UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluation of a NEMO model of the Strait of Georgia and insights into mixing and transport of the Fraser… Liu, Jie 2017

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

Item Metadata

Download

Media
24-ubc_2017_may_liu_jie.pdf [ 6.48MB ]
Metadata
JSON: 24-1.0343600.json
JSON-LD: 24-1.0343600-ld.json
RDF/XML (Pretty): 24-1.0343600-rdf.xml
RDF/JSON: 24-1.0343600-rdf.json
Turtle: 24-1.0343600-turtle.txt
N-Triples: 24-1.0343600-rdf-ntriples.txt
Original Record: 24-1.0343600-source.json
Full Text
24-1.0343600-fulltext.txt
Citation
24-1.0343600.ris

Full Text

Evaluation of a NEMO model of theStrait of Georgia and insights intomixing and transport of the FraserRiver plumebyJie LiuB.Sc., Ocean University of China, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Oceanography)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2017c© Jie Liu 2017AbstractThe goals of this modeling study of the Fraser River plume, located in the Strait of Georgia,British Columbia, are twofold. Firstly, it aims at improving the Fraser River plume properties byevaluating the model results with various available observations. Secondly, mixing and transportprocesses within the plume, driven by different forcing factors, are investigated with the improvedconfiguration to understand the plume dynamics in the model. The problems found by comparingwith ferry-based salinity data, drifter data and CTD data in the modeled Fraser River properties are:(1) too weak cross-strait velocities; (2) too strong along-strait flows; (3) too salty surface water.To fix the problems, a longer and deeper river channel was created and added into the model.The results show promising improvements with stronger cross-strait motions. Background verticaleddy viscosity was reduced from 1×10−4 to 1×10−5 m2s−1, which tends to reduce the along-straitvelocities. In addition, background vertical eddy diffusivity was reduced to 1×10−6 m2s−1 whichreduced the surface salinity. Furthermore, effects of river discharge, tides, winds and the Coriolisforce are explored on plume mixing and transport. As expected, plume size increases with increasingriver outflow. Tides are important in mixing at the river mouth and inside the river channel duringlow and moderate river flow periods with wind magnitude smaller than 5 m s−1, whereas windsbecome the dominant factor in mixing over almost the entire plume domain when wind speed isgreater than 5 m s−1. The Coriolis force strengths the northward flux across a transect north ofiiAbstractthe river mouth when winds are not strong, resulting in a fresher plume in English Bay, north ofthe City of Vancouver. This thesis provides both a guide to accurately modeling the Fraser Riverplume and insight into plume dynamics.iiiPrefaceThis thesis contains detailed numerical experiments and analysis conducted and undertaken pri-marily by the author, Jie Liu. Susan Allen was my supervisor on this project and contributed sub-stantially on suggesting modeling techniques, interpreting the results and editing the manuscript.This work is previously unpublished, although a manuscript version of Chapter 2 is prepared tosubmit for publication in the future.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 The Strait of Georgia and the Fraser River Estuary-Plume System . . . . . . . . . . 11.2 Past Modeling Studies on the Fraser River Plume . . . . . . . . . . . . . . . . . . . 41.3 The Model Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4 Vertical Turbulence Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.5 Mixing in a River Plume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9vTable of Contents1.6 Transport in a River Plume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.7 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 Plume Sensitivities to Model Parameters and Forcings . . . . . . . . . . . . . . . 172.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.1 The Model Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.2 Strategies in this Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.2.3 Data for Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.2.4 Comparison Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.2.5 Variations Attempted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.2.6 Sensitivity Studies of Plume Properties Affected by Forcing Factors . . . . . 282.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.3.1 Fraser River Plume Sensitivity to Model Parameters . . . . . . . . . . . . . . 312.3.2 Fraser River Plume Sensitivity to Forcing . . . . . . . . . . . . . . . . . . . . 362.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442.4.1 Plume and Estuary Response to Different River Geometries . . . . . . . . . 442.4.2 Effects of Vertical Eddy Viscosity and Diffusivity . . . . . . . . . . . . . . . 452.4.3 Four Important Forcing Mechanisms Reproduced by Numerical Simulations 462.4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 883.1 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 883.2 Implications and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91viTable of ContentsBibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93AppendicesA Inverse Distance Weighting Interpolation Method for Ferry-based Salinity com-parison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100B Tidal Amplitude Comparison Inside the Fraser River Channel . . . . . . . . . . 102C Extended, Deepened River Channel With Jetty . . . . . . . . . . . . . . . . . . . 103D Mean Surface Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105E Froude Number at Peak Floods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108F Coriolis Impact on Surface Salinity in the English Bay . . . . . . . . . . . . . . . 110G Centre Plume Location for the Other Two River Flow Periods . . . . . . . . . . 112H Surface Drifter Tracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115viiList of Tables2.1 List of simulations for evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562.2 List of plume physics simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572.3 Mixing and transport parameters impacted by winds and the Coriolis force for variouswind event periods during three river flow periods. Flux is across the northerntransect (Figure 2.24). Change of plume centre location is calculated with (2.15).Along-strait and cross-strait distance in the near-field region starts from the rivermouth. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58B.1 Tidal amplitude comparisons inside Fraser River between extended river (bathymetry#5), extended and deepened river (bathymetry #6) and observations . . . . . . . . . 102viiiList of Figures1.1 Salt-wedge schematic diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151.2 Observation sites and salinity profiles. . . . . . . . . . . . . . . . . . . . . . . . . . . 162.1 Domain and observation sites. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592.2 From left to right are bathymetry #2, #5 and #6, respectively. . . . . . . . . . . . . 602.3 Thalweg of bathymetry#6 in the model. . . . . . . . . . . . . . . . . . . . . . . . . . 612.4 Model forcing for June 15-29, 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622.5 Model forcing for October 1-31, 2014. . . . . . . . . . . . . . . . . . . . . . . . . . . 632.6 Model forcing for January 1-31, 2016. . . . . . . . . . . . . . . . . . . . . . . . . . . 642.7 Model forcing for May 1-31, 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652.8 Number of occurrence of salinity of 10.5 m depth at NS station. . . . . . . . . . . . . 662.9 Comparison of the 1.5 m depth model results and the ferry-based salinity alongdiagonal and south ferry route on October 8, 2014. . . . . . . . . . . . . . . . . . . . 672.10 Statistics of bias (model - data) of ferry crossing comparisons of minimum salinityvalue and location between modeled results and ferry-based salinity. . . . . . . . . . 682.11 Daily averaged fresh water fraction along the diagonal ferry route. . . . . . . . . . . 692.12 A drifter comparison with modeled surface particles. . . . . . . . . . . . . . . . . . . 70ixList of Figures2.13 Scatter diagram of the observed CTD versus modeled results and cast locations onthe map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712.14 Transect locations, instantaneous volume flux, daily integrated volume flux and saltflux. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722.15 Statistics of averaged distance between the observed drifters and various modeledparticles after one hour. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 732.16 Comparison of modeled particle tracks. . . . . . . . . . . . . . . . . . . . . . . . . . . 742.17 Salinity time series at NS station (Figure 2.1) under moderate wind conditions (5-10 m s−1) during different river discharge periods. . . . . . . . . . . . . . . . . . . . 752.18 Salinity time series at NS station (Figure 2.1) as for Figure 2.17 but under weak windconditions (0-5 m s−1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762.19 Potential energy relative to completely unmixed state (2.11) at NS station (Figure2.1) for different forcing run cases in different river flow periods. . . . . . . . . . . . 772.20 Spatial distribution of Pr (2.7) for low, moderate and high river flow periods duringweak, moderate and strong wind events. . . . . . . . . . . . . . . . . . . . . . . . . . 782.21 Instantaneous salinity along the river transect (Figure 2.2) at high tide in January,October and May. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 792.22 The surface salinity with the Coriolis force, without the Coriolis force and theirdifference (With Coriolis - without Coriolis) at instantaneous peak ebbs. . . . . . . . 802.23 Froude Number, momentum and internal wave speed at instantaneous peak ebb inthree river flow periods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 812.24 Timeseries of cumulative freshwater transport over the upper 10 m across the transectshown in the inset map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82xList of Figures2.25 Daily integral of freshwater flux for without the Coriolis force (nof), combine allforcings (all) and river and tide (nowind) in October, 2014. . . . . . . . . . . . . . . 832.26 Freshwater thickness (2.15) spatial distribution and plume centre location (2.15, star)for January, October and May period for river only case. . . . . . . . . . . . . . . . 842.27 Plume centre location (2.15, star) as well as freshwater thickness (2.15) spatial dis-tribution for weak, moderate and strong wind events in October, 2014. . . . . . . . . 852.28 Background values and modeled values of vertical eddy diffusivity and viscosity av-eraged over October 8 to 10, 2014 at VENUS Central station (Figure 2.1). . . . . . . 