UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Atmospheric recirculation during ozone episodes in the Lower Fraser Valley, B. C. Seagram, Annie F. 2014

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

Item Metadata


24-ubc_2014_november_seagram_annie.pdf [ 80.5MB ]
JSON: 24-1.0167615.json
JSON-LD: 24-1.0167615-ld.json
RDF/XML (Pretty): 24-1.0167615-rdf.xml
RDF/JSON: 24-1.0167615-rdf.json
Turtle: 24-1.0167615-turtle.txt
N-Triples: 24-1.0167615-rdf-ntriples.txt
Original Record: 24-1.0167615-source.json
Full Text

Full Text

ATMOSPHERIC RECIRCULATION DURING OZONE EPISODES IN THELOWER FRASER VALLEY, B. C.byAnnie F. SeagramB. Sc. (Hons), The University of British Columbia, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Atmospheric Science)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)October 2014c© Annie F. Seagram, 2014AbstractThe presence of thermo-topographic circulations in areas of complex terrain plays an importantrole in recirculating pollutants during periods of degraded air quality. In this modellingstudy, we seek to define and detect atmospheric recirculation in the Lower Fraser Valley(LFV), British Columbia, a region that frequently experiences episodes of degraded airquality despite its modest total emissions and relatively small population size. The WeatherResearch and Forecasting (WRF) model is used to simulate wind fields during seven severe,three-day summertime ozone episodes occurring over a period of 20 years (1985 – 2006).These episodes cover the known set of synoptic and mesoscale circulation regimes conduciveto ozone episodes in the LFV. A trajectory modelling study is devised where WRF modeloutput is used to compute trajectories using the HYbrid Single Particle Lagrangian IntegratedTrajectory (HYSPLIT) model. In order to examine pollutant transport, the starting locationof trajectories is selected to coincide with the location of maximum ozone precursor (NOx andVOCs) emissions. Number density maps generated from composite trajectory fields revealdifferent spatial distributions of trajectories by circulation regime. A generally applicablequantitative definition and objective detection algorithm for recirculation is developed, andthen applied to the modelled trajectories to identify recirculating trajectory segments (RTSs).Recirculation is detected during all episodes, though not all circulation regimes result in thesame the frequency of detection. Analysis of RTSs shows that recirculation in the LFV isspatially and temporally the same regardless of mesoscale circulation conditions. There isstrong evidence that pollutants may be “carried-over” from one day of an episode to the next,and that air parcels frequently return to their origin within less than 12 hours. Results suggestthat recirculation is primarily driven by onshore flows and mountain-valley circulations withinthe main valley floor of the LFV, and secondarily by diurnal flows within tributary valleys.This research adds to our understanding of atmospheric transport during ozone episodes inthe LFV, and provides a new framework for studying recirculation elsewhere.iiPrefaceThis thesis contains research conducted by me, Annie Seagram, a student under the supervisionof Dr. Douw Steyn, Dr. Ian McKendry, and Dr. Bruce Ainslie. The research topic wasoriginally suggested by Dr. Douw Steyn. The scope and focus of the research were thendeveloped by myself with the help of Dr. Douw Steyn. Preliminary results of the researchherein were previously published in the following conference proceedings:A. F. Seagram, D. G. Steyn, and B. Ainslie (2013). Modelled recirculation of air pollutantsduring ozone episodes in the Lower Fraser Valley, B. C. Air Pollution Modeling and ItsApplication XXII. Ed. by D. G. Steyn, P. J. H. Builtjes, and R. M. A. Timmermans. NATOScience for Peace and Security Series Series C: Environmental Security. The Netherlands:Springer, 291–295Large portions of this research source work by Ainslie and Steyn (2007) and Steyn et al.(2011), and are cited as such. This thesis represents my original work, in its entirety, withediting provided by Dr. Douw Steyn. In the future, a paper based on the results of this workwill be submitted for publication in a peer-reviewed journal.Chapter 1I wrote this chapter. Figure 1.1 was taken from Oke and Hay (1998), with permission from theauthor. My supervisors, Dr. Douw Steyn and Dr. Ian McKendry, guided me in reformulatingthe research questions and setting the scope of the research.Chapter 2All observational datasets and preliminary modelling in WRF and SMOKE were prepared andperformed by Dr. Bruce Ainslie. I performed a more detailed meteorological model evaluationfollowing the work of Steyn et al. (2011). Using the model output provided to me, I performedall subsequent post-processing, analyses, and created all figures. I was responsible for settingiiiup the HYSPLIT trajectory model, and planning and performing all model runs. I preparedall figures in this chapter, except: data for Figure 2.2 was provided by Dr. Bruce Ainslie, andFigure 2.3 was adapted from Ainslie and Steyn (2007).Chapter 3I performed all analyses and created all figures in this chapter. All committee membersprovided input to clarify the figures.Chapter 4Dr. Douw Steyn and I discussed the proposed definition for “recirculation”. I refined, devised,and optimized a suitable detection algorithm to apply to trajectory model output. The domainof interest was chosen with input from all committee members. I designed and created alloriginal data visualizations. Dr. Douw Steyn helped analyze and interpret Figures 4.12 and4.15.Chapter 5I wrote this chapter and drew all conclusions.AppendicesI created all additional figures presented in the Appendices (B, C, D). In Appendix B, Iperformed the meteorological model evaluation following the work of Dr. Bruce Ainslie. Iapplied the Sea Breeze Filter of Steyn and Faulkner (1986) to all WRF model output fields,and analyzed all wind fields.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Supplementary Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xivAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Air pollution meteorology and tropospheric ozone . . . . . . . . . . . . . . . . . 11.2 Air quality and ozone episodes in the Lower Fraser Valley . . . . . . . . . . . . 21.3 Atmospheric recirculation of air pollutants . . . . . . . . . . . . . . . . . . . . . 61.3.1 Definition and description . . . . . . . . . . . . . . . . . . . . . . . . . . 61.3.2 Evidence of recirculation in the Lower Fraser Valley . . . . . . . . . . . 71.3.3 Previous studies of recirculation . . . . . . . . . . . . . . . . . . . . . . . 81.4 Research questions and approach . . . . . . . . . . . . . . . . . . . . . . . . . . 122 Study approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.1 The Lower Fraser Valley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.1.1 Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.1.2 Description of ozone episode circulation regimes . . . . . . . . . . . . . . 15v2.1.3 Selected ozone episodes in this study . . . . . . . . . . . . . . . . . . . . 192.2 Meteorological modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.1 WRF model description and set up . . . . . . . . . . . . . . . . . . . . . 192.2.2 Overview of episode meteorology . . . . . . . . . . . . . . . . . . . . . . 212.2.3 Model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.3 Chemical emissions field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3.2 Emissions inventory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.4 Trajectory modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.4.2 HYSPLIT model description . . . . . . . . . . . . . . . . . . . . . . . . 252.4.3 Model set up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283 Trajectory analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.1 Overview of HYSPLIT modelling results . . . . . . . . . . . . . . . . . . . . . . 303.2 Trajectory sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.2.1 Qualitative spatial assessment . . . . . . . . . . . . . . . . . . . . . . . . 353.2.2 Transport deviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.2.3 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.3 Analysis of composite trajectory fields . . . . . . . . . . . . . . . . . . . . . . . 433.3.1 Spatial patterns of trajectories in relation to circulation regimes . . . . . 433.3.2 Number density maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504 Recirculation definition and analysis . . . . . . . . . . . . . . . . . . . . . . . . 524.1 A proposed definition for “recirculation” . . . . . . . . . . . . . . . . . . . . . . 524.1.1 General description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53vi4.1.2 Application to ozone episodes in the Lower Fraser Valley . . . . . . . . . 564.2 Spatial characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.2.2 Description of selected Recirculating Trajectory Segments . . . . . . . . 654.3 Temporal characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.4 The Recirculation Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734.4.1 Description of RF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734.4.2 Comparison to detection algorithm for recirculation . . . . . . . . . . . . 744.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 795.1 Major findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.2 Future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85A Station locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92A.1 Meteorological stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92A.2 Air quality monitoring stations . . . . . . . . . . . . . . . . . . . . . . . . . . . 93B Meteorological modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94B.1 WRF model configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94B.2 Model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96B.3 Assessment of thermo-topographic circulations . . . . . . . . . . . . . . . . . . 102B.3.1 Land-sea breeze . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102B.3.2 Mountain-valley breeze and slope flows . . . . . . . . . . . . . . . . . . . 106C Additional material for trajectory analysis . . . . . . . . . . . . . . . . . . . . 110C.1 HYSPLIT modelling results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110C.2 Number density maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117viiD Additional material for recirculation analysis . . . . . . . . . . . . . . . . . . . 120D.1 Episodic temporal characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . 120D.2 Relationship between diurnal wind cycle and recirculation lifetime . . . . . . . 127viiiList of Tables2.1 Selected modelled episode days and their associated circulation regimes, asdefined by Ainslie and Steyn (2007). . . . . . . . . . . . . . . . . . . . . . . . . 212.2 HYSPLIT model configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . 264.1 Observed and modelled BL heights during ozone episodes selected in this study. 58A.1 Meteorological station locations within the Lower Fraser Valley and surroundingarea. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92A.2 Select air quality monitoring stations in the Lower Fraser Valley. . . . . . . . . 93B.1 WRF model domain configuration used for all modelled episodes. . . . . . . . . 94B.2 Model evaluation results for u and v wind component directions of the 4 kmWRF domain for the 2006 episode. . . . . . . . . . . . . . . . . . . . . . . . . . 97B.3 Model evaluation results for wind speeds of the 4 km WRF domain for the 2006episode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98B.4 As in Table B.2, but for the 1.33 km WRF domain. . . . . . . . . . . . . . . . . 100B.5 As in Table B.3, but for the 1.33 km WRF domain. . . . . . . . . . . . . . . . . 101B.6 Onshore and offshore wind direction ranges by location, determined by Steynand Faulkner (1986). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102B.7 Southern Strait of Georgia mean monthly sea surface temperature (SST) forsummer months. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103B.8 SB filter results at YVR for all episodes . . . . . . . . . . . . . . . . . . . . . . 105ixList of Figures1.1 Schematic of an ozone episode in the LFV. . . . . . . . . . . . . . . . . . . . . . 51.2 Surface ozone measurements taken at the mouth of Pitt River Valley, August1-6, 1993 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.1 Map of the Lower Fraser Valley (LFV). . . . . . . . . . . . . . . . . . . . . . . . 152.2 Composite daily hodographs at YVR for circulation regimes I – IV. . . . . . . . 172.3 Composite synoptic MSLP (hPa) maps for circulation regimes I – IV. . . . . . . 182.4 WRF model domain set up for all modelled episodes. . . . . . . . . . . . . . . . 202.5 Daily hodographs at YVR for the 2006 episode. . . . . . . . . . . . . . . . . . . 222.6 SMOKE-adjusted 4 km resolution inventory for average hourly ozone precursor(NOx + VOCs) emissions during the 1985 episode. . . . . . . . . . . . . . . . . 243.1 Modelled trajectories (black lines) during the selected ozone episodes. . . . . . . 323.2 Modelled trajectories during the 1985 episode. . . . . . . . . . . . . . . . . . . . 333.3 Control (black) and test trajectories (red) modelled during the 1993 episode. . . 363.4 As in Figure 3.3, but for trajectories modelled during the 1998 episode. . . . . . 373.5 The Absolute Horizontal Transport Deviation (AHTD) for test trajectoriesduring the 1993 episode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.6 As in Figure 3.5, but for test trajectories during the 1998 episode. . . . . . . . . 393.7 As in Figure 3.5, but for the Absolute Vertical Transport Deviation (AVTD)during the 1993 episode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.8 As in Figure 3.7, but for test trajectories during the 1998 episode. . . . . . . . . 423.9 Number density maps generated from composite trajectory fields. . . . . . . . . 463.10 The number density maps by day (circulation regime) of the 1987 episode. . . . 483.11 Pearson’s r correlation matrix for daily trajectory number density maps. . . . . 49x4.1 Schematic diagram illustrating the definition of recirculation and itsimplementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.2 The selected horizontal domain of interest used to identify recirculation in theLFV in this study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574.3 Number of RTSs per trajectory of all modelled trajectories for the seven ozoneepisodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.4 Detected RTSs from modelled trajectories during ozone episodes. . . . . . . . . 614.5 Direction of departure and direction of return of RTSs, with respect to thetrajectory origin (X0). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.6 Statistics of detected RTSs by circulation regime. . . . . . . . . . . . . . . . . . 634.7 As in Figure 4.4, but for all RTSs by circulation regime. . . . . . . . . . . . . . 644.8 A RTS detected from a trajectory initiated at 1300 (local) on the second dayof the 2006 episode (June 25, 2006; hour 37 of the three-day model run). . . . . 664.9 As in Figure 4.8, but for the trajectory initiated at 1700 (local). . . . . . . . . . 674.10 As in Figure 4.9, but for the trajectory initiated at 0600 on the first day of the2006 episode (June 24, 2006; hour 6 of the three-day model run). . . . . . . . . 694.11 Temporal characteristics of detected RTSs during the 2006 episode. . . . . . . . 704.12 RTS start times (local) and lifetimes (τR), by circulation regime. . . . . . . . . 724.13 Distribution of RTS lifetimes (τR), over all seven modelled episodes. . . . . . . 734.14 Schematic diagram of Allwine and Whiteman’s (1994) Recirculation Factor. . . 754.15 Eulerian (RFE) and Lagrangian (RFL) measures of Recirculation Factor for alldetected RTSs, by circulation regime. . . . . . . . . . . . . . . . . . . . . . . . . 76B.1 Topography of the LFV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95B.2 Inner 1.33 km domain in WRF used to model the 2006 episode. . . . . . . . . . 99B.3 Frequency of criterion pass of the SB filter. . . . . . . . . . . . . . . . . . . . . 104B.4 WRF modelled surface (10 m) winds on the second day of the 2001 episode. . . 107B.5 As in B.4, but for the second day of the 2006 episode. . . . . . . . . . . . . . . 109xiC.1 Modelled trajectories during the 1987 episode. . . . . . . . . . . . . . . . . . . . 111C.2 As in Figure C.1, but for trajectories during the 1993 episode. . . . . . . . . . . 112C.3 As in Figure C.1, but for trajectories during the 1995 episode. . . . . . . . . . . 113C.4 As in Figure C.1, but for trajectories during the 1998 episode. . . . . . . . . . . 114C.5 As in Figure C.1, but for trajectories during the 2001 episode. . . . . . . . . . . 115C.6 As in Figure C.1, but for trajectories during the 2006 episode. . . . . . . . . . . 116C.7 The number density maps by day (circulation regime) of the 1985 episode. . . . 117C.8 As in Figure C.7, but for days of the 1993 episode. . . . . . . . . . . . . . . . . 117C.9 As in Figure C.7, but for days of the 1995 episode. . . . . . . . . . . . . . . . . 118C.10 As in Figure C.7, but for days of the 1998 episode. . . . . . . . . . . . . . . . . 118C.11 As in Figure C.7, but for days of the 2001 episode. . . . . . . . . . . . . . . . . 118C.12 As in Figure C.7, but for days of the 2006 episode. . . . . . . . . . . . . . . . . 119D.1 Temporal characteristics of detected RTSs during the 1985 episode. . . . . . . . 121D.2 Temporal characteristics of detected RTSs during the 1987 episode. . . . . . . . 122D.3 Temporal characteristics of detected RTSs during the 1993 episode. . . . . . . . 123D.4 Temporal characteristics of detected RTSs during the 1995 episode. . . . . . . . 124D.5 Temporal characteristics of detected RTSs during the 1998 episode. . . . . . . . 125D.6 Temporal characteristics of detected RTSs during the 2001 episode. . . . . . . . 126D.7 Idealized effect of diurnal wind cycle on resulting recirculation lifetimes (τR). . 128xiiList of Supplementary MaterialsSupplementary Materials are available from cIRcle, The University of British Columbia’sonline digital repository for research and teaching materials. Files can be found under theSupplementary Thesis Materials and Errata collection.V1. Animation of surface (10 m) wind fields modelled in WRF and trajectoriesmodelled in HYSPLIT for the 1985 episode . . . . . . . . . hysplit_1985.mp4V2. As in V1 but for the 1987 episode . . . . . . . . . . . . . . . hysplit_1987.mp4V3. As in V1 but for the 1993 episode . . . . . . . . . . . . . . . hysplit_1993.mp4V4. As in V1 but for the 1995 episode . . . . . . . . . . . . . . . hysplit_1995.mp4V5. As in V1 but for the 1998 episode . . . . . . . . . . . . . . . hysplit_1998.mp4V6. As in V1 but for the 2001 episode . . . . . . . . . . . . . . . hysplit_2001.mp4V7. As in V1 but for the 2006 episode . . . . . . . . . . . . . . . hysplit_2006.mp4xiiiList of Acronymsagl Above Ground LevelAHTD Absolute Horizontal Transport Deviationasl Above Sea LevelAVTD Absolute Vertical Transport DeviationAQ Air QualityARL Air Resources LaboratoryARW Advanced Research WRFBC British ColumbiaBL Boundary LayerCAAQS Canadian Ambient Air Quality StandardsCMAQ Community Multi-scale Air Quality modeling systemFVRD Fraser Valley Regional DistrictHYSPLIT HYbrid Single-Particle Lagrangian Integrated Trajectory modelLB Land BreezeLFV Lower Fraser ValleyLSB Land-Sea BreezeML Mixed LayerMSLP Mean Sea-Level PressurexivMV Metro VancouverNAAQO National Ambient Air Quality ObjectiveNARR North American Regional Re-analysisNOAA National Ocean and Atmospheric AdministrationNWP Numerical Weather PredictionRF Recirculation FactorRMSE Root-Mean-Square ErrorRTS Recirculating Trajectory SegmentPGF Pressure Gradient ForceSB Sea BreezeSMOKE Sparse Matrix Operator Kernel Emissions modelling systemSST Sea Surface TemperatureUSA United States of AmericaVCC Vector Correlation CoefficientWA WashingtonWRF Weather Research and Forecasting modelxvAcknowledgementsI would like to take this opportunity to extend my most sincere gratitude to those who helpedmake this work possible.First and foremost, I would like to thank my two supervisors, Dr. Douw Steyn andDr. Ian McKendry, for their continued support, encouragement, and understanding. Theirguidance during my time at UBC has been invaluable, all the way from my first years as anundergraduate student through the completion of this degree. To you both: thank you forintroducing me to and guiding me through the field of air pollution meteorology, for sharingyour passion for scientific research, and for the many insightful discussions. I am also sincerelythankful to Douw for keeping me on track and connecting me to the ITM conference, amongmany other amazing opportunities.I would also like to extend my gratitude to my other graduate committee member, Dr.Bruce Ainslie (Environment Canada), for all the data and model output he made accessibleto me, and for answering my many questions. His instructions and critical insights were muchappreciated.In my time as a graduate student, I had the pleasure of learning from Dr. Susan Allen,Dr. Phil Austin, and Dr. Roland Stull. Their instruction on course material truly deepenedmy understanding and appreciation of atmospheric sciences. I would also like to acknowledgeDr. Allan Bertram for his guidance through the CREATE-AAP program.To those listed above, it has been both a pleasure and privilege working with all of you.Financial support for this work was provided by the Natural Sciences and EngineeringResearch Council of Canada (NSERC) Collaborative Research and Training Experience –Atmospheric Aerosol Program (CREATE-AAP), and derived from the NSERC UndergraduateStudent Research Award (USRA) granted to Annie Seagram, from NSERC grants to DouwSteyn and Peter Jackson and from the British Columbia Clean Air Research Fund supportedxviby the Fraser Basin Council, Fraser Valley Regional District, and Metro Vancouver.Special thanks to all of my fellow officemates and “labmates”, past and present. I trulyenjoyed our overzealous discussions, endless hours at the whiteboard, sharing programmingtips, and innumerable coffee breaks.To my dearest friends, thank you for always being there with a reassuring “thumbs up”.To my siblings and parents, thank you for your continued support, despite being so far away.Finally, my deepest gratitude to Steven Luscher, for not only being my personal “de-bugger”,but for his endless patience, encouragement, and partnership.xvii1 IntroductionThe interaction between synoptic and mesoscale flow systems in the Lower Fraser Valley(LFV) of British Columbia (BC) determines the occurrence, severity, and spatiotemporalvariability of degraded air quality episodes (Ainslie and Steyn 2007). Situated in a coastalregion with complex terrain, slack synoptic conditions over the LFV (see Figure 2.1) canfavour the development of both the land-sea and mountain-valley breeze circulations, whichare modified by the surrounding topographic constraints. In particular, the phenomenon of“recirculation” – the process by which a polluted air mass may pass over the same site severaltimes, leading to an increase in pollutants at the site – is proposed to have significant effects onthe spatiotemporal variability of ozone in the LFV (e.g. Ainslie and Steyn 2007). This studyaims to determine if recirculation occurs in the LFV through the use of a trajectory model, inorder to better understand the patterns and processes of atmospheric transport during ozoneepisodes.1.1 Air pollution meteorology and tropospheric ozoneThough air pollution is directly related to the type and amount of emissions, it is the state ofthe atmosphere that determines whether or not air quality (AQ) conditions will be exacerbatedor alleviated. In other words, given that regional emissions generally do not change on adaily basis, the occurrence of periods of degraded AQ is governed by atmospheric conditions.The atmosphere acts to disperse, transport, destroy (remove via wet or dry deposition), andproduce (chemically transform pollutants and alter chemical reaction rates) pollutants (Okeand Hay 1998, Sillman 1993), which determines ambient air pollution concentrations measured11.2. Air quality and ozone episodes in the Lower Fraser Valleywithin