 Profiles of the east-west component of wind velocity at Queen ElizabethPark in Vancouver at a) 0900 PST, b) 1200PST, c) 1500 PST, and d) 1800PST. The solid line refers to the modelled profile; The dashed line to themeasured profile. Positive wind speed value indicates a westerly wind. . 493.8 As in Fig. 3.7 but of the north-south component of wind velocity. Positivewind speed value indicates a southerly wind. ^  50x4.9 Plan view of 15 three-dimensional grid-scale trajectories at three coastallocations and at five different heights. Each trajectory starts at 0800PST 23 August 1985 and lasts for 14 hours. Trajectory ending pointlabels indicate the five release heights of 176m, 328m, 670m1103m and 1673m above sea level (ASL) at one site in Vancouver. '6' to'10' sequentially are for trajectories released at Tsawwassen at heightsof 130m, 335m, 660m, 1094m and 1668m ASL correspondingly. '11' to`15' are for trajectories near Bellingham, USA at heights of 206m, 407m,733m, 1174m and 1783m ASL, respectively. The dots on the trajectoriesindicate the hourly trajectory positions. The background is the 200m-interval terrain contour map for the LFV and the model buffer area. . 534.10 Six three-dimensional grid-scale trajectories released about 50m above theground at three eastern locations of the LFV and at two different times. a)plan view, and b) west-east vertical slice viewed northward. 'A', IF and 'C'refer to releasing stations near Mission City, Abbostford and Chilliwack,respectively. '1', '2' and '3' are the corresponding trajectories released at0800 PST 23 August 1985 while '4', '5' and '6' are for releases at 0900PST. Each trajectory lasts for 19 hours. The dots on the trajectoriesindicate the hourly trajectory positions. The background on a) is the200m-interval terrain contour map for the LFV and the model buffer area.The background on b) refers to the terrain on the slice that cuts throughstation "C". Labels are in units of km.   54xi4.11 Seven three-dimensional grid-scale trajectories 50m above the ground re-leased at one site in Vancouver at seven one-hour intervals starting from0800 and followed for up to 19 hours. a) plan view, and b) west-east ver-tical slice viewed northward. Trajectory endpoint labels '1' to '7' indicatethe seven release times from 0800 to 1400 PST, respectively. The dots onthe trajectories indicate the hourly trajectory positions. The backgroundon a) is the 200m-interval terrain contour map for the LFV and the modelbuffer area. The background on b) refers to the terrain on the slice thatcuts through the release station. Labels are in units of km.   574.12 Locations of 15 emission stations in Vancouver that form three groups forthe trajectory study. The solid line is the coastline, and the dashed line isthe 100m terrain contour. Outer labels are in units of UTM coordinates(x 103)   584.13 Plan view of the three-dimensional grid-scale trajectories from five sourcesof the first group in Vancouver. Released near the surface at a) 0800 PST,and b) 1200 PST 23 August 1985 followed up to 19 hours. Trajectory end-point labels `1' to '5' indicate the five stations spaced from west to eastat 5km interval. The dots on the trajectories indicate hourly trajectorypositions. The background is the 200m-interval terrain contour map forthe LFV and the model buffer area. Labels are in units of km.   604.14 Same as Fig. 4.13 but the XZ projection that cuts through the secondemission group on Fig. 4.12. Viewpoint is northward. ^ 614.15 Same as Fig. 4.13 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward.  ^624.16 Plan view of the three-dimensional grid-scale trajectories from five sourcesof the second group in Vancouver. For details see caption to Fig. 4.13. .^63xii4.17 Same as Fig. 4.16 but the XZ projection that cuts through stations of thissecond group. Viewpoint is northward. ^  644.18 Same as Fig. 4.16 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward. ^ 654.19 Plan view of the three-dimensional grid-scale trajectories from five sourcesof the third group in Vancouver. For details, see Fig. 4.13.  ^664.20 Same as Fig. 4.19 but the XZ projection that cuts through the secondemission group on Fig. 4.12. Viewpoint is northward. ^ 674.21 Same as Fig. 4.19 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward.  ^684.22 Snapshot of the ozone concentration isopleths at 1500 PST on August23, 1985. Thick solid line refers to ozone contour, and thick dashed linerefers to uncertain ozone contour. Thinner solid line is the coastline, andthinner dashed line is the 100m terrain contour indicating the edge of thevalley walls. Values beside dots are the ozone concentrations at respectivestations. Inner labels refer to the corresponding latitude or longitude ofthe boundary. Outer labels are in units of UTM coordinates (x 10 3 ). . . 71A.1 Priestley-Taylor coefficient a calculated for a range of canopy resistancefor the Cabauw data set ( from McNaughton and Spriggs 1987).^.^. 84A.2 Time series for surface fluxes at the Sunset suburban site in Vancouver. a)Net radiation, b) Storage heat flux, c) Sensible heat flux, d) Latent heatflux. Solid line - modelled values. Dashed - observed.   90A.3 a) Location and land-use maps of Vancouver. b) The surface radianttemperature distribution for the same area at 13.56 (LST) on 16 August1985. The rainbow brightness temperatures in b) are in degree Celsius.From Roth et al. (1989), p1705-1707. 92A.4 a) Contour plot of surface temperatures derived from the satellite thermalimage in Fig. A.3b, and b) contour plot of surface temperatures derivedfrom the RAMS model output at 1400 PST on August 23, 1985. All plotscover the same area as depicted in Fig. A.3a. Temperatures are in degreeCelsius.  xivAcknowledgementsMy deepest gratitude goes to my supervisor, Prof. Douw Steyn, for the most valuablestudy opportunity he provided me, and for his consistent support in both academic andfinancial affairs. His timely advice, criticism and encouragement have been the importantforces to push me one step further. His friendly enthusiasm and open-minded attitudehave influenced me greatly. Thanks also to the other members of my committee, Profs.Tim Oke and Ian McKendry, for their constructive comments and guidance in the courseof the thesis research. My modelling work is much improved, thanks in no small measureto Prof. Oke's suggestion of a updated energy balance scheme.I am very grateful to my senior colleagues Drs. Peter Jackson and Xiaoming Cai fortheir help with the RAMS model. I could not have operated such a comprehensive modelsystem without their assistance. I would also like to thank Dr. Mike Moran in Atmo-spheric Environment Service of Canada for his generous supply of the RAMS compatibleparticle model and for his hands-on help with its operation. Dr. Sue Grimmond atUniversity of Indiana, USA and Dr. Jingming Chen at Remote Sensing Center, Canadawere indispensible for their help in constructing the energy balance scheme.My colleague Ms. Natalie Suzuki, carefully read through the entire thesis draft andmade important suggestions. I will remember many other students at the Departmentof Geography for their warmness and friendship.This thesis is dedicated to my beloved family in China, my wife Shuzhen Xu and mydaughter Ling Miao who will be three years old by July 25. My deepest feeling to themand my gratitude to their understanding are beyond description.x vChapter 1Introduction1.1 IntroductionThe Lower Fraser Valley (LFV) region along the British Columbia-Washington coast,extends from the Strait of Georgia in the west to the Fraser canyon in the east. Thevalley is bounded by the Coast Mountains to the north and the Cascade Range to thesoutheast (Fig. 1.1). The most densely populated and industrialized area of the LFVlies in the Greater Vancouver area with a population of approximately 1.5 million. WhileVancouver enjoys a world-wide reputation as a beautiful harbour city, it is not withoutan air pollution problem. Data obtained from the Greater Vancouver Regional District(GVRD) air quality monitoring network have revealed some startling facts. The Maxi-mum Acceptable hourly average ozone 1 concentration (82 ppb) was exceeded an averageof 160 times a year during the 1980s. Over the same period, the Maximum Tolerablehourly ozone concentration (153 ppb) was exceeded an average of four times per year.Ozone episodes are particularly frequent in summer when the synoptic weather pattern isstagnant and insolation is strong. As an extreme example, from August 1 to September30, 1988, four ozone episodes were observed. On 27 consecutive days, high ozone valueswere reported with a maximum ozone concentration of over 200 ppb occurring on Sept.3, 1988. Clearly Vancouver and the surrounding LFV are subject to ozone episodes.This is somewhat surprising since Vancouver is considerably less industrialized and has1 Unless otherwise noted, the word "ozone" will be used herein to refer strictly to the ozone in thetroposphere as opposed to ozone in the ozone layer of the stratosphere. Excessive ozone in the loweratmosphere can harm human and plant health and many building materials.1uucpnuecunqnslainidpnlinopby•uogeoupuopt Jo osn anunj Joj are sowuu uopuool aplsaq sioqwXs•sapupunoq ow Jo; sawn opnOuoi io aprupui 2u!puodsauoo ain air slam null' Tam2umuno.uns atil puu ifourA JawId Jamoi atp jo sduw asn-puul put? uoprool :I. 1 ani5LqcvChapter 1. Introduction^ 3a significantly smaller population than other large mid-latitude coastal cities, such asLos Angeles, Athens and Tokyo, which are known to experience ozone episodes. It isgenerally recognized that it is the particular combination of emissions, topography andmeteorology in the LFV that exacerbates the air pollution problem (Steyn et al., 1992;Concord, 1982; Concord, 1985).The major source of ozone precursors is motor vehicles which contribute over two-thirds of the total (Steyn et al., 1992), and most of these emissions occur in the denselypopulated western portions of the LFV. These emissions, when exposed to strong solarradiation, will produce ozone through a series of photochemical reactions (Giisten, 1986).When these conditions coincide with meteorological conditions which restrict ventilation,ozone will accumulate. Such conditions usually occur in summer. The stagnant andweak Pacific high-pressure system, characterized by low wind speeds, clear skies andsubsidence inversions, caps the valley while the valley walls act as barriers to furtherblock ventilation in the valley. Thermally-induced winds exacerbate the poor ventilationproblem through pollutant recirculation. Figure 1.2 schematically depicts such a closedatmospheric circulation system under the influence of a stagnant high-pressure system.This pattern has been suggested by meteorologists since the early 1970's (Hay and Oke,1976), but the detailed meteorological influence on the pollutant transport has yet to beexplored.To help understand the air pollution problem in the LFV, it is necessary to introducesome concepts of air pollution, particularly regarding the impact of meteorology in acoastal area such as Vancouver.1.1.1 Air Pollution and Mesoscale MeteorologyAir pollution spans a wide range of scales from local scales as exemplified by stackplumes to continental and global scales as exemplified by the multi-country radionuclide—IS.- Closed-'111E— circulationf EmissionsGeorgia Sir.Chapter 1. Introduction^ 4f■..1NCoast MrnsFigure 1.2: The essentially closed atmospheric circulation system characteristic of Van-couver while under the influence of a stagnant high-pressure system. Taken from Hayand Oke (1976).contamination experienced during the 1986 Chernobyl disaster in the former Soviet Union(Persson, Rodhe and Geer, 1987; Smith and Clark, 1988; Stern et al., 1984; Wheeler,1988). Air pollution dispersion can be very different from one scale to another. Atlocal scales, turbulent diffusion and advection dominate the dispersion of pollutants.Dispersion is limited to the lower atmosphere, and affects a fairly small radius in thevicinity of the pollutant source. At the continental and global scales, long-range transportplays a major role in the movement of pollutants. Pollutants are often transportedthrough the stratosphere and affect regions well beyond the source areas. The regionalor mesoscale is of particular interest in the present study. This scale shares some of thecharacteristics of both small and large scale problems in that turbulent diffusion andatmospheric transport are important. While emission sources are typically local, thedispersal of the pollutants is driven by regional meteorology.Coastal Mesoscale Meteorology1^otstivEo ►tumf PtCFILESH3,03mConiltant Lid Hoight135.a)b)Chapter 1. Introduction^ 5c)^d)Figure 1.3: Examples of coastal meteorology and resultant air pollution. a) plume trap-ping. b) fumigation. c) sea breeze. d) plume behaviour within the sea breeze regime.Taken from Lyons (1975).Chapter 1. Introduction^ 6Mesoscale air pollution problems are experienced in many places. Since most majorpopulation and industrial centres are located along rivers, oceans and lakes, emphasisis given to coastal meteorology and its impact on air pollution. Typified by differentialheating between land and water, coastal areas form a distinct environment in diffusingand transporting pollutants. On a cloudy day or at night, neutral to stable air movingacross warmer near-shore water before landfall will result in a shallow mixed layer nearthe surface. This shallow layer will be enhanced slightly by the mechanical mixing whenthe air travels inland. A plume emitted from a low stack near the shore will then betrapped in this shallow mixed layer (Fig. 1.3a), a phenomenon known as plume trapping(Lyons, 1975). On a sunny day, when air flows onshore, the air mass will be modified bychanges in roughness and temperature. A thermal internal boundary layer (TIBL) willform and deepen with distance downwind of the shoreline (Stull, 1989). The intersectionof a plume released from an elevated near-shore source with the TIBL will result infumigation(Fig. 1.3b).The sea/land breeze is a well-known mesoscale phenomenon in the shoreline environ-ment with onshore sea breeze in the daytime and offshore land breeze at night (Atkinson,1981). This phenomenon is a direct consequence of differential heating over land and sea.The inland sea breeze near the surface is often associated with an offshore return flowaloft(Fig. 1.3c). These two flows often form a circulation called sea breeze circula-tion which has an important influence on pollutant movement (Lyons, 1975). Pollutantsinjected in the near surface inflow may enter the return flow once being lifted in a conver-gence zone such as a sea breeze front. If the synoptic pattern is stagnant for several days,pollutants may recirculate in the coastal regions, a condition conducive to air pollutionepisodes. The varying wind direction in the vertical could also cause plumes at differentheights to move in very different directions. Figure 1.3d illustrates such phenomena. Allthese distinct features in the coastal atmosphere complicate the study of air pollutionChapter 1. Introduction^ 7near the coast.The picture is further complicated by the presence of mountains. When mountainsare present, mountain/valley winds, upslope/downslope winds and channel flows are of-ten observed (Atkinson, 1981). For mountain-valley situations, stable lapse rates canresult in a marked decoupling of the air within the valley from the upper-air flow. Thisphenomenon results in poor ventilation of the lower levels and constitutes meteorologicalconditions conducive to the long-term buildup of air pollution levels. A similar situationoccurs when an elevated inversion exists near or below ridge levels and contains a flowwithin a valley region. Such mountainous coastal regions present a complex mesoscalepicture with interactions among various terrain-induced mesoscale phenomena and syn-optic flow. Air pollution studies in such an area are indeed a big challenge. Extensivemeteorological observations, and atmospheric diffusion and transport experiments arenecessary to reveal the dynamics of mesoscale meteorology and its impact on pollutanttransport. Unfortunately field experiments are costly. Mesoscale modelling, on the otherhand, provides a convenient and economic tool to simulate and forecast the wind flow ina region. In fact, a complete picture of wind fields can only be achieved with a prognosticmesoscale model.Mesoscale and Air Pollution Modelling in Coastal RegionsA number of modelling studies have been carried out on such urbanized mountainouscoastal areas as Los Angeles, Tokyo and Athens (Chang et al., 1990; Chang et al., 1989;Chang et al., 1990; Moore et al., 1991; Moussiopoulos, 1993; Ulrickson and Mass, 1990a;Ulrickson and Mass, 1990b). A common purpose of these studies was to understand howthe regions' various mesoscale phenomena interact with each other and with large-scaleweather patterns, and how the specific local meteorological conditions determine thetransport of pollutants in and around the studied area.Chapter 1. Introduction^ 8Ulrickson and Mass(1990a, 1990b) employed the hydrostatic Colorado State Univer-sity Mesoscale