862.29 Comparison of the position of the salt-wedge tip in the estuary between predictiveresults and model results for different forcing run cases. . . . . . . . . . . . . . . . . 87C.1 Bathymetry #10 with the Steveston jetty, longer and deeper bathymetry #6 as wellas comparison between the observed drifter and modeled particles. . . . . . . . . . . 104D.1 Mean surface flows of HF radar data (left) and model mean surface currents (right,run #9, Table 2.1) during the same time period in October 1-31, 2015. . . . . . . . . 105D.2 Mean surface currents of HF radar data (green) and model results with improvedconfigurations (red) from June 22-August 25, 2016. . . . . . . . . . . . . . . . . . . 106E.1 Froude Number, momentum and internal wave speed at instantaneous peak flood inthree river flow periods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109F.1 Time series of averaged surface salinity in the English Bay. . . . . . . . . . . . . . . 111G.1 Plume centre location with wind forcing and without wind forcing in January, 2016. 113G.2 Plume centre location with wind forcing and without wind forcing in May, 2015. . . 114xiList of FiguresH.1 47 surface drifter tracks after 2.2 days released at Sand Heads. . . . . . . . . . . . . 115xiiList of AcronymsSoG Strait of GeorgiaSJDF Juan de Fuca StraitGLS Generic Length ScaleCTD Conductivity Temperature DepthIOS Institute of Ocean SciencesNEMO Nucleus for European Modelling of the OceanCGRF Canadian Meteorological Centre’s 33 km Global Atmospheric Reforecasting ModelHRDPS High Resolution Deterministic Prediction SystemROMS Regional Ocean Modeling SystemPOM Princeton Ocean ModelTKE Turbulent Kinetic EnergyxiiiAcknowledgementsI would really like to thank Susan Allen, my supervisor, who gave me such an interesting and well-structured project as well as countless encouragement and suggestions during my masters study.Susan is an organized, well-respected, humorous and knowledgeable professor who always cares themost about her graduate students. I enjoy the regular weekly individual meeting with you, and allthe guidance and assistance that you provide. I also appreciate the many conference opportunitiesI have had that you supported, which both broadened my research horizons and made me moresocial.I would like to thank my committee members, Nancy Soontiens and Mark Halverson. Thank youfor committing your time, enthusiasm and useful suggestions throughout my whole master program.In particular, Nancy, thank you for always being so patient and helpful whenever I turn to you andalways encouraging me no matter how silly the mistake I made.Thank you to Richard Pawlowicz for examining my thesis carefully and the suggestions andcomments during the physical oceanography seminar.I consider myself very lucky to be in the Salish Sea modeling group and thank you to every teammember for discussing the modeling issues during each group meeting and being in such harmony.Thanks to Ben Moore-Maley, for providing your movie of surface currents to help me understand thecirculation pattern and listening to my pre-defense as well as my pre-talk for seminar with patiencexivAcknowledgementsand many helpful tips. Also thanks to you, Doug Latornell, for teaching and fixing Mercurial andweb-related problems for me.Last but not the least, thank you to my enlightened parents for supporting me to study abroadand to my husband Tony for your consistent companionship, encouragement and love.xvChapter 1Introduction1.1 The Strait of Georgia and the Fraser River Estuary-PlumeSystemThe Strait of Georgia (hereafter SoG) is a strongly stratified, semi-enclosed ocean basin locatedbetween Vancouver Island and the mainland coast of British Columbia, Canada, constituting one ofCanada’s major ecological marine environments (Thomson, 1981). It is part of the larger Strait ofGeorgia-Juan de Fuca-Puget Sound system, collectively known as the Salish Sea. The SoG is around220 km long and 28 km wide and 155 m deep on average, with a maximum depth exceeding 400 min the central strait. The SoG connects to the Pacific Ocean via Juan de Fuca Strait (hereafterSJDF) at the south and narrow but relatively long Johnstone Strait at the north. At the south,Haro Strait is the largest passage connecting with SJDF and is a strong tidal mixing region due tothe complex bathymetry and steep topographic slopes (Thomson, 1981; Soontiens et al., 2016).Prevailing winds in the SoG are predominately from the northwest in summer, when clockwisemotion of air around the North Pacific High centred west of California dominates and southeast inwinter, which is driven by the strong anticlockwise flow of air around the Aleutian Low that developsjust south of Alaska. In general, winds are channeled along the SoG by the mountainous terrain(Stronach, 1981; Halverson, 2009). Winds are typically lighter and more variable in direction in11.1. The Strait of Georgia and the Fraser River Estuary-Plume Systemsummer when southeasterlies and southwesterlies dominate the southern Strait while northwesterliesdominate the northern Strait (Thomson, 1981).Mixed semi-diurnal and diurnal tides are found throughout most of the Salish Sea, although tidesare mainly semi-diurnal. The tidal range undergoes a bi-weekly variation due to the spring/neap(i.e., lunar phase) and tropic/equatorial (i.e., lunar declination) cycling (Thomson, 1981; Halverson,2009). The minimum daily tidal range at Pt. Atkinson is around 1.9 m while the maximum reachesabout 5.0 m (Halverson, 2009).The barotropic tides form a standing wave in the SoG, produced by a combination of northwardprogressive ocean tides that enter from the SJDF and southward reflected waves that are reflectedfrom the northern end of the SoG. The characteristics of the standing wave affect the timing ofmaximum flood and ebb within the SoG, where maximum flood occurs roughly 3 hrs before highwater and maximum ebb around 3 hrs before low water. The standing wave is more pronouncedin the northern SoG and river channels and weakens slightly southward where the progressive tidesbecomes more prevalent (Thomson, 1981).The Fraser River is the largest single source of freshwater that empties into the SoG, supplyingup to 50-85% of total freshwater in the Strait (Crean et al., 1988; Pawlowicz et al., 2007). Itoriginates in the Rocky Mountains near Jasper, Alberta, descends rapidly until Hope, BC, andstarts a 160 km journey through the Fraser valley to the SoG (Thomson, 1981). The Fraser Riverplume is formed by discharge from each of its three arms and Canoe Pass, with the southern (main)arm carrying roughly 75% of the total river flow, located 15 km south of Vancouver, with theremaining outflow divided among the North Arm (15%), the Middle Arm (5%) and Canoe Pass(5%) (Thomson, 1981). The plume thickness varies within the upper 10 m (Stronach, 1981) andwas selected as 7 m by Halverson and Pawlowicz (2011).21.1. The Strait of Georgia and the Fraser River Estuary-Plume SystemThe Fraser River estuary is considered a salt-wedge type estuary (Geyer and Farmer, 1989) (Fig-ure 1.1). The location of salt-wedge depends on river discharge and tides (Crean et al., 1988; Geyerand Farmer, 1989; Kostaschuk and Atwood, 1990). Through an observational-based parametriza-tion, Kostaschuk and Atwood (1990) predict that the tip of salt-wedge (practical salinity of 10)can propagate upstream of the mouth at Steveston by 30 km at high tides and by 22 km at lowtides during low river flow (around 700 m3s−1 at Hope). During high river flow (freshet period,flows around 8000 m3s−1 at Hope), the salt-wedge position is at the mouth at low tides and 10 kmupstream at high tides.Similar with tides in the SoG, tides in the estuary are mixed, mainly semi-diurnal. Tides in theestuary channel are influenced by the amount of freshwater carried downstream. For example, thetides can be detected at Chilliwack, 100 km upstream of Steveston during the winter period of lowflows. During the freshet, the tides cannot be detected beyond Mission, around 60 km upstream(Thomson, 1981).Geometry in the river mouth region is complex (Stronach, 1981). Beyond the mouth at Steve-ston, a rock jetty breakwater protects a 9 km channel through tidal mud flats to Sand Heads, ameteorological station. The depth of the channel is about 12 m on average (Halverson, 2009).The volume of Fraser River discharge varies considerably from year to year and also seasonally(Thomson, 1981). Annual discharge peaks in late May and early June due to snow-melt. Thismaximum, called the freshet, discharge ranges from 6000-10000 m3s−1 between 2014-2016 at Hope.River discharge decreases significantly to about 2000 m3s−1 by the end of August and reachesabout 800 m3s−1 by early December which continues until the following spring. However, winterrain storms can cause river discharge at the mouth to more than double over the span of a few days(Halverson and Pawlowicz, 2011).31.2. Past Modeling Studies on the Fraser River Plume1.2 Past Modeling Studies on the Fraser River PlumeThe Fraser River plume is distinct from most previously studied river plumes that are located onthe shelf, such as Columbia River plume, Changjiang River plume and Merrimack River plume.The Fraser River plume thickness is less than 10 m (Stronach, 1981). Compared to the ColumbiaRiver plume which is typically 20 m thick (MacCready et al., 2009), Changjiang River plume withthickness around 18 m at maximum ebb during spring tide (Wu et al., 2011), the Fraser Riverplume is thinner. The Fraser River plume surface area was estimated with a combination of ferry-based data and satellite images and it varies from around 400 km2 during low river flow periodsand to 1700 km2 during freshets, when the plume may span the SoG (Tabata, 1972; Halverson andPawlowicz, 2008). The Fraser River plume depth in the plume surface area is not a constant and itchanges in space. For instance, S2-3 station and S3 station of STRATOGEM observation projectare considered in the Fraser River plume area [Figure 1, (Halverson and Pawlowicz, 2011)], but thedifference of plume depth is about 1 m during low river discharge (<3500 m3s−1) and up to 4 mduring high river discharge (>3500 m3s−1) assuming plume salinity threshold is 25 as an example(Figure 3 in (Halverson and Pawlowicz, 2011)). Recent CTD profiles in October, 2014 in the regiondirectly influenced by the Fraser River discharge (Figure 1.2) show salinity stratification occursmostly in the upper 10 m with salinity values below about 30, which are generally in agreementwith the study of Stronach (1981).Previous modeling studies of the SoG, in particular the Fraser River plume date back to the1980s. Stronach (1981) developed an upper-layer model using a control-volume approach to simulatethe