AQ monitoring networks.Tropospheric ozone (O3) is a gaseous secondary air pollutant that can have significantnegative impacts on human health (Brauer and Brook 1997), vegetation (Krzyzanowski et al.2006), and environmental visibility (Steyn et al. 1997). Ozone is formed photochemicallythrough the nonlinear reactions of oxides of nitrogen (NOx, the total of nitrogen monoxide,NO, and nitrogen dioxide, NO2) and volatile organic compounds (VOCs). In the LFV, mobileemissions released in the western portion of the valley account for 80% of these “precursor”species (Steyn et al. 1997, Oke and Hay 1998). In the lower troposphere, the lifetime of NOx(∼ 6 h) is much shorter than those of VOCs (1 h – 5 d), while ozone has a lifetime of several daysto weeks (Sillman 1993). The formation and ambient concentration of ozone depends on theratio and reactivity of these precursor species and removal processes, in addition to the stateof the atmosphere: the amount of sunlight, ambient air temperature, wind speed, and mixedlayer (ML) depth. High temperatures enhance ozone production (Sillman 1993), while lightwind speeds and low ML depths reduce dispersion, allowing pollutants to accumulate. In theLFV, though ozone concentrations are strongly correlated with ambient air temperatures, MLdepths and wind speeds are not (Steyn et al. 1990 as cited in Joe et al. 1996). Wind patternsand wind direction are also important since they determine the direction that an air masswill travel, as well as its chemical history, and areas that pollutants could potentially impact.As a result of this multitude of chemical and meteorological influences, ozone mixing ratios(“concentrations”) vary on hourly, diurnal, hebdomadal (Pryor and Steyn 1995), seasonal, andinter-annual time scales.1.2 Air quality and ozone episodes in the Lower Fraser ValleyOzone episodes (or “exceedance days”) are periods of time when ozone concentrations atone or more of the AQ monitoring stations exceed the acceptable limits. In BC, ambientozone must not exceed levels outlined by the National Ambient Air Quality Objective(NAAQO), where the national maximum acceptable level is 82 ppb based on 1-h averagedmeasurements, and the Canadian Ambient Air Quality Standards (CAAQS, formerly the21.2. Air quality and ozone episodes in the Lower Fraser ValleyCanada Wide Standards/CWS), where the threshold is 65 ppb of 8-h averaged measurements,and achievement is based on the 4th highest annual values averaged over 3 years (CanadianCouncil of Ministers of the Environment 2000). The former is useful when examining isolatedepisodes of elevated ozone levels and issuing public advisories, whereas the latter is used todetermine if a region is meeting national long-term health protection objectives.Ozone episodes generally occur under synoptic high pressure (anticyclonic) conditionsin the summer months. The upper-level subsidence and weak forcing cause a temperatureinversion, which suppresses ML heights and limits vertical dispersion, and in response, hightemperatures, high solar radiation, and low wind speeds occur at the surface. In the LFV,these episodes typically last 1–3 days, and, in most recent years, occur only 2–3 times eachsummer (Ainslie and Steyn 2007).The climatology of ozone episodes in the LFV is well known. On the synoptic scale, theseepisodes manifest under a very narrow set of conditions, requiring the presence of a lower-level(MSLP) thermal trough in conjunction with an upper-level (500 hPa) ridge (McKendry 1994).At the mesoscale, Ainslie and Steyn (2007) found that there are four circulation regimes thatgovern ozone episodes. These regimes are distinguished by their averaged surface pressuremaps, and daily hodographs derived from surface stations: two with northerly daytime winds,and two with southerly daytime winds. A more detailed description of these circulation regimesis provided in Section 2.1.2. The authors note that these wind regimes differ most in themorning hours, which has important implications for the transport of precursor pollutantsreleased during the morning rush-hour period. Despite considerable differences between thecirculation regimes’ daily wind patterns, their composite ozone distributions are similar. Thissuggests that the spatial evolution of ozone in the LFV may be influenced by other processesthat act to build up and/or recirculate precursor pollutants.The same conditions that facilitate ozone episodes (slack anticyclonic synoptic conditions,high insolation) also give rise to thermo-topographically forced circulations, such as theland-sea breeze (LSB) circulation. The LSB is a thermally-forced mesoscale circulation thatfrequently occurs in the LFV (Steyn and Faulkner 1986) and in coastal regions worldwide (see31.2. Air quality and ozone episodes in the Lower Fraser ValleyAbbs and Physick (1992), Miller et al. (2003), and Crosman and Horel (2010) for comprehensivereviews). With its diurnally reversing wind pattern, from daytime onshore flow to nighttimeoffshore flow, it is often proposed that the LSB acts to recirculate air pollutants from one dayof an episode to the next, which further intensifies periods of degraded AQ (an illustration andreview of this behaviour is provided in Section 1.3). However, Ainslie and Steyn (2007) findthat ozone episodes are not necessarily associated with well-defined LSB circulations. Thismay be attributed to the presence of a synoptic lower-level thermal trough, which can inhibitsea breeze (SB) formation (McKendry 1994).Furthermore, the complex topography surrounding the LFV allows for mountain-valleybreezes and slope flows. These are generated in a similar manner to LSBs. In the LFV,these flows interact near mid-valley (Chen and Oke 1994, Cai and Steyn 2000), furthercomplicating the airshed’s flow pattern and spatial variations of pollutants. And, with similardiurnally-reversing winds, mountain-valley breezes and slope flows can recirculate air massesto and from the main valley floor.Not only does the steep terrain flanking the northern and southern margins of the LFV giverise to these thermo-topographic circulations, but it also acts as a physical barrier that furtherlimits net horizontal pollutant transport out of the airshed. Vertical dispersion is limited due tothe temperature inversion caused by large scale subsidence, where the maximum vertical extentof the ML is lower than the peaks of the valley side walls (Oke and Hay 1998). Essentially, onecan think of the LFV airshed as a three-dimensional “box” that effectively traps pollutants inthe region (Figure 1.1).The spatiotemporal distribution and formation of ozone in the LFV has been welldocumented in over three decades of research, through several observational, statistical,empirical, and numerical modelling studies, including two extensive field campaigns (Pacific1993 (Steyn et al. 1997), and Pacific 2001 (Li 2004)), and a major retrospective modellinginvestigation of ozone formation (Steyn et al. 2011, subsequently published as Ainslie et al.(2013) and Steyn et al. (2013)). Taken together, this body of literature helps to createa three-dimensional understanding of how ambient ozone concentrations evolve during an41.2. Air quality and ozone episodes in the Lower Fraser ValleyFigure 1.1. Schematic of an ozone episode in the Lower Fraser Valley, onSeptember 1, 1988. (A) The 3-dimensional “box” in which pollutants are trapped, withschematic of thermo-topographic circulations (arrows) (B) Typical transport and chemicaltransformations, and (C) surface isopleths of ozone (ppb) and surface wind vectors. Takenfrom Oke and Hay (1998), with permission.51.3. Atmospheric recirculation of air pollutantsepisode in response to various meteorological mechanisms and changing chemical influences.Here, the findings of these works are summarized in order to illustrate typical observationsduring ozone episodes.During the day, ozone concentrations are lowest around the downtown core of Vancouverwhich can be attributed to ozone titration (e.g. Ainslie and Steyn 2007) – the process wherefresh NO emissions react with ozone to form NO2 and oxygen. Owing to onshore winds and thetime required for ozone formation, higher ozone concentrations are typically found eastward(downwind) (Pryor and Steyn 1995, Hoff et al. 1997, McKendry et al. 1998a, Hedley et al.1997, Oke and Hay 1998), increasing towards Hope (Steyn et al. 2011) (cf Figure 1.1). Theadjoining tributary valleys are frequently subjected to ozone concentrations that are equivalentto or higher than those found in the main valley floor (McKendry et al. 1998a,b, Hedley et al.1997). Slope flows can inject pollutants above the inversion layer to form elevated layers ofpollutants (known as the “chimney effect”; Hayden et al. 1997, Hoff et al. 1997, McKendryet al. 1997, Brook et al. 2004, Strawbridge and Snyder 2004a). Pollutants from within theselayers may recirculate via morning downward mixing or fumigation, which can impact surfaceconcentrations.At night, higher concentrations of pollutants may be found over the Strait of Georgia,coincident with nighttime offshore flow (Hoff et al. 1997, Strawbridge and Snyder 2004a).Nocturnal outflow from tributary valleys carries cleaner air (low ozone, but high NO2) backinto the main valley floor. The reduction of ozone in this airmass can be attributed to ozonetitration and dry deposition along the valley sidewalls (Banta et al. 1997).1.3 Atmospheric recirculation of air pollutants1.3.1 Definition and descriptionUp to this point, the term “recirculation” has not been precisely defined. Though there exists asubstantial body of literature surrounding the topic of atmospheric recirculation of pollutants,the term is rarely defined explicitly, especially in relation to the methods used to identify its61.3. Atmospheric recirculation of air pollutantsbehaviour.Steyn (2003) relates recirculation to the LSB and mountain-valley breeze, and distinguishesbetween “horizontal” and “vertical” recirculation. He describes vertical recirculation as pollutedair that follows “a recirculating trajectory by being advected landward in the sea breeze, andsubsequently be lofted by rising air at the [sea breeze front], to return seaward in the upperreturn flow”, and horizontal recirculation as “when polluted air is carried landward by the SB[and valley winds] during daylight hours, and seaward by the land breeze [and mountain breeze]at night” (pp. 288 – 290). Note that the opposite cycle is also possible: pollutants releasedinto the nighttime land breeze (LB) may travel offshore where they reside over the adjacentbody of water before returning onshore during the next day’s daytime SB. In either case,pollutants released on one day may return the next to intensify the episode. Thus, the LSBand mountain-valley breeze circulations horizontally recirculate pollutants on the time-scaleof the circulations, i.e. 24 h.Steyn (2003) implies that slope flows govern vertical recirculation more than horizontalrecirculation, as they loft pollutants into elevated layer structures. Vertical down-mixing ofpollutants from aloft has been referred to as a pollutant “carryover” or “handover process” byothers (e.g. McKendry and Lundgren 2000).1.3.2 Evidence of recirculation in the Lower Fraser ValleyThe LFV has generally good air quality (Steyn et al. 1997, Vingarzan and Taylor 2003) due tothe relatively low population size of the region, low total emissions of pollutants (Ainslie andSteyn 2007), and relatively low industrial activity. Despite these facts, however, the region isstill subject to severe – though infrequent – ozone episodes (Ainslie and Steyn 2007, Vingarzanand Taylor 2003). The severity of ozone episodes is often attributed to strong recirculation ofpollutants within the airshed. Evidence for both horizontal and vertical recirculation in theLFV has already been indicated in previous sections (cf Section 1.2). To summarize:(1) Episodic trends in ozone concentrations suggest that recirculation acts to intensifyozone episodes. A noted trend of increasing ozone daily maxima, with low varying71.3. Atmospheric recirculation of air pollutantsdaily minima, is evident in observational data sets in the LFV (e.g. McKendry et al.1998a; Figure 1.2) and elsewhere (e.g. Evtyugina et al. 2006, St. John and Chameides1997). Since the formation of ozone is a photochemical reaction, a low concentration ofozone is expected during nighttime hours. The maximum concentration is observed afterlocal noon, often corresponding to time of maximum ambient temperatures. However,relatively consistent low nighttime values of ozone indicate that emission rates do notchange over time. If the latter is true, the increase in daily maxima suggests that there issome carryover of pollutants from one day of an episode to the next through the processof recirculation.(2) Several studies in the LFV (McKendry 1994, Pryor et al. 1995, Ainslie and Steyn 2007)point out that neither synoptic nor mesoscale meteorological conditions alone accountfor the observed spatiotemporal distribution of pollutants during ozone episodes. Theypropose that recirculation and precursor build-up is more important than the overallmesoscale conditions.(3) The same meteorological conditions conducive to ozone episodes also favour thedevelopment of thermo-topographically forced circulations, that have characteristicdiurnally-reversing wind patterns. Though the LSB is not necessarily associated withozone episodes in the LFV (Ainslie and Steyn 2007), the presence of strong onshore flowand thermo-topographic circulations can still recirculate pollutants.1.3.3 Previous studies of recirculationA priori knowledge of the methods used to study the phenomenon is essential to begin aninvestigation of this kind. The recirculation of pollutants has been shown to play a criticalrole during air pollution episodes observed worldwide. Common approaches to investigatingrecirculation include: interpreting meteorological parameters (especially wind speed anddirection) and complementary pollution measurements at AQ monitoring stations duringobservational periods (e.g. Evtyugina et al. 2006), or by using combination of observational81.3. Atmospheric recirculation of air pollutantsFigure 1.2. Surface ozone measurements taken at the mouth of Pitt River Valley, August1-6, 1993. From McKendry et al. (1998a).data and numerical modelling techniques. Meteorological and chemical modelling (e.g.Clappier et al. 2000), trajectory modelling (e.g. Ding et al. 2004), and dispersion modelling(e.g. Segal et al. 1982) are common tools used in these studies.The role of the LSB on pollutant recirculation has been found in many coastal locations,including: Los Angeles, California (Holzworth et al. 1963, Shair et al. 1982, Lu and Turco1995, Lu and Turco 1996); Houston, Texas (Banta et al. 2005); Perth, Australia (Hurley andManins 1995, Ma and Lyons 2003); Chicago, Illinois (lake breeze rather than SB; Harris andKotamarthi 2005); and several locations in Greece (Clappier et al. 2000, Grossi et al. 2000,Papanastasiou and Melas 2009), Israel (Segal et al. 1982, Tov et al. 1997, Levy et al. 2008a,b,2009, 2010), China (Cheung and Wang 2001, Wang and Kwok 2003, Ding et al. 2004, Shanet al. 2009), and across the Iberian Peninsula (Millan and Artinano 1992, Millan et al. 2000,Evtyugina et al. 2006). The results from these studies are inherently site-specific, relatingto the region’s topography and coastline orientation, as well as the region’s ozone episodeclimatology. Nonetheless, several general principles can be drawn from this body of research.For illustrative purposes, the findings of a few of these studies are discussed below.In the Los Angeles basin, early observational research on recirculation confirmed that theLSB causes a reversal (‘looping’) in the trajectory of a constant-level balloon (Holzworthet al. 1963). Shair et al. (1982) performed a tracer experiment where they released an91.3. Atmospheric recirculation of air pollutantsatmospheric chemical tracer into the nighttime LB to observe its dispersion along the LosAngeles coastline. They observed that the tracer remained over the ocean for over 12 h beforeadvecting back towards the coast by the next morning’s SB. They noted that the spatialimpacts of recirculation can be widespread, and that the onset of the SB coincides with anincrease in pollutants measured at the shoreline. Several other studies have also observed thisfinding (e.g. Millan and Artinano 1992, Grossi et al. 2000).With the advent of computer modelling came a number of coupled meteorological-chemicalmodelling studies that attempted to determine the impact of recirculation on entire airsheds.As shown by Lu and Turco (1995), Lu and Turco (1996), Clappier et al. (2000), and Grossiet al. (2000), the general meteorological and chemical trends of recirculation (as describedin section 1.3.2) are reproduced well in these modelling studies. These modelling studieshighlight the relationship between the spatiotemporal variability of pollutants and the spatialand temporal scales of the region’s mesoscale thermo-topographic circulations.Semi-Lagrangian forward and backward trajectory (“back trajectory”) models are oftenused to investigate ozone transport and recirculation, since they directly trace potentialrecirculating paths of tracer air parcels. That is to say, trajectory modelling is a convenienttool since it allows the investigator to “see” a recirculating air parcel. Back trajectorymodelling is a more common approach, since the terminal point can be used to establishsource-receptor relationships (see Fleming et al. (2012) for a review of this technique), and thepoint of arrival can be made to coincide with an AQ monitoring station from which pollutantdata are available. By attributing a trajectory’s arrival time to a measured or modelledozone concentration, many studies have found that higher ozone concentrations are linked totrajectories that (visually) recirculate (e.g. Tov et al. 1997, Zabkar et al. 2008, Davis et al. 2010,Cheung and Wang 2001). Studies employing trajectory models also show that the start timeof the trajectory determines the resulting pathway and thus the duration of the recirculationof an air parcel, since the wind field changes over time (Ma and Lyons 2003, Ding et al. 2004,Harris and Kotamarthi 2005). Some of these studies have found that although the LSB playsa role in intensifying ozone episodes, the LSB alone does not necessarily recirculate pollutants101.3. Atmospheric recirculation of air pollutants(Ma and Lyons 2003, Harris and Kotamarthi 2005, Ding et al. 2004). Rather, recirculationoccurs as a result of interacting mesoscale and synoptic conditions.Alternatively, Allwine and Whiteman (1994) propose a simple, mathematical measure ofatmospheric recirculation: the Recirculation Factor (RF). The RF is derived from stationwind data (speed and direction) at a single point in space. An overview of the RF is providedin Section 4.4. By assembling a vector progression of the wind vectors over time, the RF iscalculated from the ratio of the total wind run to the net displacement of the wind vectors,computed over the transport time, τ . By definition, however, the RF is sensitive to theselection of τ , though a value of τ = 24 h is recommended (Allwine and Whiteman 1994, Levyet al. 2008b) because that is the period of diurnal wind circulations.To summarize, first, though the use of trajectory modelling to identify recirculation inthe aforementioned studies seems promising for any future investigations, we note that theliterature lacks a precise, comprehensive, and quantitative definition of “recirculation”. Theseworks (Hurley and Manins 1995, Tov et al. 1997, Cheung and Wang 2001, Ma and Lyons2003, Ding et al. 2004, Harris and Kotamarthi 2005, among others) tend to present outputfrom trajectory models that show apparent recirculation as a loop in a trajectory. However,this analysis is highly subjective, since there is no indication of where or when a recirculationbegins or ends. As a result, this does not allow for a quantitative temporal and spatialdescription. Thus, to investigate recirculation and its behaviour through space and time, aquantitative and objective definition must be developed. Second, though there exists a measureof the RF, defining the mathematical measure of recirculation based on a transport time isnot robust. This is especially evident when compared to the aforementioned modelling studiesthat implicitly deduce recirculation spatially from trajectory output. Furthermore, the RF isnot likely to represent more than a characterization of potential flow at a single point for theLFV, an area of complex terrain with inhomogenous wind fields, where the spatial scales areconsistent with the time scales of the circulations. This necessitates a thoughtfully designed,generally applicable measure and objective criteria for recirculation.111.4. Research questions and approach1.4 Research questions and approachIn this modelling study, we seek to define and provide evidence of recirculation during seventhree-day, summertime ozone episodes that have occurred in the LFV over a period of 20 years.These episodes represent the known set of mesoscale circulation regimes of ozone episodesidentified by Ainslie and Steyn (2007), and were previously modelled and described at lengthby Steyn et al. (2011). Here, we will make use of the same modelled meteorological fields ina forward trajectory analysis. By using a trajectory model, we will focus our attention onhorizontal recirculation.In essence, this research contributes to – and ties together – a much larger body ofwork that focusses on ozone episodes in the region over the past three decades. Identifyingpossible recirculation mechanisms and (trajectory) pathways during ozone episodes is keyto understanding the dispersion patterns and the spatiotemporal variability that have beenobserved. From the perspective of individual episodes, it is important to determine how farpolluted air is transported within the LFV since this determines the distribution and build-upof pollutants for the next morning (Banta et al. 1997). In a broader sense, understandingthe downwind transport and recirculation of ozone and its precursors is imperative given thechanging spatial chemical sensitivity of the region (Ainslie and Steyn 2007, Ainslie et al. 2013)and eastward shift in the location of the ozone plume during episodes (Ainslie and Steyn2007). A study of this type will highlight the ability of the LFV airshed to trap pollutants,which is difficult to determine given the current spatial distribution of AQ monitoring stationsconcentrated in urbanized areas. To inform future abatement strategies of precursor pollutants,a more complete understanding of the potential for recirculation during ozone episodes isessential.In light of the known meteorological circulation regimes of ozone episodes in the LFV, andthe lack of definition and quantification of recirculation in general, the following set of researchquestions will steer the investigation:(1) Can we develop a generally applicable definition of ‘recirculation’?121.4. Research questions and approach(2) Are existing measures of recirculation appropriate based on Question (1)?(3) Does recirculation occur in the LFV during ozone episodes?(4) Are there particular transport pathways that are characteristic of different ozonecirculation regimes in the LFV?