Model (CSUMM) to simulate three-dimensional airflows in the Los Ange-les basin. Having documented the model's accuracy in simulations of diurnal mesoscalefeatures in the basin, they then looked at synoptic influences on mesoscale circulationsduring episode days. They concluded that light large-scale winds had little influence onstrong summertime mesoscale circulations such as sea/land breeze and mountain/valleywind circulations, whereas stronger large-scale winds exerted considerable influence onthe weak mesoscale wintertime circulations. They further investigated the interactionsbetween various mesoscale phenomena and the influence of these phenomena on pollu-tant transport using a forward parcel trajectory method. One of their findings is thatdaytime upslope flows near mountains ventilate the basin, inducing basinwide airflowand augmenting the sea breeze. However, based on their simulations, the mechanism forthe development of air pollution episodes was not clear.Moussiopoulos (1993) reviewed air pollution studies over the past decade in theGreater Athens Area (GAA), a region surrounded on three sides by mountains and on thefourth side by the sea. Observations have shown that most of the air pollution episodesare associated with the development of a sea breeze (Lalas et al., 1983) which tendsto stratify the atmosphere above Athens, thus trapping air pollutants at a relativelylow height above ground. Offshore pollutant transport by the land breeze and its re-advection into the basin by the sea breeze result in the further buildup of pollutants. Anonhydrostatic mesoscale model - MEMO was used to simulate the wind fields in Athens(Moussiopoulos et al., 1993) on episode days when a sea breeze was observed. Using thismodel output, they went on to simulate the dispersion and chemical transformation of airpollutants in the GAA with a fully vectorized photochemical dispersion model - Model forthe Atmospheric Dispersion of Reactive Species (Moussiopoulos, 1989; Moussiopoulosand Oehler, 1988). Predicted 0 3 , CO, NO and NO2 values compared favourably withChapter 1. Introduction^ 9available observations.The long-range transport (LRT) of oxidant pollutants in central Japan is believedto play an important role in regional air pollution (Chang et al., 1990; Chang et al.,1990; Kurita et al., 1985). Although the Tokyo metropolitan complex contains most ofthe emission sources in the region, observed maximum ozone concentrations are muchhigher in a mountainous region 150 km away than they are in Tokyo (Kurita et al.,1985). With the aid of an intensive meteorological and photochemical field study carriedout in central Japan in 1983, Chang et al. (1989) explored the long-range transportmechanism through a trajectory method and found that the combined roles of land/seabreeze, mountain/valley winds, steady onshore winds, strong thermal low and subsidenceinversions under a synoptic high pressure system contributed to the LRT mechanism. Anadvanced Eulerian combined transport/chemistry/removal model: The Sulfate TransportEulerian Model-II was then used to analyze LRT and the associated chemical processes.Predictions coincided well with the observations and showed clearly that sources locatedmore than 100 km from the region could have a profound impact on local air quality.1.2 Objectives1.2.1 Previous Air Pollution Studies in the LFVA number of studies on air pollution in the LFV have been carried out (Robeson andSteyn, 1990; Steyn and Faulkner, 1986; Steyn and McKendry, 1987; Steyn and McK-endry, 1988; Steyn and Oke, 1982; Steyn, Roberge and Jackson, 1990). A continuousmonitoring network for meteorological and pollutant data has been in operation for morethan a decade. Since terrain-induced mesoscale phenomena like sea/land breeze, moun-tain/valley wind and valley channel flows are often associated with air pollution episodes,a major effort has been given to the study of the LFV's specific local meteorology andChapter 1. Introduction^ 10its impact on air pollution.Mixed layer depth is known as an important surrogate for pollutant diffusion ormixing. An attempt was made by Steyn and Oke (1982) to simulate the mixed layerdepth in the LFV using an Eulerian mixed layer depth model. The model predictedtemporal variations of the mixed layer depth agreed well with the observations, andrevealed that the mixed layer depth in the LFV rarely exceeds 800 m due to the onshoreadvection of cool marine air. As an adjunct study to the observational and modellingstudies of the physical nature of the sea breeze, Steyn and Faulkner (1986) studied theclimatology of sea breezes in the LFV through examination of ten years of hourly winddata from two stations in the LFV, one near the coast and the other in the valley. Thestudy revealed that the sea breeze is a phenomenon frequently observed in the LFV. Italso showed strong interactions between the sea breeze and slope winds near the shoreline,and between sea breeze and valley winds further inland.As part of the efforts in providing a scientific basis for the GVRD's air pollutioncontrol and abatement strategies, Robeson and Steyn (1990) presented three statisticalmodels to forecast daily maximum ozone concentrations in the LFV. They concludedthat the TEMPER regression model, based on daily maximum air temperature and theprevious day's ozone concentration, provided the most accurate forecast.Steyn, Roberge and Jackson (1990) analysed the LFV's meteorology surroundinga persistent ozone episode occurring between September 1-3, 1988. They confirmed thetypical meteorological characteristics of episodes in the LFV: high temperatures, low windspeeds, shallow mixed layer depth, slack synoptic pressure gradients, strong subsidenceand the occurence of the sea breeze phenomena.Taylor (1991) also looked at the meteorology especially the synoptic conditions sur-rounding air pollution episodes in the LFV. Based on the recent work of AtmosphericEnvironment Service Pacific Region on the relationship between synoptic patterns andChapter 1. Introduction^ 11high ozone events in the LFV, Taylor concluded that synoptic conditions of a strongupper ridge over the eastern Pacific and British Columbia coupled with a surface ther-mal trough along the Washington-southern B.C. coast are necessary for elevated ozoneconcentrations in the LFV. He further pointed out that the lowered sea level pressureon the coast produced by these thermal troughs may also interfere with and restrict thedevelopment of mesoscale sea breeze. An ozone forcast method based on the crucialconnection between specific synoptic patterns and high ozone events was carried out andshowed some success in predicting ozone. He also used a statistical ozone forecast model,which links maximum daily ozone to temperature, yesterday's ozone, precipitation andmean sea level pressure differences between the coast and the interior, to forecast theozone concentrations. He argued that this may be a more promising tool to the ozoneforecast.McKendry and Li-Ting-Wai (1993) investigated further of the relationship betweenhigh ozone events in the LFV and synoptic meteorological conditions with extensive set ofozone concentration data at Port Moody and synoptic charts at both 500 hPa and meansea level for the period of 1978-92. Their studies confirmed what Taylor (1991) discoveredabout the specific synoptic meteorological conditions on the days of high ozone, i.e., a lowlevel thermal trough and upper level ridge of high pressure. The inter-annual variabilityof ozone episode days was also studied in their report.A realistic model to simulate the windfields in the LFV is an essential prerequisiteto the understanding of the etiology of severe ozone episodes during summertime andto the knowledge of the horizontal spatial structure of the pollutant concentrations inthis region. Intensive mesoscale modelling has been carried out to identify the localmeteorology in the LFV during episodes. Steyn and McKendry (1987) employed thehydrostatic CSUMM model to simulate windfields in the LFV for episode days in 1985and 1986. Quantitative and qualitative comparisons of the model output with observedChapter 1. Introduction^ 12data were also conducted for one episode day (Steyn and McKendry, 1988). Resultsshowed that in general the model performed well in simulating both the spatial andtemporal variability of the thermally-forced flows within the chosen domain. The modelperformed less well at night and had a tendency to overpredict wind speeds in the daytime.Unfortunately their modelling studies were limited to the simulation of the meteorologicalfields and did not try to investigate the relationship between the mesoscale meteorologyand pollutant transport.1.2.2 ObjectivesGiven that extensive modelling studies have been carried out on the local mete-orology, but have not considered in detail the interaction of the meteorology with airpollutant transport, the present study will expand on these previous modelling effortsand will attempt to elucidate the relationship between the mesoscale meteorology andair pollutant transport in the LFV. The major objectives of this study are the following:• Simulate the three-dimensional windfields over the LFV for a 24-hour period duringa specific summer ozone episode day using the Regional Atmospheric ModellingSystem(RAMS).The simulation will be performed with the RAMS model from Colorado StateUniversity. This is an advanced version of the CSUMM model, and should providemore realistic windfields given its non-hydrostatic formulation and more realistictreatment of physical processes. Furthermore, improved model performance shouldbe achieved by using the following considerations:1. Use 2.5km x 2.5km resolution in the horizontal so as to resolve more terrainfeatures. Steyn and McKendry (1987) used a horizontal resolution of 5km x5km in their model runs.Chapter 1. Introduction^ 132. Select the model domain to cover the Georgia Strait and part of VancouverIsland to the west and the American part of the LFV to the south in orderto include all the mesoscale phenomena in the LFV that might influence thepollutant movement within the region.3. Preserve the surface inhomogeneity in the model.4. Develop a prognostic surface temperature scheme using a semi-empirical sur-face energy balance representation to replace a pure soil model. The RAMSsoil models are not adequate to deal with the various landcovers of t he LFV.The study day of August 23, 1985 is chosen for the following reasons.1. Reliable tethersonde data from Queen Elizabeth Park in Vancouver were avail-able and showed the day exhibited classical sea breeze features, i.e., develop-ment of a shallow layer of onshore flow overlain by offshore winds during theday. Also available on this day were surface wind and temperature obser-vations obtained from an expanded wind monitoring network set up to studythermally-forced mesoscale flows in the LFV during the summer of 1985 (Steynand McKendry, 1988).2. That day was characterised by clear skies, light winds and an upper ridge aloftassociated with a surface thermal trough, a synoptic meteorological patterntypical for the ozone episode buildup in the LFV as discovered by Taylor(1991) and McKendry and Li-Ting-Wai (1993).3. It was a significant atmospheric oxidant episode day. Three out of nine ozonemonitoring stations recorded ozone concentrations in excess of 82 ppb on thestudy day and the following day. Contours depicting maximum ozone concen-trations for 23 August 1985, as shown in Fig. 1.4, indicate the severity of theChapter 1. Introduction^ 14episode.An additional reason for selecting this day is that it is the same day as studiedintensively by Steyn and McKendry (1987), Steyn and McKendry (1988). Thereforemodel output can be compared against theirs even though this is not a majorobjective of this study.• Verify the model performance with an extensive set of wind and temperature ob-servations covering the selected 24-hour period.It is critical that the model's ability to produce a realistic meteorological field bedemonstrated before any further studies based on the model output can be madewith confidence. In addition to the general qualitative evaluations, Willmott's(1985) systematic statistical model evaluation methods will be used to conduct themodel evaluation.• Investigate pollutant transport in the LFV.A Lagrangian Particle Dispersion Model (Moran, 1992) will be employed for thispurpose. Specifically, forward trajectories will be calculated to explore the hypoth-esized possibilities of heated upslope flow ventilation and pollutant recirculationsthrough sea breeze regimes. It is hoped that this study will shed light on theetiology of the air pollution episode occurring on the selected day.49.50 NChapter 1. Introduction^ 15450 500 550 600Figure 1.4: Maximum ozone concentration isopleths on August 23, 1985 in the LFV.Thick solid line refers to ozone contour, and thick dashed line refers to uncertain ozonecontour. Thinner soild line is the coastline, and thinner dashed line is the 100m terraincontour indicating the edge of the valley walls. Values beside dots are the ozone con-centrations at respective stations. Inner labels refers to the corresponding latitude orlongitude of the boundary. Outer labels are in units of UTM coordinates (x 103).Chapter 2The RAMS Model and Its Application2.1 IntroductionIn a study of the meteorological impact on regional air pollution, a detailed three-dimensional flow structure is an essential prerequisite to the understanding of the etiologyof an air pollution episode. Observations, while generally accurate and reliable, often failto provide such detailed information due to the limited observational coverage. This isespecially the case in the vertical direction. On the other hand, realistic three-dimensionalmeteorological modelling could provide us with such information at a considerably lowercost. Once validated against observations, model output could be used for many otherpurposes, e.g., studying the pollutant transport and diffusion. In this study, a three-dimensional windfield simulation will be performed for one selected ozone episode day inthe LFV with the aid of the RAMS model. The configuration of the RAMS model usedfor this study will be discussed in this chapter.2.2 RAMS ConfigurationThe RAMS model is a merging of three models that were designed to simulatedifferent atmospheric circulations. These were a non-hydrostatic cloud model and twohydrostatic mesoscale models (McNider and Pielke, 1981; Pielke, 1974a; Pielke, 1974b;Tremback et al., 1986; Tripoli and Cotton, 1982). Because of this, RAMS is a generaland flexible limited-area, finite-difference modelling system rather than a single-purpose16Chapter 2. The RAMS Model and Its Application^ 17model (Pielke et al., 1992). The atmospheric model is constructed around the full set ofprimitive dynamical equations which govern atmospheric motions, and supplements theseequations with optional parameterizations for turbulent diffusion, solar and terrestrialradiation, moist processes, sensible and latent heat exchange between the atmosphereand multiple soil layers, the kinematic effects of terrain, and cumulus convection (Walkoand Tremback, 1991). This model system is suitable for the study of thermally-forced,terrain-induced mesoscale phenomena.Table 2.1 lists the important parameters and options of RAMS used in this study.In addition, the choice of some important parameters, the input data, and an importantmodification that was made to the RAMS' surface temperature prediction and surfaceflux calculations, will be discussed.2.2.1 Model Domain and Grid StructureGrids in the model use the Arakawa type C grid stagger. Scalar variables are assignedto the centres of grid boxes while velocity components are designed to lie in the midpointsof the grid-box faces or sides normal to their direction of motion.As real topography is involved in this modelling study, a terrain following or "sigma-Z" coordinate system (Gal-Chen and Somerville, 1975) is chosen to permit more efficientuse of computer resources, and to simplify the application of lower boundary conditions.A decision has to be made first on the size of a model domain and the resolution forboth horizontal and vertical direction. Ideally the model domain should include as manyof the meteorological effects on the air pollution episodes in the LFV, and the resolutionshould be fine enough to resolve most of the terrain features and surface inhomogeneitiesin the valley. Yet the constraint of the available computer resources should also betaken into account. The LFV has its north wall, the Coastal Mountains, merge withits southeast wall, the Cascade Range, in the east. The only opening in the LFV isChapter 2. The RAMS Model and Its Application^ 18Table 2.1: RAMS meteorological model configuration used in the LFV simulations.Model Characteristic Option UsedModel mode non-hydrostaticNumerical schemeNested gridsCoordinate systemGrid dimensionsHorizontal spacingVertical spacingHorizontal advectionVertical advectionTime differencingTime stepnot usedterrain-following100 x 68 x 26(complete domain)72 x 44 x 26(domain of interest)2.5km x 2.5km100m at surface, 1.15 stretch