plume although sufficient calibration and verification were not done due to lack of observations.This model did not include wind stress. Based on Stronach (1981), Royer and Emery (1985)further calibrated the same model with measured salinity from ferry data (Royer and Emery, 1982),41.2. Past Modeling Studies on the Fraser River Plumeadded wind stress and found the effect of the varying river discharge on salinity in the model wasminimal. They improved model performance by modifying entrainment velocities to be dependenton stratification and reducing the newly included wind stress term for regions with high salinity.They found plume salinity along the northern ferry route, which is between Horseshoe Bay andNanaimo (Royer and Emery, 1982), was more sensitive to changes in the entrainment velocity dueto the stronger influence of discharge fluctuations but the salinity of southern ferry route betweenTsawwassen and Swartz Bay was more affected by the change of wind stress. However, the northernend of the SoG was not included, which would influence the freshwater circulation (Royer and Emery,1985). Marinone et al. (1996) used a three-dimensional baroclinic model which was forced by tides,winds and runoff, to study the residual current in the central SoG generated by these forcing agentsand by interaction among them. They discovered that the one-year mean residual at the first layer(0-5 m) showed the river discharge enters the Strait and is deflected northward by the Coriolis force.Wind stress pushes this layer northward as a slab. They defined the residual current as the flowremaining after low pass filtering with a 2-day cut-off. This residual contains both a time-averagedand a fluctuating component. The total residual maximum speed can reach 0.4 m s−1 at the surface.Time-averaged residual speed is about 0.1 m s−1 at the surface away from the river mouth. Morerecently, Masson and Cummins (2004) evaluated time-averaged salinity in the surface freshwaterplume, including the Fraser River plume, produced by the Princeton Ocean Model (POM) withobserved CTD salinity. They found the surface plume is more trapped in the south compared tothe observations, and they argued this problem was caused by inadequate representation of runoffin the northern Strait and also inaccurate salinity values at the northern end of SoG.In summary, past modeling studies discovered some plume characteristics, such as the modeledaveraged residual current speed in the central SoG is about 0.1 m s−1 at the surface away from the51.3. The Model Configurationriver mouth (Marinone et al., 1996). However, there are a number of limitations in the previousstudies. One example is that runoff in the northern Strait is not represented appropriately in thestudy of Masson and Cummins (2004), which affects the surface plume location. In this thesis, athree-dimensional baroclinic ocean model, which solves the three-dimensional hydrostatic equationsof motion for an incompressible fluid under the Boussinesq approximation on a structured compu-tational grid (Madec, 2012; Soontiens et al., 2016), will be used to simulate the Fraser River plume.The focus will be on improving small-scale dynamical features with high spatial and temporal vari-ability and understanding the impact of forcings on plume dynamics, such as river flows, tides,winds and the Coriolis force. The ultimate goal is to predict the plume movement and properties.1.3 The Model ConfigurationIn this study, the Nucleus for European Modelling of the Ocean (NEMO) framework version 3.4 in itsregional configuration (Madec, 2012) has been employed to model the whole Salish Sea. The modeldomain extends from the SJDF at the west to Puget Sound at the south to Johnstone Strait at thenorth. The curvilinear orthogonal numerical domain has been divided into 398 by 898 grid pointshorizontally and 40 levels vertically, with around 500 m resolution horizontally. Vertical spacing isunequal. It is 1 m for the upper 10 m and increases smoothly to a maximum of 27 m starting fromaround 150 m to the bottom (Soontiens et al., 2016). The purpose of this spacing is to better resolvethe surface plume structure. The initial model used a crude representation of the Fraser River. Itwas 10 km long, up to 2 km wide and depths were set to 4 m. This representation is not realistic.As will be shown, lengthening and deepening the river channel affects tidal propagation in theriver, and thus influences the cross-strait velocity at the river mouth. Similar river treatment wasperformed for the Columbia River in Regional Ocean Modeling System (ROMS), where a straight61.4. Vertical Turbulence Schemeriver channel was added and extended 300 km to the east (MacCready et al., 2009). This rivergeometry was considered as the most realistic one for tidal energy dissipation (MacCready et al.,2009).This NEMO model in the Salish Sea region was spun up for 15.5 months, starting on September16, 2002, with initial uniform horizontal conditions for temperature and salinity from a CTD castin the middle of the SoG taken in September, 2002. Velocity was initialized at zero. The atmo-spheric forcing from 2002 to 2003 is the Canadian Meteorological Centres 33 km global atmosphericreforecasting model (CGRF) (Smith et al., 2014). Temperature, salinity and sea surface height atthe boundaries are climatologies, as well as tides and climatology river discharge (Soontiens et al.,2016). Using the September 7, 2003 temperature and salinity field, the model was restarted fromzero velocities on September 10, 2014. Since then, the model has run operationally producing dailynowcasts for the current day and up to two 30 hr forecasts across the whole domain, with forcing ofdaily Fraser River runoff, eight tidal constituents at the west and north open boundaries and dailyoperational model winds employing the western Canada component of the High Resolution Deter-ministic Prediction System (HRDPS, https://weather.gc.ca/grib/grib2_HRDPS_HR_e.html).1.4 Vertical Turbulence SchemeThe surface currents in the plume are sensitive to two model physical parameters: vertical eddyviscosity and diffusivity. These two turbulence coefficients are used to model the vertical turbulentfluxes of momentum and tracers. The coefficients are calculated by the k − ε configuration of thegeneric length scale (hereafter GLS) turbulence scheme (Umlauf and Burchard, 2003) and then themaximum of these values and the background vertical eddy viscosity and diffusivity parametersvalues are used.71.4. Vertical Turbulence SchemeThe GLS turbulence closure scheme computes the vertical eddy viscosity and diffusivity using(1.1). The mixing length l and turbulent kinetic energy k are calculated through a set of differentialequations and the stability functions Cµ and C′µ are derived from the Reynolds stress equations anddepend on shear and buoyancy numbers. Consequently, the turbulent coefficients are calculatedand updated using the equations below:νt = Cµ√kl,κt = C′µ√kl(1.1)where νt and κt are vertical viscosity and diffusivity, respectively (Reffray et al., 2015). Mostfrequently used closures available in the GLS scheme are : k − kl, k − ε and k − w (Reffray et al.,2015). In a comparison of several turbulence closures, including k − ε, k − w and k − kl, Reffrayet al. (2015) found that the k − ε configuration had the smallest salinity bias in a modelling studycovering the North Pacific Ocean during the period June 2010-June 2011. The k − ε, k − w andk − kl closures are known as two-equation models because the differential equation for turbulentkinetic energy is solved alongside a second differential equation that determines the mixing lengthscale.Alternatively, single equation models, like the Turbulent Kinetic Energy (hereafter TKE) model,uses an algebraic expression for the mixing length scale and a differential equation for the turbulentkinetic energy (Reffray et al., 2015). Comparing GLS turbulence scheme with TKE turbulencescheme, Reffray et al. (2015) found that GLS is more accurate in reproducing the mixed layer depthdeepening of an initially linear stratified fluid subjected to a stationary surface stress than the81.5. Mixing in a River PlumeTKE closure. As a consequence, k− ε closure of GLS turbulence scheme was employed to computeturbulence parameters of vertical eddy viscosity and diffusivity in this study.The background vertical eddy viscosity and diffusivity are set to ensure a minimum level ofmixing (Reffray et al., 2015). Typical values used in process studies are:νb = 1.2× 10−4m2s−1,κb = 1.2× 10−5m2s−1,(1.2)where νb is background vertical eddy viscosity and κb is background vertical eddy diffusivity(Madec, 2012). These two values must be greater than those associated with molecular viscos-ity of ∼10−6 m2s−1 and molecular diffusivity of ∼10−9 m2s−1 for salinity (Madec, 2012). Actualvalues are the maximum of values calculated from the k−ε configuration of GLS turbulence schemeand the corresponding background values that are set.1.5 Mixing in a River PlumeA river plume impacts shelf circulation and ecosystem health by its salinity dilution and transportprocesses. Dilution is driven largely by vertical mixing (Horner-Devine et al., 2015). In a singlesystem, the mixing of the plume and the associated plume structure can change with forcing, suchas river discharge, tidal range, winds and the Coriolis force (Horner-Devine et al., 2015). Threeinteracting dynamical regions have been previously identified provided that the width of river mouthis small relative to the Rossby deformation radius: the estuary, near-field and far-field (Hetland,2005). The Rossby deformation radius is defined as below:91.5. Mixing in a River PlumeRd =√g′hf,g′= g∆ρρ0(1.3)where h is the depth at the river mouth, g is acceleration of gravity, ρ0 is the reference density,and ∆ρ is the difference in density between the plume and the layer below and f is the Coriolisparameter.The estuary and near-field are characterized by intense shear-mixing of freshwater and salt waterand the near-field is also referred to as a tidal plume in regions with large tides (Horner-Devine, 2009;Horner-Devine et al., 2015). Far-field refers to the region beyond the near-field where wind inducedmixing is dominant in vertical mixing processes. Wind stress generally causes lower turbulencelevels than are observed in the near-field (Houghton et al., 2009), but since wind usually acts overthe entire plume domain, the net effect due to the spatial area of active mixing makes wind drivenmixing important (Horner-Devine et al., 2015).In this thesis, the terms near-field and far-field are defined according to the importance of tidesor winds in mixing, with the near-field characterized as the tide mixing dominated region, whilefar-field refers to the wind dominated mixing