(5) What are the spatial and temporal characteristics of recirculation in the LFV underdifferent circulation regimes?An overview of the ozone episodes selected for this study and their circulation regimes aredescribed in Chapter 2. The study approach and suite of numerical models used are thenoutlined, and intermediate modelling results are discussed. A trajectory modelling exercise isperformed, and the subsequent analysis is separated into two portions. First, the total set oftrajectories generated for all ozone episodes is presented in Chapter 3. Then, a definition for“recirculation” is developed and examined in Chapter 4. This study concludes in Chapter 5,with a review of the original research objectives. Future research avenues are also discussed.132 Study approachIn this section, the four mesoscale circulation regimes of ozone episodes in the LFV, and theseven ozone episodes used in this study are described. In order to investigate atmospherictransport and recirculation in the LFV, a combination of numerical meteorological, chemical,and trajectory models will be used. Meteorological modelling of all episodes is performedin WRF. After evaluating the wind velocity fields, the output from WRF is used to computetrajectories using the HYSPLIT model. Of critical importance is the selection of an appropriatestarting location for the simulated trajectories. To select a point that will align with theresearch questions and reflect the LFV’s atmospheric chemistry, an emissions inventory ofozone precursors modelled by SMOKE is employed. Note that this is a kinematic approach.2.1 The Lower Fraser Valley2.1.1 SettingThe LFV is situated on the west coast of BC, straddling the Canada-USA border along the49th parallel (Figure 2.1). It is roughly triangular in shape: the main valley bottom is flankedby the North Shore Mountains (a local name for the Coast Mountains) to the north, NorthCascade Mountains to the south, and the Salish Sea (comprising the Strait of Georgia andthe Strait of Juan de Fuca) to the west. Though the LFV is relatively flat throughout, themountain ranges rise steeply from the valley floor, to upwards of 2000 m asl. The LFV iswidest near the coast, stretching 100 km across in the north-south direction, and narrowsdramatically to the east, to only a few kilometers wide near Hope. Numerous tributary valleys142.1. The Lower Fraser Valley-124° -123° -122° -121°+48.85°+49.35°+49.85°CANADAStrait of Georgia      USAStrait of Juan de FucaNorth Cascade MountainsNorth Shore MountainsStave LakePitt LakeHarrisonLakeHoweSoundIndian ArmVancouverIslandSurreyAbbotsfordVancouverYVRBellinghamSquamishHopeChilliwack06001200180024003000Elevation (m)Figure 2.1. Map of the Lower Fraser Valley (LFV). The roughly-triangular main valley floorof the LFV is delineated by dashed lines. Cities are marked as circles. YVR (black triangle) isthe Vancouver International Airport meteorological station. Other major geographic featuresare labelled. All heights are in m asl.intersect the main valley floor, most notably the Indian Arm, Pitt Lake, Harrison Lake, andStave Lake valleys along its northern edge.The LFV airshed encompasses Metro Vancouver (MV) and the Fraser Valley RegionalDistrict (FVRD) (both in BC), and Whatcom County (Washington). The majority of theLFV’s population of over 2 million resides in the metropolitan city of Vancouver. Other largecities and geographic features are shown in Figure Description of ozone episode circulation regimesAinslie and Steyn (2007) used a two-stage hierarchical clustering method on 134 days (from1984 – 2003) of wind data from the Vancouver (YVR) and Abbotsford (YXX) meteorological152.1. The Lower Fraser Valleystations to identify the mesoscale circulation regimes of ozone episodes. The selected days wereozone “exceedance days”, where high ozone concentrations were caused by local photochemicalproduction (identified by < 2.5 mm rainfall at YVR, and maximum daily temperature greaterthan 24.7◦C at YXX). The analysis yielded four circulation regimes (clusters), which arebroadly classified into two groups: those with northerly daytime winds and stronger pressuregradient force (PGF) toward the coast (Type I and III), and those with southerly daytimewinds and weaker PGF toward the coast (Type II and IV)1. The composite daily hodographsand MSLP maps (derived from reanalysis data) for each cluster are shown in Figures 2.2and 2.3, respectively. For the first group, the strong PGF causes morning winds to originatefrom the Strait of Georgia, and Type III exhibits stronger wind speeds and more variablemorning wind directions than Type I. For the second group, weak PGF causes morning windsto originate from the Strait of Juan de Fuca, and Type IV has more pronounced nighttimeeasterly wind components than Type II. The hodograph of Type II is most similar to a “pure”SB, whereas Type IV closely resembles a “stratus surge” and signifies the end of an ozoneepisode.All of the regimes are characterised by a thermal trough extending northward into theLFV, which mirrors the results of McKendry’s (1994) synoptic typing analysis. The SBfilter (Steyn and Faulkner 1986; Appendix B.3.1) was applied to all ozone exceedance days,revealing that 5.7%, 65.7%, 7.5%, 33.3% of days that were Types I through IV, respectively,had well-developed SBs. Thus, not all exceedance days coincide with “pure” SB circulations.As aforementioned, an analysis of the ozone distributions across the LFV’s monitoringnetwork during these exceedance days revealed that, regardless of the circulation regime, thecomposite spatial ozone distributions are similar, suggesting the influence of build up andrecirculation processes.It is important to note that the circulation regime hodographs at YVR (Figure 2.2) do notall indicate the diurnally-reversing wind regime characteristic of recirculation (Section 1.3.1).This can be explained by: (1) the reversing wind direction of individual episodes being lost to1Note that the cluster numbering system of Steyn et al. (2011) (or Steyn et al. 2013 and Ainslie et al. 2013)is adopted here, which differs from that of the original work of Ainslie and Steyn (2007).162.1. The Lower Fraser ValleyNNEESESSWWNW246 m/s000612 1824Type I NESW246 m/s000612 18 24Type IINESW246 m/s0006121824Type III NESW246 m/s0006121824Type IVFigure 2.2. Composite daily hodographs at YVR for circulation regimes I – IV. Hourlypoints are marked (crosshairs), and times are labelled in UTC every 6 hours. The cross atthe origin gives the average standard deviation of the u and v wind components. Adaptedfrom Steyn et al. (2011) following Ainslie and Steyn (2007).172.1. The Lower Fraser ValleyFigure 2.3. Composite synoptic MSLP (hPa) maps for circulation regimes I – IV. The grayshaded area encompasses the LFV. Adapted from Ainslie and Steyn (2007).182.2. Meteorological modellingaveraging in the cluster’s composite hodograph, or (2) the build-up/recirculation of pollutantsbeing the result of some mechanism that is not observed at the station.2.1.3 Selected ozone episodes in this studyAs previously stated, this study focuses on seven summertime ozone episodes that haveoccurred in the LFV over a period of 20 years (1985 – 2006). All episodes spanned at least fivedays, though only three days were investigated from each. Steyn et al. (2011) use these episodesin their retrospective analysis of ozone formation, which attempts to explain the effect of thereduction in precursor emissions in the LFV and the resulting ozone concentration fields duringthe episodes. These seven episodes are a carefully selected subset of all past episodes over the20 year period, and were initially chosen because: (1) they recur in roughly uniformly-spacedintervals over the 20 year time period, (2) they were previously studied, modelled, or hadancillary observational data against which modelled output could be compared, and (3) theyrepresent all four mesoscale circulation regimes of ozone episodes, described in the previoussection.In this study, it is the latter point that supports the current research framework. And, itis the small number of circulation regimes that allows us to restrict this study to only a fewepisodes in order to investigate the possible variations in transport and recirculation pathwaysover the known set of mesoscale conditions (which addresses Questions 4 and 5), rather thanconduct full multi-year model runs (as suggested by Steyn et al. 2008). Finally, it should benoted that since these episodes were already modelled and evaluated by Steyn et al. (2011),the model output is readily available.2.2 Meteorological modelling2.2.1 WRF model description and set upThe meteorology during the selected ozone episodes was modelled by Steyn et al. (2011) usingthe Weather Research and Forecasting (WRF) model version 3.1 (Skamarock et al. 2008).192.2. Meteorological modellingThe Advanced Research WRF (WRF-ARW) model is a public domain, limited-area numericalweather prediction (NWP) model. The model simulates 3-dimensional meteorological fieldsover time by solving the fully compressible, non-hydrostatic flux-form Euler equations.For each episode, WRF was configured with three two-way nested computational domainscentered over the LFV (Figure 2.4) at 36 km, 12 km, and 4 km horizontal grid spacing(Table B.1), and with 48 vertical sigma levels (terrain-following mass vertical coordinates).Fields were output at 1 h intervals. Each run was initialized at 1800 UTC using NorthAmerican Regional Re-analysis (NARR) fields at 32 km resolution, and lasted 96 h (allowingfor model-spin up). Only three full days (72 h) from each episode were retained for furtheranalysis. Nudging was applied to the outer two domains, but not to the inner 4 km domain.Further model configuration details are described in Steyn et al. (2011) and Steyn et al. (2013).+20°+40°+60° +60°+80° +80°-140° -120° -100°CANADAPacific OceanUSAFigure 2.4. WRF model domain set up for all modelled episodes, with horizontal gridspacings of 36 km (outer thick black square), 12 km (gray rectangle), and 4 km (thin blackrectangle).202.2. Meteorological modellingThe WRF model uses an Arakawa-C grid staggering to horizontally discretizemeteorological variables. For all subsequent analyses, wind velocities were post-processedusing the ARWpost program to obtain their values at the mass point (center) of the grid cells.This is also the location of all thermodynamic variables.2.2.2 Overview of episode meteorologyThree full consecutive days (72 h) from each modelled episode were selected and classified intoone of the four mesoscale circulation regimes (Table 2.1). Note that some episodes exhibit asingle circulation regime over their entire duration (the 1987, 1993, 1995, and 2001 episodes),whereas others do not (the 1985, 1998, and 2006 episodes). The circulation regime of eachday of the episodes should give the reader a general idea of the wind patterns at YVR, so theywill not be elaborated upon here. An example of the daily hodographs at YVR for the 2006episode are shown in Figure 2.5; the strong westerly and northerly wind components at YVRduring the three days of this episode gave rise to a I-I-III circulation regime pattern.Table 2.1. Selected modelled episode days and their associated circulation regimes, asdefined by Ainslie and Steyn (2007).Year Episode dates Circulation regime1985 July 19-20-21 I-IV-IV1987 August 25-26-27 IV-IV-IV1993 August 2-3-4 I-I-I1995 July 17-18-19 III-III-III1998 July 25-26-27 II-III-II2001 August 10-11-12 II-II-II2006 June 24-25-26 I-I-IIIA detailed description of the meteorological conditions of the modelled ozone episodesis provided in Steyn et al. (2011). All episodes displayed characteristic synoptic slackgradient high pressure conditions, with a lower-level trough and upper-level ridge, except the1993 episode. This episode had uncharacteristic meteorological conditions, with upper-levelshortwave ridges and troughs propagating over the region, leading to unstable convective212.2. Meteorological modellingNNEESESSWWNW246810 m/sDay 1NNEESESSWWNW246810 m/sDay 2NNEESESSWWNW246810 m/sDay 3Figure 2.5. Daily hodographs at YVR for the 2006 episode. Modelled winds are black, andobservations are gray. Hourly values are represented by crosshairs. The circle and squaredenote 0000 and 2300 local time, respectively.conditions. Thunderstorms were observed along the northern edge of the LFV on August 4,1993 (Banta et al. 1997, McKendry et al. 1998a).An analysis of the full wind fields and occurrence of thermo-topographic circulationsis provided in Appendix B.3. None of the modelled episode days were SB days, thoughmountain-valley breezes and slope flows were prevalent on all episode days.2.2.3 Model evaluationA model evaluation is a critical and necessary component of any modelling study. Here, theprimary goal was to determine if the WRF output would be fit (1) for use by a trajectory modeland (2) for exploring recirculation in the LFV during the meteorological conditions of ozoneepisodes. Hence, we focused our attention on the accuracy of the modelled u and v wind fieldsof the 4 km domain. The wind fields were interpolated to the meteorological station locationsin the LFV (Table A.1), and then compared to their corresponding observational time series,obtained from Environment Canada (http://climate.weather.gc.ca/). Details of the modelevaluation are provided in Appendix B.2. Overall, statistical scores were mainly satisfactory,indicating that the 4 km domain modelled fields are suitable for exploring pollutant (tracer)transport and recirculation in the LFV.222.3. Chemical emissions field2.3 Chemical emissions field2.3.1 OverviewThe location of precursor emission sources can help inform where to initiate trajectorieswithin the LFV (Section 2.4.3). Steyn et al. (2011) used the Sparse Matrix Operator KernelEmissions (SMOKE) modelling system version 2.5 (Houyoux and Vukovich 1999) to createadjusted emissions inventories for each of the seven ozone episodes based on existing emissionsinventories provided by MV (see Steyn et al. 2011 for details). The SMOKE modellingsystem converts emissions inventories into gridded fields at the temporal and spatial resolutionrequired by atmospheric chemistry models. The SMOKE inventories include a multitude ofchemical species from biogenic, mobile, area source, and point source emissions.2.3.2 Emissions inventoryThe SMOKE model-adjusted emissions inventory of total ozone precursor emissions (NOx +VOCs 2) for the 1985 episode is mapped in Figure 2.6. This represents the emissions at thebeginning of the study period spanned by the seven episodes. This field was generated byaveraging all time steps of the episode’s emission fields at 4 km horizontal grid spacing, thensumming over the lowest four model layers (on average, up to 173.23 m agl ± 1.74 m agl).Only the Tuesday-Wednesday-Thursday emissions were used in order to avoid any hebdomadalvariations (the “weekend effect”).There are a few prominent sources of emissions located around Bellingham (BLI) (black,red, and dark orange grid squares) that can be attributed to local oil refineries and aluminumsmelters (i.e. point sources of emissions). The relatively low emissions along the coastrepresented by pale orange colours, extending from south of BLI up to YVR, arise from trafficsources along a major highway route (the Interstate-5 corridor). Otherwise, the location ofthe precursor emission sources is intimately tied to urban areas located throughout the LFV,2Emissions inventories for the following VOC species were available from SMOKE output: acetylaldehyde,propionaldehyde and higher aldehydes, benzene, ethene, ethane, ethanol, formaldehyde, isoprene, olefine,parraffin, terpene, toluen, xylene, and various other monoalkyl and polyalkyl aromatics.232.3. Chemical emissions fieldFigure 2.6. SMOKE-adjusted 4 km resolution inventory for average hourly ozone precursor(NOx + VOCs) emissions during the 1985 episode. Only Tuesday-Wednesday-Thursdayvalues have been included. Emissions < 0.5 metric t/h are omitted. For reference, selectedmajor cities are denoted by black triangles: Bellingham (BLI), Vancouver (YVR), Surrey(T15), Abbotsford (YXX), Chilliwack (T12), and Hope (YHE). The open white triangleindicates the location of maximum non-point source emissions.especially near Surrey (T15), Abbotsford (YXX), Chilliwack (T12), and Hope (YHE) (TableA.2). However, note that the highest emissions near Vancouver (surrounding the white trianglein Figure 2.6), located just southeast of the downtown urban center, is not related to an areaor point source of emissions. Rather, this maximum arises from mobile sources in an area ofhigh daily traffic congestion.Though there has been a noted shift in the location of precursor emission in the LFV dueto shifting population density (Pryor and Steyn 1995, Joe et al. 1996, Ainslie and Steyn 2007,Steyn et al. 2013), the SMOKE model-adjusted emissions inventory for the 2006 episode, atthe end of the study period, resulted in a nearly-identical location of maximum precursor242.4. Trajectory modellingemissions (not shown). For this reason, the 1985 emissions inventory was used to examine allother episodes, which subsequently simplifies the trajectory analysis that follows (see Section2.4.3).2.4 Trajectory modelling2.4.1 BackgroundTrajectories have been used for numerous applications in the atmospheric sciences, includingthe study of recirculation (Section 1.3.3). Trajectories are paths that moving objectsfollow through space over time. In the case of atmospheric trajectories, these objects areinfinitesimally small, imaginary parcels of air that are advected by 3-dimensional wind fields. Itis important to note that a trajectory is purely kinematic, and trajectory models only consideradvection of an air parcel by the wind fields alone, ignoring the effects of turbulence anddeposition. Hence, one can think of a trajectory as an atmospheric tracer that approximatesthe centerline of travel of a dispersive air mass.2.4.2 HYSPLIT model descriptionIn this study, trajectory modelling was performed using the HYbrid Single Particle LagrangianIntegrated Trajectory (HYSPLIT) model version 4 (Draxler and Hess 1997, Draxler et al.1999). HYSPLIT is distributed by the National Ocean and Atmospheric Administration(NOAA) Air Resources Laboratory (ARL). As its name indicates, HYSPLIT uses asemi-Lagrangian framework and Improved Euler-Cauchy integration method to compute atrajectory (Draxler and Hess 1997). HYSPLIT is run offline, where wind velocity fields aresupplied by meteorological model output. Hence, the spatial and temporal ‘accuracy’ of theHYSPLIT trajectory output depends on that of the input modelled meteorological fields. Fora simple model run, a single trajectory is computed both horizontally and vertically in space,given an initial position and height within the model domain. When a trajectory exits themodel domain or the model top, the computation is terminated. A trajectory that ‘intersects’252.4. Trajectory modellingthe (model) ground is advected along the surface.HYSPLIT does not explicitly account for boundary layer (BL) processes to compute atrajectory. In particular, if a trajectory is initiated within the BL, the trajectory may not berepresentative of the “true” path of an air parcel due to strong turbulent mixing (Stohl 1998).These effects can be included by using the HYSPLIT particle dispersion model instead. Inthis study, however, the time scale of the vertical mixing is much shorter than the transporttime of the air parcel; that is, the scale of the trajectories is representative of the scale of the(mesoscale) circulations. Hence, a trajectory model is an appropriate tool given the mesoscalecirculation regimes to be investigated.2.4.3 Model set upFor each ozone episode, the meteorological fields modelled in WRF at 4 km horizontal gridspacing and 1 h intervals served as input to the HYSPLIT trajectory model. For each modelrun, the same model configuration parameters were used, namely: starting location, releaseheight above ground, start time, total run time, and vertical motion (advection) calculation.Though a trajectory only acts as a passive tracer for atmospheric motion, considerable thoughtmust be given to the configuration parameters in order to ‘simulate’ the characteristics of anozone episode and provide guidance to the research questions. The selected configurationparameters are summarized in Table 2.2 and explained in the subsections that follow.Table 2.2. HYSPLIT model configuration. The start date of each episode corresponds tothe dates in Table 2.1. The trajectory frequency is determined by the temporal resolutionof the input meteorological fields from WRF.Parameter ConfigurationStart longitude λ0 −123.0753Start latitude φ0 + 49.2645Start altitude z0 (m agl) 10Start hour PDT (UTC) 00 (07)Total run time h 72Trajectory frequency trajectories / time step 1 / 1 h262.4. Trajectory modellingRationaleMany previous studies used back trajectories initiated from AQ monitoring station locationsto examine recirculation during periods of degraded air quality. Back trajectories are useful tostudy transport and determine potential source areas of pollutants. In the LFV, summertimeelevated ozone levels are generally attributed to local emissions and local photochemicalproduction (Vingarzan and Taylor 2003). Hence, in this study, trajectories were initiatedfrom within the LFV and run forward in time from the site of maximum local emissions(discussed below). And, since the focus of this research is only to identify potential pollutanttransport pathways and recirculation in the LFV (as opposed to using a trajectory modelto explain pollutant concentrations at particular locations over the course of selected ozoneepisodes), forward trajectories are appropriate.The modelled trajectories were fully 3-dimensional. This mode of trajectory computationis the most useful and representative of the “true” transport of an air mass (Stohl et al.1995, Stohl et al. 2001). Since mesoscale wind regimes during ozone episodes are oftendiurnally-reversing, the resulting path and behaviour of a trajectory is sensitive to its starttime. In accordance with previous research (e.g. Harris and Kotamarthi 2005), trajectorieswere initiated quasi-continuously for the duration of the simulation.Selection of a starting locationThe SMOKE-based emissions inventory for the 1985 ozone episode was used to select thestarting location (also referred to as the initial or departure point) of the trajectories for allepisodes. Because the spatial density of emissions and location of maximum emissions didnot significantly change from the 1985 (Figure 2.6) to the 2006 (not shown) episode, the samestarting location was used for all modelled episodes.The starting location for trajectories was chosen to coincide with the point of “averagemaximum” precursor emissions. Using the spatial distribution of NOx and VOCs, the averagecoordinates of the 10 grid cells of highest emissions within the LFV were used. Point sourceswere excluded from this ranking in order to better represent areas of mobile source emissions.272.5. SummaryNote that the location of maximum NOx and maximum VOC emissions is nearly identical(not shown). All trajectories were initiated near the ground to emulate mobile emissions, i.e.the release of ozone precursors. The starting location, X0, was then:X0 = (λ0, φ0, z0) = (−123.0753,+49.2645, 10)where λ is longitude (in decimal degrees, East is positive), φ is latitude (in decimal degrees,North is positive), and z is height (m agl).The LFV’s daily wind patterns are spatially inhomogeneous, especially owing to theproximity of steep terrain and developing slope flows. Spatial variations of wind patternswill cause different spatial distributions of trajectories from different initial locations over thecourse of the simulations. Since the selected single starting location was interpolated fromother data sources, the sensitivity of the resulting trajectory pathways to the choice of X0 willbe addressed in Section 3.2.ConfigurationThe total model run time for all episodes was 72 h (i.e. the length of the episode), starting onthe first selected day of each episode at 0000 local time (PDT). A new trajectory was initiatedevery 1 h (i.e. the resolution of the output WRF fields), totaling 72 trajectories of differingmaximum possible duration (age). The method of vertical motion calculation was set to thedefault used by the meteorological model. This means that vertical advection was calculatedusing omega (ω = dPds , in Pa s−1), rather than using the WRF vertical velocity fields (w, inm s−1) directly.2.5 SummaryThe seven ozone episodes used in this study cover all four of mesoscale circulation regimes,and thus the known set of meteorological conditions conducive to ozone episodes. Thisprovides a simplified framework for examining transport and recirculation that may contribute282.5. Summaryto degraded AQ conditions in the LFV. The meteorology of the seven episodes werepreviously modelled by Steyn et al. (2011) using WRF. All episodes had high pressuresynoptic conditions, with prevalent onshore winds, and mountain-valley breeze and slope flowcirculations throughout the LFV. A model evaluation of the surface wind fields indicated thatthe simulated meteorological fields were suitable for use in subsequent trajectory modellingin HYSPLIT. Of critical importance was the selection of a starting location for the simulatedtrajectories. To take into account the LFV’s chemistry and the aims of the research questions,a SMOKE-adjusted emission field of ozone precursor species was used to select a startinglocation that coincided with the location of maximum ozone precursor emissions arising fromprimarily mobile sources. The trajectory model output will be analyzed in Chapters 3 and 4.293 Trajectory analysisTrajectories modelled by HYSPLIT during the seven ozone episodes are investigated. First,the sensitivity of the resulting spatial distributions of trajectories in response to choice of initialstarting location is examined. The primary focus of the trajectory analysis is on compositetrajectory fields at the end of each day (24 h) and each episode (72 h). The composite fieldsare used to identify preferred transport pathways of trajectories and the degree of spatialsimilarity between trajectory fields belonging to the same circulation regime.3.1 Overview of HYSPLIT modelling resultsTrajectories modelled by HYSPLIT during each of the seven ozone episodes are shown in Figure3.1. These composite trajectory plots are end ‘snapshots’ of all 72 trajectories modelled duringeach episode (for full animations, see videos V1 – V7). As such, they reveal the total spatialdistribution of the trajectories over the course of the entire modelled period. Trajectoriesbeginning at 0000 and 1200 local time are highlighted in blue and red, respectively, and areonly meant to serve as visual guides.There are two important factors to consider when interpreting the mapped trajectories:• The total modelled path of the trajectories extends much further than the bounds of themap used. However, since we are concerned with the effects of transport and recirculationwithin the LFV, the mapped area restricts our focus to regional transport.