factor,maximum spacing 2000m, 19km topsecond-order leapfrogsecond-order forwardforward-backward time-split40 s long, 8 s shortParameterizationsRadiationMoist processesHorizontal diffusionVertical diffusionSurface layerMahrer and Pielke's shortwave, longwave schemespassive water vapour only: no condensationSmagorinsky's deformation K (coefficient=0.5)Smagorinsky's deformation K (coefficient=0.5)Louis schemeBoundary conditionsLateral boundariesUpper boundaryBottom boundaryKlemp-Lilly radiativerigid topsemi-empirical surface energy balance scheme toprognose surface temperature for various landuse types.Initialization horizontally homogeneousOther aspectsStart timeSimulation length0500 PST, August 23, 198524 hChapter 2. The RAMS Model and Its Application^ 19on the west side facing Georgia Strait and Vancouver Island. The series of mountainchains outside the valley make this V-shaped valley an essentially isolated region. Withthe synoptic weather pattern usually being a stagnant high-pressure system during asevere ozone episode, air pollution buildup in the valley is mainly orchestrated by theterrain-induced mesoscale flows in the region. Based on this rationale, the model domainis selected to cover the whole LFV and a small area outside the LFV (Fig 1.1). Alsobecause of the static nature of synoptic conditions during ozone episodes, nested gridsare not used. This simplifies the simulation and avoids the possible distortion of flowscaused by the interactions between different grids.Horizontal DiscretizationWhile nested grids are not necessary for this study, problems arise for this limited-area modelling study due to the influence of the artificially imposed lateral boundaryconditions on the flows generated in the model domain. A common technique to circum-vent this problem is to extend the boundaries as far away from the domain of interest aspossible. Since a stretched grid spacing scheme in the horizontal is not available in theRAMS model, a buffer area with constant grid spacing was added outside the domain ofinterest. The terrain height at each grid point in the buffer area has been assigned tobe the same as that of the domain of interest boundary point it directly faces with theexception at the southwest corner where real smoothed terrain is imposed instead. Themodel domain of interest covers a rectangular area from 48.5 degrees latitude south to49.5 degrees latitude north, and from 121.5 degrees longitude east to 124 degrees longi-tude west. Horizontal grid spacing has been set to 2.5 km on both sides. That makes ahorizontal grid of 100 x 68 points covering a total model area of 250km by 170km withthe domain of interest consuming 72 x 44 points and an equivalent area of 182km by 110km (Fig. 2.1). The buffer area occupies a band 28 and 24 grids wide in East-West, and20Chapter 2. The RAMS Model and Its Application-50. 0 .-100.80.60.40.20.12000.-20.- 40.• e- 600.011111111111111111111111111111111111111111111111111111^11 III^(1111111150.1^1111 111111111100.- 80.Figure 2.1: Smoothed terrain contour map for the model domain that includes the modeldomain of interest (inside the dashed line) and the buffer area (outside the dashed line).Contour interval is 100 metres. Outer labels are in units of km. Symbol *refers toVancouver; -A- to Tsawwassen; 11 to Bellingham; .7., to Abbotsford; * Chilliwack; toMission City and Ly to Pitt Lake.North-South directions, respectively outside the domain of interest.Vertical DiscretizationVertical discretization uses stretched grid spacing. This stretched vertical spacingenables more detailed vertical structures of the mesoscale flows to be presented as a resultof finer spacing near the surface, and enables the top of the domain to be placed wellbeyond the mesoscale flow active zone, with coarser grid spacing in the upper levels. Gridspacing of 100 metres, stretch ratio of 1.15 and a maximum grid spacing of 2000 metresChapter 2. The RAMS Model and Its Application^ 21are assigned to this study. A total of 26 vertical levels are used. The top of the domainis therefore as high as 19 km. The resultant vertical discretization is listed in Table 2.2.Vertical grid spacing smaller than 100-metre is not permitted according to Mahrer (1984),who showed that too fine a vertical resolution may lead to a numerically inconsistentapproximation of the horizontal pressure gradient terms. To avoid this problem, heshowed that, at the surface, the smallest vertical grid interval must satisfy the requirementthat Azi > AzG, where Az1 is the vertical grid spacing in the first model layer and AzG isthe elevation difference between two horizontally adjacent grid points. In this study, somegrid points have elevation differences of greater than 100 metres between their adjacentpoints (maximum AzG is less than 200 metres). Therefore vertical spacing should be atleast 100 metres for this study.2.2.2 Initialization and Initial DataAs discussed in Chapter 1, August 23, 1985 is chosen for this modelling study. Thiswas a significant episode day characterized by clear skies and light winds aloft associ-ated with a broad high-pressure region over the British Columbia-Washington coastalzone and a thermal trough at the surface (Fig. 2.2 and Fig. 2.3). With this synopticpattern persisting over the day of study as can be easily seen on Figs. 2.2 and 2.3,ventilation in the LFV is restricted and high ozone concentrations may accumulate quitefavourably. This invariant nature of the synoptic meteorological conditions also allows usto conveniently impose a constant synoptic weather condition for the simulation period.A horizontally homogeneous scheme is adopted to initialize the simulation. It isrecognized that this initialization scheme leaves room for improvement as it assumesthe identical initial wind and temperature profiles at each grid point. In reality, thisis hardly true, given the fact that various surface inhomogeneities exist in the LFV. AChapter 2. The RAMS Model and Its Application^ 22Table 2.2: RAMS vertical discretization and initial conditions. Z denotes scalar levelsand ZZ denotes vertical velocity levels in the RAMS Arakawa C stagger in meters. p, 0,q, u and v are initial pressure (in Pa), potential temperature (in K), water vapour mix-ing ratio (in gkg'), east-west and north-south wind component (in ms -1 ) respectively,corresponding to scalar levels Z.N ZZ Z p 0 q u v1 0. -45.0 102638.0 287.19 8.0 -.14 -.202 100. 48.3 101518.3 287.19 8.0 -.14 -.203 215. 155.5 100246.4 289.39 8.1 -.45 -.364 347. 278.8 98807.3 291.03 8.0 -.55 -.455 499. 420.6 97178.2 292.12 7.8 -.65 -.316 674. 583.7 95334.0 292.86 7.5 -.70 -.147 875. 771.3 93249.6 293.77 7.1 -.71 -.018 1110. 987.0 90903.8 295.83 6.2 -.89 -.199 1370. 1235.0 88279.9 299.03 4.6 -1.08 -.0210 1680. 1520.3 85348.8 300.12 4.2 -1.31 .0011 2030. 1848.3 82083.2 302.58 4.0 -1.68 .0012 2430. 2225.6 78469.0 305.23 3.8 -2.22 .0013 2900. 2659.4 74491.6 308.38 3.6 -2.65 .0014 3440. 3158.3 70135.5 310.70 3.4 -3.00 .0015 4050. 3732.1 65387.2 312.83 3.2 -3.00 .0016 4760. 4391.9 60255.6 315.82 2.9 -3.00 .0017 5570. 5150.7 54765.6 318.69 2.4 -3.00 .0018 6510. 6023.3 48965.6 323.05 1.4 -3.00 .0019 7580. 7026.8 42943.5 329.07 0.8 -3.00 .0020 8820. 8180.8 36807.2 335.99 0.2 -3.00 .0021 10200. 9507.9 30682.1 343.96 0.1 -3.00 .0022 11900. 11034.1 24718.3 353.12 0.0 -3.00 .0023 13800. 12798.5 19051.9 362.92 0.0 -3.00 .0024 15800. 14755.6 14047.1 372.70 0.0 -3.00 .0025 17800. 16763.2 10072.0 384.49 0.0 -3.00 .0026 19800. 18763.2 7066.1 396.48 0.0 -3.00 .00Chapter 2. The RAMS Model and Its Application^ 23a)b)Figure 2.2: a) 500 mb chart for August 23, 1985 1200Z(0400 PST on August 23). Unitsof contours are in dm. b) Corresponding sea level pressure chart. Units of contours arein mb.Chapter 2. The RAMS Model and Its Application^ 24a)b)Figure 2.3: a) 500 mb chart for August 24, 1985 OZ(1600 PST on August 23). Units ofcontours are in dm. b) Corresponding sea level pressure chart. Units of contours are inmb.Chapter 2. The RAMS Model and Its Application^ 25more accurate approach would be the use of a variable initialization scheme if sufficientsounding stations were available.This initialization deficiency is presumed to be remedied within a few hours afterthe simulation is started as the model quickly adjusts itself to the strong daytime solarheating. The simulation starts near sunrise at 0500 PST. Vertical profiles of potentialtemperature and specific humidity at initialization were derived from a composite of the1200Z radiosonde sounding at Quillayute (200km to the southeast of Vancouver) and atethersonde sounding (0545 PST) from Queen Elizabeth Park in Vancouver. The inputinitial wind profile is based on the Queen Elizabeth Park tethersonde sounding, theQuillayute sounding as well as a pilot balloon sounding from Vancouver InternationalAirport, and was adjusted by an Ekman spiral wind profile. The determination of aproper initial wind field proves to be far more difficult than that of the initial temperaturefield because of the more variable nature of the wind field across the valley. However,Pielke (1984) argued that if the ratio of the advection and horizontal pressure gradientforce /0 = U2 /RST, (U is the characteristic horizontal velocity, ST is the representativemagnitude of the horizontal temperature variations across the mesoscale system andR is the gas constant for dry air ( 287 JK- 1 kg- 1 )), is much smaller than unity, anaccurate wind initialization is less important than that for temperature. Such is the casein this study, as the wind speed is of the order of only several metres per second andthe horizontal temperature difference as large as 10 degrees between land and water.This means that Jo is far less than unity, thus justifying the initialization method andprocedure. Table 2.2 lists these initial profiles together with the vertical discretization.2.2.3 Boundary ConditionsAs Moran (1992) pointed out, boundary conditions are an integral and importantcomponent of a limited-area model since time-dependent boundary conditions must beChapter 2. The RAMS Model and Its Application^ 26specified in order to close any initial-boundary value problem. While specification of thebottom boundary conditions is relatively straightforward, specification of the lateral andupper boundary conditions has been a thorny problem since the earliest days of numericalweather prediction (Charney, Fjortoft and Neumann, 1950; Oliger and Sundstrom, 1978;Platzman, 1954; Platzman, 1979). This is not too surprising considering the artificialityof conceptually 'cutting out' a finite volume of the atmosphere. The model boundariesmust then be perfectly transparent, i.e., 'open', both to flow 'information' travelling fromthe atmospheric exterior into the model interior by advection or wave propagation andto flow structures or disturbances generated within the model domain and propagatingaway from their source. Ideally, the lateral and upper boundaries should not be reflectiveor refractive in any way nor should they themselves be the source of any disturbances ornoise.Lateral Boundary ConditionsIt is very difficult, if not impossible, to make the lateral boundaries totally opengiven the artificial nature of the boundary conditions. It is therefore always desirableto move the lateral boundaries as far from the region of interest as possible. Followingthis logic, this study adds a buffer area just outside the domain of interest in order tominimize the effects of the lateral boundaries on the flows.RAMS offers a variety of lateral boundary condition options, including cyclic condi-tions in either or both horizontal directions, three different radiative boundary conditionson the normal velocity component and radiative, zero-gradient, or constant conditionson other variables, and two schemes to incorporate large-scale boundary time tendencies.Cyclic conditions are not appropriate in this real topography problem since the values ofthe dependent variables at one boundary of the model domain are assumed identicallyequal to the values at the other end. Radiative boundary conditions on the other handChapter 2. The RAMS Model and Its Application^ 27are more commonly used in mesoscale modelling because they are designed to make vari-ables at the lateral boundaries change in value so as to minimize the reflection of outwardpropagating perturbations back into the model domain.Klemp and Lilly's (1978) lateral boundary conditions are selected from three availableoptions in the model, and are applied to all variables. This boundary condition hasthe advantage that the predicted vertical profile of momentum at the boundary is alinear combination of the previous flow at the boundary and the flow at the vertical gridcolumn adjacent to the boundary. Unlike Orlanski's (1976) or Klemp and Wilhelmson's(1978) radiation condition, large imbalances of horizontal momentum flux integrated inthe vertical which lead to domain scale pressure trends do not occur. This seems to beespecially valuable in simulations of flow over topography where a mountain barrier tendsto rearrange the vertical profile of horizontal momentum severely.Upper Boundary ConditionsThe top of the mesoscale model, as with the lateral boundaries, should be movedas far as possible from the region of significant mesoscale disturbance. In this study,the top of the vertical domain has been placed as high as 19 kilometres, well inside thestratosphere. Such a setup provides a deep layer of stable thermodynamic stratificationwhich could effectively minimize the vertical advection or propagation of mesoscale dis-turbances from reaching the top of the domain. Therefore the choice of a proper upperboundary condition becomes less important in this case; rather, economics and efficiencyoccupy more weight in making the choice. Among four available options provided byRAMS, only "wall on top" and the "Klemp-Durran radiative" upper boundary condi-tions could be used for the nonhydrostatic mode. "Wall on top" is selected as the topupper boundary condition for this study. Sensitivity studies showed that model studiesusing the " wall on top" option presented no appreciably different results from that withChapter 2. The RAMS Model and Its Application^ 28the Klemp-Durran radiative upper boundary condition, while consuming significantlyless CPU time.Bottom Boundary ConditionsThe realistic specification of bottom boundary conditions is important when terrainforcing is significant, as it is expected to be in this study. The imposed kinematic bottomboundary conditions used are the "no slip" condition in the tangential direction and zerovelocity in the normal direction. For the surface temperature and surface fluxes, theRAMS model has two multi-level soil models to predict surface temperature and soilmoisture for 12 classes of soil and uses surface-layer similarity theory to diagnose surfacefluxes. As much of the LFV is urbanized or covered in vegetation, applying a pure soilmodel to various complex land surfaces is inappropriate. In this study, soil models werereplaced with a semi-empirical energy balance model which is able to predict the surfacetemperature and surface fluxes for five typical landuse types in the LFV. The details ofthis model will be discussed later. As a first step, terrain, and surface properties for eachlanduse type will be discussed below.Terrain and Surface Properties Terrain height data were obtained from a 1km dataset developed by Ministry of Energy Mines and Resources, Canada. The datawere first averaged to 2 km using 4 points (2 x 2), and then interpolated to 2.5 kmusing an inverse distance weighting method. The resultant dataset is smoothed by abinomial filter to eliminate wavelengths less than two grid lengths. This smoothingprocedure is necessary to maintain computational stability within the numerical scheme.The smoothed terrain contour map for the domain of interest is displayed in Fig. 2.1,together with the buffer area discussed before. With this resolution, the principle featuresof the regional terrain are retained.Chapter 2. The RAMS Model and Its Application^ 29Table 2.3: Roughness Length and Albedo for Each Land-use Type.Land-use Type Urban Suburban Rural Agricultural ForestRoughness (m) 1.8 0.6 0.5 0.25 1.5Albedo 0.15 0.18 0.19 0.23 0.15One important surface inhomogeneity is the water and land contrast. Thereforea 2.5km x 2.5km resolution "percentage-of-land" data file was built up for the modeldomain from the UTM grid system maps.In addition, various surface inhomogeneities over land should be identified and in-cluded in the model application if simulations are to be in correspondence with obser-vations. Five broad landuse categories representing urban and suburban environments,forested mountain regions, and rural and agricultural areas have been recognized in themodel region as being shown in Fig. 1.1. Two important surface properties, aerodynamicroughness length and albedo are assigned for each landuse type through survey, observa-tions (roughness for forested and suburban areas in the LFV) (Lee, 1992; Steyn, 1980),and consultation with experts and recognized values for similar landuse types (Lewisand Carlson, 1989; Oke, 1989; Steyn and McKendry, 1988; Wieringa, 1991). Table 2.3displays these