region.Most past studied river plumes are located on the shelf, for example, the Columbia River,Changjiang River and Merrimack River. Mixing studies on these river plumes are common and theplume is usually divided into those dynamical regions discussed above. However, as for the FraserRiver plume, mixing induced by tides and winds for the Fraser River plume which lies within asemi-enclosed ocean basin, can be different. Tides in the SoG are relatively strong, especially in101.5. Mixing in a River Plumethe Haro Strait region (Thomson, 1981). Therefore, mixing generated by tides is likely to responddifferently than on the shelf, where the tides are only significant near the river mouth. The winddriven circulation in the SoG might also behave differently from the Ekman dominated flow drivenby winds in shelf systems (Halverson and Pawlowicz, 2008).Past studies on mixing of the Fraser River plume are numerous. Halverson and Pawlowicz(2008) studied the impact of river discharge and tides on the Fraser River plume based on four-years(2003-2006) of ferry data. The effect of river discharge on plume salinity is that plume salinity is aquasi-linear function of river discharge on time scales of 25 days. However, the fresh water residencetime, defined as the total volume of fresh water divided by input of river flow is about 2 days(Halverson and Pawlowicz, 2011), which is roughly independent of river discharge. Tides were foundto impact plume salinity over three time scales: semi-diurnal, diurnal and fortnightly (Halversonand Pawlowicz, 2008). The plume is fresher at lower-low water than higher-high water. Plumesalinity fluctuates in phase with the daily maximum tidal height, being fresher during neap tidesand saltier during spring tides on fortnightly scales (Halverson and Pawlowicz, 2008). Entrainmentis a principal physical process that causes the transfer of salt water into the fresh water in theestuary. By examining the entrainment process along a transect in the Fraser River estuary andusing nitrate as a tracer to study entrainment, Yin et al. (1995a,b,c) found the entrainment wasaffected by river discharge, tides and winds. Specifically, more nitrate was entrained during higherriver discharge than lower river discharge. This result was partially explained by the seaward freshwater spreading that allows a wider area for nitrate transfer. More entrainment occurs duringspring tides than neap tides. Wind is a major factor in increasing nitrate concentration in theupper layer by increasing entrainment. Recent studies quantified the entrainment velocities in theFraser River plume and estuary. Halverson and Pawlowicz (2011) estimated entrainment velocities111.6. Transport in a River Plumeof a few mm s−1 from the entrainment flux using a quasi-steady salinity budget. These entrainmentvelocities were calculated by assuming the mixing primarily occurs along the salt-wedge estuary orin the near-field plume. They are in agreement with MacDonald and Geyer (2004), who estimatedvertical velocities on the order of 1-4 mm s−1. Shear instability was investigated in the Fraser Riverestuary based on echo sounder observations (Geyer and Farmer, 1989; Tedford et al., 2009). Tedfordet al. (2009) also pointed out that the majority of the observed mixing in the Fraser estuary wasdue to shear instabilities at the pycnocline, although mixing due to bottom generated turbulence isobserved during late ebb during low discharge periods.1.6 Transport in a River PlumeTransport in a plume on the shelf away from the river mouth is driven mainly by the plume’sbuoyancy, wind stress and ambient coastal currents and is strongly modified by Earth’s rotation(Horner-Devine et al., 2015). For the Fraser River plume, tides should also be considered as surfacetidal currents can reach 0.5 m s−1 (Foreman et al., 1995; Halverson and Pawlowicz, 2016), potentiallycomposing a large part of the total surface velocities. Variable tidal rapids occur near and withinthe narrow and shallow passages leading into the Strait, with maximum flood and ebb flows up to5 m s−1 (LeBlond, 1983).In the Strait of Georgia, wind stress was found to drive the highest surface currents relative toriver forced and tidal forced flows based on HF radar data averaged over the period from August2012 to May 2013 (Halverson and Pawlowicz, 2016). The wind-driven flow speeds are approximately3-5% of the wind speeds. The dominance of wind-forcing in the subtidal surface currents over otherfactors in the SoG was discovered in multiple drifter studies, all of which indicated that the surfaceflow primarily follows the wind (Thomson, 1981). Drifters go southward in the case of northwesterly121.6. Transport in a River Plumewinds, indicating the wind can overcome the tendency for the buoyant outflow to turn northwardunder the Coriolis force.An unsteady growing anticyclonic recirculation offshore of the river mouth referred to as thebulge, is commonly reported by studies for plumes on the shelf (Oey and Mellor, 1993; Kourafalouet al., 1996; Garvine, 2001; Fong and Geyer, 2002; Horner-Devine et al., 2015). It forms when thewidth of river mouth is small relative to the Rossby deformation radius (1.3) and under sufficientlylow wind conditions (Horner-Devine et al., 2015). The main role of a bulge on plume transportis to accumulate a fraction of between 25% and 75% of the river discharge and thus reduce inthe transport of fresh water away from the river mouth region in the coastal currents (Fong andGeyer, 2002; Horner-Devine et al., 2006). However, in the Fraser River plume, the bulge featureis rarely observed. Halverson and Pawlowicz (2016) reported an anticyclonic turning seen in themean surface current of HF radar data averaged from August 2012 to May 2013. Although thisstructure shows some resemblance to the bulge pattern, it is not clear if this pattern fits with classicbulge model described by Horner-Devine (2009) and Horner-Devine et al. (2015). Halverson andPawlowicz (2016) found the mean HF radar surface flow was inconsistent with the gradient-windrelationship determined by Horner-Devine et al. (2015). Under the gradient-wind relationship, thepressure gradient is balanced by both the centrifugal and the Coriolis forces. A second reason couldbe that the west boundary of the semi-enclosed Strait of Georgia may not be favourable for bulgeformation. Nonetheless, the mean surface flow field averaged over the period from August 2012 toMay 2013, forced by the Fraser River observed with HF radar data is characterized as a jet-likeoutflow near the river mouth, flanked by two counter-rotating eddies, with the bulge-like eddy atthe north and a cyclonic eddy at the south (Halverson and Pawlowicz, 2016). The mean flow speednear the river mouth was 0.14 m s−1, while the flow further away from the mouth was no more than131.7. Research Questions0.05 m s−1 (Halverson and Pawlowicz, 2016).1.7 Research QuestionsIt is interesting but also challenging to realistically model the Fraser River plume velocity andsalinity fields. The reason why it is challenging is that plume motion is sensitive to many forcingfactors that can vary significantly in time. In addition, it is necessary to understand mixing andtransport processes impacted by various forcing agents in the Fraser River plume.The objectives of this study are to answer the following research questions:1. How well do the modeled flows and salinity compare to the observations?2. What geometry of the Fraser River should be used in the NEMO model and how sensitiveare the salinity and surface currents in the plume to the geometry of Fraser River estuary and theregion around its mouth?3. How do vertical eddy viscosity and diffusivity affect the plume properties?4. What is the importance of tides, winds and the Coriolis force on the mixing of the plumeand how does the importance vary with different river flows?5. What is the importance of tides, winds and the Coriolis force on the transport of the plumeand how does the importance vary with different river flows?141.7. Research QuestionsFigure 1.1: Salt-wedge schematic diagram. Diagram shows a vertical cross-section from the mouthof the river (on the left) to upstream. The depth shown is about 10 m and the length shown isabout 15 km. Salt water below the freshwater that penetrates into the river is the salt-wedge.151.7. Research QuestionsFigure 1.2: Observation sites and salinity profiles in October 2014, from Institute of Ocean Sciences,Joanne Breckenridge and Evgeny Pakhomov at University of British Columbia. Left panel: CTDobservations shown (stars). Right panel: salinity profiles at those observation sites. Color of thesite and the salinity profile correspond. Three different observations (teal color) are carried outat the same location during October 2014 by Joanne Breckenridge. Salinity stratification occursmostly in the upper 10 m with salinity values below about 30.16Chapter 2Plume Sensitivities to ModelParameters and Forcings2.1 IntroductionThe Salish Sea NEMO nowcast system has run daily since October 24, 2014. This operationalmodeling system employs the NEMO model to run the Salish Sea domain daily and the hourlyresults of temperature, salinity, sea surface height and currents are published on the web (https://salishsea.eos.ubc.ca/nemo/results/). One of the results of this modeling system is stormsurge prediction (±10 cm) in the Strait of Georgia, which can be accessed through the web (https://salishsea.eos.ubc.ca/storm-surge/forecast). More details of how this nowcast modeling isconfigured is available in the Introduction 1.3.Several discrepancies were found in the Fraser River plume salinity and surface currents of thenowcast results compared to different observations. First, by comparing to near-surface observationsavailable from a ferry that crosses the Strait, it was found that the modeled minimum salinity valuealong the ferry track was higher than the ferry observations. Second, the position of the minimumsalinity along the ferry track was found to be far to the east in the model results. This comparisonsuggests that the cross-strait velocities are too weak in the modeled surface currents. The same172.1. Introductiondiscrepancy was seen comparing to available surface drifter tracks. Third, modeled surface currentsgenerally have too strong southward along-strait flows compared to the drifters. Therefore, thisthesis has a goal of improving the modeled Fraser River plume surface currents and salinity byimproving these modeled plume properties by adjusting model parameters.In addition to determining the sensitivity of the plume to model parameters, the sensitivityof the plume (including its mixing and transport) to variation in its forcings, such as the winds,tides, river discharge and the Coriolis force, is also