• It is misleading to display trajectories on maps that have higher topographic resolutionthan that of the native meteorological model. Due to grid cell discretization, the303.1. Overview of HYSPLIT modelling resultstopography in WRF is smoothed – valley bottoms are filled in, and mountain peaksare flattened (Figure B.1a). To emulate this smoothing, in all subsequent figures thetopography is contoured in 600 m intervals, starting from sea level (Figure 2.1 andB.1b).First, it is evident that the paths of trajectories are guided by the topography of the LFV.Second, a qualitative visual assessment of the composite trajectory fields in Figure 3.1 revealsperceivable differences and similarities between the spatial distribution of trajectories duringdifferent episodes. For example, trajectories during the 1985 (Figure 3.1a) and 1987 (Figure3.1b) episodes travelled almost exclusively northward from the origin. In contrast, trajectoriesduring the 1998 (Figure 3.1e) and 2001 (Figure 3.1f) episodes exhibited much more zonal(east-west) movement and transport up Howe Sound.Recall that trajectories were modelled using HYSPLIT’s 3-dimensional mode – that is,the trajectories as plotted in Figure 3.1 travelled both horizontally (north-south, east-west)and vertically in space over time. Though this study focusses on horizontal transport andrecirculation, the heights of the trajectories are important when considering ambient chemistryat the surface. An example of the horizontal and vertical paths of trajectories for the1985 episode is provided below (Figure 3.2). Trajectories that began at nighttime generallyremained near the surface until sunrise, often below 500 m agl (cf. Figures C.1 – C.6). Incontrast, trajectories that were initiated during or continued into the daytime rapidly reachedheights up to ∼1000 m agl. This behaviour reflects the evolution of the BL, from shallownighttime stable layer to deep daytime ML. Furthermore, trajectories generally followed theterrain of topographic features (not shown).A further discussion of composite trajectory fields and general horizontal transportpathways follows in Section 3.3.313.1.OverviewofHYSPLITmodellingresults(a) 1985 I - IV - IV (b) 1987 IV - IV - IV (c) 1993 I - I - I(d) 1995 III - III - III (e) 1998 II - III - II (f) 2001 II - II - II-124° -123° -122°+49°+49.5°(g) 2006 I - I - IIIFigure 3.1. Modelled trajectories (black lines) during the selected ozone episodes. All trajectories are initiated from the origin(black triangle). Trajectories initiated at 0000 and 1200 local time are highlighted in blue and red, respectively. The circulationregimes of each day of the episode are labelled above each plot.323.1. Overview of HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure 3.2. Modelled trajectories during the 1985 episode. Trajectories starting at 0000and 12000 local time are highlighted in blue and red, respectively. Trajectory positions aremarked every 6 h (black points). (top) The origin of the trajectories is denoted by the blacktriangle. (bottom) The heights of the trajectories within the mapped domain are plotted assolid lines. The heights of the trajectories extending past the plotted domain are extendedwith dashed lines.333.2. Trajectory sensitivity analysis3.2 Trajectory sensitivity analysisBefore continuing, the trajectory ‘error’ that arises from the choice of X0 must be examined.All trajectory computations are susceptible to errors. These errors arise from several sources:inaccuracies in the input modelled wind fields and estimations of vertical velocity (w),truncation errors (from the order of terms retained in the numerical integration scheme),and interpolation errors (from estimating wind velocities between grid points and betweenmodel time steps) (see Stohl 1998 for further details). Furthermore, as Stohl (1998) notes, ina physical sense, air parcels in the real atmosphere are not infinitesimally small; they occupyfinite volumes. In divergent flows, they may be distorted over time. Hence, trajectories becomeless representative of an air parcel’s “true” path as time progresses. Errors in trajectory positionof up to 20% of the total travelled distance are common (Stohl 1998).All types of errors discussed above are unavoidable, and most cannot be easily quantified.However, one source of error that can be quantified – and of most importance in this study – isthe uncertainty of trajectory paths associated with the choice of trajectory starting position,X0. This is especially important owing to the fact that X0 was inferred and interpolated fromother data sources (Section 2.4.3). Small changes to X0 in an inhomogeneous wind field couldpotentially lead to an amplification of position deviations over time, and effectively changethe overall interpretation of modelled trajectory fields.However, to some degree, the choice of X0 must be constrained to ‘simulate’ the release ofozone precursors. VaryingX0 horizontally by large distances would no longer be representativeof the maximum emission source location. Thus, X0 was varied horizontally by one grid spacing(4 km) to the north, east, south, and west. Hence, trajectories modelled using the originalchoice of X0 were used as a control against which test trajectories initiated 4 km from X0were compared.Differences in modelled trajectories were assessed both qualitatively, by visually examiningthe spatial distribution of composite trajectory fields, and quantitatively, using a statisticalapproach. Rather than examine the sensitivity of trajectories to X0 during all seven343.2. Trajectory sensitivity analysismodelled episodes, two episodes were selected for analysis: the 1993 and 1998 episodes.The uncharacteristic meteorology of the 1993 episode and the poor meteorological modellingperformance of the 1998 episode should resolve an upper limit on the spatial deviations oftrajectories in response to the choice of X0.3.2.1 Qualitative spatial assessmentControl and test sets of trajectories for the 1993 and 1998 episodes are plotted in Figures3.3 and 3.4, respectively. For the 1993 episode, the most noticeable spatial deviations arosewhen the starting location was moved 4 km southward: trajectories did not travel northward,and were more promptly pushed to the south out of the valley (Figure 3.3c). In contrast, thetrajectories during the 1998 episode appear most sensitive to shifting the starting location 4km northward (Figure 3.4b): a larger proportion of trajectories were swept into the tributaryvalleys of the North Shore Mountains.3.2.2 Transport deviationsStohl et al. (1995) present a set of equations after Rolph and Draxler (1990) to quantifyspatial deviations between modelled trajectories. The Absolute Horizontal and AbsoluteVertical Transport Deviations (AHTD and AVTD, respectively) are treated separately. Foran ensemble of N trajectories,AHTD(t) =1NN∑n=1{[Λn(t)− λn(t)]2 + [Φn(t)− φn(t)]2}12AVTD(t) =1NN∑n=1|Zn(t)− zn(t)|where AVTD has been adapted here for height (rather than pressure), the control trajectoriesare a function of (Λ, Φ, Z), and test trajectories are a function of (λ, φ, z), at time t.353.2. Trajectory sensitivity analysis(a) E (b) N-124° -123° -122° -121°+49°+49.5°(c) S (d) WFigure 3.3. Control (black) and test trajectories (red) modelled during the 1993 episode.Control trajectories are initiated from the large black triangle (X0), and test trajectories areinitiated from the small black triangle, 4 km away from X0 to the (a) East, (b) North, (c)South, and (d) West.AHTDThere are two ways to examine the AHTD: by hour elapsed from the start of the episode,i.e. by start time of each trajectory in local time (Figures 3.5a and 3.6a), or by the age ofthe trajectory (Figures 3.5b and 3.6b). In either case, as expected, AHTD increased overtime. For the 1993 episode, there were dramatic horizontal deviations for trajectories thatstarted or continued between the early morning hours and noon on each day of the episode.This can be attributed to the increase in morning wind speeds and the timing of sudden winddirection shifts in different areas of the LFV. This effect was less apparent for the 1998 episode363.2. Trajectory sensitivity analysis(a) E (b) N-124° -123° -122° -121°+49°+49.5°(c) S (d) WFigure 3.4. As in Figure 3.3, but for trajectories modelled during the 1998 episode.trajectories.When examined by trajectory age, the AHTD after 24 h was ∼125 km and ∼75 km forthe 1993 and 1998 episodes, respectively. At times, the hourly standard deviation of AHTD(error bars in Figures 3.5b and 3.6b) was quite large, up to 57.50 km. When compared tothe AHTD results by elapsed hour as described above, the large standard deviations are notsurprising because trajectory positions of a given age are associated with trajectory positionsat different times.373.2. Trajectory sensitivity analysis0 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0100200300AHTD (km)(a)0 6 12 18 24 30 36 42 48 54 60 66 72Age (h)0100200300AHTD (km)(b)Figure 3.5. The Absolute Horizontal Transport Deviation (AHTD) for test trajectoriesduring the 1993 episode, (a) by start time of each trajectory and (b) by trajectory age. Thethick black line is the mean AHTD by hour. Error bars indicate one standard deviation.383.2. Trajectory sensitivity analysis0 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0100200300AHTD (km)(a)0 6 12 18 24 30 36 42 48 54 60 66 72Age (h)0100200300AHTD (km)(b)Figure 3.6. As in Figure 3.5, but for test trajectories during the 1998 episode.393.2. Trajectory sensitivity analysisAVTDSimilar to the trends in AHTD, the largest vertical deviations occurred for trajectories thatstarted or continued during morning hours (Figures 3.7 and 3.8). This corresponds to thedevelopment of the ML, when vertical transport rapidly increases via turbulent mixing. Therewere some instances when the AVTD reached 800 m, though the mean AVTD was ∼ 200 m±110 m over both episodes.3.2.3 InterpretationThe resulting spatial distribution of trajectories were most susceptible to displacements northand south of X0. This is not surprising, given the proximity of the North Shore Mountains andthe prevalence of small-scale variations in mountain-valley breezes and slope flows (AppendixB.3.2). As found in other studies, both the AHTD and AVTD increased nearly linearly withtime (e.g. Rolph and Draxler 1990). Morning test trajectory positions deviated the most fromcorresponding control trajectory positions, both horizontally and vertically. This may haveimportant implications on the transport of ozone precursor pollutants since a large portionare released during the morning rush-hour period.However, it is important to consider that, though AHTD and AVTD may be large (reachingseveral hundred kilometers and several hundred meters, respectively), these deviationstypically occur when trajectories are far from X0 and out of the LFV entirely. These far-fieldimpacts somewhat confound the sensitivity analysis.Overall, though there are perceivable differences in spatial distributions of trajectoriesgiven different starting locations, these differences are relatively small. That is, the choice ofX0 does not change the overall “picture” of transport during these modelled ozone episodes.403.2. Trajectory sensitivity analysis0 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0200400600800AVTD (m)(a)0 6 12 18 24 30 36 42 48 54 60 66 72Age (h)0200400600800AVTD (m)(b)Figure 3.7. As in Figure 3.5, but for the Absolute Vertical Transport Deviation (AVTD)during the 1993 episode.413.2. Trajectory sensitivity analysis0 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0200400600800AVTD (m)(a)0 6 12 18 24 30 36 42 48 54 60 66 72Age (h)0200400600800AVTD (m)(b)Figure 3.8. As in Figure 3.7, but for test trajectories during the 1998 episode.423.3. Analysis of composite trajectory fields3.3 Analysis of composite trajectory fields3.3.1 Spatial patterns of trajectories in relation to circulation regimesBecause X0 is fairly close to YVR (only ∼11.25 km away, fewer than 3 grid squares in theinner 4 km WRF domain), circulation regime hodographs at YVR (Figure 2.2) can be usedas guidance for interpreting the composite trajectory fields. To simplify the discussion thatfollows, the convention “YYYY-n” to refer to a day of an episode will be adopted, e.g. thefirst day of the 1985 episode is 1985-1.In Section 3.1, it was briefly remarked that there is visual similarity between the endsnapshot of trajectories from the 1985 and 1987 episodes (Figure 3.1a and 3.1b). Similaritiesbetween trajectory distributions of these two episodes can be largely explained by theircirculation regimes. Two days of the 1985 episode and all three days of the 1987 episodewere of Type IV circulation. This regime is dominated by winds with southerly component,with strong south-easterly veering to south-westerly daytime winds (Figure 2.2d). Hence,trajectories travelled predominantly northward, many up the west side of the North ShoreMountains within Howe Sound. Only four trajectories on 1985-1 (visible in Figure 3.1a)travelled southward, which can be attributed to that day’s Type I circulation regime, withpredominantly north-westerly to westerly winds.Trajectories during the 1998 (Figure 3.1e) and 2001 (Figure 3.1f) episodes also frequentlytravelled up Howe Sound, but overall exhibited much more zonal (east-west) movement withinthe main valley floor of the LFV. Again, we can relate these spatial distributions to thecirculation regimes that occurred. Both episodes mainly exhibited a Type II circulation, whichhas a prominent easterly (‘offshore’) to westerly (‘onshore’) diurnal wind shift. This confinedthe majority of the trajectory movement to the northern edge of the LFV, where trajectorieswere then guided by adjacent tributary valleys. The majority of southward trajectories duringthe 1998 episode occurred during 1998-2, coincident with a Type III circulation.All three days of the 1995 episode were a Type III circulation. The trajectories of thisepisode (Figure 3.1d) spread furthest east within the LFV, and relatively few trajectories433.3. Analysis of composite trajectory fieldstravelled up Howe Sound. This comes as no surprise since Type III circulations exhibit strongwesterly daytime winds at YVR.The spatial distribution of trajectories over the course of the 2006 episode (Figure 3.1g)were similar to those of the 1998 and 2001 episodes, but with a higher number of southwardtrajectories. Though the 2006 episode had both Type I and III circulations, both of theseregimes have northerly component wind directions, which accounts for the high frequencyof southward trajectories. These northerly winds, originating from the Strait of Georgia,effectively channel trajectories between Vancouver Island and the LFV mainland. Recallthat a similar pattern of southward trajectories were observed on 1985-1, a day with Type Icirculation as well.Finally, the spatial distribution of the trajectories during the 1993 episode is remarkablydifferent compared to those of all other episodes. Though all three days of the episode were aType I circulation, the composite trajectory field (Figure 3.1c) scarcely resembles those of otherType I circulation days (1985-1, 2006-1, and 2006-2). In other words, compared to the analysesabove, the spatial distribution of trajectories during the 1993 episode does not resemble theexpected distribution given its circulation regime at YVR. Again, this is likely due to theuncharacteristic meteorological conditions that occurred. As aforementioned, thunderstormswere observed along the northern edge of the LFV. The presence of outflow winds from thisstorm system, and throughout the episode, prevent the trajectories from traveling northwardinto tributary valleys.3.3.2 Number density mapsAn alternative way to examine the composite trajectory fields is to generate a number densitymap, where the number of trajectories crossing through a two-dimensional area is counted.The purpose of generating number density maps was (1) to highlight areas where trajectoriesfrequently travelled (which is difficult to visually decipher based on the composite fields as inFigure 3.1), and (2) to obtain a statistical measure of similarity between composite trajectoryfields belonging to different circulation regimes. To generate these maps, the trajectories443.3. Analysis of composite trajectory fieldswere counted as they passed through each of the 4 km × 4 km (16 km2) grid boxes ofthe innermost WRF domain that was used for trajectory modelling in HYSPLIT. Only theportions of trajectories within the mapped area of Figure 3.1 were considered, since it is notimportant to compare trajectory positions once they have diverged in space far from the origin(Section 3.2.3). The resulting number densities include trajectories at all heights. Given thattrajectories generally remained below 1000 m agl (Figures C.1 – C.6), the maps represent a‘surface’ number density of trajectories. To address the two goals above, number density mapswere generated for entire episodes (i.e. for composite trajectory fields), and for individual days(i.e. by circulation regime).By episode (composite trajectory fields)Number density maps of end snapshots for each episode are plotted in Figure 3.9. Note thatthe colour scale is logarithmic to enhance lower trajectory “frequencies”. A salient featurecommon to all episodes is a maximum trajectory frequency east of the origin, reaching 70–81trajectories counted within grid cells closest to X0. Trajectory number densities greater than72 indicate that some trajectories crossed twice or more through the same grid cell (sincethere are 72 trajectories in total). Of course, this eastward maximum is due to the persistentonshore winds (at YVR) during all episodes (Section 2.2.2).For each episode, several distinct areas of high trajectory frequency, or “preferredpathways”, are apparent. Extending the general spatial analysis from the previous section,the number density maps reveal the following features:• During the 1985 episode (Figure 3.9a), trajectories frequently travel northeastward upthe Pitt River Valley and over the North Shore Mountains. In contrast, during the 1987episode (Figure 3.9b), there is a preferred trajectory pathway to the northwest up theStrait of Georgia. Common to both of these episodes is a preferred pathway up HoweSound, along the west side of the North Shore Mountains.453.3.Analysisofcompositetrajectoryfields1 2 4 6 10 20 40 60 80× 16 km−2(a) 1985 I - IV - IV (b) 1987 IV - IV - IV (c) 1993 I - I - I(d) 1995 III - III - III (e) 1998 II - III - II (f) 2001 II - II - II-124° -123° -122° -121°+49°+49.5°(g) 2006 I - I - IIIFigure 3.9. Number density maps generated from composite trajectory fields. The colormap (log scale) indicates the numberof trajectories crossing each 4× 4 km2 (16 km2) grid square. The origin of the trajectories is represented by the black triangle.The circulation regime of each day of the episode is labelled above each plot.463.3. Analysis of composite trajectory fields• Preferred trajectory pathways during the 1998 and 2001 episodes are around the NorthShore Mountains, namely: within Indian Arm and along the Pitt River Valley, and upHowe Sound. A distinct ‘hook’ shaped (or U-shaped) preferred pathway is apparent inthis region (yellow through red hues in Figures 3.9e and 3.9f).• Strong westerly winds during the 1995 episode (Figure 3.9d) cause the majority oftrajectories to travel eastward, but trajectories are moderately frequent just offshoreof Vancouver, and within the Pitt River Valley.• During the 2006 episode (Figure 3.9g), there is distinct preferred trajectory pathwaythrough the Pitt River Valley and towards Stave Lake.