two surface property values for each landuse type.Surface Temperature and Fluxes As mentioned earlier, an effort has been madeto develop a realistic, yet semi-empirical surface energy balance model to predict surfacetemperature and surface fluxes for each landuse type. This model is made possible bythe recent developments and extensive observational studies of the surface energy bal-ance in this region (de Bruin, 1983; de Bruin and Holtslag, 1982; Camuffo and Bernardi,1982; Cleugh, 1990; Deardorff, 1978; Grimmond, Cleugh and Oke, 1991; McCaughey,1985; McCumber, 1980; McNaughton and Spriggs, 1987; Monteith, 1981; Oke, 1989). AChapter 2. The RAMS Model and Its Application^ 30complete description of the model and the specific surface properties used in the modelfor each landuse type are given in Appendix A. In brief, this model adopts Mahrer andPielke's (1977b) short- and long-wave radiation parameterization scheme to calculate thenet radiation for every 600 seconds (15 times the long time step). It relates the storageheat flux AQs directly to net radiation Q* by an objective hysteresis model (Grimmond,Cleugh and Oke, 1991), and the latent heat flux QE to the difference between net ra-diation and the storage heat flux through the Priestley and Taylor equation (Priestleyand Taylor, 1972). The remaining term, the sensible heat flux QH, is calculated throughLouis'(1979) analytic functions. Ground surface temperature is then solved in this sur-face energy balance equation through iteration. This surface temperature is, in turn,used to obtain the needed fluxes. Care has been taken to make this model compatiblewith the rest of the RAMS model code.Chapter 3Results and Model ValidationThe model is evaluated based on a comparison between modelled outputs and ob-servations. Particular focus is placed on the model's ability to reproduce the observedforced mesoscale flows in the LFV.A 24-hour simulation was carried out for the period starting at 0500 Pacific StandardTime (PST) 23 August 1985 (near sunrise). Prior to the scheduled run, dynamic initial-ization periods of one-hour, two-hour and four-hour were applied, as suggested by Steynand McKendry (1988) in order to achieve realistic nocturnal flow features at sunrise.Unfortunately the results showed that the predicted nocturnal winds were too strong.Therefore dynamic initialization was not used in the final model run. Model output wasrecorded every hour. The model run was performed on an IBM RS/6000, and consumedabout 24 hours of CPU time.Extensive observations were carried out in the LFV during the period of interest(Steyn and McKendry, 1988). The observation network is displayed in Fig. 3.1. Sur-face winds and directions at a height of 10 m above ground level (AGL) were monitoredcontinuously at 25 locations on the valley floor. Eight of the stations operated by theUniversity of British Columbia also monitored temperature with identical shielded ther-mistors installed at the same height. In addition, direct measurements of the energybalance components Q* (surface net all-wave radiation) and QH (sensible heat flux) wereavailable until midafternoon from a suburban location. The storage heat flux AQ s wascalculated using a hysteresis model developed by Grimmond, Cleugh and Oke (1991)31500 550 600i^ 1^ i^ i450ac‘iNrU,anemometers onlyanemometers and thermometersacoustic sounderstethersondeenergy budgetaa_a0co_coU)Chapter 3. Results and Model Validation^ 32Figure 3.1: Observation network for August 23 1985 in the LFV. The solid line is thecoastline. The dashed line is the 100 metre terrain contour, showing the edge of thevalley wall. Outer labels are in units of UTM coordinates (x10 3 ).Chapter 3. Results and Model Validation^ 33while QE (the latent heat flux) was calculated from residuals. Vertical variations in windspeed and direction, temperature and humidity were derived from tethersonde soundingscarried out at Queen Elizabeth Park, Vancouver during the daytime. Soundings weremade continuously and took about one hour for an ascent and descent. Three acousticsounders located in the valley were also used to monitor the vertical thermal structureand to give an indication of mixed layer depth.Evaluations of model performance are made both quantitatively and qualitatively.Comparison between this model performance and that of Steyn and McKendry (1988)has also been conducted briefly. Model evaluation is necessary in order to establish thecredibility of the RAMS output and enable us to use the validated output to performfurther investigations.3.1 Quantitative Evaluation of Model PerformanceWillmott (1985) recommended that quantitative evaluation of model performancebe based on the following groups of parameters:• Observed and modelled means and standard deviations;• Total, systematic, and unsystematic root mean squared differences (RMSD, RMSD S ,RMSDu ) between observed and modelled values (scalars or vectors). Total RMSDrepresents the total difference between modelled values and observations. The sys-tematic RMSD represents model linear bias while the unsystematic RMSD repre-sents model precision. These parameters are defined as (Willmott, 1982; Willmott,1985):NRMSD = [N Ip i —2=10.5(3.1)Chapter 3. Results and Model Validation 34RMSDs =N^0.51 E 'p i - oid^(3.2)1^N0.5RMSDu = E^- P=1 2^(3.3)Iwhere N is the number of stations; p i and o f are modelled and observed variablesrespectively; p i is the ordinary least square estimate of p = a + bo i , where aand b are the intercept and slope). Areal weight has been assumed to be unity foreach station.• The Index of Agreement d, which is defined as:EL 113, oz12 d = 1^ (3.4)E liv=1(113z — 6 1+1 02 — 6 1) 2where ö is the simple mean of the elements in o. This dimensionless index has atheoretical range of 0.0 (for no agreement) to 1.0 (for perfect agreement).It must be pointed out that this statistical method assumes that all observations areerror-free and that all errors come from model simulations. Therefore it is very importantto filter out the erroreous observational data before applying the statistics.According to Willmott (1981), this set of statistical measures is necessary and suffi-cient for the complete assessment of model performance. These validation statistics havebeen applied in atmospheric modelling by Steyn and McKendry (1988), Ulrickson andMass (1990b) and Jackson (1993). This study will employ the same method to com-pare RAMS surface wind and temperature fields, with observed surface data, covering24 hours during which a sea-breeze circulation was observed in the LFV. All data arehourly means. Before comparison, modelled fields were extrapolated to the observationheight of 10 metres AGL from the lowest model level of 48.3 metres. Businger's profilefunctions (Businger et al., 1971) are used to make the extrapolation. Friction velocityMonin-Obukhov length L, scaling temperature O. and surface potential temperature 00Chapter 3. Results and Model Validation^ 35are extracted from the model results and applied to Businger's profile functions togetherwith specified roughness length zo. The extrapolation is made within the RAMS modelrun.3.1.1 Surface Wind EvaluationFigures 3.2a and b display time series of observed and modelled surface mean winddirection and speed on August 23, 1985. From Fig. 3.2a, it can be seen that the modelledaverage wind direction is in close agreement with that observed throughout the dayexcept near sunrise (transition period from land breeze to sea breeze) and near midnight(transition period from sea breeze to land breeze), when winds in the valley are verylight (less than 1 ms'). Wind direction is almost persistently westerly during the day,indicating the prevalence of the sea breeze. Modelled average wind speeds follow theobserved pattern quite closely in most hours with a maximum of about 3 ms' occurringat about 1500 PST (Fig. 3.2b). The relative magnitudes also show good agreement butwith a slight tendency for the model to overpredict the wind speed around the peak andafter 2100 PST at night.Figures 3.3a-c show, respectively, the wind standard deviation, root mean square dif-ferences and Index of Agreement. The modelled wind standard deviation is less thanthe observed value by up to 1 ms -1 until 1900 PST. This is to be expected as mod-elled surface winds represent the winds over a volume of 2.5km x 2.5km x 48.3m (laterextrapolated to 10m) and do not include subgrid scale effects, while observations mustbe significantly affected by local surface conditions. At night, however, the transitionfrom sea breeze to land breeze and from upslope to downslope make the wind field inthe valley very variable. This effect, together with the slight overprediction of modelledwind speed, may be the cause of the larger modelled standard deviation at night. Thetotal RMSD in Fig. 3.3b indicates that the model capably simulated the observed windChapter 3. Results and Model Validation^ 36field, with total RMSD values no greater than 2 ms -1 thoughout the simulation. Thesystematic component RMSDS is the major contributor to the total RMSD. The smallunsystematic component RMSDU indicates good precision. The Index of Agreementis another important tool in evaluating model performance and is especially useful inmaking cross-comparions between models (Steyn and McKendry, 1988; Willmott, 1985).Figure 3.3c shows that except for the first two hours of the simulation when the modelwas adjusting itself to the solar radiation and overcoming the inexact initialization, theindex stays consistently around 0.6 with an average of 0.58.Steyn and McKendry (1988) studied the same event in the LFV with a hydrostaticColorado State University Mesoscale Model (CSUMM) and used the same statisticalmethods to evaluate their model results. For the wind field evaluation, their results hadan Index of Agreement ranging between 0.41 to 0.61 and with a average of 0.51. Clearlythe present modelling study has an improved overall performance.3.1.2 Surface Air Temperature EvaluationSurface heating is an important driving force in inducing sea breezes and slopewinds. Its realistic representation by the model will influences the success of wind fieldsimulations. A direct index of such surface heating is the near surface air temperature(10 metres AGL).Figure 3.4a presents the evolution of observed and modelled near-surface air tem-perature, and shows relatively close agreement between them, with peak temperaturesoccurring at 1600 PST in both cases. Modelled temperatures tend to underestimate theobserved values in the daytime by up to 3 degrees and overestimate the observed values atnight. A comparison of standard deviations in Fig. 3.4b reveals that the observed fieldshave greater spatial variation than the modelled fields, presumably because the modelonly dealt with five broad landuse types in the LFV while in reality every station mayChapter 3. Results and Model Validation^ 376^8 10 12 14 16 18 20 22 24TIME (PST)a)b)d-coCVT06^8 10 12 14 16 18 20 22 24TIME (PST)Figure 3.2: The evolution of modelled and observed: a) average wind direction, b) averagewind speed for August 23, 1985.A^Modelled0 Observed■-A AChapter 3. Results and Model Validation^ 38a) -6 8 10 12 14 16 18 20 22 24TIME (PST)b)c)6^8 10 12 14 16 18 20 22 24TIME (PST)6^8 10 12 14 16 18 20 22 24TIME (PST)Figure 3.3: Statistic comparison of modelled and observed wind as time series of a) stan-dard deviation of wind speed, b) total (RMSD), systematic (RMSDs) and unsystematic(RMSDu) root mean square deviations and c) index of agreement for August 23, 1985.Q ModelledO Observed.•-Chapter 3. Results and Model Validation^ 39 0 -22• NCDE o_■• .^•^V, ,1; Ole 4141,^h,07,1 ••■•-.1.A. aQ ModelledO Observed06^8 10 12 14 16 18 20 22 24TIME (PST)a)b)6^8 10 12 14 16 18 20 22 24TIME (PST)Figure 3.4: The evolution of the modelled and observed: a) average temperature, b)standard deviation of temperature for August 23, 1985.RMSDA RMSDso RMSDu11tiz—INNIROK.a.^•b)00CD000Pi'ait4linChapter 3. Results and Model Validation^ 4006^8 10 12 14 16 18 20 22 24TIME (PST)a)6^8 10 12 14 16 18 20 22 24TIME (PST)Figure 3.5: The evolution of the modelled and observed: a) total (RMSD), systematic(RMSDs) and unsystematic (RMSDu) root mean square deviations of temperature, b)index of agreement for August 23, 1985.Chapter 3. Results and Model Validation^ 41be exposed to a different surface environment. The less variable modelled wind fields, aspreviously mentioned, may also be the cause for the smaller variation in the modelledtemperature fields. It is worthy of mention that the modelled standard deviation tendsto increase with time and equals the observed value at 2000 PST, thereafter decreas-ing. The pattern of the RMSD and the clear dominance of the systematic RMSDS inFig. 3.5a show that the discrepancy between modelled and observed temperatures comesmostly from the underestimation of the model near the peak and from the overestimationat night. Again, model precision remains high as the unsystematic RMSD remains lowthroughout. The Index of Agreement (d) (Fig. 3.5b) is low near sunrise and at nightbut remains higher than 0.55 during most hours with a maximum index of 0.79 at 1800PST when the modelled temperature mean matches the observed one exactly and whenthe RMSD is a minimum. Steyn and McKendry's (1988) temperature Index of Agree-ment ranged from 0.11 to 0.74, with a mean of only 0.34. This study has a mean d fortemperature of 0.54, with a range of 0.31 to 0.79. This is another strong indication thatRAMS performs a much better job than CSUMM.3.2 Qualitative Evaluation of Model PerformanceA visual comparison of the observed and modelled wind fields is an important sup-plement to the above quantitative evaluations. It gives a direct picture of how well themodelled field matches the observed one. In the following two sections, surface wind andvertical profile visualization will be presented.3.2.1 Surface WindsTwo major differences between the modelled and observed surface wind fields mustbe taken into account when comparing these fields. Firstly, modelled surface wind fieldsChapter 3. Results and Model Validation^ 42represent instantaneous "snapshots" while observed wind fields are hourly means validfor the previous hour. Secondly, modelled wind fields are collected at the lowest modellevel of 48.3 metres above ground whereas the wind observations were made at 10 metresAGL.This complicates the comparison but does not result in significant discrepancies forthis visual comparison. The magnitude of the modelled wind should be slightly smallerat 10 metres but its direction should be much the same. The instantaneous modelledwind field should be similar to the hourly averaged one as long as the winds are notchanging drastically over the course of an hour.Figures 3.6a-h depict a series of surface wind fields from 0900 PST to 2300 PST onAugust 23 1985 at 2-hour intervals. Observed wind vectors are represented by bold arrowswhich overlay the modelled wind fields. As can be seen in Fig. 3.6a, thermally-drivensea breeze and upslope flows have been built up 4 hours after sunrise yet are still veryweak. As time progresses, the westerly sea breeze from the Strait of Georgia strengthenswhile upslope winds also increase in strength and flow up the valley walls (Figs. 3.6c-h).These two winds merge at some point and constitute an integral wind field. Channelflow blowing eastward along the valley also forms as a result of the interactions amongthe valley wind, the pumping effect of slope winds on both the southeast and northeastwalls, and the westerly sea breeze. Around 1500 to 1700 PST, one or two hours after themaximum surface heating, sea breeze and upslope winds come to their maximum (Figs.3.6d and e). Thereafter the sea breeze continues with diminishing strength until around2100 PST when it almost ceases and strong downslope winds come down from valleywalls to drive the wind seaward. Weak land breezes then start to build, while a streamof southeastward wind continues to flow over the Strait of Georgia.In the daytime, both the magnitude and direction of the model surface winds showgood agreement with observations for most of the 25 stations. They do not coincideChapter 3. Results and Model Validation^ 43OCDLf)O03•gtLC)b)CDCDcoU.)a)Co450 500 550 600Figure 3.6: (a-h) The observed and modelled surface winds at a) 0900 PST and b) 1100PST of August 23, 1985. Bold arrows are the observed winds. The dashed line is the100 metre terrain contour, refering to the edge of the LFV wall. The solid line is thecoastline. Outer labels are in units of UTM coordinates (x103 ). Inner labels are latitudeor longitude values for the boundaries. The spacing between vector tails represents 7.50U)0lAOOOctLOOtoOOLC)OMa)48.50 N48.50 NChapter 3. Results and Model Validation^ 4449.50 Na)00t -'LOb)450 500 550 60049.50 N450^500^550^600I iFigure 3.6: c) 1300 PST and d) 1500 PST of August 23, 1985.Oco _InInONt- _CoO_49.50 NOcoTr —toOCoO_toOO _•trOco_coCAOco _toChapter 3. Results and Model Validation^ 4549.50 Na)OO _CAb)450 500 550 600450^500^550^600Figure 3.6: e) 1700 PST and f) 1900 PST of August 23, 1985.450 500 550 600OCO-L)OCD-Co49.50 N/ /^/ / /^/ I/ ; .1 /^/ /Ovi-v-\\\\^\ \*I •v •\\\\\ ‘‘.. N.N.'84 - /^""_toONI- -to 