determined. Insights into how plume mixingand transport vary with forcing factors are helpful for plume prediction. Winds and tides areboth important in plume mixing, but understanding which region of the plume is dominated bytide-induced mixing or by wind-induced mixing increases our knowledge of river plume dynamics.River discharge generally increases the plume area, however, whether the amount of river dischargechanges the plume location or not remains unclear. The Coriolis force is important for river plumesthat occur on the shelf, but the impact of the Coriolis force on modeled movement of a plume locatedin a small scale semi-enclosed basin, such as the Strait of Georgia, has not been well studied before.Hence, a model study of the Fraser River plume and how mixing and transport are influenced bythese forcing factors is needed.Here, two model sensitivity studies are conducted. First, model sensitivities to river bathymetriesand turbulence parameters are investigated by evaluating the modeled surface currents and salinityof the plume with available observations. Second, by employing the improved configuration fromthe evaluation, model sensitivities to forcings: the winds, tides, river flow and the Coriolis force,are studied to understand plume mixing and transport.In the first part of this chapter, baseline run conditions, various observational data and compar-ison methods between model results and observations are described. Several bathymetric variations182.2. Methodsand sensitivity experiments using varying turbulence parameters are conducted with a goal of im-proving the plume properties compared to the baseline. Evaluation of the results from the baselineand the variations are analyzed. The improved configuration is determined.In the second part of this chapter, based on the improved configuration, plume dynamics in themodel are studied. In particular, how mixing and transport processes in the plume are affectedby river discharge, tides, winds and the Coriolis force are studied. Numerical simulations withand without each of the forcings above are performed. Plume mixing and transport metrics aredeveloped to quantify and analyze mixing extent and freshwater transport.2.2 Methods2.2.1 The Model DescriptionThe model employed in this study is the Nucleus for European Modelling of the Ocean (NEMO)version 3.4 in a regional configuration (Madec, 2012) for the Salish Sea region (Figure 2.1). Thenumerical grid is 398 by 898 by 40 grid cells, with a horizontal grid spacing of around 440 m by500 m. The vertical z-levels are unequally spaced, with 1 m vertical grid spacing down to about10 m in depth and a maximum grid spacing of 27 m starting from around 150 m to the bottom.Eight tidal constituents (K1, O1, P1, Q1, M2, K2, N2, and S2) are forced at two open boundariesthat connect to the Pacific Ocean, the western boundary at the mouth of Juan de Fuca Straitand the northern boundary in Johnstone Strait (Soontiens et al., 2016). All the model results foranalysis in this study are hourly averages. In order to limit large changes in depth across grid cells,bathymetry over the entire domain was smoothed so that ∆h/h ≤ 0.8, where ∆h is the differencein depth between two adjacent grid cells, and h is their average depth (Soontiens et al., 2016).192.2. MethodsCompared with the study of Soontiens et al. (2016), differences for the baseline simulation for thisstudy are:(1) Lateral eddy viscosity of 15 m2s−1 was used rather than 20 m2s−1.(2) The western Canada component of the High Resolution Deterministic Prediction System (HRDPS),a nested 2.5 km resolution atmospheric model provided by Environment Canada (https://weather.gc.ca/grib/grib2_HRDPS_HR_e.html), was employed instead of the Canadian Meteorological Cen-tre’s 33 km global atmospheric reforecasting model (CGRF) wind products (Smith et al., 2014;Soontiens et al., 2016). The temporal resolution of the operational wind is hourly.(3) Daily Fraser River runoff at Hope (https://wateroffice.ec.gc.ca/report/report_e.html?type=realTime&stn=08MF005) was used rather than climatology data. For all other rivers andcontributions to the Fraser River downstream of Hope, climatology was used.(4) The background vertical eddy viscosity and diffusivity were set to the same values in thebaseline run as Soontiens et al. (2016): 1× 10−4 and 1× 10−5 m2s−1, respectively. The NEMO modelselects the maximum values of the vertical eddy viscosity and diffusivity between background valuesand the values that are calculated by the k − ε configuration of the GLS turbulence scheme. Weevaluate variations in background vertical viscosity and diffusivity.(5) The initial bathymetry we employed here is the same as Soontiens et al. (2016), which con-tains a short river channel for the Fraser River (left panel, bathymetry #2, Figure 2.2). We evaluate202.2. Methodsvariations in bathymetry.The model equation of state implicitly assumes that model salinities are equivalent to observedPractical Salinities on the Practical Salinity Scale 1978 (Unesco, 1981).2.2.2 Strategies in this StudyFor modeled plume properties sensitivities to model parameters, two criteria were considered tochoose simulation periods: 1) the availability of observational data and 2) a variety of forcingconditions. Three types of data were available for October 1-31, 2014: drifter data, ferry-basedsalinity data on a track crossing the plume and CTD observations. June 15-29, 2015 was selectedbecause it is during the Fraser River freshet period and thus quite different than the Octoberperiod. Also, ferry-based salinity data is available to evaluate against modeled salinity during thissimulation period. Finally, October 1-31, 2015 was chosen since measurements of surface currentare available for this period.It is important to evaluate the salinity and surface currents in the modeled Fraser River plumebecause these two variables reveal plume physical characteristics. Magnitudes of surface cross-straitvelocities and along-strait flows near the river mouth directly reflect features of surface currents inthe Fraser River plume. Plume position will be used as this variable is also related to surfacecurrents and it reflects the surface currents.In order to evaluate the modeled plume salinity, the minimum salinity value and fresh waterfraction along the ferry track were chosen. These variables will be related to the total plumesalinity, and can be directly compared with observations. To evaluate the modeled plume position,the location of the minimum salinity value along the ferry track was adopted since this variable212.2. Methodsindicates the plume location along the ferry route in general. In addition to evaluating the modeledplume results against observations, it is useful to determine the sensitivity of the modeled plumeresults with different configurations. For example, the volume flux and salinity time series at a rivermouth transect with different river bathymetries were analyzed because they provide another viewof the strength of cross-strait velocities and salinity values, respectively.In the plume sensitivity to forcing section, the simulations were divided into January 1-31, 2016(low river flow), October 1-31, 2014 (moderate river flow) and May 1-31, 2015 (high river flow)periods. These three periods were chosen because they allow us to investigate the effects of theamount of river flow on mixing and transport processes within the plume.To study the mixing processes within the Fraser River plume, the total amount of mixing inthe top 10 m of the water column within the plume is used to determine the total mixing a waterparcel has experienced. Furthermore, the total mixing induced by different forcing factors willbe estimated. A goal of this thesis is to study the relative importance of tide- and wind-inducedmixing in the Fraser River plume and how it varies with the river flow. Understanding the relativeimportance of tides versus winds in mixing in the Fraser River plume is novel as past studies ofmixing in river plumes focused on coastal shelf regions instead of a semi-enclosed ocean basin.To investigate the transport processes within the Fraser River plume, the amount and thedirection of freshwater transport under different forcing factors are determined. In addition, it isinteresting to investigate how the fresh water plume moves around due to winds and river flows.222.2. Methods2.2.3 Data for Model EvaluationFerry-based Observational Salinity DataTwo passenger ferries run by BC Ferries Inc. were instrumented by Ocean Networks Canada(ONC). These ferries travel between Tsawwassen and Duke Point (diagonal route) and Tsawwassenand Swartz (southern route) (Figure 2.1), each crossing the Fraser River plume to different extents.They make four round trips during weekdays and fewer on the weekends. Each trip takes around2 hours (1 hour and 35 minutes) to cover the complete transects of around 70 (44 km) for thediagonal (southern) route (Figure 2.1). Seawater near-surface salinity was derived by ONC frommeasurements of temperature and conductivity every 10 seconds using a SeaBird ThermosalinographSeaKeeper in the FerryBox. The effective water sampling depth is approximately 2 m (Wang, 2015).In the model, grid points are at 1.5 m depth and 2.5 m depth and so 1.5 m depth was selected tocompare against the ferry data.Drifter DataA total of nine Microstar GPS drifters, with a drogue 1.22 m tall centred at a depth of 1 m (Ohlmannet al., 2005), were released on October 8th, 2014, at three different locations around Sand Heads(Figure 2.1), by Mark Halverson and Rich Pawlowicz, University of British Columbia. All thedrifters were released during ebb tides and drifted 7-29 hours before recovery.CTD DataFive Conductivity Temperature Depth (CTD) profiles obtained by Peter Chandler from Institute ofOcean Sciences (IOS) in October, 2014, in the central SoG were used (Figure 2.1). Three other CTDcasts conducted by Joanne Breckenridge and Evgeny Pakhomov, University of British Columbia,232.2. Methodsmeasured at the same location near the mouth of the Fraser River during different days in Octoberwere also used (pink star, Figure 2.1).Water Level and River Discharge DataHourly real-time water level data from June 15-29, 2015, at Steveston, Deas Island Tunnel, NewWestminster and Mission were acquired from Environment Canada (https://wateroffice.ec.gc.ca/search/searchRealTime_e.html).2.2.4 Comparison MethodsComparison with Ferry Salinity DataAll ferry tracks were visually checked to confirm they included the full transect across the Strait anddid not include data obtained while the ferry was sitting in port or outside the Strait. The InverseDistance Weighting Interpolation method is applied to interpolate gridded model values onto theobservational points on the