• A broad area of maximum trajectory frequency for the 1993 episode (Figure 3.9c) iswithin the main valley bottom of the LFV, extending from northwest to southeast.Overall, the preferred pathways for trajectories are along Howe Sound and the tributaryvalleys of the North Shore Mountains (Indian Arm, Pitt River Valley, and Stave Lake inparticular). It is also interesting to note that trajectories did not travel frequently towardsthe eastern portion of the LFV (i.e. towards Chilliwack and Hope).By day (circulation regime)To provide a statistical measure of similarity between trajectory fields, trajectories from eachepisode were (1) separated by day in order to isolate trajectories associated with one circulationregime from another, then (2) truncated at midnight of each day, regardless of their starttime. Trajectories initiated in the morning are therefore overrepresented in these daily numberdensity maps.The resulting number density maps for the daily trajectory fields reveal differences betweendays with identical circulation regimes. For example, though all three days of the 1987 episodeare a Type IV circulation, areas of higher trajectory frequency are evident west ofX0 on 1987-1and 1987-2, but east of X0 on 1987-3 (Figure 3.10; see Figures C.7 – C.12 for other episodes).473.4. Discussion-124° -123° -122° -121°+49°+49.5°(a) Day 1 - IV (b) Day 2 - IV (c) Day 3 - IVFigure 3.10. The number density maps by day (circulation regime) of the 1987 episode.The origin of the trajectories is represented by the black triangle. The colormap scale is asin Figure 3.9. The circulation regime of each day is indicated above each plot.The Pearson product-moment correlation coefficient, r, was used to quantify thesedifferences. Each of the daily number density maps were correlated to one another. Theresulting r values compose a 21 × 21 (7 episodes with 3 days each) symmetric correlationmatrix, whose diagonal is unity (r = +1.0) (Figure 3.11). No values of r were less than zero.Generally, the highest values of r were found between days with Type I, II, and III circulationregimes. The lowest values of r were found between number density fields of 1993-2 andall other days, and between days that were a Type IV circulation and any other circulationregime. For example, the lowest association was found between 1987-2 (Type IV) and 2006-3(Type III) (r = 0.18). In contrast, the highest correlations were found between 2006-2 (TypeI) and 1995-2 (Type III) (r = 0.86), and 1995-1 (Type III) and 1995-2 (Type III) (r = 0.84).3.4 DiscussionSimilarities between spatial distributions of trajectories belonging to the same circulationregime based on hodographs at YVR substantiate that wind fields are similar as well. Thatis, the spatial distribution of trajectories are geometrically similar within circulation regimes.The prevalence of westerly wind components during the Type I, II, and III circulationsresults in a high degree of association (high r values) between trajectory fields. Of these483.4. DiscussionFigure 3.11. Pearson’s r correlation matrix for daily trajectory number density maps. Onlythe lower triangle of the matrix is shown since the entire matrix is diagonally symmetric.Values of r = +1.0 are plotted in black, and values of −1 ≤ r < 0.35 are gray. The labelsalong the x and y axes as YYYY-n-CR denote the year of the episode, day of the episode,and circulation regime of that day.circulation regimes, the Type II circulation is the only regime with easterly wind componentsas well. These pre- and post-midnight easterly winds appear to govern the frequency oftrajectories that travel up Howe Sound, giving rise to the high frequency ‘hook’ in thenumber density maps of days with this circulation regime. The lack of southward trajectoriesduring days of Type IV circulation leads to very low association (low r values) betweenType IV trajectory distributions and all other circulation regimes. This reflects the TypeIV circulation’s characteristic southeasterly winds, and lack of westerly wind components.Finally, the visual discrepancy between trajectories during the 1993 episode and all otherepisodes from the composite trajectory snapshot and number density map is reflected in the493.5. Summarylow values of r between the episode days and all others, and reflects the uncharacteristicmeteorological conditions.Taken together, composite trajectory fields and their corresponding number density mapsindicate that different circulation regimes give rise to similar preferred trajectory pathways.These “hotspots” are intimately related to the topography surrounding the LFV, and include:Howe Sound, Squamish, and within the tributary valleys of the North Shore Mountains. Itis reasonable to postulate that AQ may be most impacted at these hotspots by emissionsfrom downtown Vancouver. However, it is important not to conflate areas of high trajectoryfrequency with expected areas of high ozone concentrations. Though precursor pollutants maybe advected in a given direction and trajectories may frequently pass over certain areas, thisdoes not necessarily mean that ozone will form in these areas as well. Rather, trajectoriesinitiated from a location of high precursor emissions (i.e. from X0) reflect the advection of theprecursor emissions alone. The high number of trajectories that travel within the Pitt RiverValley agree with observations of polluted air masses advecting to and from the main valleyfloor from this tributary valley. It is also evident that precursor emissions released from X0 arelikely not significantly contributing to degraded AQ conditions in the eastern portion of theLFV during ozone episodes, namely at: Abbotsford, Chilliwack, Hope, and within HarrisonLake valley.3.5 SummaryTrajectories simulated by the HYSPLIT model over all seven ozone episodes were examined.First, the sensitivity of the resulting spatial distributions of trajectories in response to choiceof initial starting location, X0, was investigated for two episodes. Overall, the choice of X0did not affect the interpretation of transport during the modelled ozone episodes. Next, thespatial distribution of trajectories were found to be geometrically similar within circulationregimes. Number density maps generated from composite trajectory fields were used toidentify preferred transport pathways around the North Shore Mountains, and the degree ofspatial similarity between trajectories belonging to the same circulation regime. The spatial503.5. Summarydistribution of trajectories during Type IV circulation conditions were the most distinct,given the regime’s characteristic southeasterly winds, and lack of westerly wind components.Furthermore, the lack of visual and statistical association between the trajectories of the 1993episode and those with the same circulation regime (Type I) clearly highlight the unusualsynoptic and mesoscale conditions that prevailed during this episode.514 Recirculation definition and analysisIn this chapter, an objective and generally applicable definition for “recirculation” is developed.The algorithm for detecting recirculation from trajectories is then applied to the trajectoriesmodelled by HYSPLIT during each of the seven ozone episodes. In doing so, start andend positions and times for the portion of trajectories that recirculate are isolated, andspatiotemporal characteristics of recirculating trajectories are then examined within circulationregimes. Finally, results from applying the proposed definition herein are compared to Allwineand Whiteman’s (1994) Recirculation Factor.4.1 A proposed definition for “recirculation”As discussed in Section 1.3.3, a logical, quantitative definition that incorporates objectivecriteria for recirculation is required. This definition is formulated such that it takes intoaccount the qualitative description of recirculation (Section 1.3.1) and its governing physicalprocess (advection). Most importantly, an a priori definition for horizontal “recirculation”must be constructed such that it:(a) can be used under any meteorological conditions,(b) can be applied to either observational data or model output,(c) can be used for any type trajectory (forward or back trajectories, isobaric, 3-dimensional,etc),(d) is independent of study location,524.1. A proposed definition for “recirculation”(e) incorporates the spatial scales of the study region and/or research goals,(f) is not determined by or restricted to any time scales (such is the case in Allwine andWhiteman’s (1994) definition), and(g) does not depend on any particular atmospheric chemistry.In effect, a definition for “recirculation” should be a purely kinematic view of a flow field asit changes through time. It is left to the investigator to then interpret the results in relationto their particular research needs.4.1.1 General descriptionIn the simplest of terms, horizontal “recirculation” is defined in the following manner:RecirculationA trajectory is said to have externally recirculated at the point where it is aat minimum distance from its origin, if and only if: (a) the trajectory left andreturned to the horizontal domain of interest, and (b) the point is below a criticalheight. If the trajectory fails to leave the domain but does re-approach its origin,and given that (b) is also true, then the trajectory has internally recirculated.The portion(s) of the trajectory that recirculates is referred to as a “recirculatingtrajectory segment” (RTS).The basic algorithm to find recirculating trajectory segments requires only the trajectorypositions (latitude, longitude, and height) over time, and a selected domain and criticalheight. A general discussion and illustrative examples follow below. Note that a recirculatingtrajectory is necessary but not sufficient to attribute increasing pollutant trends (as observedat a single location) to recirculation during periods of degraded AQ (Section 1.3), as furtherspatiotemporal atmospheric chemistry information would be required.534.1. A proposed definition for “recirculation”Finding the closest pointFirst, the distance between each trajectory position, Xn, and the trajectory’s origin, Xi, iscalculated. Either a flat-surface (for short distances) or ellipsoidal-surface (for long distances)formula to calculate the distances from Xn to Xi can be used. A recirculating trajectorysegment (RTS) will have a point that is furthest (maximum distance) from the origin, Xn,max,and a point that is closest (minimum distance) to the origin, Xn,min, the latter of whichdefines the recirculation (Figure 4.1). To determine Xn,min, no interpolation is used along thetrajectory, i.e. only the original trajectory positions are considered. The total time elapsedbetween start and end of recirculation defines the recirculation lifetime, τR. To filter outsmall-scale recirculation, the distance between Xn,max and Xn,min should be greater thanresolution of the tool used to construct the trajectory (e.g. accuracy of sounding balloonpositions, grid spacing of the meteorological fields used for trajectory modelling). Finally, notethat by the definition above, a single trajectory may have more than one RTS: the endpointof one RTS is potentially the start of the next, regardless of how close to the origin Xn,min is,so long as it is in the horizontal domain and below the critical height. An illustration of thiscase is provided below (Figure 4.1).Selecting a domain and critical heightOnce a trajectory has re-approached its origin, the horizontal domain of interest and criticalheight define a volume that determines whether the recirculating trajectory is relevant to theinvestigation. Hence, the domain and critical height must be carefully selected to reflect thescope and purpose of research. Both may be determined either geometrically, physically, orstatistically. For example, a study that employs back trajectories to investigate daytime airpollution at an urban AQ monitoring station may select a domain that encompasses the areaof representativeness of the station, and a critical height within the ML. Furthermore,• The domain may be any geometric shape and, to some extent, may be any size. Theshape will largely depend upon the region’s geography and topography, and intent ofstudy.544.1. A proposed definition for “recirculation”Figure 4.1. Schematic diagram illustrating the definition of recirculation and itsimplementation. This definition is based on trajectories. (a) Three trajectories are initiatedfrom the origin, Xi (black triangle). Each exhibits at least one point that is furthest fromthe origin, Xn,max (crosshairs), and at least one point that is closest to the origin, Xn,min(open circles). A rectangular domain (thick dark grey line) is selected. The first trajectoryis made up of three continuous sections (thick black, thin black, and dashed black lines),where the first two of these sections (segments 1 and 2 ) internally recirculate. Over time,the displacement of the trajectory positions to its origin is plotted as in (b). The lifetime ofthese recirculations are τR1 and τR2, respectively. The remainder of the trajectory (dashedblack line) is discarded. The second trajectory (thick gray line) illustrates one externallyrecirculating segment (segment 3 ). The third trajectory (thin gray line) consists of twosections that comprise one recirculating segment since only the final Xn,min is within thedomain of interest (segment 4 ).554.1. A proposed definition for “recirculation”• The critical height may vary in space and time, but must be known over the entiredomain and study period.• Xi must be within the horizontal domain, but may be above, below, or at the (initial)critical height.Other considerations and illustrative examplesFor illustrative purposes, consider recirculation in only the horizontal direction by assumingthat all trajectory positions are below the selected critical height, as required. A schematicillustration of RTSs is presented in Figure 4.1, where the idealized cases of an internal (segment1) and external (segment 3) recirculation are drawn. A secondary internal recirculation(segment 2) is also illustrated, originating from the endpoint of the first.An additional consideration must be given to externally recirculating trajectories: if thetrajectory exits the domain and eventually returns, then this entire portion of the trajectory istaken as one recirculating segment, regardless of the path of the trajectory outside the domain(segment 4 in Figure 4.1). That is to say, there may be many local closest points to the originoutside of the domain, but the RTS is determined by the Xn,min that is inside the domain.4.1.2 Application to ozone episodes in the Lower Fraser ValleyTo apply the above definition and detection algorithm of recirculation to the current study, thedomain and critical height were configured based on known meteorological and atmosphericchemical conditions during ozone episodes in the LFV.DomainThe main goal of this research is to identify whether or not recirculation occurs in the LFVduring ozone episodes. Accordingly, a large domain was used, as opposed to focussing on areassurrounding particular sources or AQ monitoring stations. The domain was selected so thatit covers the ‘chemically active’ area of the LFV (Figure 4.2), which was roughly determinedfrom previous observational and modelling studies that show the location of maximum ozone564.1. A proposed definition for “recirculation”isopleths and the spatial extent of high ozone concentrations (see Steyn et al. 1990, Ainslieet al. 2009, Ainslie et al. 2013). The four sides of the quadrilateral domain encompass theentire main valley floor and the spatial coverage of the LFV’s AQ monitoring network (exceptstations located in Howe Sound). Furthermore, since tributary valleys are frequently subjectto high pollutant concentrations (Section 1.2), sides A and B of the domain (Figure 4.2) extendto the north end of Pitt Lake and around Stave Lake. Sides A, B, and C traverse the peaksof the surrounding mountain ranges. Side D bisects the Strait of Georgia, where SB systemsmay converge from opposite coasts and thus may limit net horizontal transport of pollutantsfrom the LFV offshore.-124° -123° -122° -121°+49°+49.5° A BCDFigure 4.2. The selected horizontal domain of interest used to identify recirculation in theLFV in this study. The domain is a quadrilateral enclosed by sides A, B, C, and D (thickgray lines). The origin of the trajectories, X0, is denoted by the black triangle.574.1. A proposed definition for “recirculation”Critical heightAs discussed in Section 1.1, ML height is an important parameter in predicting AQ conditions:reduced ML heights limit vertical dispersion and can enhance ozone concentrations at thesurface. Thus, a critical height corresponding to the afternoon maximum ML height duringthe seven ozone episodes was used.Due to the proximity of the ocean and surrounding topographic constraints, ML depthsvary strongly in time and space within the LFV (e.g. Batchvarova et al. 1999). Underanticyclonic synoptic conditions and clear skies over land, ML heights are low in the morningand in coastal areas, and high during daytime hours and inland. Observational and modellingstudies during the selected ozone episodes indicate that ML heights reach a maximum of 600– 1300 m agl in the afternoon at mid-valley locations (Table 4.1). The numerical models usedin these studies tend to overestimate ML heights (e.g. Steyn et al. 2011). At night, the depthof the stable BL can be quite shallow, reaching only ∼100 m.Table 4.1. Observed and modelled BL heights during ozone episodes selected in this study.The BL height range indicates the minimum and maximum observed/modelled heights, andincludes both nighttime and daytime values over the entire study period. The Pitt RiverValley is abbreviated by “PRV”.EpisodeBL height Method ofLocation(s) Referencerange (m agl) determination1985 350 – 1000 Acoustic sondes,tethersondesCoastal,mid-valleyHedley and Singleton (1997)1993 100 – 1300 Tethered and freeflying balloons, lidarCoastal,mid-valley, PRVHoff et al. (1997),Batchvarova et al. (1999)1998 < 620 Modelled Inland Moisseeva (2011)2001 < 1200 Lidar Mid-valley Strawbridge and Snyder(2004b)2006 < 100 – 350 Modelled PRV Steyn et al. (2011)Because of the complex spatial and temporal variations in ML height, a single criticalheight for all episodes of 800 m agl for the entire domain over the entire modelling periodwas used. This value is representative of the average maximum limit of daytime ML heightsduring ozone episodes, as well as the depth of SB circulations (Steyn 1998). The same value584.2. Spatial characteristicswas used for nighttime hours, as it should represent the extent of the nighttime residual layer.Pollutants within this layer typically mix downward as the ML begins to develop the nextmorning.Finally, it is worthwhile noting that the volume enclosed by the horizontal domain andcritical height is similar to the hypothetical “box” where pollutants are trapped in the LFV(cf. Figure 1.1).4.2 Spatial characteristics4.2.1 OverviewThe recirculation detection algorithm herein was applied to trajectories modelled byHYSPLIT, and recirculating trajectory segments (RTSs) were identified. Over the total setof trajectories of all seven episodes (7 × 72 trajectories = 504 trajectories total), 297 (∼59%)trajectories did not recirculate, and 139 (∼28%) recirculated only once (see Figure 4.3). Themaximum number of recirculations by a single trajectory was four, which occurred only twice(0.4% of trajectories).0 1 2 3 4Number of RTSs per trajectory050100150200250300N0~9.9~19.8~29.8~39.7~49.6~59.5%Figure 4.3. Number of RTSs per trajectory of all modelled trajectories for the seven ozoneepisodes. The left scales gives counts (N), and the right scale gives approximate percentageof the total dataset (504 trajectories total).594.2. Spatial characteristicsRecirculation was detected during all episodes, totaling 293 RTSs. These RTSs are mappedby episode in Figure 4.4. When compared to their corresponding number density maps ofcomposite trajectory fields, paths of RTSs tend to coincide with preferred trajectory pathways(cf. Figure 3.9), except for those during the 1985 and 1987 episodes (dominant Type IVcirculations). That is, recirculation neither occurred within Howe Sound nor over the NorthShore Mountains. Moreover, RTS pathways exhibited primarily zonal movement. It was oftenthe case that trajectories initially headed eastward (90◦), only to recirculate and return fromeast of the origin (i.e. with a direction of return between 0◦ − 180◦; Figure 4.5).Though recirculation was detected during all episodes, the frequency of detection wasnot equal. Before further examining this finding, RTSs were grouped by circulation regime.This procedure is justified given that trajectories within circulation regimes exhibit similarspatial properties (Section 3.3.2), so it is assumed that trajectories that recirculate withinregimes do as well. Each RTS was labelled according to the prevailing circulation regimethat occurred during its lifetime. From this, it is made clear that recirculation was notdetected equally between circulation regimes. RTSs were detected least frequently duringType IV circulation conditions (Figure 4.6a,b), when trajectories travelled northward over theNorth Shore Mountains. Conversely, trajectories recirculated most frequently during TypeI conditions. Recall that this regime has prominent onshore winds over the entire diurnalperiod.The RTSs are mapped by circulation regime in Figure 4.7. Regardless of circulation regime,RTSs were confined to the northern edge of the LFV. During Type I, II, and III circulationconditions, trajectories recirculated several times within tributary valleys (e.g. examine theRTSs in the vicinity of Pitt River Valley and Stave Lake in Figure 4.7a; refer to Figure 2.1 forlocations; see also Section 4.2.2). Trajectories recirculated much further inland (eastward) thanoffshore (westward). Very few RTSs travelled above the critical height during their lifetime(Figure 4.6d), and most internally recirculated (Figure 4.6c). Because of the proximity of thewestern domain edges to X0, almost all external recirculations occurred on this side of thedomain (e.g. Figure 4.7c).604.2.Spatialcharacteristics(a) 1985 I - IV - IV (b) 1987 IV - IV - IV (c) 1993 I - I - I(d) 1995 III - III - III (e) 1998 II - III - II (f) 2001 II - II - II-124° -123° -122° -121°+49°+49.5°(g) 2006 I - I - IIIFigure 4.4. Detected RTSs from modelled trajectories during ozone episodes. Only those portions of the trajectories whichwere found to have recirculated are featured, and the remaining portions of the trajectories have been discarded. The initialorigin of the trajectories (black triangle) and domain of interest (thick gray lines) are also plotted. The circulation regimes ofeach day of the episode are labelled above each plot.614.2. Spatial characteristics0°45°90°135°180°225°270°315°90 o180 o270 o360 oFigure 4.5. Direction of departure and direction of return of RTSs,with respect to thetrajectory origin (X0). The azimuthal scale (angle from center of plot) gives the direction ofdeparture (in degrees from North), and the range rings (distance from center of plot) givedirection of return (in degrees from North). When the departure and return directions areequal, points will lie along the thick red line. The range of the departure angle ±45◦ and±90◦ are highlighted in dark and light red, respectively.624.2. Spatial characteristics0255075100N(a) Unique trajectories that recirculated0255075100N(b) Number of RTSs1 2 3 4Circulation Regime0255075100%(c) Internal RTSs1 2 3 4Circulation Regime0255075100%(d) RTSs below critical heightFigure 4.6. Statistics of detected RTSs by circulation regime. (a) The number of uniquetrajectories that recirculated at least once. (b) The number of RTSs detected. (c) Thepercentage of RTSs that internally recirculated within the horizontal domain of interest. (d)The percentage of RTSs that remained below the critical height for their entire lifetime.634.2. Spatial characteristics(a) Type I (b) Type II-124° -123° -122° -121°+49°+49.5°(c) Type III (d) Type IVFigure 4.7. As in Figure 4.4, but for all RTSs by circulation regime.644.2. Spatial characteristics4.2.2 Description of selected Recirculating Trajectory SegmentsIt is essential to keep in mind that RTSs are not fixed objects; their paths trace the movementof imaginary air parcels through space over time. Hence, it is difficult to decipher the timescales of trajectories and RTSs from composite maps. In this section, select individual RTSsare examined in more detail.Take for example the trajectory beginning from X0 just after noon on 2006-2 (Figure 4.8).The trajectory initially travels inland, then up the Pitt River Valley. As the sun sets (∼2000local), wind directions reverse from up- to down-valley, and the trajectory reverses its path aswell. The detected RTS ends nearly at the same position that it began.In contrast, the trajectory beginning at 1700 on the same day does not travel as far inland(Figure 4.9). Since it begins later in the day, there is less time between its start time andthe time at which wind directions reverse, so it returns towards X0 much sooner. Then, asecond recirculation begins, where the trajectory continues in the offshore direction during theremaining nighttime hours, before returning towards downtown Vancouver in the morning, aswind directions reverse again. Both detected RTSs remained below the critical height.The two examples described above reflect the highly idealized description of horizontalrecirculation as described by Steyn (2003) (Section 1.3.1): a recirculating trajectory travelsonshore and up-valley during daytime hours, then offshore and down-valley during nighttimehours. This finding indicates that the proposed definition of recirculation herein is able tocapture this idealized behaviour. Furthermore, the spatiotemporal characteristics of the RTSsagree with observations during ozone episodes of high pollutants found east of Vancouverand within tributary valleys during the day, and over the Strait of Georgia at night (Section1.2). The difference between the inland travel distance of the two RTSs described above andthus their resulting paths over time highlight the necessity of quasi-continuously initiatingtrajectories over the course of the modelled episodes.There are, of course, frequent RTSs that do not entirely correspond to the idealized modelof “recirculation”. Two such consecutive RTSs are illustrated in Figure 4.10. Much like thatin Figure 4.8, the trajectory travels inland during daytime hours, and recirculates within the654.2. Spatial characteristics-124° -123° -122° -121°+49°+49.5° 19010 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0500100015002000Height (m asl)00 06 12 18 00 06 12 18 00 06 12 18 00Hour (local)Figure 4.8. A RTS detected from a trajectory initiated at 1300 (local) on the second dayof the 2006 episode (June 25, 2006; hour 37 of the three-day model run). Daytime hours(0700–2000) are denoted by red circles, and nighttime hours (2100–0600) are denoted byopen blue circles. (top) The trajectory origin (black triangle) and domain of interest (thickgray lines) are also plotted. (bottom) The critical height (800 m agl) is represented by thedashed black line, and terrain elevation (m asl) under the trajectory positions is shaded ingray.664.2. Spatial characteristics-124° -123° -122° -121°+49°+49.5°2306120 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour02505007501000Height (m asl)00 06 12 18 00 06 12 18 00 06 12 18 00Hour (local)Figure 4.9. As in Figure 4.8, but for the trajectory initiated at 1700 (local). The startposition of the secondary RTS is denoted by the open black triangle.674.3. Temporal characteristicsPitt River Valley. Once the air parcel re-enters the horizontal domain of interest the nextmorning (0900 local), it is located about 44.7 km away from X0. At this point, a secondrecirculation beings as the air parcel travels back up the ridge line of the Pitt River Valley,away from X0. The second RTS ends at night (2000 local), as the air parcel descends into themain valley floor once more. The paths of these RTS are clearly driven by diurnally reversingwinds throughout the LFV, but they are spatially complex.4.3 Temporal characteristicsWithin the context of ozone episodes, examining the temporal characteristics (start time, endtime, and τR) of RTSs be can used to:(1) help explain pollutant trends at AQ monitoring stations,(2) determine if RTSs reveal potential day-to-day “carryover” of pollutants (and thuspotential increasing pollutant trends),(3) determine if recirculation occurs more frequently at particular times of day, and(4) reveal underlying dynamic mechanisms that lead to recirculation and that arecharacteristic of wind fields during poor AQ conditions.The first application listed above is beyond the scope of this research, though is typicallyaddressed in other studies of pollutant recirculation (e.g. Hurley and Manins 1995, Ding et al.2004, Evtyugina et al. 2006).To investigate potential “carryover” of ozone precursors, the relationship between start andend times of RTSs was examined. For illustrative purposes, start and end times are plottedfor RTSs detected during the 2006 episode, in Figure 4.11 (see Figures D.1 – D.6 for otherepisodes). A few notes on the construction of this particular graphic:• The central plot area appears triangular, since end time cannot occur before start time.