5Oco -to1 1^1 \11 11^N. N. \^• / / /\^/I:^1\^\^\t8.43 Nl \^N1 1• I 1- // /OCD _-toONI- -toI^1^1\ 1 \\^t49.50 N/ I^/ /1^/ /O4-to'84 •Octo/^•• NN. N.OTr -toco-O to ' / /1\ • 10 1 1\ 1\1 1\,,,k N1 1Chapter 3. Results and Model Validation^ 46q)a)450^500^550^600Figure 3.6: g) 2100 PST and h) 2300 PST of August 23, 1985.Chapter 3. Results and Model Validation^ 47as well at night when the model simulated flows along the Strait of Georgia that aresignificantly stronger than observed (Fig. 3.6g and h). This problem was also encounteredby Steyn and McKendry (1988) and may be due to the over-simplified surface heat fluxrepresentations over the sea. Their daytime simulations also suffered problems such asoverestimating the strength of the sea breeze and unrealistically strong downslope windsalong the northwest wall in the daytime. As can be seen in Figs. 3.6a-h, these twoproblems have been alleviated in this study.3.2.2 Vertical ProfilesIn an attempt to investigate the model's ability to simulate vertical wind profiles,this section conducts a vertical profile comparison at a single urban site in Vancouverwhere tethersonde measurements were made to provide instantaneous observed verticalprofiles. It should be understood that to expect exact duplication of the instantaneouswind profiles is not practical, considering that even the instantaneous observed profilestook about 15 minutes to complete, and that randomness was inevitably involved inmonitoring the wind. Therefore, profile patterns and approximate magnitudes are themain focus.Figures 3.7a-d display the observed and modelled wind profiles of the east-west com-ponents at Queen Elizabeth Park for the hours 0900, 1200, 1500 and 1800 PST. Generalpatterns of the modelled profiles are seen to match those of the observed ones surprisinglywell in all cases. Moreover, sea breeze and return flow aloft are clearly simulated andthe modelled height of zero wind is roughly in agreement with that observed, especiallywhen the sea breeze is strong (Fig. 3.7b and c). However, it is also obvious that themodel overpredicts the strength of the observed sea breeze and return flow. A glimpseof the north-south wind component (Fig. 3.8a-d) shows that this component is verysmall compared with east-west component, indicating the dominance of the westerly seaChapter 3. Results and Model Validation^ 48breeze. While the small and variable north-south component makes the exact simulationvery difficult, the simulated patterns do agree in gross term with the observed ones.Steyn and McKendry (1988) also performed a comparison of vertical profiles usingthe same observations. Results from this study demonstrate a clear improvement overtheir simulations. While one single-point vertical profile validation is not sufficient toidentify the success of the model simulation in the vertical direction, it does give us anindication of the RAMS' capability to simulate the vertical structure.3.3 ConclusionsThis chapter shows quantitative and qualitative comparisons between observationsand RAMS model output for the episode day of August 23, 1985 in the LFV. The resultsshowed that, overall, RAMS is able to simulate real-world mesoscale flows in complexterrain. Model results show general agreement with the observations made on this day.Comparison with Steyn and McKendry's model performance evaluation reveals that theRAMS model performance was better in almost every respect.E -6-6^-4^-2^0^2^4U component (m/s)0900 PSTa)CMo000 --6^-4^:2^0 2^4^6U component (m/s)b)0LC)CV00CM1200 PSTLO^ LOCV1500 PSTCV01800 PSTChapter 3. Results and Model Validation^ 49c) d)-6^-4^-2^0^2^4^6^-6^-4^-2^0^2^4^6U component (m/s) U component (m/s)Figure 3.7: Profiles of the east-west component of wind velocity at Queen Elizabeth Parkin Vancouver at a) 0900 PST, b) 1200PST, c) 1500 PST, and d) 1800 PST. The solidline refers to the modelled profile; The dashed line to the measured profile. Positive windspeed value indicates a westerly wind.-6^-4^:'2 0^2 4V component (m/s)-6^-4^-2^0^2^4V component (m/s)Chapter 3. Results and Model Validation^ 50:6 :4 :2 0^2^4 6V component (m/s) b)0o_0(NJ0900 PST00_U,000 -00 -tf)0 --4 -2^0^2^4^6V component (m/s)d)Figure 3.8: As in Fig. 3.7 but of the north-south component of wind velocity. Positivewind speed value indicates a southerly wind.Chapter 4Trajectory StudyThis chapter studies the pollutant transport in the LFV with the RAMS mesoscaleLagrangian particle dispersion model (MLPDM). The validated meteorological fields inthe LFV as described in Chapter 3 are used as input to the MLPDM model.4.1 The RAMS Lagrangian Particle Dispersion ModelMLPDM is compatible with the RAMS meteorological model, and has been codedas a component of the RAMS visualization and analysis (VAN) postprocessing package.It is based on the MLPDM developed by McNider (1981) and uses a discrete form ofthe Langevin equation. Meteorological information is supplied by RAMS: these meteo-rological fields are used both to specify the resolved grid-scale flow field and to parame-terize subgrid-scale turbulence quantities required by the MLPDM (Moran, 1992). Moredetailed descriptions of the formulation of this model and its application are given inMoran(1992), McNider et al. (1988) and Pielke (1984)The particle model can be used simply as a Lagrangian analysis package to calculateparcel trajectories based on gridded wind fields from the RAMS meteorological analysisfiles. It can also simulate one-particle, subgrid-scale atmospheric diffusion by releasing,and then tracking, a set of passive, independent tracer particles from one or more sources.Differential advection and shear-enhanced dispersion may occur due to grid-scale resolvedvelocities (Moran, 1992).The trajectory mode in the MLPDM model is used in the present study. The MLPDM51Chapter 4. Trajectory Study^ 52turbulence parameterization is turned off and particles are advected solely by the time-varying grid-scale RAMS wind fields. RAMS wind fields were available every hour of thesimulation and linear interpolation was employed in time (the model does this automati-cally) to ensure a gradual transition between consecutive wind fields. The MLPDM timestep was set to 15 s for calculating these 'grid-scale' trajectories.4.2 General Pollutant Transport Investigation in the LFVA survey of the general patterns of the pollutant transport in the LFV is first con-ducted. Forward trajectories of fifteen particles, released at three near-coastal sites, andat five different heights, are displayed in Figure 4.9. Their paths are remarkably differ-ent. Based on the meteorological fields depicted in Fig. 3.6, it appears that the lowerthree trajectories at each station are influenced by the mesoscale flows while the uppertwo trajectories at any of the three locations (Trajectories '4','5','9','10','14'and '15'), aredriven out of the LFV mainly by easterly synoptic winds. The upper-level transport ofpollutants does not appear to assist in the development of a pollutant episode and sowill not be investigated further in this study. It is apparent from Fig. 4.9 that in thelower levels, not all source areas contribute to pollutant buildup in the LFV. Trajectories`11,`12','13' originating over Bellingham in the United States leave the LFV in less than14 hours. The same is true for trajectory '8' which travelled well out of the LFV. Twoother trajectories ('6' and '7') released from Tsawwassen headed southeast and up theslope of the Cascade Range, giving no indication of being returned to the valley. Thethree trajectories demonstrating the potential for recirculation were all from Vancouver( '1','2' and '3').The survey of the general patterns of pollutant transport is extended to the easternLFV. This time all particles are released 50 metres AGL at three eastern locations on80.60.40.20.0.- 20.- 40.-60.-80.Chapter 4. Trajectory Study^ 53-100 - 90 0 90 100Figure 4.9: Plan view of 15 three-dimensional grid-scale trajectories at three coastallocations and at five different heights. Each trajectory starts at 0800 PST 23 August1985 and lasts for 14 hours. Trajectory ending point labels '1','2','3','4','5' indicate thefive release heights of 176m, 328m, 670m 1103m and 1673m above sea level (ASL) at onesite in Vancouver. '6' to '10' sequentially are for trajectories released at Tsawwassen atheights of 130m, 335m, 660m, 1094m and 1668m ASL correspondingly. '11' to '15' are fortrajectories near Bellingham, USA at heights of 206m, 407m, 733m, 1174m and 1783mASL, respectively. The dots on the trajectories indicate the hourly trajectory positions.The background is the 200m-interval terrain contour map for the LFV and the modelbuffer area.a)80. -c60. -40 .-20 . -C-40. ^- 60. E^20.-100. -50. 0. 50. 100.3.0 -2.52.0 -1.5 -1.0 -0.5 -1^111111111111^1111111111111111111111110.0Chapter 4. Trajectory Study^ 54- 80 • -C11111111y1111111111111111111 i11111111111111111111111 11 1^1111111 lily 1111111- 1100.^-50.^0. 1 50.1111111111111111100.b)Figure 4.10: Six three-dimensional grid-scale trajectories released about 50m above theground at three eastern locations of the LFV and at two different times. a) plan view,and b) west-east vertical slice viewed northward. 'A', 'B' and 'C' refer to releasingstations near Mission City, Abbostford and Chilliwack, respectively. '1', '2' and '3' are thecorresponding trajectories released at 0800 PST 23 August 1985 while '4', '5' and '6' arefor releases at 0900 PST. Each trajectory lasts for 19 hours. The dots on the trajectoriesindicate the hourly trajectory positions. The background on a) is the 200m-intervalterrain contour map for the LFV and the model buffer area. The background on b)refers to the terrain on the slice that cuts through station "C". Labels are in units of km.Chapter 4. Trajectory Study^ 55the valley floor. In addition, two releases are made at 0800 PST and 0900 PST onAugust 23 1985 at each station to see if there are any distinct differences resulting fromrelease time. Figures 4.10a and b depict the resultant trajectories in a plan view and awest-east vertical slice, respectively. The three particles released at 0800 PST are similarto each other in terms of the respective trajectories and destinations after 19 hours.Each of these trajectories leaves the valley. Releases made at 0900 PST produce similartrajectory patterns, with the exception of station 'A'. Unlike trajectory '1', released at0800 PST, which later moves with the upper-level southwestward flow, trajectory '4' iscaught in downslope winds at about 1900 PST and is then driven back to the valleyfloor. It is possible that pollutants released around station 'A' after 0900 PST on theday of study may assist in episode buildup because they stay in the valley. On the otherhand, emissions released around stations 'B' and 'C' may not be important in episodeformation.The following conclusions may be drawn from these preliminary trajectory studies.1. High-altitude emission sources are not the major concern in episode buildup sincethey usually move with the synoptic wind in the upper level and eventually leavethe valley.2. Low-level emissions from the Vancouver area and its eastern neighbours are thepossible areas which may contribute significantly to the episode because of theirlocations in relation to air flows that recirculate pollutants.3. Other parts of the LFV do not seem to contribute as much to the elevated airpollution levels on and after the study day.4. A different release time may result in a very different pollutant transport in thevalley.Chapter 4. Trajectory Study^ 56Further pollutant transport investigation will therefore focus on the pollutant emissionin the Vancouver area and on the influence of the different release time on the pollutanttransport.4.3 Intensive Investigation of Pollutant Transport in the LFVTo investigate in more detail how different release times affect pollutant transport,a near-surface source in the Vancouver area will be examined first. Figures 4.11a and bgive plan and elevation (northward) views of seven trajectories released at seven one-hourintervals beginning at 0800 PST. These two pictures reveal two categories of pollutanttransport based on release time. If released before 1200 PST as in trajectories '1' to '4',pollutants first move up the Coastal Mountains through the combined effect of both thestrong sea breeze and upslope winds. They are able to travel high enough to be capturedin the sea breeze return flow and the assisting easterly geostrophic wind at about 1700PST, and are driven westward at a height of about 800 meters above ground. On theother hand, if released on or after 1200 PST, pollutants do not seem to have enoughtime and momentum to climb as high along the mountain slopes before being capturedby downslope winds at about 2000 PST. They are then driven back to the valley, 30to 50km to the southeast of their origins. Emissions trapped in these parcels will likelycontribute to pollution levels of the following day.4.3.1 Dual Release-time Study of Pollutant Transport in VancouverHaving observed a large difference in pollutant transport for different release times ata single site in Vancouver, it is of interest to see if this is true for the whole Vancouver area.In order to carry out an experiment, a rectangular area of 25km x 15km in Vancouveris selected. Emission sources are evenly distributed in a regular 5 x 3 array, with 5km-50.-100. 0.111111150.80.60.40.20. -'-0 .-20.- 40.- 60.  I^11111 1 111111111100.- 80 . -s-r111111111111111111111111111y11111111111111111111 1 11Chapter 4. Trajectory Study^ 57a)b)-100. -50. e. 50. 100.Figure 4.11: Seven three-dimensional grid-scale trajectories 50m above the ground re-leased at one site in Vancouver at seven one-hour intervals starting from 0800 and followedfor up to 19 hours. a) plan view, and b) west-east vertical slice viewed northward. Tra-jectory endpoint labels '1' to '7' indicate the seven release times from 0800 to 1400 PST,respectively. The dots on the trajectories indicate the hourly trajectory positions. Thebackground on a) is the 200m-interval terrain contour map for the LFV and the modelbuffer area. The background on b) refers to the terrain on the slice that cuts throughthe release station. Labels are in units of km.Chapter 4. Trajectory Study^ 58Figure 4.12: Locations of 15 emission stations in Vancouver that form three groups forthe trajectory study. The solid line is the coastline, and the dashed line is the 100mterrain contour. Outer labels are in units of UTM coordinates (x 10 3 ).spacing (Fig. 4.12).Results are analyzed in terms of each east-west 5 x 1 array. The most southerlyarray is designated as the first group; the middle array is the second group; and themost northerly array is the third group. A paired study is conducted for each group bysetting two different release times; one at 0800 PST, the other at 1200PST. In the end,inter-group comparison is made to see if there is any notable difference in the pollutanttransport due to different source locations.Figures 4.13 to 4.15 show, from three different view points the trajectories of particlesreleased from five evenly-spaced near-surface sources of the first group at a) 0800 PSTChapter 4. Trajectory Study^ 59and b) 1200 PST on 23 August 1985 and followed for 19 and 17 hours, respectively. InFigs. 4.13-4.15a, three out of five particles released at 0800 PST are later captured in thereturn flow of both the sea breeze and upslope winds, but do not seem to travel very far.The other two particles numbered '1' and '2', are caught in the downslope wind alongPitt Lake and are driven back down towards the valley. This means that for this studygroup, even pollutants released in the early morning can be recirculated in the valley.For releases at 1200 PST, all trajectories are returned to the valley floor to the east ofthe source region and most appear to converge in the vicinity of Abbostford.Figures 4.16 to 4.18 depict similar information to that in Figures 4.13 to 4.15, but fortrajectories of the second group. For the 0800 PST release, Figs. 4.16-4.18a show onlyone trajectory '3' is captured in the sea breeze return flow, but still remains in the valley.The other four trajectories are clearly returned to the valley. For the 1200 PST releasecase, not surprisingly, all trajectories return to regions east of Vancouver. Trajectorynumber '5' travels up and down Pitt Lake, suggesting that this valley might be affectedby a pollutant episode.Results from the third group are depicted in Figs. 4.19-4.21. Except for two trajec-tories numbered '4' and '5' from the morning release that travel southward to the Straitof Georgia, all other trajectories remain in the valley.4.3.2 Summary of the Dual Release-Time StudyContrary to initial investigations conducted at the beginning of the section, morningemissions may also have a substantial influence on the recirculation pollutants. Afternoonemissions appear to reduce the air quality in the LFV to a greater extent; every trajectorysimulated for the study day returns to the east of Vancouver without exception, resultingin a carry-over of pollutants to the following day. The spatial locations among emissionsin the Vancouver area do not affect the pollutant transport pattern significantly.b)80.60.40.20.0.