ferry routes (Appendix A). Linear interpolation between hourly modelresults is used to interpolate to the time of the ferry observations.Comparison with Drifter DataModel particle trajectories were calculated using Ariane (Blanke and Raynaud, 1997) based onhourly average output from the model. Model particles were forced to remain in the uppermostmodel grid box, nominally 0.5 m depth. Individual drifter-particle track comparisons were made.Particles are released at the same time and same position for the same duration as the observeddrifters. Particles are also released every hour, at the drifter position at the corresponding time, andtracked for one hour. Based on this, statistics of the separation between the drifters and modeled242.2. Methodsparticles after every hour were derived. Furthermore, particle tracks offset by 0.5 hour and by 500 mrelative to the initial released time and position of the drifter were also calculated.Comparison with CTD DataFor each CTD cast, model results for the grid point closest to the cast were averaged over a UTCday and the minimum and maximum hourly-average salinity values over the day were found. Scatterdiagrams of the observed CTD versus model results were made.Comparison with Water Level DataA time series from June 15 to 29, 2015, was selected to calculate the maximum, minimum andaveraged modeled tidal amplitudes at Steveston, Deas Island Tunnel, New Westminster and Mission(locations on Figure 2.2), and compared to those calculated from observations at the correspondingstations.2.2.5 Variations AttemptedTo test the sensitivity to model parameters, a set of 11 runs are described (Table 2.1). Four differentbathymetries were evaluated.Short River Channel (baseline, bathymetry #2)In the baseline case (Figure 2.2), the river channel is 10 km long by 2 km wide with a closedboundary at the head of the river. The river channel and banks are uniformly 4 m deep.252.2. MethodsExtended River Channel (bathymetry #5)An extended river channel (Figure 2.2) was based on the measured bathymetry of the Fraser River(Canadian Hydrographic Chart #3492) from Gravesend Reach to New Westminster with depthat New Westminster of 13 m. Beyond this location, the model channel turns to the north andextends for another 40 km (Figure 2.2) and then runs eastward along the model grid to the edgeof model domain. The depth after New Westminster is set to 13 m. The total length of the riverchannel is about 76 km from the river mouth at Steveston, which is over 7 times longer than thatof bathymetry #2. Based on the chart, the minimum width of the river channel is only one gridcell, approximately 500 m.Extended and Deepened River Channel (bathymetry #6)Bathymetry #5 has weak tides in the lower Fraser River. Bathymetry #6 was created based onbathymetry #5, by deepening the region around the river mouth by 5 m. This region in the Straitis around 10 km along-strait by 4.5 km cross-strait. In addition, the area from the river mouthand into the river until Gravesend Reach (Figure 2.2) was also deepened by 5 m; the length ofthis region is around 10 km (Figure 2.3). Note that some grid cells in the river mouth region werealready modified in bathymetry #5, and they were not further deepened in bathymetry #6. Thisgeometry was generated to correct the tidal amplitudes in the Fraser River estuary.Vertical Eddy Viscosity and DiffusivityModel results with lower background vertical eddy viscosity of 1× 10−5 m2s−1 and both lowerbackground vertical eddy viscosity and lower background eddy diffusivity of 1×10−6 m2s−1 wereevaluated (Run # 4 and 5, Table 2.1).262.2. MethodsSensitivity Runs for Model EvaluationFifteen-day sensitivity experiments, from June 15 to 29, 2015 (Run #1a, 2 and 3a, Table 2.1, Figure2.4), were performed to examine whether the extended river channel (bathymetry #5) and extendedand deepened river channel (bathymetry #6) improve the surface velocity and tidal amplitudesinside the river channel and to further investigate how sensitive the plume properties are to thegeometry of the river.Run #1a is initialized from the operational nowcast, and thus started from non-zero velocities.Run #2 and #3a started with zero velocities and same temperature and salinity values used tostart Run #1a, that is operational results from June 14, 2015. In the added river channel grid cells,temperature values are set to 14◦C , salinity values are set to 0 east of Deas Island Tunnel and 1west of it.One month sensitivity experiments, from October 1 to 31, 2014 (Runs #1b, 3b, 4 and 5, Table2.1, Figure 2.5) were made to compare the model results of the three different bathymetries and oflower background vertical eddy viscosity and lower background vertical eddy diffusivity, respectively,with various available observations. Run #6 and 7 (Table 2.1) were designed to investigate theeffects of bathymetry without winds. Run #8 was conducted to evaluate the impact of adding arepresentation of the Steveston Jetty (Appendix C). All these runs were initialized with temperatureand salinity from the operational nowcast on September 25, 2014 and run for 5 days to spin up themodel before analysis start for October 1, 2014 (Table 2.1).All these runs (Runs #3b, 4 and 5, Table 2.1) started from zero velocities except Run #1b,which is initialized from the operational nowcast.Another one month sensitivity experiment, from October 1 to 31, 2015 (Run #9, Table 2.1)was performed with the extended and deepened river channel (bathymetry #6), lower background272.2. Methodsvertical eddy viscosity and lower background vertical eddy diffusivity, respectively, to evaluate themean surface currents with HF radar data (Appendix D). It also started from zero velocities. Initialtemperature and salinity were from the operational model results from September 18, 2015.2.2.6 Sensitivity Studies of Plume Properties Affected by Forcing FactorsPlume Physics RunsA total of fifteen runs were conducted to study how plume salinity stratification and fresh watertransport are influenced by: the river runoff, tides, winds and the Coriolis force. These fifteensimulations are divided into three river flow periods: low, moderate and high (Table 2.2, Figure 2.6,2.5 and 2.7). All the sensitivity runs were started at least five days prior to the river flow periodof interest with zero initial velocities, temperature and salinity from the operational model, clima-tology sea surface heights, temperature and salinity at the boundaries. The model Fraser Riverdischarge is realistic and daily. Five different runs were done for each river flow category:1. Combined all forcing case. In this run, river flows, tides, winds and the Coriolis force areall present, which can be considered the most realistic simulation.2. Only river discharge case. The model is only forced with daily river flow, without any otherexternal forcings and includes the Coriolis force.3. River + tides case. The model is forced with all the forcing conditions, except the opera-tional winds.282.2. Methods4. River + winds case. The model is forced with all the forcing conditions, except the tides.5. No Coriolis force case. The model is forced with all the forcing conditions, and the Coriolisforce is set to zero.Weak, Moderate and Strong Wind Period SelectionFor each of the simulations during low, moderate and high river flow period, different wind magni-tude periods, are chosen in order to study the impact of the wind on plume mixing and transportprocesses. A weak wind period is that when over 90% of the wind magnitudes during the selectedperiod is weaker than 5 m s−1. A moderate wind period is selected when the wind speed within therange of 5-10 m s−1 accounts for more than 40% of the total time. In addition, for the moderatewind period, to make sure the wind magnitude is mainly less than 10 m s−1, wind component withmagnitudes larger than 10 m s−1 should constitute no more than 5% of the overall wind valuesduring the selected period. A strong wind period occurs when wind speed larger than 10 m s−1occupying over 20% of the selected wind period. In addition to the criteria for each wind perioddiscussed above, for the moderate and strong wind events, wind direction prior to the event isconsidered. Wind periods are not selected if the winds are strong in the opposite direction to thewind in the event period. This last requirement is to avoid the wind lag effect as much as possible.292.2. MethodsDefining the PlumeIt is necessary to identify the plume in the model output. The edge of the plume is defined using athreshold surface salinity (Halverson and Pawlowicz, 2011):Sthresh = Sref − Soffset (2.1)where the Sref is the 2.5 km by 2.5 km spatially averaged surface-level (nominally 0.5 m depth)salinity over the northwest corner of the subdomain (Figure 2.1). This region is close to the areawhere Halverson and Pawlowicz (2011) selected Sref for their calculation. For each of January,2016, October, 2014 and May, 2015, the monthly average of the combined all forcing case is used.Soffset is selected to be a linear function of the reference salinity:Soffset = 4.8− 0.14× Sref (2.2)(Halverson and Pawlowicz, 2011).Note the subdomain (Figure 2.1) which is used to analyze plume impacted by forcings has somelimitations. Halverson and Pawlowicz (2016) and Tabata (1972) observed that plume can go furthersouth than the southern boundary of the selected subdomain. However, due to computational andstorage limitations, this subdomain was not extended further south.302.3. ResultsUpper Layer Depth, Reference Salinity and Reference Density SelectionThe upper layer depth h is selected to be 10 m as we expect that the plume is mostly trapped withinthe top 10 m (Stronach, 1981) and indeed we find that no more than 1% of the time does the plumereach 10 m depth during three months based on salinity criteria (2.1) at NS station (Figure 2.1,2.8). The reference salinity value is chosen to be 30 since this is a typical mean salinity value in therange of 50 to 100 m depth in STRATOGEM observations [Figure 3.1, (Riche, 2011)], and we foundthis value corresponds with the mean salinity at 50 m depth in October 2014 over the northwestcorner of the subdomain (Figure 2.1). The reference density of 1023 kg m−3 selection is based ontemperature of 10◦C and salinity of 30. Similarly, this value of reference salinity represents themean density at 50 m depth in October 2014 over the northwest corner of the subdomain (Figure2.1).2.3 Results2.3.1 Fraser River Plume Sensitivity to Model ParametersEvaluation of Operational Results – Runs #1a and #1bIn addition to the methods introduced in Methods (2.2.4) to compare model results with ferry-baseddata, three other metrics are used for quantitative comparisons: the minimum salinity value, thelocation of the minimum salinity value and