• A third axis of τR emerges at 45◦ between the x and y axes (dashed gray lines).684.3. Temporal characteristics-124° -123° -122° -121°+49°+49.5°12180006 1521030 6 12 18 24 30 36 42 48 54 60 66 72Elapsed hour0500100015002000Height (m asl)00 06 12 18 00 06 12 18 00 06 12 18 00Hour (local)Figure 4.10. As in Figure 4.9, but for the trajectory initiated at 0600 on the first day ofthe 2006 episode (June 24, 2006; hour 6 of the three-day model run).694.3. Temporal characteristics• Multiple RTSs belonging to different trajectories can begin and end at the same time.If this is the case, multiple points will be plotted on top of one another. The histogramson the top and right sides of the central plot are used for clarification.0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour122436486010I I III14IIIIIFigure 4.11. Temporal characteristics of detected RTSs during the 2006 episode. (center)Start and end times (elapsed model hour) of RTSs. Note that hours 0, 24, and 48 correspondsto 0000 (local) on the first, second, and third day of the modelled episodes. The green shadedarea represents periods of time where a RTS started on one day of the episode, and returnedon the next (“carryover”). The axis at 45◦ (dashed gray lines) are the RTS lifetimes, τR (h).(top) Histogram of RTS start times. (right) Histogram of RTS end times. The circulationregimes of each day of the episode are labelled in each histogram, and each day is separatedby a dotted black line. Rush hour periods (0600-0900) are highlighted in black on bothhistograms.704.3. Temporal characteristicsOf the 61 RTSs detected during the 2006 episode, 49 (80.3%) began on one day of theepisode and ended on the next, i.e. the air parcel represented by the trajectory was “carriedover”. RTS lifetimes varied between 3 h and 28 h. During the circulation regimes that prevailedduring this episode, it appears as though trajectories recirculated regardless of their start times(Figure 4.11, top histogram). However, RTS end times appear to have a more restricted rangeof times (Figure 4.11, right histogram). The majority of RTSs terminated in the morning,between 0000 and 0900, and often before the morning rush hour period (0600 – 0900).Structure is apparent within the data plotted in Figure 4.11. This structure is betterrevealed by plotting RTS start times and corresponding values of τR. Once again, theresulting temporal data were examined within circulation regimes, as shown in Figure 4.12.During Types I, II, and III circulation regimes, trajectories initiated at any hour of the dayrecirculated, and the distribution of start times is roughly bimodal (Figure 4.12a,b,c, tophistograms). The distribution of RTS start times for the Type IV regime is distinctly U-shaped(Figure 4.12d, top histogram), and resembles the distribution of the other three regimes only inthe early morning (0000–0900) and late evening (1800–2300) hours. More trajectories initiatedafter the morning rush-hour period recirculated than those initiated before. For all circulationregimes, τR is heavily skewed towards shorter lifetimes. The total set of RTS lifetimes (Figure4.13) reveals that the majority of trajectories (70%) typically recirculated within < 12 h.Overall, trajectories initiated in the early morning and daytime hours (∼ 0400–1600) hadlonger recirculation lifetimes (≥ 12 h), and those before and after this time period had veryshort recirculation lifetimes (< 12 h). Hence, there is a sawtooth-like oscillation of τR overtime. This behaviour can be explained by reversing wind directions over the course of onediurnal period. That is, trajectories released just after (before) a switch in wind direction (e.g.from offshore to onshore) will have a long (short) time to travel before the wind switches backand carries the air parcel towards the origin. An illustrative example is provided in AppendixD.2. The asymmetry between the peak and trough of the sawtooth structures in Figure 4.12results from stronger daytime than nighttime wind speeds.714.3. Temporal characteristics061218243036τ R (h)0 6 12 18Start time (local)061218243036τ R (h)0 6 12 18Start time (local)20(a) Type I3020(b) Type II2520(c) Type III2020(d) Type IV15Figure 4.12. RTS start times (local) and lifetimes (τR), by circulation regime. For eachsubplot (a) – (d): (top) Histogram of start times. The rush hour period (0600-0900) ishighlighted in black. (right) Histogram of RTS lifetimes.724.4. The Recirculation Factor0 3 6 9 12 15 18 21 24 27 30 33 36τR  (h)0102030%Figure 4.13. Distribution of RTS lifetimes (τR), over all seven modelled episodes.4.4 The Recirculation FactorAt present, the use of the RF (Allwine and Whiteman 1994) is not widespread in peer-reviewedscientific literature. However, since it is the only quantitative and objective measure ofdetecting recirculation, it deserves some attention here. The RF is a simple, geometric measureof the potential for air to recirculate at a single location. It is derived from wind data (speedand direction) at a site, and hence it is an Eulerian measure. Rather than examine the RFover time at X0 (as done by others), the goal here is to see if the RF is appropriate and/orequivalent to the definition of recirculation outlined in this study.4.4.1 Description of RFWind speed and direction data from a single point are decomposed into u (east-west) and v(north-south) components. By assembling a vector progression of u and v wind vectors overtime, RF is defined as:RF (τ) = 1−LS734.4. The Recirculation Factorwhere S is the wind run (sum of the magnitude of all wind vectors), and L is the netdisplacement of the assembled wind vectors (Figure 4.14). These measures are computedover a selected period of transport time, τ , for all time steps i = 0, . . . , N − (τ/T ) in the dataseries, where N is the number of data points and T is the time interval between data points.The result is a series of RF values over time. A value of τ = 24 h is typically used, sinceit represents the duration of one complete wind or heating cycle, and is independent of thestart time of integration (for an idealized sinusoidal wind rotation; Levy et al. 2008a,b). WhenRF = 0, no recirculation has occurred. When RF = 1, the point of interest is said to havestrong recirculation potential3. That is, a hypothetical air parcel returns exactly to its originwithin a period of τ hours, which gives L = 0 and thus RF = 1. This situation only holds fora homogeneous wind field: RF is a measure of recirculation potential at a single point, ratherthan the transport of a pollutant plume through space.Finally, to determine if a site is prone to recirculation, the computed RF must meet orexceed a critical value, RFc, that determined by the average of all computed RF values overthe period of study (Allwine and Whiteman 1994).4.4.2 Comparison to detection algorithm for recirculationTo assess the suitability of Allwine and Whiteman’s (1994) mathematical definition, values ofRF calculated from (1) wind vector timeseries at X0 (interpolated from WRF model output)and (2) the 293 detected RTSs were compared. For (1), RF values follow directly from Allwineand Whiteman’s (1994) definition, and are thus Eulerian measures of recirculation (RFE). For(2), RF was calculated from the total distance along RTSs (S) and net displacement from theirstart to end position (L). Resulting RF values are thus Lagrangian measures of recirculation(RFL). To compare the two measures, RFE was integrated from corresponding RTS starttimes over a period of τ = τR hours. Thus, a fixed value of τ = 24 h was not necessarilyused for each computation. This modification has the advantage of better reflecting “true”recirculation lifetimes that are often < 12 h (Figure 4.13). Finally, all RFE values were3As Allwine and Whiteman (1994) note, the RF equation has a mathematical singularity when S = 0. Thissingularity will not physically occur, however, since when S = 0, L = 0 as well, i.e. there is no displacementfrom the origin.744.4. The Recirculation Factor00000300060009001200150018002100LSFigure 4.14. Schematic diagram of Allwine and Whiteman’s (1994) Recirculation Factor.Wind timeseries (at hourly intervals; local times are labelled) are taken from a single point(black triangle), and a progressive vector diagram is created (black arrows) over a specifiedtransport time (here, τ = 24 h). The resulting wind run (S) and net transport distance (L)are illustrated.averaged (given differing values of τ), which gave RFc = 0.29.Overall, no correlation between RFE and RFL was found (r = 0.21) (Figure 4.15).Differences between the two measures did not depend on circulation regime or τR (not shown).Furthermore:• The Recirculation Factor based on the winds at the origin alone (RFE) tended tounder-predict the “strength” of recirculation, i.e. RFE < RFL.• When RFE ≈ 0, RFL ranged between 0 and 1. This is a clear indication of inhomogeneityof the wind field over space and time: while RFE ≈ 0 indicates constant wind directionsover τR at X0, higher values of RFL over the same period imply that wind directionschange downwind.754.5. Discussion• When RFc is used determine if the site (X0) is prone to recirculation, RFE ≥ RFc only125 times. Thus, the site is only prone to recirculation during 42.7% of the time periodsexamined. In contrast, RFL exceeds RFc 270 times (92.2%). By calculating RF fromRTSs, it is evident that the LFV is much more prone to recirculation than values of RFEindicate.0.0 0.2 0.4 0.6 0.8 1.0RFE0. IType IIType IIIType IVFigure 4.15. Eulerian (RFE) and Lagrangian (RFL) measures of Recirculation Factor forall detected RTSs, by circulation regime. A one-to-one line (solid gray) and critical valueRFc = 0.29 (dotted gray) are plotted for reference.4.5 DiscussionThe frequency of recirculating trajectories within ozone circulation regimes seems to indicatethat not all mesoscale conditions have the same capacity to recirculate pollutants in theLFV. It is interesting to note that the lowest frequency of recirculation occurred during764.5. Discussionthe regime that signifies the termination of a multi-day ozone episode (Type IV), and thehighest frequency of recirculation occurred during the regime that is least associated withSB circulations (Type I) (Section 2.1.2). Further analyses should be performed to determineif this finding is statistically robust. However, all RTSs, regardless of circulation regime,exhibited dominant zonal movement. Earlier analysis of the wind fields indicated that SBswere not observed on any modelled day (Appendix B.3.1), thus the prevalence of recirculatingtrajectories suggests that recirculation in the LFV is primarily driven by onshore flows andmountain-valley circulations within the main valley floor, and secondarily by diurnal flowswithin tributary valleys. Indeed, there is strong evidence that pollutants released from thedowntown Vancouver area (X0) can recirculate several times within the LFV and throughtributary valleys (e.g. Figure 4.10), and that recirculation is driven by reversing wind directions(Figures 4.12 and D.7).Overall, though spatial distributions of trajectories are different between circulationregimes (Figure 3.11), RTSs are spatially and temporally similar. This is somewhat surprisinggiven the differences between the daily hodographs of circulation regimes. However, thissuggests that all identified RTSs, regardless of circulation regime, are merely samples of thesame population of possible recirculating trajectories, just as all mesoscale circulation regimesare samples of the same population of synoptic conditions.In contrast to the idealized model of recirculation that acts over a 24 h period, RTSlifetimes are often less than 12 h. Note that this is on the scale of the chemical lifetime ofozone precursors. Thus, RTSs may indeed play a role in the build-up of precursor pollutantsand exacerbate AQ conditions during ozone episodes.Finally, the lack of structure and correlation between RFE and RFL is a clear indicationof the lack of general applicability of Allwine and Whiteman’s (1994) Recirculation Factor.Though the RF is simple since it only requires wind data at one point (which is oftenreadily available for stations), assessing the potential for recirculation based on this measurealone warrants criticism, especially in areas of complex terrain, such as the LFV. This is anobvious drawback that seems to have been ignored in other studies (see Fast and Zhong 1998,774.6. SummaryCastell et al. 2008, Charabi and Al-Yahyai 2011, Perez et al. 2012). As opposed to the RF,the detection algorithm for recirculation herein has the advantage of providing recirculationlifetimes, which can be useful when considering chemistry, as discussed above. Note that for anidealized sinusoidal wind cycle (Figure D.7), the detection algorithm for recirculation hereinidentifies strong recirculation within short lifetimes (e.g. τR = 1 h), whereas the RF wouldindicate low recirculation “strength” since the entire length of the trajectory is considered overτ = 24 h. Values of RFL < RFc also indicate that the recirculation factor, overall, is not agood diagnostic for recirculation.4.6 SummaryAn objective and generally applicable definition for horizontal “recirculation” was developed.The definition requires trajectory data and a carefully selected horizontal domain of interestand critical height that reflect the scope and purpose of research. To apply the definition anddetection algorithm to trajectories modelled during ozone episodes in the LFV, a domain ofinterest covering the entire ‘chemically active’ region and a critical height corresponding tothe afternoon maximum inland ML height were selected. Recirculation was detected duringall episodes; approximately 40% of the trajectories over all episodes recirculated at leastonce. However, recirculation did not occur equally between mesoscale circulation regimes,with a low frequency during Type IV conditions. The paths of RTSs were often confinedto the northern edge of the LFV. The temporal characteristics of detected RTSs providestrong evidence that pollutants may be “carried-over” from one day of an episode to thenext. Trajectories initiated after the morning rush-hour period recirculated more often thanthose initiated before. The total set of recirculation lifetimes, τR, is heavily skewed towardsshorter lifetimes, but τR is shorter in the early morning and late evening hours, and muchlonger during the day. Overall, regardless of circulation regime, RTSs have similar spatial andtemporal characteristics. Finally, a comparison between the results of the proposed detectionalgorithm for recirculation and the measure defined by Allwine and Whiteman (1994) suggeststhat their Recirculation Factor is not an appropriate measure for recirculation in the LFV.785 ConclusionThe occurrence and spatiotemporal variability of ozone episodes in the Lower Fraser Valley ofBritish Columbia is strongly influenced by the presence of interacting synoptic and mesoscaleflow systems. Ozone episodes occur under a small set of synoptic and mesoscale conditionswith weak forcing, with high solar radiation, high temperatures, and low wind speeds. Giventhat the LFV has a relatively small population size and a lack of heavy industry, the severityof ozone episodes is often attributed to strong recirculation of pollutants coincident withthermo-topographically forced mesoscale circulations.In essence, the purpose of this modelling study was to examine the prevalence ofatmospheric recirculation in the LFV during ozone episodes, then examine its spatial andtemporal characteristics within the known set of mesoscale circulation regimes. A rigorousstudy approach was devised, where trajectory analyses were performed with the HYSPLITmodel using WRF model output from seven, three-day summertime ozone episodes occurringover a period of 20 years, representative of the four mesoscale circulation regimes conduciveto ozone episodes. HYSPLIT trajectory model runs were configured such that the startingposition of all trajectories coincided with the location of maximum ozone precursors, informedfrom a SMOKE-adjusted emissions inventory. An objective algorithm to detect recirculationbased on trajectory model output was formulated, and then applied to the modelledtrajectories.795.1. Major findings5.1 Major findingsTo summarize the major findings of this study, we return to the original set of researchquestions:(1) Can we develop a generally applicable definition of ‘recirculation’?A quantitative, rational definition and algorithm for detecting horizontal recirculation wasdeveloped and applied to trajectory data. The definition developed herein was formulatedsuch that it is independent of the type of trajectory data, study location, meteorological andchemical conditions, and time scale. The definition depends on a spatial scale that reflectsthe scope and purpose of the research, which is restricted by a horizontal domain of interestand critical height. Essentially, the algorithm detects the point at which an air parcel alongits trajectory is at a minimum distance from its origin. The resulting recirculating trajectorysegment (RTS) is thus objectively determined, and finite start and end positions and times ofrecirculation are obtained.From this generally applicable definition, note that detection of a RTS is a necessary butnot sufficient requirement to explain increasing trends of chemical species (e.g. daily ozonemaxima) during episodes of degraded AQ at single locations. The investigator must interpretthese results in conjunction with additional AQ data, in light of their research needs.(2) Are existing measures of recirculation appropriate based on Question (1)?Though only briefly mentioned at the beginning of this research (see Section 1.3.3), anexisting measure, or rather, method of detecting recirculation is by simply visually inspectingtrajectories for ‘looping’ patterns (e.g. as done by Holzworth et al. 1963, Ding et al. 2004 ,Harris and Kotamarthi 2005, Ma and Lyons 2003). Given the objective definition followingfrom Question (1), this measure and method of detecting recirculation is clearly insufficient,and does not allow for any sort of quantitative assessments.Alternatively, Allwine and Whiteman (1994) proposed a mathematical measure of805.1. Major findingsatmospheric recirculation, the Recirculation Factor (RF), determined from wind vectors ata single fixed point in space, over a predetermined transport time. Several factors renderthis measure of recirculation inappropriate. First, to define recirculation based on a fixedtime scale (e.g. 24 h) contradicts the visual spatial assessment described above, and does notadhere to the objective criteria of the proposed definition herein. This is especially true giventhat recirculation can occur over a wide range of times, often less than 12 h. Second, theRF cannot be used in all study locations. A comparison between RF values derived from thedefinition herein and those calculated from the trajectory starting location revealed that theRF should not be used in the LFV due to the presence of complex flow systems, which is oftenneglected in other works.(3) Does recirculation occur in the LFV during ozone episodes?Yes, atmospheric recirculation does occur in the LFV under meteorological conditions of ozoneepisodes. The frequency of recirculation appears to be driven by daily mesoscale circulationregimes, as defined by Ainslie and Steyn (2007). Recirculation is common when wind directionshave westerly components at YVR (Types I, II, and III), but is fairly rare when south-easterlywinds dominate (Type IV). Hence, not all mesoscale conditions during ozone episodes havethe same capacity to recirculate pollutants in the LFV.