-40.-60.-80 . -"-tiiiiinviiiiiiiimiliftl(1miltIIIH111111iyi-100.^-50.^0.1111 III III 11111100.111111111111 111 II 11150.Chapter 4. Trajectory Study^ 60a)80.60.20.- 20.- 40.Iit^s^iitityitttiw -60. ^-100.^-50.^0. 50.^100.Figure 4.13: Plan view of the three-dimensional grid-scale trajectories from five sourcesof the first group in Vancouver. Released near the surface at a) 0800 PST, and b) 1200PST 23 August 1985 followed up to 19 hours. Trajectory end-point labels `1' to '5'indicate the five stations spaced from west to east at 5km interval. The dots on thetrajectories indicate hourly trajectory positions. The background is the 200m-intervalterrain contour map for the LFV and the model buffer area. Labels are in units of km.a) 1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.21111111111111111111111111111111111111111111111li11111111i 1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 --0.0Chapter 4. Trajectory Study^ 610.0^Himulim11111111H1H1111111^IIII111111. 11111111111111111111111111111111111111IIIIIIIII-100.0^-50.0 0.0^50.0^100.0-100.0^-50.0^0.0^50.0^100.0b)Figure 4.14: Same as Fig. 4.13 but the XZ projection that cuts through the secondemission group on Fig. 4.12. Viewpoint is northward.-80.0^-60.0^-40.0^-20.0^0.0 20.0 40.0^60.0^80.00 . 0 ^1111111111.4-1-1-4-111111111111111111111111^1111111111111111111111a)Chapter 4. Trajectory Study^ 62b)1.21.0 -0.8 -0.6 -0.4 -0.2 0.0 ^11111111,11.4-1-4-4-111Itilliwillillilltil -80.0^-60.0^-40.0^-20.0^0.0^20.0^40.0^60.0^80.0Figure 4.15: Same as Fig. 4.13 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward.0. -1 1 1 1 1 11- 80. - -`tilititlyttilittititittlimittititittitttititttyl-100.^-50.^0.- 60.11111 111 11111150.1111111)11111111100.60.40.20.0.80.60. -=40. -E20.-ca)b)Chapter 4. Trajectory Study^ 63111./1111111111111111111111y1111111111111111111?1^11^1111111^11111 1111111^I^itillyttlittli-100.^- 50.^0. 50. 100.Figure 4.16: Plan view of the three-dimensional grid-scale trajectories from five sourcesof the second group in Vancouver. For details see caption to Fig. 4.13.100.01111111111111111111111111 I11111111111111111111111111111110.0^50.0a)-100.0^-50.00.01.6 -1.41.2 -1.0 -0.8 -0.6 -0.4 -0.2 -r""0.0Chapter 4. Trajectory Study^ 64b)- 100.0^-50.0^0.0^50.0^100.0Figure 4.17: Same as Fig. 4.16 but the XZ projection that cuts through stations of thissecond group. Viewpoint is northward.1.2 -1.0 -0.8 -0.6 -0.4 -0.20.0-80.0 -60.0 -40.0 -20.0^0.0^20.011111111m111111111140.0^60.0^80.0Chapter 4. Trajectory Study^ 651 .2 -1 .0 -0.8 -a)b)0.6 -0.4 -0.20.0 m 1111111'1111111111-80.0^-60.0^-40.0^-20.0^0.0^20.0^40.0^60.0^80.0Figure 4.18: Same as Fig. 4.16 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward.1111 11111111111150.I^111111111111111100.80.60.40.20.II-100.^-50.^0.III100.80. -'--60.40.20.0-40. ^-60. ^H I-100.^-50.^0.ittyttlint50.Chapter 4. Trajectory Study^ 66a)b)Figure 4.19: Plan view of the three-dimensional grid-scale trajectories from five sourcesof the third group in Vancouver. For details, see Fig. 4.13.111111111.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 -1-0.0^imiii1 1 1111111111HIIIIJ1111111-100.0^-50.0 0.0^50.0^100.0Chapter 4. Trajectory Study^ 67a)b)1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 111111^1111111111111_11111111111 , 111111111I1111 . 11111111111HW_LilL_-100.0^-50.0 0.0^50.0^100.0Figure 4.20: Same as Fig. 4.19 but the XZ projection that cuts through the secondemission group on Fig. 4.12. Viewpoint is northward.a)1.2 -40.0 60.0 80.0-80.0 -60.0 -40.0 -20.0^0.0^20.0Chapter 4. Trajectory Study^ 681 .0 -0.8 -0.6 -0.4 -0.2 - 40.0 ^0.0-80.0 -60.0 -40.0 -20.0 0 .0 40.020.0 60.0 80.0b)Figure 4.21: Same as Fig. 4.19 but the YZ projection that cuts through the stationnumbered '3' on Fig. 4.12. Viewpoint is westward.Chapter 4. Trajectory Study^ 694.4 Discussions and ConclusionsThis trajectory study starts with a survey of the general patterns of the pollutanttransport in the LFV by deploying six emission source stations across the LFV and bytracking particles released from these stations at different heights as well as at differenttimes. Results show that particles at high levels do not affect much the air quality of theLFV as they usually travel with the sea breeze return flow and the easterly geostrophicwinds out of the valley. Yet for emissions at lower levels, it appears that only thosein Vancouver area and its easterly neighbouring area are of primary concern as theytravel in a recirculating form within the valley while at most other stations, pollutantstravel out of the valley. Different release time also show some importance in affecting theway particles travel. These preliminary conclusions provide us with guidelines to focusthe further trajectory studies on the Vancouver area, the near-surface emissions and thedifferent release time.A look at trajectories produced by particles released at a near-surface emission stationin Vancouver at seven different times almost makes one convinced that two distinctpollutant transport categories can be defined by release time. Morning emission releasesbefore 1200 PST are all captured in the return flow of both sea breeze and upslope wind,and then travel westward out of the LFV. Whereas afternoon emission releases on orafter 1200 PST are all driven back to the valley by downslope winds. This image islater proved not exactly true for other emission stations in the Vancouver region by threesets of dual-release-time trajectory studies conducted in the Vancouver area. While theresults from the studies do confirm that particles released in the afternoon recirculatein the valley, they do not agree very well with the results from the previous study at asingle point in Vancouver. Trajectories from morning particle releases mostly stay in thevalley, even though there are a few which do not. It is therefore clear that even morningChapter 4. Trajectory Study^ 70emissions in Vancouver exacerbate the air pollution problem of the LFV on the studyday and after.When sea breeze flows onshore in summertime, one would intuitively suppose thatpollutants emitted in Vancouver area would travel directly to the east and may veryprobably arrive at Chilliwack (the eastmost city in the LFV) in the afternoon. Thistrajectory study does not support this hypothesis. The strong upslope winds over theCoastal Mountains have deflected the pollutant paths, and the combined sea breeze andupslope flows make many pollutants released in Vancouver travel northeastward up thenorthern mountains in the daytime and southwestward at night by the downslope winds.Many pollutants are never able to reach Chilliwack during the day, although they mayin the following day.This trajectory study can at least partly explain the high ozone concentration patternobserved in the LFV on the study day and the following day. Ozone episode occurredon 23th and 24th of August in 1985. A maximum ozone concentration contours in theLFV on the study day is presented in Fig. 1.4. Figure 4.22 gives a snapshot of theozone concentration contours at 1500 PST on the study day. Due to the sparsity ofmonitoring stations, the contour lines should be used as an approximation. The highestozone concentration zone shown on both figures lies in a broad area from Port Moody toChilliwack along the base of the northeast valley wall-the Coastal Mountains. Recallingthose trajectory plots in this study, one would discover that the destinations of mostof the trajectories which remain in the valley, are to the east of Vancouver near thenortheast valley wall, coinciding with this high ozone area. This discovery suggests thatthe high ozone area result mainly from the emissions in the Vancouver area and pollutantrecirculations induced by the mesoscale flows on the study day.It can therefore be concluded that this trajectory study reveals the great influence ofmesoscale flows on pollutant transport in the LFV. Pollutant recirculations are shownChapter 4. Trajectory Study^ 71aOO_8a- OOO450 500 550 600Figure 4.22: Snapshot of the ozone concentration isopleths at 1500 PST on August 23,1985. Thick solid line refers to ozone contour, and thick dashed line refers to uncertainozone contour. Thinner solid line is the coastline, and thinner dashed line is the 100mterrain contour indicating the edge of the valley walls. Values beside dots are the ozoneconcentrations at respective stations. Inner labels refer to the corresponding latitude orlongitude of the boundary. Outer labels are in units of UTM coordinates (x 103).Chapter 4. Trajectory Study^ 72clearly to occur frequently because of the specific mesoscale flows, and this pollutantrecirculation may be the cause for the continued ozone episode in the following day.4.5 Significance and SpeculationsThis is the first of such attempts to investigate the pollutant transport in the LFV withcomprehensive model studies. Before this study, there were only speculations about howthe sea breeze and other mesoscale flows would influence the pollutant transport thatleads to the ozone episodes in the LFV. This study presents a series of visualizations thatdemonstrate the interactions among mesoscale flows and the resultant pollutant trans-port patterns. Pollutant recirculations are detected in various trajectories and suggestthe linkage to the elevated ozone concentrations on August 23, 1985 and the followingday. This study day is characterized with the slack high pressure synoptic system, north-easterly or easterly weak synoptic wind, strong solar radiation and prevalence of thethermally forced mesoscale flows. It is possible to conclude that under similar meteoro-logical conditions to 23 August 1985, severe air pollution would likely occur as pollutantscan not be easily ventilated out of the valley but recirculate within the valley. Pollutanttransport for other meteorological conditions can be explored with the same methodol-ogy. Sufficient pollutant transport studies like this study under different meteorologicalscenarios would provide invaluable information to the air quality control and managementin the LFV.A multi-day pollutant transport study is recommended to reveal the relationshipbetween the air pollution episode and the mesoscale meteorology. Unfortunately thisstudy was unable to track pollutants for more than one day because the RAMS modelcould not provide realistic meteorological fields beyond one day. A nested-grid RAMSconfiguration would allow us to update the synoptic condition and may be a solution toChapter 4. Trajectory Study^ 73this problem. Future inclusion of the atmospheric diffusion process and incorporationwith a photochemical model will enable the present model study to explore the temporaland spatial pollutant concentration structures in the LFV.Appendix ASurface Energy Balance and Surface TemperatureA.1 IntroductionIn a mesoscale atmospheric model the bottom is the only boundary that has physicalsignificance. Moreover, it is the differential gradient of the dependent variables along thissurface that generates many mesoscale circulations and that has a pronounced influenceon the remaining mesoscale flows. Because of the crucial importance of this boundaryfor the mesoscale atmospheric systems, it is essential that it be represented as accuratelyas possible.Among those important variables necessary to represent this boundary, wind velocityis least difficult to determine and can be reasonably set to zero in all three directions atthe roughness height. However, good determination of the surface temperature and thesurface heat fluxes requires substantial effort. Various schemes for the specification ofsurface temperature have been presented in the past. They range widely in complexityfrom a sinusoidally-varying surface temperature to the "force-restore"method (Blackadar,1976; Deardorff, 1978), to multi-level soil models (Mahrer and Pielke, 1977b; Trembackand Kessler, 1985), and to vegetation parameterizations (Deardorff, 1978; McCumber,1980). Both multi-level soil models mentioned above have been adopted in the RAMSmesoscale model and are designed to deal with up to 12 different kinds of soil ranging fromsand to peat. While these soil models are adequate to deal with bare soil energy balances,this is only a limited case since much of the world is covered with vegetation. Neglecting74Appendix A Surface Energy Balance and Surface Temperature^75the presence of vegetation incurs errors of up to a factor of two in evapotranspiration(Pielke, 1984). Moreover, even accompanied by a vegetation parameterization method assuggested by Deardorff (1978) or McCumber (1980), there are still difficulties in treatingsuch complex land-use types as urban or suburban which contain mostly buildings andconcrete with little presence of vegetation and bare soil. In view of the complexities of thisproblem, a different approach has been tried in this study. It relates the storage heat flux/.Qs directly to net radiation Q* by an objective hysteresis model (Grimmond, Cleughand Oke, 1991) and the latent heat flux QE to the difference between net radiation andthe storage heat flux through the Priestley and Taylor equation (Priestley and Taylor,1972). The remaining term, the sensible heat flux QH, is calculated through Louis'(1979)analytic functions. Ground surface temperature can then be solved in the surface energybalance equation through iteration. The principle of this model and the methods usedare discussed in the following text.A.2 The ModelHeat fluxes at the ground surface are related through an energy budget balanceequation:Q* - QH - QE - AQS =0^ ( A.1)Each term in the above equation is discussed individually in the following. Finally, ascheme to compute the surface temperature will be presented.A.2.1 The Net Radiation FluxMahrer and Pielke's(1977b) short- and long-wave radiation parameterization schemefor clear sky has been adopted in this study. The following is a description of this scheme.Appendix A Surface Energy Balance and Surface Temperature^76Short-wave RadiationThe diurnal variation of the solar flux on a horizontal surface at the top of theatmosphere is computed fromS = So cos Z^ (A.2)withcos Z = sin 0 sin 6 + cos 0 cos 6 cos 7b^(A.3)where So is the solar constant, Z is the solar zenith angle, ck is the latitude, 6 is thesolar declination and 0 is the solar hour angle. At the surface the solar radiation isobtained by using two empirical functions. The first empirical transmission functionincludes molecular scattering and absorption by permanent gases such as oxygen, ozone,and carbon dioxide. This function (denoted in the following as G), originally presentedby Kondratyev (1969) and modified by Atwater and Brown (1974) to account for theforward Rayleigh scattering, is given by0.000949p + 0.051 )1/2 1G = 0.485 + 0.515 [1.041 — 0.16(^ (A.4)cos Zwhere p is pressure in mb.The second empirical function is from McDonald (1960) and accounts for the absorptivityof water vapour0.3a,,, = 0.077 I r(z) [ cos Zwhere r is the optical path length of water vapour above the layer z. It is given astopr(z) = f pqdz(A.5)(A.6)where p is the density of water vapour and q is the specific humidity. The net short-waveradiative flux at the surface isSo cos Z(1 — A)(G — a tu ) cos Z > 0Rs =^ (A.7)0 cos Z < 0Appendix A Surface Energy Balance and Surface Temperature^77where A is the albedo.The solar radiative heating rates are computed for the absorption of short-wave energyby water vapour only and are given byaT so cos Z^r(z) -0.7 dr= 0.0231[cos(A.8)at ) s pcp^Z dzLong-wave RadiationLong-wave radiation and atmospheric heating due to its flux divergence are calcu-lated for each time step. Carbon dioxide and water vapour are considered as emitters oflong-wave radiation. The path length for water vapour(Ar3 ) is computed for each layerfrom the surface to the top of the model byAr = (P3+1 —7)3)^(A.9)3^ g3The path length for CO2 (Acj ) isOct = —0.4148239(p3 +1 — /3.7)^(A.10)After these increments are obtained they are summed from the first level to the ith levelto give the total path length, for water vapour and carbon dioxide, respectively:ri = E Arj ,j=1c1 =E Ocj (A.11)The emissivity for water vapour was derived from data of Kuhn (1963) and are given inJacobs and Brown (1974).( ) =0.113log i0 (1^12.63r1j )0.104 log io 7.13 + 0.4400.121 log lo r13 + 0.4910.146 loglo r ij + 0.5270.161 loglo r i + 0.5420.136 log lo r 23 + 0.542log lo rij < —4loglo rij < —3loglo r1j < —1.5loglo r ij < —1log lo rij < 0log lo ri > 0(A.12) Appendix A Surface Energy Balance and Surface Temperature^78where rii = Iri — ri I is the optical path length between the ith and jth levels.Kondratyev's(1969) emissivity function for carbon dioxide in the formEc02(i,j) = 0.185 [1 — exp(-0.3919Ic i — ci 10.4)]^(A.13)is used, and finally the emissivity at each level is given byE(i,j) = cr(i,j) + 6.2(i,j)^ (A.14)Using the above emissivity functions we have for the downward and upward fluxes at alevel Ntop-1Rd (N) = E 2(7-21+1 + Tjfi) [f(N, j 1) — c(N, j)] crTt',,p (1 — f(N, top))^(A.15)j=NandN-1Ru (N) = E a^+ 7-34) [e(N, j) —^j 1)] + o-V(1 — f(N, 0))^(A.16)3 =1where TG is the ground surface temperature in degrees Kelvin.When N = 1, Rd (1) is the downward long-wave flux to the surface RL . The radiativecooling at each layer is computed fromaT1 (Ru (N + 1) — Ru (N) Rd(N) — Rd (N + 1)) ( )N = pcp^z(N + 1) — z(N)Terrain Effects on Radiation(A.17)Kondratyev (1969) gives the following expression for the solar radiation on a slant surfaceSs/ = So cos i^ (A.18)where i is the angle of incidence of solar rays on the inclined surface, andcos i = cos a cos Z -1- sin a sin Z cos(0 — ?7).