fresh water fraction along a complete diagonal ferrytransect. Location of the minimum salinity value here is the longitude of the minimum salinityvalue along a single ferry track. This metric indicates how far the plume has moved offshore fromthe river mouth. Freshwater fraction, Fa, for every available ferry track is calculated through the312.3. Resultsformula below:Fa =1L∫ L0S0 − S(x)S0dx (2.3)where the reference salinity S0 is taken to be 30, S(x) is the salinity value at position x at 1.5 mdepth on the ferry route, and x is the distance along the ferry track.Eighty-six ferry crossings were used to compare to the model (e.g., Figure 2.9). A typicalcomparison from October 8, 2014, shows modeled salinity values for Run #1b in a similar range asthe ferry-based salinity for both routes, ranging between 14 and 28 for the diagonal route and from20 to 30 for the south route during this moderate river flow period (Figure 2.9). Lower values areseen on the diagonal route as it cuts the plume closer to its centre. However, the locations of themodeled minimum salinity value along the ferry track are closer to the mainland coast, indicatingtoo weak cross-strait velocities (Figure 2.10). Furthermore, the modeled minimum salinity alongthe diagonal track is always higher. The integrated fresh water fraction (2.3) is generally too smallalthough there are some examples when the model has a higher fresh water fraction comparedto that observed (Figure 2.11). For example, on June 21, 26 and 27, the baseline has a higherfreshwater fraction than the ferry observations.The drifter-particle comparison (e.g., Figure 2.12) confirms the problem of too weak cross-straitflows in Run #1b. The modeled particle does not go far enough offshore. Averaged modeled particlevelocity during ebb tide is about 0.4 m s−1 while drifter’s velocity is 0.6 m s−1.All of these features are robust as revealed from the statistics based on ferry data (Figure 2.10).In addition, the model has too strong along-strait surface flows (e.g., Figure 2.12).322.3. ResultsEvaluation of Bathymetric ChangesWith a longer and deeper river channel (Run #3a and #3b, bathymetry #6), the location of theminimum salinity value along the diagonal and south track moves further offshore compared tothe baseline (Run #1a and #1b, bathymetry #2) (Figure 2.10 and example on Figure 2.9). Theposition of the plume is improved about 50% from the statistics of 86 ferry crossings in June and 23ferry crossings in October (Figure 2.10). The drifter trajectories show slightly stronger cross-straitflow, although the along strait flow remains too strong (e.g., Figure 2.12).Overall, the extended river channel (Run #2, bathymetry #5) reduces the values of the plumesalinity, making it closer to the observations (Figure 2.10). However, this channel does not generatestronger cross-strait velocities (Figure 2.10).Compared to most ferry tracks, the model has too little integrated freshwater along the diagonalroute (Figure 2.11). This result is consistent with the minimum salinity value being too high in themodel (Figure 2.10) as well as a saltier near-surface plume in general compared to observed CTDcasts (Figure 2.13). The bathymetry variations caused little change in the fresh water fraction.Generally, both the observed and modeled tidal amplitudes decay in the river channel withdistance from the river mouth (Table B.1, Appendix B). Tidal amplitude at Steveston agrees withobservations for both the extended river channel (Bathymetry #5) and the extended and deepenedriver channel (Bathymetry #6) (Table B.1, Appendix B). However, tidal amplitudes at Deas IslandTunnel and New Westminster are too small in bathymetry #5, whereas they are slightly overesti-mated in bathymetry #6. Tidal amplitude at Mission is too high in bathymetry #6, but this isunlikely to impact the plume itself.332.3. ResultsThe volume flux V , through a cross-section is defined as:V =∫∫udydz (2.4)where u is the velocity perpendicular to the transect, and y is along-transect distance and theintegral is over the whole transect to the bottom or just the top 4 m. An integral over the just top4 m depth is selected because it is the original depth (bathymetry #2) of this transect.The salinity integral SI , is defined as:SI =∫∫Sdydz (2.5)where S is the salinity, and the integral is evaluated across the river channel and over the top 4 mor the full water column.The volume flux (2.4) across and salinity integral (2.5) along transects at the river mouth andinside the river channel between longer and deeper river channel (Run #7, bathymetry #6) andshort channel (Run #6, bathymetry #2) were calculated (Figure 2.14). Large differences of theinstantaneous volume fluxes occur at the river mouth transect (Figure 2.14 b) between these tworiver channels, and the magnitude of peak instantaneous volume flux at peak ebbs with the longerand deeper river channel is nearly three times larger than the short shallow channel. Indeed, thevolume flux over only the upper 4 m exceeds that of short river channel by a factor of two. Acrossthe transect inside the river channel, the variation of volume flux with the tides is very weak withthe original bathymetry (Figure 2.14 f), reflecting the closeness of this transect to the original river342.3. Resultshead in the model. The net daily volume flux over the full depth is weaker than that integrated onlyover the top 4 m at the river mouth transect (Figure 2.14 c) whereas the net daily outflow volumeflux over full depth is stronger than that integrated over only the top 4 m at the transect inside theriver channel (Figure 2.14 g). Generally, the total volume flux is not affected by the bathymetrychanges (Figure 2.14 c and g).The salinity integral (2.5) at both transects is much higher with the extended and deepenedriver channel (bathymetry #6) (Figure 2.14 d and h), which is consistent with a saltier signalfrom the ferry comparisons (Figure 2.10) presumably caused by stronger mixing effect. This saltiersignal with the extended and deepened river channel (bathymetry #6) arises because a salt-wedgeintrusion usually occurs in this deeper river.Evaluation of the Reduced Viscosity and Diffusivity CasesReducing the background viscosity and reducing both the background viscosity and diffusivityimproved the model salinity compared to observations along the ferry routes (Figure 2.10 andexample in Figure 2.9) by reducing the near-surface salinity.Variation of the viscosity and diffusivity has little impact on salinity profiles (Figure 2.13). Inthe top 10 m, most of the modeled results are saltier than observations by up to 5 practical salinityunits; the difference is less significant at deeper depths and the model is too fresh in the bottomlayer.For the example shown, modeled particle tracks show substantial improvements with loweredviscosity, especially in the along-strait movements (Figure 2.12). Changes in diffusivity make onlysmall changes to the drifter track (Figure 2.12). Although improvements were seen in this drifter-particle trajectory comparison (Figure 2.12), the general ability of the model to track surface drifters352.3. Resultswas not improved from the statistics of averaged distance between the drifters and modeled particlesafter one hour (Figure 2.15). The cause can be two-fold. One could be that the tidal currents inthe model are not accurate enough, leading to inaccurate tidal variability in the surface flows. Theother reason might be the randomness of the drifters, which could suggest chaotic features in thereal SoG.To illustrate the spatial and temporal variation of particle trajectories, particles offset by 500 mand by half an hour relative to the initial released position and time of the corresponding drifterwere done (Figure 2.16). There are large variations of the modeled surface flows in both timeand space. Maximum spread among the nine particles in each case occurs at the final hour. Themaximum separation distances between the modeled trajectories are 3.7 km, 6.2 km, 17.1 km and16.4 km for run #1b, #3b, #4 and #5, respectively. Although the general patterns of the particletracks in each run case does not vary much, modeled trajectories of the particles released half anhour late give the best qualitative results.2.3.2 Fraser River Plume Sensitivity to ForcingMixingTo investigate the extent of mixing within the plume, a ratio was employed (Masunaga et al., 2016):Rmix =SSurf< S >(2.6)where SSurf is the surface salinity and < S > is the mean salinity over the top 10 m of the watercolumn (< S > never has a zero value in this study). The larger the value of Rmix, the more362.3. Resultsintense the mixing that has occured in the water column. Specifically, Rmix = 1 indicates that thewater column is fully mixed vertically while Rmix = 0 occurs when pure fresh water exists at thesurface.Recall the near- and far-field plume region in Introduction 1.5. The major difference betweenthese two regions is the relative importance between tides and winds in inducing mixing. Therefore,we can operationally define the near- and far-field plume region based on the Rmix ratio fromdifferent scenarios:Pr =Rmix(river + winds)−Rmix(river + tides)Rmix(river + winds+ tides)(2.7)If the value of Pr is greater than 0.1, it shows mixing induced by winds is much greater than thetides. This region is defined as the far-field, where winds dominate mixing. Otherwise, this regionbelongs to near-field plume. We select 0.1 as the threshold rather than 0 because greater than 0.1shows the dominance of wind in mixing although 0.1 is somewhat arbitary.Potential energy, employed to measure shear-induced mixing relative to the fully mixed state,is defined as:PE FullMixed =∫ 10m0(ρ− ρ)gzdz,ρ =∫ 10m0 ρdz10 m(2.8)One can also define an unmixed system (two-layer). To conserve mass:372.3. ResultsρD = ρminl + ρmax(D − l),D = 10 m,ρmin = 999.7 kgm−3,ρmax = 1023.0 kgm−3(2.9)where ρmin is the density of a surface layer of depth l calculated with salinity 0 and temperature10◦C, and ρmax is the density of the lower layer of depth D − l calculated with salinity 30 andtemperature 10◦C. D is the total depth. Equation (2.9) can be solved for l. Then the potentialenergy of this unmixed system is:Unmixed =−g(ρmin − ρ)l22− (ρmax − ρ)(D − l)g × (l + D − l2) (2.10)By subtracting (2.10) from (2.8), the potential energy relative to a completely unmixed state,PE UnMixed, can be calculated.PE Unmixed = PE FullMixed− Unmixed (2.11)NS station is always within the plume region (Figure 2.17, 2.18) according to the salinity criteria(2.1). The fresh water, mainly from the Fraser River, enters into the SoG and at NS station, thewater body with salinity below the salinity criteria threshold (2.2), is mostly trapped in the top5 m for the river only case (Figure 2.17 a, e and i, Figure 