(4) Are there particular transport pathways that are characteristic of different ozone circulationregimes in the LFV?This research question was addressed in Chapter 3, by qualitatively and quantitativelyanalyzing the spatial distributions of trajectories and resulting number density fields.Trajectories distributions within circulation regimes are spatially similar, with a high degreeof statistical association. Trajectory distributions between circulation regimes are perceivablydifferent, but share preferred transport pathways along Howe Sound and within tributaryvalleys of the North Shore Mountains.815.1. Major findings(5) What are the spatial and temporal characteristics of recirculation in the LFV underdifferent circulation regimes?Though the total composite trajectory fields differ between circulation regimes, RTSs arespatially and temporally similar. Recirculating trajectories were confined to the northern edgeof the LFV, often heading to and returning from the eastern portion of the valley. Frequentrecirculation occurred within tributary valleys under Type I, II, and III conditions. Thisresult is in agreement with other observational and modelling studies (e.g. McKendry et al.1998a,b), and highlights the role that tributary valleys play on redistributing pollutants withinthe LFV. The spatial characteristics of detected RTSs correspond to the conceptual definitionand idealized description of recirculation (see Steyn 2003).RTSs began at any time of day, but often ended in the early morning the next day. Thissuggests there is indeed recirculation that contributes to day-to-day “carryover” of pollutants.The majority of recirculation lifetimes, τR, were less than 12 h, which is on the scale of thechemical lifetime of ozone precursor species. Under Type I, II, and III circulation conditions,τR oscillates in a sawtooth-like pattern over one diurnal period. This reflects the relationshipbetween trajectory start time and timing of wind reversal. Under Type IV conditions, thesame pattern is apparent only in the early morning and and late evening, as if the cycle istruncated during daytime hours. The lack of recirculation during this period coincides withstrong southerly daytime winds at YVR.Overall, spatial and temporal similarities between RTSs of differing circulation regimes issomewhat surprising, given the statistically distinct circulation regime hodographs at YVR.This suggests that the influence of smaller-scale thermo-topographic flows (in particular,mountain-valley breeze and slope flows, but not necessarily SBs; Appendix B.3), occurringunder all four mesoscale conditions, play a more important role in determining the transportand recirculation of pollutants than the mesoscale flow patterns alone. This findingconfirms the claims by McKendry (1994) and Ainslie and Steyn (2007) on the importanceof sub-mesoscale processes leading to precursor build-up during ozone episodes.825.2. Future research5.2 Future researchAn interesting finding of this study was that, for the most part, trajectories initiatednear downtown Vancouver during ozone episodes did not travel to the eastern portionof the LFV. This seems to suggest that high ozone concentrations observed in theAbbotsford-Chilliwack-Hope region during exceedance days are likely not strongly influencedby precursor emissions originating from this area. Further modelling studies are needed todetermine the source of emissions affecting this NOx-limited region, and to implement effectiveVOC and NOx control strategies.One major limitation of this study is the use of a single starting location to modeltrajectories and detect recirculation. Though this starting location was carefully selectedto represent a point of high precursor emissions, the complexity of transport and recirculationwithin the entire LFV cannot be fully addressed using this point alone. It would be interestingto extend the current analysis to other locations in the LFV, and create a heat map of“recirculation potential” by counting the number of RTSs originating from each location foreach modelled episode, which can be easily automated. A more complete understanding ofthe entire airshed’s ability to recirculate polluted air masses will improve our understandingof the spatial distributions of pollutants that have been observed, and may help inform futureozone abatement strategies.The primary goal of this research was to determine if recirculation occurs in the LFV duringozone episodes. What has yet to be addressed is the relative importance of recirculation onspatiotemporal distributions of ozone and precursor species. This task could be addressedby conducting a similar trajectory modelling exercise using back trajectories, in combinationwith statistical source-receptor trajectory analysis methods (e.g. see Ashbaugh et al. 1985,Hsu et al. 2003).This study has revealed that recirculation frequently occurs in the LFV during mostmesoscale circulation regimes, and has provided evidence of pollutant carryover. Furtherchemical modelling studies are needed to better understand and isolate the effects of precursor835.2. Future researchemissions released on one episode day on the next day’s AQ conditions.Finally, the author would like to encourage others to use the definition and detectionalgorithm for recirculation proposed in this work. Applying this definition and detectionmethod to trajectory datasets from other study locations in studies with similar overall researchgoals will help confirm its utility.84ReferencesAbbs, D. J., Physick, W. L. (1992). Sea-breeze observations and modelling: a review. AustralianMeteorological Magazine, 41, 7–19.Ainslie, B., Steyn, D. G. (2007). Spatiotemporal Trends in Episodic Ozone Pollution in theLower Fraser Valley, British Columbia, in Relation to Mesoscale Atmospheric CirculationPatterns and Emissions. Journal of Applied Meteorology and Climatology, 46 (10),1631–1644.Ainslie, B., Steyn, D. G., Reuten, C., Jackson, P. L. (2013). A Retrospective Analysis ofOzone Formation in the Lower Fraser Valley, British Columbia, Canada. Part II: Influenceof Emissions Reductions on Ozone Formation. Atmosphere-Ocean, 51 (2), 170–186.Allwine, K., Whiteman, C. (1994). Single-station integral measures of atmospheric stagnation,recirculation and ventilation. Atmospheric Environment, 28 (4), 713–721.Ashbaugh, L. L., Malm, W. C., Sadeh, W. Z. (1985). A residence time probability analysis ofsulfur concentrations at grand Canyon National Park. Atmospheric Environment, 19 (8),1263–1270.Banta, R. M., Shepson, P. B., Bottenheim, J. W., Anlauf, K. G., Wiebe, H. A., Gallant, A.,Biesenthal, T., Olivier, L. D., Zhu, C. J., McKendry, I. G., Steyn, D. G. (1997). Nocturnalcleansing flows in a tributary valley. Atmospheric Environment, 31 (14), 2147–2162.Banta, R. M., Senff, C. J., Nielsen-Gammon, J., Darby, L. S., Ryerson, T. B., Alvarez, R. J.,Sandberg, S. P., Williams, E. J., Trainer, M. (2005). A bad air day in Houston. Bulletinof the American Meteorological Society, 86 (5), 657–669.Batchvarova, E., Cai, X., Gryning, S.-E., Steyn, D. (1999). Modelling Internal Boundary-Layerdevelopment in a region with a complex coastline. Boundary-Layer Meteorology, 90 (1),1–20.Biggs, W. G., Graves, M. E. (1962). A Lake Breeze Index. Journal of Applied Meteorology,1 (4), 474–480.Brauer, M., Brook, J. R. (1997). Ozone personal exposures and health effects for selectedgroups residing in the Fraser Valley. Atmospheric Environment, 31 (14), 2113–2121.Brook, J. R., Strawbridge, K. B., Snyder, B. J., Boudries, H., Worsnop, D., Sharma, S.,Anlauf, K., Lu, G., Hayden, K. (2004). Towards an understanding of the fine particle85Referencesvariations in the LFV: integration of chemical, physical and meteorological observations.English. Atmospheric Environment, 38 (34), 5775–5788. issn: 1352-2310.Cai, X. M., Steyn, D. G. (2000). Modelling study of sea breezes in a complex coastalenvironment. Atmospheric Environment, 34 (18), 2873–2885.Canadian Council of Ministers of the Environment (2000). Canada-wide standards forparticulate matter (PM) and ozone. Technical Report. Quebec City, Quebec, Canada, p. 10.Castell, N., Mantilla, E., Millan, M. M. (2008). Analysis of tropospheric ozone concentrationon a Western Mediterranean site: Castellon (Spain). Environmental Monitoring andAssessment, 136 (1-3), 3–11.Charabi, Y., Al-Yahyai, S. (2011). Integral assessment of air pollution dispersion regimes inthe main industrialized and urban areas in Oman. Arabian Journal of Geosciences, 4 (3-4),625–634.Chen, J. M., Oke, T. R. (1994). Mixed-layer heat advection and entrainment during the seabreeze. Boundary-Layer Meteorology, 68 (1-2), 139–158.Cheung, V. T. F., Wang, T. (2001). Observational study of ozone pollution at a rural site inthe Yangtze Delta of China. Atmospheric Environment, 35 (29), 4947–4958.Clappier, A., Martilli, A., Grossi, P., Thunis, P., Pasi, F., Krueger, B. C., Calpini, B., Graziani,G., Bergh, H. van den (2000). Effect of sea breeze on air pollution in the Greater AthensArea. Part I: Numerical simulations and field observations. Journal of Applied Meteorology,39 (4), 546–562.Crosby, D. S., Breaker, L. C., Gemmill, W. H. (1993). A Proposed Definition for VectorCorrelation in Geophysics: Theory and Application. Journal of Atmospheric and OceanicTechnology, 10 (3), 355–367.Crosman, E. T., Horel, J. D. (2010). Sea and Lake Breezes: A Review of numerical studies.Boundary-Layer Meteorology, 137 (1), 1–29.Davis, R. E., Normile, C. P., Sitka, L., Hondula, D. M., Knight, D. B., Gawtry, S. P.,Stenger, P. J. (2010). A comparison of trajectory and air mass approaches to examineozone variability. Atmospheric Environment, 44 (1), 64–74.Ding, A., Wang, T., Zhao, M., Wang, T., Li, Z. K. (2004). Simulation of sea-land breezes anda discussion of their implications on the transport of air pollution during a multi-day ozoneepisode in the Pearl River Delta of China. Atmospheric Environment, 38 (39), 6737–6750.Draxler, R. R., Hess, G. D. (1997). Description of the HYSPLIT4 modeling system. TechnicalMemorandum 224. Silver Spring, MD: NOAA Air Resources Laboratory, p. 24.Draxler, R. R., Stunder, B., Rolph, G., Taylor, A. (1999). HYSPLIT4 User’s Guide. NOAATechnical Memorandum ERL ARL, 230, 35.86ReferencesEvtyugina, M. G., Nunes, T., Pio, C., Costa, C. S. (2006). Photochemical pollution undersea breeze conditions, during summer, at the Portuguese West Coast. AtmosphericEnvironment, 40 (33), 6277–6293.Fast, J. D., Zhong, S. (1998). Meteorological factors associated with inhomogeneousozone concentrations within the Mexico City basin. Journal of Geophysical Research:Atmospheres, 103 (D15), 18927–18946.Fleming, Z. L., Monks, P. S., Manning, A. J. (2012). Review: Untangling the influence ofair-mass history in interpreting observed atmospheric composition. Atmospheric Research,104-105, 1–39.Grossi, P., Thunis, P., Martilli, A., Clappier, A. (2000). Effect of sea breeze on air pollutionin the Greater Athens Area. Part II: Analysis of different emission scenarios. Journal ofApplied Meteorology, 39 (4), 563–575.Harris, L., Kotamarthi, V. R. (2005). The characteristics of the Chicago Lake breeze and itseffects on trace particle transport: Results from an episodic event simulation. Journal ofApplied Meteorology, 44 (11), 1637–1654.Hayden, K. L., Anlauf, K. G., Hoff, R. M., Strapp, J. W., Bottenheim, J. W., Wiebe, H. A.,Froude, F. A., Martin, J. B., Steyn, D. G., McKendry, I. G. (1997). The vertical chemicaland meteorological structure of the boundary layer in the Lower Fraser Valley duringPacific ’93. Atmospheric Environment, 31 (14), 2089–2105.Hedley, M., Singleton, D. L. (1997). Evaluation of an air quality simulation of the Lower FraserValley – 1. Meteorology. Atmospheric Environment, 31 (11), 1605–1615.Hedley, M., McLaren, R., Jiang, W. M., Singleton, D. L. (1997). Evaluation of an air qualitysimulation of the Lower Fraser Valley – 2. Photochemistry. Atmospheric Environment,31 (11), 1617–1630.Hoff, R. M., Harwood, M., Sheppard, A., Froude, F., Martin, J. B., Strapp, W. (1997). Useof airborne lidar to determine aerosol sources and movement in the Lower Fraser Valley(LFV), BC. Atmospheric Environment, 31 (14), 2123–2134.Holzworth, G. C., Kauper, E. K., Smith, T. B. (1963). Some observed low-level air trajectoriesover Los Angeles, California. Monthly Weather Review, 91 (8), 387–392.Houyoux, M. R., Vukovich, J. M. (1999). Updates to the Sparse Matrix Operator KernelEmissions (SMOKE) modeling system and integration with Models-3. The EmissionsInventory: Regional Strategies for the Future, 1461.Hsu, Y.-K., Holsen, T. M., Hopke, P. K. (2003). Comparison of hybrid receptor models tolocate PCB sources in Chicago. Atmospheric Environment, 37 (4), 545–562.Hurley, P. J., Manins, P. C. (1995). Meteorological modeling on high-ozone Days in Perth,Western-Australia. Journal of Applied Meteorology, 34 (7), 1643–1652.87ReferencesJoe, H., Steyn, D. G., Susko, E. (1996). Analysis of trends in tropospheric ozone in the LowerFraser Valley, British Columbia. Atmospheric Environment, 30 (20), 3413–3421.Krzyzanowski, J., McKendry, I. G., Innes, J. L. (2006). Evidence of elevated ozoneconcentrations on forested slopes of the Lower Fraser Valley, British Columbia, Canada.Water, Air, and Soil Pollution, 173, 273–287.Levy, I., Dayan, U., Mahrer, Y. (2008a). A five-year study of coastal recirculation andits effect on air pollutants over the East Mediterranean region. Journal of GeophysicalResearch-Atmospheres, 113 (D16), D16121.Levy, I., Dayan, U., Mahrer, Y. (2008b). Studying coastal recirculation with a simplifiedanalytical land-sea breeze model. Journal of Geophysical Research-Atmospheres, 113 (D3),D03104.Levy, I., Mahrer, Y., Dayan, U. (2009). Coastal and synoptic recirculation affecting airpollutants dispersion: A numerical study. Atmospheric Environment, 43 (12), 1991–1999.Levy, I., Dayan, U., Mahrer, Y. (2010). Differing atmospheric scales of motion and their impacton air pollutants. International Journal of Climatology, 30 (4), 612–619.Li, S.-M. (2004). A concerted effort to understand the ambient particulate matter in the LowerFraser Valley: the Pacific 2001 Air Quality Study. Atmospheric Environment. The Pacific2001 Air Quality Study, 38 (34), 5719–5731.Lu, R., Turco, R. P. (1996). Ozone distributions over the Los Angeles basin: Three-dimensionalsimulations with the SMOG model. Atmospheric Environment, 30 (24), 4155–4176.Lu, R., Turco, R. (1995). Air pollutant transport in a coastal environment – II. 3-Dimensionalsimulations of sea-breeze and mountain effects. Journal of the Atmospheric Sciences,51 (15), 2285–2308.Ma, Y. M., Lyons, T. J. (2003). Recirculation of coastal urban air pollution under a synopticscale thermal trough in Perth, Western Australia. Atmospheric Environment, 37 (4),443–454.McKendry, I. G. (1994). Synoptic circulation and summertime ground-level ozoneconcentrations at Vancouver, British-Columbia. Journal of Applied Meteorology, 33 (5),627–641.McKendry, I. G., Lundgren, J. (2000). Tropospheric layering of ozone in regions of urbanizedcomplex and/or coastal terrain: a review. Progress in Physical Geography, 24 (3), 329–354.McKendry, I. G., Steyn, D. G., Lundgren, J., Hoff, R. M., Strapp, W., Anlauf, K., Froude, F.,Martin, J. B., Banta, R. M., Olivier, L. D. (1997). Elevated ozone layers and verticaldown-mixing over the Lower Fraser Valley, BC. Atmospheric Environment, 31 (14),2135–2146.88ReferencesMcKendry, I. G., Steyn, D. G., Banta, R. M., Strapp, W., Anlauf, K., Pottier, J. (1998a).Daytime photochemical pollutant transport over a tributary valley lake in southwesternBritish Columbia. Journal of Applied Meteorology, 37 (4), 393–404.McKendry, I. G., Steyn, D. G., O’Kane, S., Zawar-Reza, P., Heuff, D. (1998b). Lowertropospheric ozone measurements by light aircraft equipped with chemiluminescent sonde.Journal of Atmospheric and Oceanic Technology, 15 (1), 136–143.Millan, M. M., Artinano, B. (1992). Mesometeorological cycles of air pollution in the IberianPeninsula, Contract EV4V-0097-E: Report. Commission of the European Communities.Millan, M. M., Mantilla, E., Salvador, R., Carratala, A., Sanz, M. J., Alonso, L., Gangoiti,G., Navazo, M. (2000). Ozone cycles in the western Mediterranean basin: Interpretationof monitoring data in complex coastal terrain. Journal of Applied Meteorology, 39 (4),487–508.Miller, S. T. K., Keim, B. D., Talbot, R. W., Mao, H. (2003). Sea breeze: Structure, forecasting,and impacts. Reviews of Geophysics, 41 (3), 1011.Moisseeva, N. (2011). Episodic Ozone Pollution in the Lower Fraser Valley, BC: A Tale of ThreeEpisodes. Undergraduate Honours Thesis. The University of British Columbia, Departmentof Earth and Ocean Sciences. Available from: http://circle.ubc.ca/handle/2429/35067(retrieved 08/02/2011).Oke, T. R., Hay, J. E. (1998). The climate of Vancouver. The University of British ColumbiaDepartment of Geography.Papanastasiou, D. K., Melas, D. (2009). Climatology and impact on air quality of sea breezein an urban coastal environment. International Journal of Climatology, 29 (2), 305–315.Perez, I. A., Sanchez, M. L., García, M. A., Pardo, N. (2012). Spatial analysis of CO2concentration in an unpolluted environment in northern Spain. Journal of EnvironmentalManagement, 113, 417–425.Pryor, S. C., Steyn, D. G. (1995). Hebdomadal and diurnal cycles in ozone time-series fromthe Lower Fraser Valley, BC. Atmospheric Environment, 29 (9), 1007–1019.Pryor, S. C., McKendry, I. G., Steyn, D. G. (1995). Synoptic-scale meteorological variabilityand surface ozone concentrations in Vancouver, British-Columbia. Journal of AppliedMeteorology, 34 (8), 1824–1833.Rolph, G. D., Draxler, R. R. (1990). Sensitivity of three-dimensional trajectories to thespatial and temporal densities of the wind field. Journal of Applied Meteorology, 29 (10),1043–1054.Seagram, A. F., Steyn, D. G., Ainslie, B. (2013). Modelled recirculation of air pollutants duringozone episodes in the Lower Fraser Valley, B. C. Air Pollution Modeling and Its ApplicationXXII. Ed. by D. G. Steyn, P. J. H. Builtjes, R. M. A. Timmermans. NATO Science forPeace and Security Series Series C: Environmental Security. The Netherlands: Springer,291–295.89ReferencesSegal, M., Pielke, R. A., Mahrer, Y. (1982). Evaluation of onshore pollutant re-circulation overthe Mediterranean coastal area of Central Israel. Israel Journal of Earth Sciences, 31 (1),39–46.Shair, F., Sasaki, E., Carlan, D., Cass, G., Goodin, W., Edinger, J., Schacher, G. (1982).Transport and dispersion of airborne pollutants associated with the Land Breeze Sea Breezesystem. Atmospheric Environment, 16 (9), 2043–2053.Shan, W., Yin, Y., Zhang, J., Ji, X., Deng, X. (2009). Surface ozone and meteorologicalcondition in a single year at an urban site in central-eastern China. EnvironmentalMonitoring and Assessment, 151 (1-4), 127–141.Sillman, S (1993). Tropospheric ozone: The debate over control strategies. Annual Review ofEnergy and the Environment, 18 (1), 31–56.Skamarock, W. C., Klemp, J. B., Dudhia, J., Gill, D. O., Barker, D. M., Duda, M. G., Huang,X.-Y., Wang, W., Powers, J. G. (2008). A Description of the Advanced Research WRFVersion 3. NCAR/TN-475+STR. Technical Report. Boulder, Colorado, USA: NationalCenter for Atmospheric Research: Mesoscale and Microscale Meteorology Division, p. 113.St. John, J. C., Chameides, W. L. (1997). Climatology of ozone exceedences in the AtlantaMetropolitan Area: 1-hour vs 8-hour standard and the role of plume recirculation airpollution episodes. Environmental Science & Technology, 31 (10), 2797–2804.Steyn, D. G. (1998). Scaling the vertical structure of sea breezes. Boundary-Layer Meteorology,86 (3), 505–524.Steyn, D. G. (2003). Mesoscale circulations and regional air pollution. Air Pollution Processesin Regional Scale. Ed. by D. Melas, D. Syrakov. NATO Science Series 30. SpringerNetherlands, 281–294.Steyn, D. G., Faulkner, D. A. (1986). The climatology of sea-breezes in the Lower FraserValley, BC. Climatological Bulletin, 20 (3), 21–39.Steyn, D. G., Bottenheim, J. W., Thomson, R. B. (1997). Overview of tropospheric ozone inthe Lower Fraser Valley, and the Pacific ’93 field study. Atmospheric Environment, 31 (14),2025–2035.Steyn, D. G., Ainslie, B., Reuten, C., Jackson, P. L. (2013). A retrospective analysis of ozoneformation in the Lower Fraser Valley, British Columbia, Canada. Part I: Dynamical modelevaluation. Atmosphere-Ocean, 51 (2), 153–169.Steyn, D., Ainslie, B., Kaminski, J. W., McConnell, J. C., Martilli, A., Neary, L. (2008). Theuse of meso-scale atmospheric circulation types as a strategy for modelling long-term trendsin air pollution. Air Pollution Modeling and Its Application XIX. Ed. by C. Borrego, A. I.Miranda. NATO Science for Peace and Security Series Series C: Environmental Security.Springer Netherlands, 145–153.Steyn, D. G., Ainslie, B., Reuten, C., Jackson, P. L. (2011). A Retrospective analysis of ozoneformation in the Lower Fraser Valley, B. C. Technical Report. The University of British90ReferencesColumbia, Department of Earth and Ocean Sciences. Available from: http://circle.ubc.ca/handle/2429/36587 (retrieved 08/09/2011).Steyn, D. G., Roberge, A. C., Jackson, C., Branch, B. C. W. M. (1990). Anatomy of an extendedair pollution episode in British Columbia’s Lower Fraser Valley. Technical Report. TheUniversity of British Columbia, Department of Geography.Stohl, A. (1998). Computation, accuracy and applications of trajectories - A review andbibliography. Atmospheric Environment, 32 (6), 947–966.Stohl, A, Haimberger, L, Scheele, M. P., Wernli, H (2001). An intercomparison of results fromthree trajectory models. Meteorological Applications, 8 (2), 127–135.Stohl, A., Wotawa, G., Seibert, P., Kromp-Kolb, H. (1995). Interpolation errors in Wind Fieldsas a Function of Spatial and Temporal Resolution and Their Impact on Different Types ofKinematic Trajectories. Journal of Applied Meteorology, 34 (10), 2149–2165.Strawbridge, K. B., Snyder, B. J. (2004a). Daytime and nighttime aircraft lidar measurementsshowing evidence of particulate matter transport into the Northeastern valleys of the LowerFraser Valley, BC. Atmospheric Environment, 38 (34), 5873–5886.Strawbridge, K. B., Snyder, B. J. (2004b). Planetary boundary layer height determinationduring Pacific 2001 using the advantage of a scanning lidar instrument. AtmosphericEnvironment, 38 (34), 5861–5871.Tov, D. A., Peleg, M., Matveev, V., Mahrer, Y., Seter, I., Luria, M. (1997). Recirculation ofpolluted air masses over the east Mediterranean coast. Atmospheric Environment, 31 (10),1441–1448.Vingarzan, R., Taylor, B. (2003). Trend analysis of ground level ozone in the greaterVancouver/Fraser Valley area of British Columbia. Atmospheric Environment, 37 (16),2159–2171.Wang, T., Kwok, J. Y. H. (2003). Measurement and analysis of a multiday photochemicalsmog episode in the Pearl River delta of China. Journal of Applied Meteorology, 42 (3),404–416.Warner, T. T. (2010). Numerical Weather and Climate Prediction. Cambridge UniversityPress.Zabkar, R., Rakovec, J., Gabersek, S. (2008). A trajectory analysis of summertime ozonepollution in Slovenia. Geofizika, 25 (2), 179–202.91A Station locationsA.1 Meteorological stationsTable A.1. Meteorological station locations within the Lower Fraser Valley and surroundingarea. The following abbreviations are used: Automatic Weather Reporting System (AWRS),International Airport (IA), Meteorological Aeronautical Presentation System (MAPS). Alllocations are in British Columbia, Canada.Stn Latitude LongitudeElevationLocation(m asl)YXX 49.033 -122.367 58 Abbotsford IAYVR 49.183 -123.167 2 Vancouver IAYHE 49.367 -121.483 39 HopeWEZ 48.783 -123.050 7 Saturna Island MAPSYYJ 48.650 -123.433 19 Victoria IAYQQ 49.717 -124.900 24 ComoxWEL 49.217 -123.800 5 Entrance Island AWRSWQC 49.250 -124.833 53 Port AlberniWSP 48.383 -123.917 21 Sheringham Point AWRSWGP 50.300 -122.733 204 Pemberton AWRSYDC 49.467 -120.510 700 Princeton AirportYPW 49.833 -124.500 130 Powell River AirportYAZ 49.083 -125.767 24 Tofino AirportWDR 48.420 -123.230 17 Discovery IslandYBL 49.950 -125.267 106 Campbell River AirportWLY 50.233 -121.583 229 LyttonWEB 49.383 -126.550 7 Estevan PointWVF 49.100 -123.300 1 Sand Heads CS92A.2. Air quality monitoring stationsA.2 Air quality monitoring stationsTable A.2. Select air quality monitoring stations in the Lower Fraser Valley.Station ID Latitude Longitude Location12 49.156 -121.94 Chilliwack15 49.133 -122.694 Surrey East29 49.370 -121.500 Hope International Airport31 49.190 -123.152 Vancouver International Airport93B Meteorological modellingB.1 WRF model configurationDetails of the WRF model configuration for the three nested grids for all modelled episodesare provided in Table B.1. Discretized topography used in the WRF 4 km domain is plottedin Figure B.1a.Table B.1. WRF model domain configuration used for all modelled episodes.DomainGrid spacing Grid squareNotes(km) dimensionsd01 36 100 × 100 Initialized at 1800 UTC with 32 km NARR fieldsd02 12 76 × 97d03 4 178 × 109 Model evaluation; input to HYSPLIT modeld04 1.33 73 × 124 2006 episode simulation only94B.1. WRF model configurationFigure B.1. Topography of the LFV. All elevations are in m asl. (a) Topography used inWRF for the 4 km domain. (b) Real topography of the LFV, in 600 m contours.95B.2. Model evaluationB.2 Model evaluationIt is imperative to determine if the WRF model output is fit for the purposes of this study:to study atmospheric transport and recirculation in the LFV during ozone episodes, using atrajectory model.The first step was to determine how well the modelled wind timeseries of speed anddirection agreed with their corresponding observational timeseries, since the modelled windfields must be sufficiently accurate for use in offline modelling to realistically capture pollutant(tracer) advection. The wind fields were interpolated to the meteorological station locationsin the LFV (Table A.1), and then compared to their corresponding observational time series.Steyn et al. (2011) (see also Steyn et al. 2013) performed this part of the evaluation for allseven episodes, and the resulting statistical scores were mainly satisfactory. There was somedifficulty modelling the winds during the 1993 episode, which can be attributed to unstableatmospheric conditions, and the wind directions during the 1998 episode. The WRF modelsystematically underestimated wind speeds over all episodes. For example, the three dailyhodographs for the 2006 episode (Figure 2.5) highlight the ability of WRF to capture winddirections, but underestimate wind speeds by up to 4 m/s. Further evaluation results for the4 km domain of the 2006 episode can be found in Table B.2 and B.3.The second step was to determine if the resolution of the innermost WRF model domain,at 4 km, could capture the small-scale thermo-topographic wind circulations that likely occurduring summertime ozone episodes. When configuring a meteorological model, the horizontalgrid spacing should be chosen to