^(A.19)Appendix A Surface Energy Balance and Surface Temperature^79Here Z is the zenith angle, a is the slope angle, and 77 are solar and slope azimuths.The latter three are defined as:azG 2 aZGa^tan" [(^(^ )2ay[cos b sin 01sin -1 sin Ztan" [ azG l azG^7r77 ay ax (A.20)(A.21)(A.22)where z_G is the ground height. For a slant surface the solar and infrared radiation willbe modifiedcos iRslsi -= Rs cos Z^(A.23)and&I s/ = RL cos a^ (A.24)The net all-wave radiation Q* on the surface is therefore equal to:Q* = RsIsi + RL J s, — crT^ (A.25)A.2.2 The Storage Heat FluxSince the surface in the real world is not usually well defined and homogeneous at thespatial scales of interest, it is much useful to introduce the concept of a surface volume.The top of the volume is to just above roof/tree/vegetation level and the base to a depthin the ground where no net vertical heat transport takes place over the period of interest(Oke and Cleugh, 1987). Then the storage heat flux (AQs) is defined to account forthe latent and sensible heat changes in, as well as the heat conduction into or out of, asurface volume.This study adopts an objective hysteresis storage heat flux model developed by Grim-mond, Cleugh and Oke (1991) to parameterize heat storage change in terms of the surfaceAppendix A Surface Energy Balance and Surface Temperature^80all-wave radiation which forces the energetics of the system, and the nature of the surfacecover. This model is essentially empirical but its form has both theoretical and physi-cal support. The performance of the model has been validated against suburban data(Grimmond, Cleugh and Oke, 1991; Roth, 1991; Roth and Oke, 1993). It was originallydeveloped for a study of urban environments. Since an objective method to characterizethe surface is used, it can also be applied generally to other areas.The Hysteresis Storage Heat Flux ModelFor a certain surface material, the storage heat flux can be related to the net all-waveradiation Q* (Camuffo and Bernardi, 1982) as:*AQs = aiQ*^OQa2^a3 (A.26)where a l , a 2 , a3 are coefficients for the surface with the units of dimensionless, hour andWm -2 , respectively. The second term involving '91* describes the rate and direction ofchange of Q*, and is included to reproduce the observed out-of-phase relationship betweenAQs and Q* (the peak of AQs precedes that of Q* by one or more hours). To representthe storage heat flux for any specific land-use type, which may consist of several differentsurface materials, Grimmond, Cleugh and Oke (1991) developed a composite equationwhich weights the role of each surface according to their plan coverage in the area understudy, so that:aQ*^,AQs = E, i { a liQ a2i at + a3iii=1A simplified form of this equation could be written as:aAQs = ClQ* + C 2 atQ* +C3(A.27)(A.28)where (io, is the fraction of the area covered by the ith surface. a 11 , a 2„ a3, are coefficientsfor the ith surface, and C 1 , C2, C3 refer to the composite coefficients for the study areaAppendix A Surface Energy Balance and Surface Temperature^81Table A.1: Summary of Coefficients for Some Surface Materials.Land cover AuthorRegression coefficientsal a 2 (h) a3 (Wm -2 )Greenspace/OpenShort grassGrasslandBare soilDoll et al.(1985)Clarke et a/.(1971)Novak(1981)0.320.330.380.540.030.56-27.4-11.0-27.3RooftopVancouverUppsalaYap(1973)Taesler(1980)0.170.440.100.57-17.0-28.9PavedConcreteAsphaltDoll et al.(1985)Narita et a/.(1984)0.810.360.480.23-79.9-19.3CanyonN-S canyon Nunez(1974) 0.32 0.01 -27.7ForestMixed forest McCaughey(1985) 0.11 0.11 -12.3(grid box in RAMS). By using this equation, AQ s in any distinct area can be calculatedas long as an inventory of the dimensions of any buildings, the distribution of surfacematerials and other information necessary to characterize the study area is provided.The coefficients for several surface materials representative of the present study area areavailable from various observations, including observations made in the Vancouver area.However there are some surface materials representative of rural and farmland areasof the LFV that are not adequately characterized by existing observations. Under thesecircumstances, combinations of the known coefficients for other surface materials are used.Here, consideration is given to materials having similar characteristics such as moisturecontent. These approximations are used when assigning coefficients for greenspace inrural areas and for agricultual areas.Coefficients resulting from observational studies for several surface materials are listedAppendix A Surface Energy Balance and Surface Temperature^82Table A.2: Summary of Coefficients for Various Land-use Types.Land-use type Land coverWeightingfactorRegression coefficientsal a2 (h) a3 (Wm -2 )Urban Rooftop 0.20 0.17 0.10 -17.0Paved 0.20 0.58 0.35 -49.6Canyon 0.60 0.32 0.01 -27.7Composite Coefficients 0.34 0.10 -29.9Suburban Greenspace 0.33 0.32 0.37 -29.2Rooftop 0.13 0.17 0.10 -17.0Paved 0.11 0.58 0.35 -49.6Canyon 0.43 0.32 0.01 -27.7Composite Coefficients 0.33 0.18 -29.2Rural Vegetation 0.90 0.34 0.38 -21.9Rooftop 0.05 0.17 0.10 -17.0Paved 0.05 0.58 0.35 -49.6Composite Coefficients 0.35 0.36 -23.0Agricultural Vegetation 1.00 0.15 0.03 -11.0Composite Coefficients 0.15 0.03 -11.0Forest Forest 1.00 0.11 0.11 -12.3Composite Coefficients 0.11 0.11 -12.3in Table A.1. Table A.2 contains information about the weighting factors (percentageof specific surface type in an area) and coefficients of the five land-use types definedfor the LFV in summertime. The assignment of the weighting factors in each land-usetype is made possible with the aid of field surveys and aerial photographs. Experts wereconsulted before values of the coefficients for each land-use type are finally set.In all cases, coefficients for paved surfaces have been taken as averages of concrete andasphalt. Greenspace values for suburban land-use takes its coefficients from the averageof dry and irrigated grass lawn since this may be more representative than using onlyone moisture status. For rural areas, coefficients are estimated as averages of short grass,grassland and bare soil (Grimmond 1992, personal communication). In agricultural areas,only coefficients from moist grassland are applied since the condition of the grass is veryAppendix A Surface Energy Balance and Surface Temperature^83close to that in the agricultural area.A.2.3 The Latent Heat FluxPriestley and Taylor (1972) argued that for a wet area at the scale of the grids usedin computer solutions of numerical weather-forcasting models, radiant energy receiptmust dominate over advective effects in controlling evaporation. Following this idea,they developed an equation for evaporation from wet regions asQE = s s^ (CI AQs)+ -y(A.29)where QE is the latent heat flux and a is an empirical coefficient. A value for a of1.26 was found to fit well the data from several sources. s is the slope of the saturationspecific humidity curve, s = dq* /dT = cp/A 0.0004 (gwat„/ gair )K - 1 isthe psychrometric constant. A (Jkg- 1 ) is the latent heat of vaporization of water andcp = 1004 Jkg'K' is the specific heat of dry air at constant pressure. dq*/dT is givenby the Clausius-Clapeyron equation:dq*Aq*dT = 0.622^^RT2(A.30)wheree*q* = 0.622T,^ ( A.31)e = (0.6112kPa)exp (17.67(T — 273.16))^(A.32)T — 29.66 and q* is the saturation specific humidity in kgkg- 1 . Temperature T is in degrees Kelvin.P and e*, both expressed in terms of kPa, are the atmospheric pressure and saturationvapour pressure, respectively. R is the gas constant for dry air, and is equal to 287.04ugs -2K--1 .Although the Priestley-Taylor equation was developed mainly to deal with evapo-ration in a well-watered region, it may be used more generally to ascribe evaporationS^IS r1. 1,1^I^1. 11111."‘E'—01.67. 4.^—10X`4115000—„ .1.1^.1 11/1111^ 11..1] 10 100 1000Canopy Resistance (s rn - ')1.5 ------ ^—+lox- a=1.26Appendix A Surface Energy Balance and Surface Temperature^84Figure A.1: Priestley-Taylor coefficient a calculated for a range of canopy resistance forthe Cabauw data set ( from McNaughton and Spriggs 1987).even in water-constrained areas if the Priestley-Taylor coefficient (hereafter referred toas the P&T coefficient) a can be adjusted to reflect the evaporative condition of the area(de Bruin, 1983; de Bruin and Keijamn, 1979; McNaughton and Spriggs, 1987). Mc-Naughton and Spriggs (1987) found from experiments at Cabauw that estimation of amay be made from canopy or surface resistance alone. Figure A.1 shows the relationshipbetween the calculated P&T coefficient a and canopy resistance. Canopy resistance isa parameterized variable accounting for water loss of an area treating it as if it were asingle hypothetical leaf with characteristics related to diurnal and seasonal variations.Based on this idea, with the aid of experience and some experimental data of canopyresistance, a set of summertime P&T coefficients are assigned to five different land-usetypes in the LFV in Table A.3. These values are only appropriate to the conditionsapplying on the day in question.The effect of advection have also been considered. Uncertainties are inevitably in-volved in the process. The assigned P&T coefficients are considered reasonable to reflectAppendix A Surface Energy Balance and Surface Temperature^85Table A.3: Summary of P&T Coefficient for Each Land-use Type.Land-use Type Urban Suburban Rural Agricultural ForestP&T Coef. 0.2 0.8 1.0 1.26 1.0evaporation conditions for different land-uses in summertime. This difference in P&Tcoefficients is necessary to induce a non-uniform temperature and wind field across theLower Fraser Valley.A.2.4 The Sensible Heat FluxThe sensible heat flux is calculated through an independent channel rather than asa residual in the surface energy balance. The scheme of Louis' (1979) analytic functionsmethod is adopted. It is based on Monin and Obukhov (1954) similarity theory andmakes use of Businger's functions for the flux-profile relationships in the surface layer.The sensible heat flux may be written: Q H = — pCp U * 0., where u * and 0* are the scalingvelocity (or friction velocity) and temperature, respectively. If the roughness length z o ,potential temperature 0 and wind speed u above the surface, and temperature differencebetween surface and air AO are provided, u * and 0* can be found through the flux-profile relationships. The following is what Louis conceived to get u * and 0* . From theflux-profile relationships:u*2 = a 2 u 2 F,i( — ,RiB)zou * O* = 1.35a 2 uA0Fh(-2. 1 R113)zowhere Rig is the bulk Richardson number.(A.33)(A.34)a 2 = K 2 1(1nz I zo ) 2g z AOR2B = ^0u 2(A.35)(A.36)Appendix A Surface Energy Balance and Surface Temperature^86and is is von Karman's constant (0.35). g is the gravitational acceleration (9.8 ms -2 ).Having the numerically computed curves for the functions a 2Fm and a 2Fh , Louis usedanalytical formulae to fit these curves:For unstable cases (RiB < 0)Fm9.4R2B^= 1  ^ (A.37)1 + cm1Ria 1 1159.4RiB^Fh = 1^ (A.38)1 + chlRiBwhere^= 69.56a 2 (-5-) 1 / 2^(A.39)zo^c h = 49.82a 2 ( 2±-)V 2^(A.40)zo For stable cases (RiB > 0)Fm = Fh = 1/(1 + 4.7R,B ) 2^(A.41)In this manner, both vertical eddy fluxes of momentum w'u' = u* and sensible heat w'O' =—u,,O,, could be obtained without going through complicated iterative computations.A.3 Numerical ProceduresIt must be appreciated that until an appropriate surface temperature at one timestep is decided, all terms in the surface energy balance equation remain unsettled. Thesurface temperature is computed by a Newton-Raphson iterative solution to the surfaceenergy balance. The computational procedure is as follows.1. Compute Q* and Q H using the surface temperature of the last time step, and thencompute AQs and QE.Appendix A Surface Energy Balance and Surface Temperature^872. Let F(TG) be equal to the sum of the terms on the left side of equation A.1 andget F(TG) value. If IF(TG)I is not less than some small value (set to 10 -3 ), thenapply the Newton-Raphson iteration process in the formTG -1-1. v _F ( TG)/t (TG ) ( A.42)Here F l (TG ) is the derivative of F(TG) with respect to TG. The four heat flux termsare then recalculated.3. Repeat step 2, until the desired accuracy of F(TG ) has been reached. This ensuresthat the energy balance equation (A.1) is closed. The final correct heat fluxes for thecurrent time step can then be calculated with the last iterated surface temperatureTG.A.4 Energy Balance and Surface Temperature at NightAt night, surfaces exhibit very different behaviour to that in the daytime in termsof the partitioning of the net radiation into the three flux terms. In light wind cases,the sensible and latent heat fluxes usually become insignificant energy sinks or sourcesso the storage heat flux tends to approximately balance the loss of net radiation fromthe surface. In view of this, the preceding model for the daytime is not applicable topredict the nighttime surface temperature and to determine the surface layer fluxes. Toovercome this problem, a formula given theoretically by Brunt (1941) to calculate thenighttime surface temperature is adopted. The Brunt formula relates the temperaturedeviation from sunset to the half power function of time, the net radiation at sunset andthe thermal admittance for the studied surface.ATt_ s = —2 Q 1/2 (A.43)71-1/2 tiAppendix A Surface Energy Balance and Surface Temperature^ 88Where Q* is from the value at sunset, ATt, is the decrease of temperature in time t fromsunset and it is the thermal admittance. This relation has proved to be very successfulin rural areas with an estimated thermal admittance of about 1600 Wm -2s 1 / 2K -1 (Okeand Maxwell, 1975). It is not as successful for urban areas, where a linear relation seemsto be more appropriate (Oke and Maxwell, 1975). Since the nocturnal stable atmospheredoes not respond to the surface as sensitively as in the daytime, it is justified to use thisfunction in urban and suburban cases as an approximation. From studies of the thermaladmittance for individual materials (Oke, 1981), thermal admittances as high as 2100and 1700 Wm -2 s 1 / 2K -1 are assigned to urban and suburban land-use types respectivelyin this study, while 1600 Wm-281/2K-1 is used for other land-use types.After surface temperature is computed, Louis' flux-profile function is used to calculatethe necessary surface fluxes u., 0* and R* (scaling specific humidity), thus completingthe surface heat budget and surface layer parameterization.A.5 Model EvaluationThe applicability of this daytime prognostic-empirical energy balance model to the presentstudy is evaluated by comparing the predicted surface energy fluxes with observations.On the day of interest, hourly mean net radiation and sensible heat fluxes were measuredat a suburban site in Vancouver during daylight hours. Storage heat fluxes and latentheat fluxes were subsequently obtained through parameterization in the objective hys-teresis storage model and as a residual, respectively. The same four fluxes have also beenextracted from the RAMS model output.Figures A.2a-d depict the time series of these four fluxes at Sunset on August 23 1985for both modelled results and observed data. The modelled net radiation in Fig. A.2aoverestimates the observed values by up to 75 Wm -2 . This consequently results in aAppendix A Surface Energy Balance and Surface Temperature^89slight overestimation of the other modelled heat fluxes. However the trends of Q*, AQ sand QH are simulated quite successfully. QE is predicted well in the morning, before1000 PST, but the abrupt decrease in the flux in the afternoon is not well-simulated,presumably because of the model's poor representation of the mechanism that gives riseto the observed trend. Comparing Figs. A.2c and d, it can be seen that the continuousincrease in the observed sensible heat fluxes QH in the afternoon may be responsible forthe observed afternoon decrease in QE.One of the major purposes of this energy balance model is to predict the surfacetemperature. Therefore, it is important to prove that the model does provide a rea-sonable surface temperature field. Unfortunately there is a paucity of temperature dataavailable for the day of interest. Instead a satellite thermal image 1 covering mainly theGreater Vancouver area at 13.56 Local Standard Time (LST) 2 on August 16 1985 wasused to obtain what are likely to be similar surface temperature patterns, which werethen compared with modelled results. The area of satellite coverage and the satellitesurface temperature picture are shown in Figs. A.3a and b, respectively. August 16 wasa cloudless summer day. The satellite picture, from Roth, Oke and Emery (1989), wasfirst scanned and later converted back to temperature data. Figure A.4a is the surfacetemperature contour map of the satellite image and Fig. A.4b is the contour map of themodelled surface temperature at 1400 PST on August 23 1985. Although a direct com-parison of the exact magnitudes between these two fields is not appropriate, similarityof the solar heating makes it possible to compare the respective temperature patterns.The similarity between the modelled and measured fields is quite obvious, although themodelled temperature field has less structure. The model was able to distinguish differ-ences in surface heating between the different land-use types even with only five land-use'Comes from NOAA-7, -8 and -9 which carry an Advanced Very High Resolution Radiometer(AVHRR). The ground resolution is 1.1 km and the swath width is 2700 km.2 Local Standard Time in Vancouver is about 10 minutes ahead of Pacific Standard Time in summerAppendix A Surface Energy Balance and Surface Temperature^90a) NE 8b6^8^10^12^14^16^18^20Time (PST)b).01Ncn0c\j0006c)N 0E00c\IoO8^10^12^14Time (PST)16^18^206^8^10^12^14^16^18^20Time (PST)d)...^• .. ... • ...... •-.6^8^10^12^14Time (PST)16 18 20Figure A.2: Time series for surface fluxes at the Sunset suburban site in Vancouver. a)Net radiation, b) Storage heat flux, c) Sensible heat flux, d) Latent heat flux. Solid line- modelled values. Dashed - observed.Appendix A Surface Energy Balance and Surface Temperature^91classes specified. It can be concluded from the above two comparisons that this surfaceenergy balance scheme is able to simulate the general time-varying patterns of the fourheat fluxes, and is able to predict reasonable surface temperature field. Yet the schemeleaves room for improvement in its overestimation of net radiation.Figure A.3: a) Location and land-use maps of Vancouver. b) The surface radiant tem-perature distribution for the same area at 13.56 (LST) on 16 August 1985. The rainbowbrightness temperatures in b) are in degree Celsius. From Roth et al. (1989), p1705-1707.Appendix A Surface Energy Balance and Surface Temperature^93a)b)Figure A.4: a) Contour plot of surface temperatures derived from the satellite thermalimage in Fig. A.3b, and b) contour plot of surface temperatures derived from the RAMSmodel output at 1400 PST on August 23, 1985. All plots cover the same area as depictedin Fig. A.3a. Temperatures are in degree Celsius.BibliographyAtkinson, B. W., 1981. Meso-scale Atmospheric Circulations, Academic Press, Toronto,p. 495.Atwater, M. A. and P. S. Brown, 1974: Numerical Calculation of the Latitudinal Variationof Solar Radiation for an Atmosphere of Varying Opacity , Journal of AppliedMeteorology, 13, 289-297.Blackadar, A. K., 1976: Modeling the Nocturnal Boundary Layer, Symp. Atmos. Turbul.Diffus. Air Qual. Prepr. 3rd.de Bruin, H. A. R., 1983: A Model for the Priestley-Taylor Parameter a , Journal ofClimate and Applied Meteorology, 22, 572-578.de Bruin, H. A. R. and A. A. M. Holtslag, 1982: A Simple Parameterization of theSurface Fluxes of Sensible and Latent Heat During Daytime Compared with thePenman-Monteith Concept, Journal of Applied Meteorology, 21, 1610-1621.de Bruin, H. A. R. and J. Q. Keijamn, 1979: The Priestley-Taylor Evaporation ModelApplied to a Large Shallow Lake in the Netherlands, Journal of Applied Meteo-rology, 18, 898-903.Brunt, D., 1941. Physical and Dynamical Meteorology, Cambridge University Press, Lon-don, p. 428.Businger, J. A., J. C. Wyngaard, Y. Izumi and E. F. Bradley, 1971: Flux Profile Relation-ships in the Atmospheric Surface Layer, Journal of Atmospheric Science, 28, 181 -189.Camuffo, D. and A. Bernardi, 1982: An observational Study of Heat Fluxes and theRelationship with Net Radiation, Boundary Layer Meteorology, 23, 359-368.94Chang, Y. S., G. R. Carmichael, H. Kurita, T. Kitada and H. Ueda, 1990: DiagnosticEvaluation of the Transport and Gas Chemistry Components of the STEM-IImodel, Atmospheric Environment, 24A(11), 2715-2737.Chang, Y. S., G. R. Carmichael, H. Kurita and H. Ueda, 1989: The Transport and For-mation of Photochemical Oxidants in Central Japan, Atmospheric Environment,23(2), 363-393.Chang, Y. S., W-C. Shin, G. R. Carmichael, H. Kurita and H. Ueda, 1990: Evaluationof the Effect of Emission Reductions on Pollutant Levels in Central Japan, Envi-ronmental Science Technology, 24, 1355-1366.Charney, J. G., R. Fjortoft and J. von Neumann, 1950: Numerical Integration of theBarotropic Vorticity Equation, Tellus, 2, 237-254.Clarke, R. H., A. J. Dyer, P. R. Brook, D. G. Reid and A. J. Troup, 1971: The WangaraExperiment-Boundary Layer Data, Division of Meteorological Physics, C.S.I.R.O.,Australia.Cleugh, H. A., 1990: Development and Evaluation of Suburban Evaporation Model: aStudy of Surface and Atmospheric Controls on the Suburban Evaporation regime,The University of British Columbia, Vancouver, Canada, Ph.D. Thesis.Concord Scientific Corporation(CSC), 1982: Vancouver Oxidant Study, Air Quality Ana-lyis 1978-1981, Prepared for Environment Canada, Environmental Protection Ser-vice.  1985: Vancouver Oxidant Study, Air Quality Analyis Update 1982-198$, Pre-pared for Environment Canada, Environmental Protection Service.95Deardorff, J. W., 1978: Efficient Prediction of Ground Surface Temperature and Moisture,with Inclusion of a Layer of Vegetation, Journal of Geophysical Research, 83(c4),1889-1903.Doll, D., J. K. S. Ching and J. Kaneshiro, 1985: Parameterisation of Subsurface Heatingfor Soil and Concrete Using Net Radiation Data , Boundary Layer Meteorology,32, 351-372.Gal-Chen, T. and R. C. J. Somerville, 1975: On the Use of a Coordinate Transformationfor the Solution of the Navier-Stokes Equations, J. Comp. Phys., 17, 241-249.Grimmond, C. S. B., 1992: Personal Communication.Grimmond, C. S. B., H. A. Cleugh and T. R. Oke, 1991: An Objective Urban Heat Stor-age Model and Its Comparison with Other Schemes, Atmospheric Environment,25B(3) , 311-326.Glisten, H., 1986: Formation, Transport and Control of Photochemical Smog. Air Pollu-tion, 0. Hutzinger, Ed., The Handbook of Environmental Chemistry, Volume 4Part A, pp. 53-106.Hay, J. E. and T. R. Oke, 1976. The Climate of Vancouver, Tantalus Research Limited,Vancouver, p. 48.Jackson, P., 1993: Gap Winds in a Fjord: Howe Sound, British Columbia, The Universityof British Columbia, Vancouver, Canada, Ph.D. Thesis.Jacobs, C. A. and P. S. Brown, 1974: IFYCL Final Report, Vol. IV Three-dimensionalResults, The Center for the Environment and Man, CEM Rep. No. 4131-509d,Hartford, Connecticut.Klemp, J. B. and D. K. Lilly, 1978: Numerical Simulation of Hydrostatic MountainWaves, Journal of the Atmospheric Sciences, 35, 78-107.96Klemp, J. B. and R. B. Wilhelmson, 1978: The Simulation of Three-Dimensional Con-vective Storm Dynamics, Journal of the Atmospheric Sciences, 35, 1070-1096.Kondratyev, K. Y., 1969: Radiation in the Atmosphere, Academic Press, New York.Kuhn, P., 1963: Radiometeorsonde Observations of Infrared Flux Emissivity of WaterVapour, Journal of Applied Meteorology, 2, 368-378.Kurita, H., K. Sasaki, H. Muroga, H. Ueda and S. Wakamatsu, 1985: Long-Range Trans-port of Air Pollution under Light Gradient Wind Conditions, Journal of Climateand Applied Meteorology, 24, 425-434.Lalas, D. P., D. N. Asimakopoulos, D. G. Deligiorgi and C. G. Helmis, 1983: Sea BreezeCirculation and Photochemical Pollution in Athens, Greece, Atmospheric Envi-ronment, 17, 1621-1632.Lee, X., 1992: Personal Communication.Lewis, J. E. and T. N. Carlson, 1989: Spatial Variations in Regional Surface EnergyExchange Patterns for Montreal, Quebec, The Canadian Geographer, 33, No. 3,194-203.Louis, J-F., 1979: A Parametric Model of Vertical Eddy Fluxes in the Atmosphere, Bound-ary Layer Meteorology, 17, 187-202.Lyons, W. A., 1975: Turbulent Diffusion and Pollutant Transport in Shoreline Environ-ments. Lectures on Air Pollution and Environmental Impact Analysis, pp. 136-208.Mahrer, Y. and R. A. Pielke, 1977a: The Effects of Topography on the Sea and LandBreezes in a Two-Dimensional Numerical Model, Monthly Weather Review, 105,1151-1162.97^ , 1977b: A Numerical Study of the Airflow over Irregular Terrain, Beitriige zurPhysik der Atmosphiire, 50, 98-113.McCaughey, J. IL, 1985: Energy Balance Storage Terms in a Mature Mixed Forest atPetawawa, Ontario - A Case Study, Boundary Layer Meteorology, 31, 89-101.McCumber, M. C., 1980: A Numerical Simulation of the Influence of Heat and MoistureFluxes upon Mesoscale Circulations, University of Virginia, Charlottesville, USA,Ph.D. Thesis.McDonald, J., 1960: Direct Absorption of Solar Radiation by Atmospheric Water Vapour,J. Meteorol., 17, 319-328.McKendry, I. G. and M. Li-Ting-Wai, 1993: A Synoptic Climatology of Ozone Con-centrations at Port Moody, Lower Fraser Valley, B.C., Report prepared for AirResources Branch, British Columbia Ministry of Environment, Lands and Parks,37 pp..McNaughton, K. G. and T. W. Spriggs, August, 1987: An Evaluation of the Priestleyand Taylor Equation and the Complementary Relationship Using Results from aMixed-Layer Model of the Convective Boundary Layer, IAHS Publication No. 177Estimation of Areal Evapotranspiration , Vancouver, B. C., Canada.McNider, R. T. and R. A. Pielke, 1981: Diurnal Boundary-Layer Development over Slo-ping Terrain, Journal of the Atmospheric Sciences, 38, 2198-2212.Monin, A. S. and A. M. Obukhov, 1954: Basic Regularity in Turbulent Mixing in theSurface Layer of the Atmosphere, Akad. Nauk. S.S.S.R. Trud. Geofiz. Inst., Tr.,24.Monteith, J. L., 1981: Evaporation and Surface Temperature, Quarterly Journal RoyalMeteorological Society, 107 , 1-27.98Moore, G. E., S. G. Douglas, R. C. Kessler and J. P. Killus, 1991: Identification andTracking of Polluted Air Masses in the South-Central Coast Air Basin, Journalof Applied Meteorology, 30, 715-732.Moran, M., 1992: Mesoscale Dispersion Modelling, The Colorado State University, Bou-lder, U. S. A., Ph.D. Thesis.Moussiopoulos, N., 1989. Mathematische Modellierung mesoskaliger Ausbreitung in derAtmosphare, VDI-Verlag,Dusseldorf, p. 307. , 1993. The Athens Experience, Monterrey, Mexico, Air Pollution '93 Conference,p. 128.Moussiopoulos, N., T. Flassak, P. Prahm and D. Berlkowitz, 1993: Simulations of theWind Field in Athens with the Nonhydrostatic Mesoscale Model MEMO, Envi-ronmental Software(submitted).Moussiopoulos, N. and W. Oehler, 1988: Numerical Simulations of the PhotochemicalAir Pollution Levels in Athens, Greece, Environmental Software, 3, 105-109.Narita, K., T. Sekine and T. Tokuoka, 1984: Thermal Properties of Urban SurfaceMaterials-Study on Heat Balance at Asphalt Pavement, Geog. Rev. Japan, 57A,639-651.Novak, M. D., 1981: The Moisture and Thermal Regimes of a Bare Soil in the LowerFraser Valley during Spring, The University of British Columbia, Vancouver,Canada, Ph.D. Thesis.Nunez, M., 1974: The Energy Balance of an Urban Canyon, The University of BritishColumbia, Vancouver, Canada, Ph.D. Thesis.Oke, T. R., 1981: Canyon Geometry and the Nocturnal Urban Heat Island: Comparisonof Scale Model and Field Observations, Journal of Climatology, 1, 237-254.99^ , 1989: The Urban Energy Balance, Progress in Physical Geography, 12, 471-508.Oke, T. R. and H. A. Cleugh, 1987: Urban Heat Storage Derived as Energy BudgetResiduals, Boundary Layer Meteorology, 39(3), 233-245.Oke, T. R. and G. B. Maxwell, 1975: Urban Heat Island Dynamics in Montreal andVancouver, Atmospheric Environment, 9, 191-200.Oliger, J. and A. Sundstrom, 1978: Theoretical and Practical Aspects of Some InitialBoundary Value Problems in Fluid Dynamics, SIAM J. Appl. Math., 35, 419-446.Orlanski, I., 1976: A Simple Boundary Condition for Unbounded Hyperbolic Flows, J.Comput. Phys., 21, 251-269.Persson, C., H. Rodhe and L. -E. D. Geer, 1987: The Chernobyl Accident - a Meteoro-logical Analysis of how Radionuclides Reached and Were Deposited in Sweden,Ambio, 16, 20-31.Pielke, R. A., 1974a: A Comparison of Three-Dimensional and Two-Dimensional Numer-ical Predictions of Sea Breezes, Journal of the Atmospheric Sciences, 31, 1577-1585.  1974b: A Three-Dimensional Numerical Model of the Sea Breeze over SouthFlorida, Monthly Weather Review, 102, 115-139. , 1984: Mesoscale Meteorological Modeling, Academic Press.Pielke, R. A., W. R. Cotton, R. L. Walko, C. J. Tremback, W. A. Lyons, L. D. Grasso,M. E. Nicholls, M. D. Moran, D. A. Wesley, T. J. Lee and J. H. Copeland, 1992:A Comprehensive Meteorological Modeling System - RAMS, Meteorol. Atmos.Phys., 49, 69-91.100Platzman, G. W., 1954: The Computational Stability of Boundary Conditions in Nu-merical Integration of the Vorticity Equation, Arch. Met. Geophys. Biokl., A7,29-40.^ , 1979: The ENIAC Computations of 1950 - Gateway to Numerical Weather Pre-diction, Bull. Amer. Meteor. Soc., 60, 302-312.Priestley, C. H. B. and R. J. Taylor, 1972: On the Assessment of Surface Heat Flux andEvaporation, Monthly Weather Review, 106, 81--92.Robeson, S. M. and D. G. Steyn, 1990: Evaluation and Comparison of Statist^ForcasiModels for Daily Maximum Ozone Concern rations, Almo,p11( vie I.'nvironia(24B(2), 303-312.Roth, M., 1991: Turbulent Transfer Characteristics over a Suburban Surface, The Uni-versity of British Columbia, Vancouver, Canada, Ph.D. Thesis.Roth, M. and T. Oke, 1993: Comparison of Modelled and "Measured" Heat Storage inSuburban Terrain, Atmospheric Environment(in press).Roth, M., T. Oke and W. J. Emery, 1989: Satellite-Derived Urban Heat Islands fromThree Coastal Cities and the Utilization of such Data in Urban Climatology, Int.J. Remote Sensing, Vol. 10, NO. 11, 1699-1720.Smith, F. B. and M. J. Clark, 1988: Wet and Dry Deposition of Chernobyl Releases.Nature, 322, 245-249.Stern, A. C., R. W. Boubel, D. B. Turner and D. L. Fox. 1981: Ftm(him< OW., o/ :byPollution, Academic Press.Steyn, D. G., 1980: Turbulence, Diffusion and the Daytime Mixed Layer Depth overa Coastal City, The University of British Columbia, Vancouver, Canada, Ph.D.Thesis.101Steyn, D. G., M. Bovis, M. North and 0. Slaymaker, 1992: The Biophysical EnvironmentToday. Vancouver and Its Region, G. Wynn and T. R. Oke, Eds., UBC Press,University of British Columbia, Vancouver, BC, Canada.Steyn, D. G. and D. A. Faulkner, 1986: The Climatology of Sea-Breezes in the LowerFraser Valley, B.C., Climatological Bulletin, 20(3), 21-39.Steyn, D. G. and I. G. McKendry, 1987: Sea Breeze Dynamics in the Lower Fraser Valley,B. C., Geography Department Report of UBC.  1988: Quantitative and Qualitative Evaluation of a Three-Dimens ional MesoscaleNumerical Model Simulation of a Sea Breeze in Complex Terrain, Monthly WeatherReview, 116, 1914-1926.Steyn, D. G. and T. R. Oke, 1982: The Depth of the Daytime Mixed Layer at Two CoastalSites: A Model and Its Validation, Boundary-Layer Meteorology, 24, 161-180.Steyn, D. G., A. C. Roberge and C. Jackson, 1990: The Anatomy of an Extended AirPollution Episode in British Columbia's Lower Fraser Valley, Report preparedfor Waste Management Branch, British Columbia Ministry of Environment, 55pp.PP-.Stull, R. B., 1989: An Introduction to Boundary Layer Meteorology, Atmospheric SciencesLibrary, Kluwer Academic Publishers, Dordrecht, Holland.Taesler, R., 1980: Studies of the Development and Thermal Structure of the UrbanBoundary Layer in Uppsala, Part I: Experimental Program; and Part II: Data,Analysis and Results, Report 61, Met. Instit., Uppsala University, Uppsala.Taylor, E., 1991: Forcasting Ground-Level Ozone in Vancouver and the Lower FraserValley of British Columbia, Scientific Services Division, Atmospheric EnvironmentService, Pacific Region, Environment Canada, Report PAES-91-3, 8pp pp..102Tremback, C., G. Tripoli, R. Arritt, W. R. Cotton and R. A. Pielke, Nov., 1986: The Re-gional Atmospheric Modeling System, Proceeding of International Conference onDevelopment and Application of Computer Techniques to Environmental Studies(ENVIROSOFT 86), Los Angeles, U.S.A..Tremback, C. J. and R. Kessler, June 17-20, 1985: A Surface Temperature and Mois-ture Parameterization for use in Mesoscale Numerical Models, 7th Conference onNumerical Weather Prediction, AMS, Montreal, Quebec, Canada.Tripoli, G. J. and W. R. Cotton, 1982: The Colorado State University Three-DimensionalCloud/Mesoscale Model 1982. Part I: General Theoretical Framework and Sensi-tivity Experiments , J. de Rech. Atmos., 16, 185-220.Ulrickson, B. L. and C. F. Mass, 1990a: Numerical Investigation of Mesoscale Circulationsover the Los Angeles Basin. Part II: Synoptic Influence and Pollutant Transport,Monthly Weather Review, 118, 2162-2184.^ 1990b: Numerical Investigation of Mesoscale Circulations over the Los AnglesBasin. Part I: A Verification Study, Monthly Weather Review, 118, 2138-2161.Walko, R. L. and C. J. Tremback, 1991: The RAMS Model Users's Guide (draft), ASTeR,Inc. Fort Collins.Wheeler, D. A., 1988: Atmospheric Dispersal and Deposition of Radioactive Materialfrom Chernobyl, Atmospheric Environment, 22, 853-863.Wieringa, J., July 1991: Updating the Davenport Roughness Classification, 8th Interna-tional Conf. on Wind Engineering, London, Ontario, Canada.Willmott, C. J., 1981: On the Validation of Models, Physical Geography, 2(2), 168-194.^ , 1982: Some Comments on the Evaluation of Model Performance, Bulletin ofAmerican Meteorological Society, 63(11), 1309-1313.103^, 1985: Statistics for the Evaluation and Comparison of Models, Journal of Geo-physical Research, 90(c5), 8995-9005.104

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0086265/manifest

Comment

Related Items