adequately represent the smallest meteorological features ofinterest; at least 5-10 grid cells are usually required to do so (e.g. Warner 2010). Duringozone episodes, the presence of a SB may require ≤ 4 km grid spacing in order to captureits small-scale features and dynamics. To address whether a finer horizontal resolution wasrequired, the evaluation statistics of the 4 km domain were compared to those of a 1.33 kmdomain for the 2006 episode (also modelled by Steyn et al. 2011; see Figure B.2 and TableB.1).96B.2. Model evaluationTable B.2. Model evaluation results for u and v wind component directions of the 4 kmWRF domain for the 2006 episode, where O are observations, M are modelled values, σ isthe standard deviation, VCC is the Vector Correlation Coefficient, and N is the number ofpoints in the time series. See Table A.1 for coordinates of each station (Stn). The stationsshaded in gray are within the 4 km WRF domain, but not within the 1.33 km domain.Stn N 〈Ou〉 〈Mu〉 〈Ov〉 〈Mv〉 σOu σOv σMu σMv VCCm/s m/s m/s m/s m/s m/s m/s m/sYXX 77 1.19 1.44 0.88 0.38 1.57 0.91 1.91 1.16 0.63YVR 95 4.12 4.06 -3.15 -2.01 2.59 2.23 2.56 1.38 0.71YHE 84 0.72 0.77 -0.31 0.35 4.00 2.37 1.26 0.93 0.66WEZ 94 1.32 3.50 -0.58 -3.91 1.13 2.06 1.44 2.99 0.36YYJ 83 -0.78 0.61 0.22 -1.22 2.10 1.13 1.16 1.74 0.54YQQ 96 1.60 1.29 -3.98 -3.43 1.67 1.01 2.38 1.75 0.69WEL 97 7.13 7.85 -4.09 -4.03 3.06 3.31 2.58 1.44 0.76WQC 80 -0.19 -1.64 0.23 -0.54 0.90 1.11 0.85 1.30 0.15WSP 87 1.55 1.67 -1.25 -0.65 4.18 2.96 1.67 1.20 0.68WGP 81 0.60 -0.08 -0.04 -0.48 0.92 0.45 0.46 0.76 0.39YDC 65 0.30 0.08 -0.06 -0.26 1.46 0.88 1.15 0.74 0.23YPW 74 3.57 3.07 -0.48 -1.76 1.43 1.28 0.68 1.09 0.71YAZ 46 2.86 2.78 -0.00 -1.11 2.62 2.03 1.87 1.39 0.64WDR 90 1.51 1.22 -1.18 -0.56 2.09 1.81 2.41 2.76 0.63YBL 60 1.32 1.07 -2.49 -1.93 3.10 1.44 2.02 1.26 0.71WLY 73 -0.97 0.08 1.76 1.29 0.58 0.51 2.10 2.21 0.27WEB 97 8.18 4.73 -3.47 -3.79 2.27 3.03 1.38 2.19 0.55WVF 97 5.80 6.08 -3.03 -3.44 4.18 3.15 2.41 2.45 0.7897B.2.ModelevaluationTable B.3. Model evaluation results for wind speeds of the 4 km WRF domain for the 2006 episode, where O are observations,M are modelled values, σ is the standard deviation, N is the number of data points in the time series; m and b are the slope andintercept of the line of best fit between the O and M values, respectively; R2 is the coefficient of determination; RMSE is theRoot-Mean-Square Error, with systematic (RMSEs) and unsystematic (RMSEu) components; IOA is the Index of Agreement,MAE is the Mean Absolute Error, NMAE, is the Normalized Mean Absolute Error, MBE is the Mean Bias Error, and NMBEis the Normalized Mean Bias Error. See Table A.1 for coordinates of each station (Stn). The stations shaded in gray are withinthe 4 km WRF domain, but not within the 1.33 km domain.Stn N 〈O〉 〈M〉 σO σM m b R2 RMSE RMSEs RMSEu IOA MAE NMAE MBE NMBEm/s m/s m/s m/s m/s m/s m/s m/s m/s m/s m/s m/sYXX 97 2.04 1.82 1.55 0.88 0.27 1.28 0.47 1.39 1.16 0.78 0.62 1.13 55.22 0.22 10.76YVR 97 5.70 4.79 2.62 2.00 0.68 0.93 0.89 1.55 1.24 0.93 0.88 1.27 22.30 0.90 15.87YHE 97 3.51 2.37 1.85 1.21 0.39 1.01 0.59 1.89 1.61 0.98 0.57 1.55 44.21 1.14 32.58WEZ 97 2.09 5.63 0.94 2.82 0.93 3.69 0.31 4.44 3.54 2.68 -0.55 3.74 178.55 -3.54 -168.96YYJ 97 2.02 2.18 1.19 1.17 0.11 1.95 0.12 1.58 1.07 1.16 0.43 1.24 61.32 -0.16 -7.76YQQ 97 4.59 3.85 2.34 1.60 0.52 1.44 0.77 1.69 1.34 1.02 0.80 1.28 27.96 0.74 16.20WEL 97 8.51 8.88 3.34 3.51 0.83 1.80 0.79 2.24 0.67 2.14 0.88 1.69 19.85 -0.37 -4.38WQC 97 0.94 2.48 0.68 0.92 0.31 2.19 0.23 1.85 1.61 0.90 -0.53 1.63 173.20 -1.54 -164.37WSP 97 3.76 3.13 2.75 1.87 0.42 1.53 0.62 2.25 1.71 1.46 0.74 1.89 50.29 0.64 16.90WGP 97 0.91 0.86 0.60 0.62 0.61 0.31 0.59 0.55 0.24 0.50 0.77 0.44 48.09 0.05 5.13YDC 97 1.00 1.23 1.17 0.74 0.11 1.12 0.17 1.29 1.07 0.73 0.45 1.03 102.84 -0.23 -22.70YPW 97 2.78 3.37 2.02 1.44 0.51 1.95 0.72 1.53 1.15 1.01 0.78 1.23 44.18 -0.59 -21.17YAZ 53 3.41 2.75 2.11 2.36 0.75 0.19 0.67 1.94 0.84 1.74 0.79 1.56 45.72 0.66 19.28WDR 97 3.10 3.14 1.78 1.59 0.24 2.39 0.27 2.04 1.35 1.53 0.57 1.67 53.73 -0.04 -1.13YBL 65 3.87 2.67 2.24 0.99 0.30 1.50 0.68 2.10 1.97 0.72 0.53 1.60 41.40 1.20 31.08WLY 97 1.89 1.98 1.74 1.98 0.15 1.69 0.14 2.45 1.48 1.96 0.45 1.85 97.64 -0.09 -4.87WEB 97 8.96 6.20 2.39 3.52 1.13 -3.89 0.77 3.58 2.78 2.26 0.59 2.96 33.07 2.76 30.79WVF 97 7.18 7.35 3.81 3.28 0.77 1.79 0.90 1.68 0.88 1.44 0.94 1.21 16.85 -0.17 -2.3398B.2. Model evaluationAt all stations, the Vector Correlation Coefficient (VCC) (Crosby et al. 1993) of the windvectors either remained the same or decreased when the 1.33 km domain was used (TableB.4). For wind speeds, differences between the 4 km and 1.33 km evaluation results werenot uniform across the domain (Table B.5). While wind speeds at inland locations (YXX,YVR, YHE) showed a small decrease in the Root-Mean-Square Error (RMSE), those locatedadjacent to or in the Strait of Georgia (WEZ, YYJ, WVF) showed considerable increases.It is unclear whether the increase in RMSE at these coastal and marine stations should beattributed to increased grid resolution or to possible spurious numerical errors introduced bythe proximity of the edges of the computational domain (Figure B.2).+49°+50°-124° -123° -122° -121°WABCYXXYVRYHEWEZYYJWVFFigure B.2. Inner 1.33 km domain (gray rectangle) in WRF used to model the 2006episode. Meteorological stations are represented by red triangles. See Table A.1 for fullstation names.99B.2. Model evaluationTable B.4. As in Table B.2, but for the 1.33 km WRF domain.Stn N 〈Ou〉 〈Mu〉 〈Ov〉 〈Mv〉 σOu σOv σMu σMv VCCm/s m/s m/s m/s m/s m/s m/s m/sYXX 77 1.19 1.37 0.88 0.42 1.57 0.87 1.91 1.12 0.62YVR 95 4.12 4.38 -3.15 -2.22 2.59 2.31 2.56 1.55 0.71YHE 84 0.72 1.35 -0.31 -0.81 4.00 3.78 1.26 1.10 0.66WEZ 94 1.32 4.57 -0.58 -3.65 1.13 2.26 1.44 2.38 0.32YYJ 83 -0.78 0.76 0.22 -1.24 2.10 1.10 1.16 1.64 0.44WVF 97 5.80 6.45 -3.03 -3.58 4.18 2.95 2.41 2.43 0.78Overall, the model evaluation revealed that a finer horizontal grid resolution (1.33 km)is not necessarily better at resolving winds and associated small-scale dynamic structures inthe LFV. Compared to the evaluation results of the 4 km domain, the 1.33 km modelledfields only modestly improved wind speed accuracy at inland stations, but decreased windspeed accuracy at coastal sites, and decreased wind direction accuracy at all stations. Hence,the WRF fields modelled at 4 km horizontal spacing were used to investigate transport andpossible recirculation during ozone episodes.100B.2.ModelevaluationTable B.5. As in Table B.3, but for the 1.33 km WRF domain.Stn N 〈O〉 〈M〉 σO σM m b R2 RMSE RMSEs RMSEu IOA MAE NMAE MBE NMBEm/s m/s m/s m/s m/s m/s m/s m/s m/s m/s m/s m/sYXX 97 2.04 1.66 1.55 0.90 0.33 0.98 0.58 1.32 1.10 0.73 0.67 1.05 51.58 0.38 18.60YVR 97 5.70 5.14 2.62 2.23 0.74 0.90 0.88 1.39 0.87 1.08 0.91 1.10 19.26 0.56 9.83YHE 97 3.51 3.84 1.85 1.62 0.42 2.36 0.48 1.81 1.12 1.42 0.67 1.51 43.04 -0.33 -9.39WEZ 97 2.09 6.08 0.94 2.60 0.46 5.13 0.16 4.77 4.02 2.56 -1.11 4.21 201.08 -3.99 -190.58YYJ 97 2.02 2.01 1.19 1.33 0.16 1.69 0.14 1.66 1.01 1.32 0.45 1.37 67.93 0.01 0.56WVF 97 7.18 7.68 3.81 3.16 0.73 2.46 0.88 1.91 1.16 1.52 0.92 1.33 18.59 -0.50 -6.92101B.3. Assessment of thermo-topographic circulationsB.3 Assessment of thermo-topographic circulationsTo determine whether thermo-topographic circulations occurred during the modelled episodedays, an assessment of the WRFmodelled fields was done both quantitatively and qualitatively.Identifying these circulations is necessary to further understand the complexity of the LFV’swind patterns and resulting trajectory pathways.B.3.1 Land-sea breezeSteyn and Faulkner (1986) designed an objective set of criteria to detect SBs in the LFV. This“SB filter” is based on observed meteorological data taken from YVR and YXX. To separateSB days from non-SB days, there are three categories of criteria that must be passed:(1) “Wind requirement”: There must be a diurnal reversal of flow observed at the surface,such that:(a) The majority of hourly winds from 0300-1000 (local) are offshore.(b) The winds are onshore for at least two consecutive hours in the period 0800-2000.(c) The wind did not blow onshore for a majority of hours from 2100-0100.The range of directions designated to onshore flow varied by station location (Table B.6).Table B.6. Onshore and offshore wind direction ranges by location, determined by Steynand Faulkner (1986).Stn onshore offshoreYVR 210◦ – 320◦ 30◦ – 150◦YXX 170◦ – 250◦ 0◦ – 120◦(2) “Thermal requirement”:The thermal forcing must overcome the synoptic forcing, since SBs are thermally-drivencirculations. This is determined by the Lake-Breeze Index (Biggs and Graves102B.3. Assessment of thermo-topographic circulations1962), which gives a balance of inertial (represented by wind speeds) versus buoyant(represented by land-sea temperature contrast) forces:LBindex =U2cp ∆T< 3.0∆T = Tland − TSSTwhere cp = 1003 J kg−1 K−1 is the specific heat of air at constant pressure, Tlandis the highest maximum daily temperature at YVR or YXX, TSST is the sea surfacetemperature (SST) of the Strait of Georgia, and U is the mean wind speed during the3 h period prior to reversal from offshore to onshore. Monthly average SSTs are usedwhen observations from a buoy station at Porlier Pass ((λ, φ) = (−123.4177,+49.1190))are not available (Table B.7).Table B.7. Southern Strait of Georgia mean monthly sea surface temperature (SST) forsummer months. Taken from Steyn and Faulkner (1986).Month TSST (◦C)June 12.2July 13.9August 13.3(3) “Sunshine requirement”:There must be at least a total of 6 h of sunshine hours observed between YVR and YXX.This requirement also ensures that circulations are thermally driven.All three criteria must be satisfied for a day to be considered a SB day. For each episodeday, the SB filter was applied to the modelled timeseries of wind speed and wind direction(at YVR and YXX), and ambient air temperature (determined from Tland). The SSTs wereobtained from WRF model output as well, and interpolated to Porlier Pass. The SST waskept constant during the WRF model runs, thus a daily average could not be determined. Toassure that this would not pose any shortcomings, the monthly average values of SST in the103B.3. Assessment of thermo-topographic circulationsStrait of Georgia (Table B.7) were also used for comparison. Finally, WRF does not outputa field for sunshine hours. Instead, the downwelling shortwave radiation at the surface (2 m),K↓, was used as a surrogate measure. To determine if it is sunny at a given hour (versus cloudyand/or nighttime), the modelled K↓ interpolated to YVR and YXX must pass a threshold ofK↓ ≥ 375 W m−2. This threshold is representative of daytime K↓ values in Vancouver (Okeand Hay 1998). A range of thresholds was tested (200 W m−2 ≤ K↓ ≤ 470 W m−2), whichdid not affect overall results.By applying the SB filter, all modelled days were identified as non-SB days (Table B.8).Using either the interpolated SST or the historical monthly average SST did not have any effecton the thermal criterion or overall results. The diurnal wind reversal requirement presentedthe most stringent criteria for identifying SB days (Figure B.3). For the most part, winddirections were consistently onshore during the day, but did not switch from or to offshore atnight. Again, the modelled daily hodographs during the 2006 episode (Figure 2.5) typify thisbehaviour. If the wind did not switch from off- to onshore, the LBindex could not be calculatedsince U is obtained before the reversal.1a 1b 1c 2 336912151821#020406080100%Figure B.3. Frequency of criterion pass of the SB filter. The criterion (labelled on theabscissa) are denoted as in Table B.8. Note: total possible passes is N = 21 (21 daysevaluated).It is evident that there was persistent daytime onshore flow during all episode days at YVR,and winds usually remained onshore at nighttime as well. Hence, a diurnal wind reversal did104B.3. Assessment of thermo-topographic circulationsTable B.8. SB filter results at YVR for all episodes. A pass (1) or fail (0) of each ofthe three criterion are given. The criterion are: (1a) A majority of nighttime (0300 – 1000local) winds are offshore, (1b) Two consecutive hours of daytime winds are onshore, (1c) themajority of winds directions are not onshore shortly after sunset (2100 – 0100 local), (2)The LBindex is greater than 3.0, and (3) There are at least a total of 6 hours of sunshine asobserved between YVR and YXX. The column “SB” indicates if an SB was detected.Episode Day(1)(2) (3) SBa b c19851 0 1 0 0 1 02 0 1 1 1 1 03 0 1 1 1 1 019871 1 0 1 0 1 02 1 0 1 0 0 03 0 1 0 1 1 019931 0 1 0 0 1 02 0 1 0 0 1 03 0 1 0 1 1 019951 0 1 0 0 1 02 0 1 0 0 1 03 0 1 1 1 1 019981 0 1 0 1 1 02 0 1 0 1 1 03 1 1 0 1 1 020011 0 1 1 0 1 02 1 1 0 1 1 03 0 1 1 0 1 020061 0 1 0 1 1 02 0 1 0 0 1 03 0 1 0 0 1 0105B.3. Assessment of thermo-topographic circulationsnot occur. Because of this, none of the episode days were ‘true’ SB days. It is possible thatthe criteria for identifying SB days is too restrictive, but it is more likely that the expectednighttime offshore winds were masked by strong onshore flows channelled between VancouverIsland and the LFV mainland.B.3.2 Mountain-valley breeze and slope flowsThere exists no formal set of objective criteria (such as the SB filter of Steyn and Faulkner1986) to determine the occurrence of mountain-valley breezes or slope flows. As such, themodelled surface (10 m) wind fields for all episode days, output at 6 h intervals (0000, 0600,1200, 1800 local time) were qualitatively examined to detect the presence of these flows.Given the slack synoptic conditions and non-SB conditions (Section B.3.1) during all episodes,identifying wind reversals in tributary valleys is likely sufficient to identify mountain-valleybreezes and slope flows. Four general regions of the LFV were examined for these flows: (1)the western portion of the North Shore Mountains, from Howe Sound to Pitt River Valley, (2)the eastern portion of the North Shore Mountains, including Stave Lake and Harrison Lakevalley, (3) the narrow, eastern portion of the LFV, extending roughly from Chilliwack to Hope,and (4) the main valley bottom of the LFV, from the coast to mid-valley (these regions aredemarcated in Figure B.4, top-left panel).For illustrative purposes, the modelled surface (10 m) wind fields for the second day(August 11) of the 2001 episode are shown in Figure B.4. At 0000, wind speeds are generallylight throughout the entire LFV, with complicated up and down-valley flows (with respectto individual valleys). By 0600, wind speeds remain light, with down-valley flows apparentfrom the North Shore Mountains (regions 1 and 2). There is also flow divergence in the mainvalley bottom of the LFV (region 4) at this time. At 1200, the wind directions are evidentlyup-valley and upslope for all tributary valleys along the northern edge of the LFV (regions1 and 2), and up-valley from the main valley bottom towards Hope (region 3). This patterncontinues to 1800, when wind speeds increase and the flows in the main valley bottom takeon a more westerly component.106B.3. Assessment of thermo-topographic circulations1 234(a) 0000 (b) 0600-124° -123° -122° -121°+49°+49.5°(c) 120010 m/s(d) 1800Figure B.4. WRF modelled surface (10 m) winds on the second day of the 2001 episode.Wind fields are provided at 0000, 0600, 1200, and 1800 local time. Focus regions 1 – 4 aredelineated by red dashed lines (see text).107B.3. Assessment of thermo-topographic circulationsFor comparison, the same model output for the second day (June 25) of the 2006 episodeis shown in Figure B.5. A similar complicated flow pattern at 0000 is observed, with areas ofconvergent and divergent flow in the main valley bottom (region 4). At 0600, wind speeds arestrong along the North Shore Mountains, with distinct downslope and down-valley flows fromall tributary valleys and from Hope (regions 1, 2, and 3). The winds weaken by 1200, whereweak upslope and upvalley flows are evident in the western tributary valleys (region 1), butdown-valley flows persist within the Harrison Lake valley and from Hope (regions 2 and 3).The winds in the main valley bottom are light and westerly, causing an area of convergencenear Chilliwack. By 1800 the flow pattern is similar to that of the second day of the 2001episode, where all tributary valleys exhibit up-valley flows, and there is up-valley flow fromthe main valley bottom towards Hope (regions 3 and 4).From the above, it is evident that both mountain-valley breeze and slope flow circulationsoccur during the modelled episode days. Similar results were observed for all other modelleddays (not shown). The two examples above highlight the temporal and spatial variability ofwind direction switch from down- to up-valley flows in the tributary valleys.108B.3. Assessment of thermo-topographic circulations1 234(a) 0000 (b) 0600-124° -123° -122° -121°+49°+49.5°(c) 120010 m/s(d) 1800Figure B.5. As in B.4, but for the second day of the 2006 episode.109C Additional material for trajectory analysisC.1 HYSPLIT modelling resultsComposite trajectory plots (end ‘snapshots’) of all 72 trajectories modelled for each episodeare provided in Figures C.1 – C.6 below. They reveal the total spatial distribution of thetrajectories over the course of the entire modelled period.110C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.1. Modelled trajectories during the 1987 episode. Trajectories starting at 0000and 12000 local time are highlighted in blue and red, respectively. Trajectory positions aremarked every 6 h (black points). (top) The origin of the trajectories is denoted by the blacktriangle. (bottom) The heights of the trajectories within the mapped domain are plotted assolid lines. The heights of the trajectories extending past the plotted domain are indicatedby dashed lines.111C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.2. As in Figure C.1, but for trajectories during the 1993 episode.112C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.3. As in Figure C.1, but for trajectories during the 1995 episode.113C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.4. As in Figure C.1, but for trajectories during the 1998 episode.114C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.5. As in Figure C.1, but for trajectories during the 2001 episode.115C.1. HYSPLIT modelling results-124° -123° -122°+49°+49.5°0 12 24 36 48 60 72Age (h)0500100015002000Height (m agl)00 12 00 12 00 12 00Hour (local)Figure C.6. As in Figure C.1, but for trajectories during the 2006 episode.116C.2. Number density mapsC.2 Number density mapsNumber density maps were generated by counting trajectories as they passed through each ofthe 4 km grid boxes of the innermost WRF domain (Figures C.7 – C.12). Trajectories wereseparated by day in order to isolate trajectories associated with one circulation regime fromanother. Trajectories were counted over all heights.-124° -123° -122° -121°+49°+49.5°(a) Day 1 - I (b) Day 2 - IV (c) Day 3 - IVFigure C.7. The number density maps by day (circulation regime) of the 1985 episode.The origin of the trajectories is represented by the black triangle. The colormap scale is asin Figure 3.9. The circulation regime of each day is indicated above each plot.-124° -123° -122° -121°+49°+49.5°(a) Day 1 - I (b) Day 2 - I (c) Day 3 - IFigure C.8. As in Figure C.7, but for days of the 1993 episode.117C.2. Number density maps-124° -123° -122° -121°+49°+49.5°(a) Day 1 - III (b) Day 2 - III (c) Day 3 - IIIFigure C.9. As in Figure C.7, but for days of the 1995 episode.-124° -123° -122° -121°+49°+49.5°(a) Day 1 - II (b) Day 2 - III (c) Day 3 - IIFigure C.10. As in Figure C.7, but for days of the 1998 episode.-124° -123° -122° -121°+49°+49.5°(a) Day 1 - II (b) Day 2 - II (c) Day 3 - IIFigure C.11. As in Figure C.7, but for days of the 2001 episode.118C.2. Number density maps-124° -123° -122° -121°+49°+49.5°(a) Day 1 - I (b) Day 2 - I (c) Day 3 - IIIFigure C.12. As in Figure C.7, but for days of the 2006 episode.119D Additional material for recirculationanalysisD.1 Episodic temporal characteristicsBelow are the “carryover diagrams” of RTS start and end times, by episode. See Section 4.3for an explanation of the following graphics (Figures D.1 – D.6).120D.1. Episodic temporal characteristicsFigure D.1. Temporal characteristics of detected RTSs during the 1985 episode. (center)Start and end times of RTSs. Note that hours 0, 24, and 48 corresponds to 0000 (local) onthe first, second, and third day of the modelled episodes. The green shaded area representsperiods of time where a RTS started on one day of the episode, and returned on the next(“carryover”). The axis at 45◦ (dashed gray lines) are the RTS lifetimes, τR (h). (top)Histogram of RTS start times. (right) Histogram of RTS end times. The circulationregimes of each day of the episode are labelled in each histogram, and each day is separatedby a dotted black line. Rush hour periods (0600-0900) are highlighted in black on bothhistograms.121D.1. Episodic temporal characteristics0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour12243648604IV IV IV4IVIVIVFigure D.2. As in Figure D.1, but for temporal characteristics of detected RTSs duringthe 1987 episode.122D.1. Episodic temporal characteristics0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour12243648604I I I8IIIFigure D.3. As in Figure D.1, but for temporal characteristics of detected RTSs duringthe 1993 episode123D.1. Episodic temporal characteristics0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour12243648606III III III8IIIIIIIIIFigure D.4. As in Figure D.1, but for temporal characteristics of detected RTSs duringthe 1995 episode124D.1. Episodic temporal characteristics0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour12243648606II III II10IIIIIIIFigure D.5. As in Figure D.1, but for temporal characteristics of detected RTSs duringthe 1998 episode125D.1. Episodic temporal characteristics0 6 12 18 24 30 36 42 48 54 60 66 72Start hour061218243036424854606672End hour12243648606II II II10IIIIIIFigure D.6. As in Figure D.1, but for temporal characteristics of detected RTSs duringthe 2001 episode126D.2. Relationship between diurnal wind cycle and recirculation lifetimeD.2 Relationship between diurnal wind cycle and recirculationlifetimeFor illustrative purposes, suppose the daily wind cycle is a perfectly sinusoidal, one-dimensionalwave (repeating infinitely in time; Figure D.7a), whose period is 24 h and whose phase isshifted such that the maximum onshore (positive) wind speed occurs in the afternoon, and themaximum magnitude of the wind speed (4 m/s) is representative of the maximum wind speedsobserved during ozone circulation regimes (see Figure 2.2). If an air parcel is released from theorigin (x = 0, Figure D.7b) at t = 0900 (red circle), the wind direction will remain onshore(positive u, positive x-direction) for the 12 hours that follow and the parcel will travel onlyin the positive x-direction. As the wind direction reverses, the parcel travels in the negativex-direction. Within 24 h, the parcel will return to the origin. Thus, based on the detectionalgorithm for recirculation, the air parcel recirculates within τR = 24 h (Figure D.7c). Incontrast, if an air parcel is released from the origin at t = 0800 (open blue circle), the winddirection is initially offshore but quickly switches to onshore. Accordingly, the parcel travels inthe negative x-direction for a very small distance before continuing in the positive x-directionfor the remaining 24 h cycle. The lifetime of the recirculation as determined by the detectionalgorithm is then τR = 1 h. When air parcels are released for all possible times within the24 h period, the resulting recirculation lifetimes oscillate in a regular sawtooth-like wave overtime.127D.2. Relationship between diurnal wind cycle and recirculation lifetimeFigure D.7. Idealized effect of diurnal wind cycle on resulting recirculation lifetimes (τR).(a) An idealized sinusoidal wind cycle, in one dimension (u). Two hypothetical trajectorystart times (red and open blue circles) are plotted. (b) The corresponding paths of thetrajectories through space. The maximum and minimum distances from the origin aredenoted by crosshairs and open black circles, respectively. (c) The change of τR over onedaily wind cycle.128


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items