2.18 a, e and i). For river only case,382.3. Resultsstrong stratification due to river outflow restricts the mixing processes (a, e and i of Figure 2.17and 2.18). However, potential energy generated by shear-induced mixing by the river is about 600 Jm−2 in these three months (Figure 2.19), which illustrates the non-negligible role of the river itselfin creating shear mixing.The stratification varies with different discharge rates of the Fraser River. In general, stronglystratified water with salinity values below 12 can reach around 4 m depth in early May (Figure2.17 i and Figure 2.18 i), when the Fraser River discharge is around 4000 m3 s−1, while water withthe same salinity is confined in the upper 1 m when Fraser River discharge is around 900 m3s−1 inJanuary (Figure 2.17 a and Figure 2.18 a).Periodic, tidally modulated vertical mixing structures are simulated by the model (Figure 2.17b, f and j, Figure 2.18 b, f and j). Without winds, the upper 6 m of the water column is stratifiedduring low and moderate river flow periods (Figure 2.17 b and f), which is similar to the riveronly forcing case. However, without winds in May, the mixing pattern induced by the tides iscomparable to the all forcing case (Figure 2.17 j). Under weak wind conditions, the salinity mixingand restratification driven by the tides (Figure 2.18 b, f, and j) is similar to that with all forcingfactors (Figure 2.18 d, i and l). During moderate and strong wind events in low and moderate riverflow periods, the tide-induced mixing region is tightly constrained near the mouth (Figure 2.20 d, g,e and h) while the near-field region expands substantially during the high river flow period (Figure2.20 f and i), to include NS station. In all flow periods, at the river channel near the mouth, salinityis around 4-8 fresher at 4 m depth by adding tides and adding tides restricts the salt-wedge positionto within 15 km from the mouth (Figure 2.21 b, e and h).At NS station, wind-induced vertical mixing is relatively strong compared to that induced bytide under moderate wind conditions during low and moderate river flow periods, where the surface392.3. Resultssalinity can reach 24 (Figure 2.17 g) and the mixing patterns are comparable between the river andwinds and the all forcing case. On average, shear-mixing induced by winds at NS station is strongerthan that due to the tides (Figure 2.19). The wind-dominated mixing region grows considerablywithin the plume area during low and moderate river flow periods when the magnitude of wind isgreater than 5 m s−1 (Figure 2.20 d, e, g and h). Under weak wind conditions, wind-induced mixingregion is limited during low and moderate river flow periods (Figure 2.20 a and b) while it grows inhigh river flow period (Figure 2.20 c). The effects of wind-driven mixing are generally weaker thantides at NS station for all river discharges under weak wind conditions. Inside the river channel,the salt-wedge (10 salinity contour) can propagate up to New Westminster with river and windsforcing and low discharge (Figure 2.21 a). The intrusion distance decreases under moderate andhigh discharge (Figure 2.21 d and g). Compared to that with river and tides or with all forcing,mixing is less intense in the river channel with river and winds.For all three river flow conditions, with the Coriolis force, we see fresher water northward ofthe Fraser River mouth along the mainland BC coast (g, h and i of Figure 2.22); the freshwatercontinues into English Bay near the city of Vancouver. Without the Coriolis force, the plume isgenerally wider (d, e and f of Figure 2.22).For the all forcings case, the effects of river discharge, tides, winds and the Coriolis force aresuperimposed. However, the relative strengths of mixing are distinctive across three different riverflow periods. During high river discharge in May (Figure 2.17 l and Figure 2.18 l), mixing is weakeraccording to Rmix (2.6) than under low and medium river discharge.The relative size of the instantaneous momentum compared to the internal wave speed is the402.3. Resultsupper layer Froude number, defined operationally here as:Fr =∆U√g′h(2.12)where∆U =√(us − ub)2 + (vs − vb)2us =∫ h0 udzh, ub =∫ Hh udzH − h ,vs =∫ h0 vdzh, vb =∫ Hh vdzH − h ,g′=g(ρb − ρs)ρ0,ρs =∫ h0 ρdzh, ρb =∫ Hh ρdzH − h ,ρ0 = 1023.0 kgm−3,g = 9.8 ms−2(2.13)u and v are cross-strait and along-strait velocities, respectively. us and vs represent the averagedupper layer cross-strait and along-strait velocities while the upper layer depth h is determined asthe location where ∂ρ/∂z is maximum. In the other words, the depth of the pycnocline. ub andvb are the averaged lower layer (from upper layer depth h to the bottom depth H) cross-straitand along-strait velocities. Averaged upper layer density ρs and lower layer density ρb are foundsimilarly; Density ρ is calculated from temperature and salinity based on Pond and Pickard (1983).If the value of Froude number is greater than a critical number of approximately 1, this indicates the412.3. Resultsflow is internally supercritical. Conversely, when internal wave speed is larger than the speed of theoutflow, Froude number is smaller than a critical number of approximately 1, which is subcriticalflow.At peak ebb in a given day, the active high momentum region with maximum speed in excess of1 m s−1 is mainly concentrated toward the southern boundary of the subdomain where the barotropictides are strong, as well as in the river mouth, where the jet-like plume discharges into the Strait(Figure 2.23 d,e and f). The strong barotropic tides at the southern boundary of the subdomain aredue to relatively shallow water and not related to the plume. Therefore, the internally supercriticalflow when Froude number is larger than order of 1 occurs predominately at where tides are strongin the Strait rather than the region of near-field plume. The internal wave speed is generally greaterin May than in January and October because of the fresher upper layer (Figure 2.23 g, h and i).The Froude number at peak floods shows similar results with these at peak ebbs, except that themomentum is slightly weaker at peak floods for all three river flow periods (Appendix H).TransportThe freshwater transport through an across-strait transect of 10 m thickness is defined as:Qfcc =∫∫ 10m0v∆SS0dzdx, (2.14)where S0 denotes the ambient water salinity, here defined to be 30. ∆S = S0 − S is the differencebetween the ambient salinity and salinity within the plume, and v is the along-strait velocity.It is clear to see that with only river forcing, the integrated freshwater transport is southward andhas the maximum southward flux (Figure 2.24). With river and tides, the direction of cumulative422.3. Resultsfreshwater flux is south by the end of the month (Figure 2.24), but the amount is only 20-30%of that with only river forcing. The Coriolis force has a non-negligible impact on the freshwatertransport, with much more freshwater transported to the south without the Coriolis force (Figure2.24), indicating the importance of the northward velocity that the Coriolis force generates. Largerfluctuations of freshwater flux with time are found without the Coriolis force compared to scenarioswith the Coriolis force. Northward freshwater transport occurs under the combine all forcings caseonly (Figure 2.24).The location of the plume centre is defined as the “centre of mass” of the freshwater thickness:x0 =∫∫xf(x, y)dxdy∫∫f(x, y)dxdy,y0 =∫∫yf(x, y)dxdy∫∫f(x, y)dxdy,f(x, y) =∫ 10m0∆SS0dz(2.15)where x and y are the along-strait and cross-strait directions, respectively, and f(x, y) is the ver-tically integrated freshwater amount at each grid cell, representing the freshwater thickness of theupper 10 m. Note that the valid domain for this calculation excludes the Fraser River channel, En-glish Bay, as well as Howe Sound (Figure 2.1), so that the impact of other river plumes is minimized.Effects of winds and river discharge on the plume centre location will be investigated.The plume centre (2.15) moves northward up to 3 km by increasing discharge from low to highriver flows (Figure 2.26). Under moderate and strong southeasterlies, the position of plume isadvected northward while under weak wind forcing, the plume centre does not move significantly.432.4. DiscussionThe maximum plume centre displacement can reach 12 km (Figure 2.27, Table 2.3).2.4 Discussion2.4.1 Plume and Estuary Response to Different River GeometriesA longer river channel (bathymetry #5) does not seem to generate more cross-strait flow at the rivermouth of the Fraser River. It is likely due to the tidal amplitudes inside the river channel not beingproduced realistically, indicating an unrealistic dissipation of tidal energy. The problem is shown bythe weak tidal amplitudes at Deas Island Tunnel and New Westminster stations compared to theobservations from EC (Table B.1). By deepening the river mouth region (bathymetry #6), tidalamplitudes at these two stations are corrected and meanwhile, cross-strait flows (Figure 2.10) areimproved compared to the longer river channel (bathymetry #5), which demonstrates the influenceof tidal energy on total cross-strait currents at the river mouth.In bathymetry #6 we see that net daily outflow volume flux integral over top 4 m is strongerthan that integrated over full depth (Figure 2.14 c). This is due to the reversal in the direction ofthe currents with depth caused by the salt-wedge intrusion, where below 4 m depth the inflow thatoccurs during floods is much stronger than the outflow that occurs during ebb tides.Bathymetric changes do not affect net daily outflow volume flux integrated over full depth(Figure 2.14 c and g) because the volume of freshwater input to the South Arm in the model is thesame between different bathymetries. As a result, changing to a longer and deeper river channel(bathymetry #6) only changes instantaneous volume flux, rather than daily integral volume flux.The saltier plume produced by a longer and deeper river channel (bathymetry #6) at the rivermouth transect (Figure 2.14 d) is mainly caused by the deepened bathymetry, which allows salt-442.4. Discussionwedge propagation into the river channel and with vertical mixing, results in a relatively saltiervertical water column. This also explains the higher minimum salinity value with bathymetry #6compared to the baseline case (Figure 2.10).Overall, cross-strait velocities are improved with this longer and deeper river channel (bathymetry#6). However, large discrepancies still remain, such as too strong along-strait flows compared toobserved drifter tracks (e.g., Figure 2.12), and a saltier plume compared to the ferry observations(Figure 2.10).2.4.2 Effects of Vertical Eddy Viscosity and DiffusivityWith reduced background eddy viscosities and diffusivities, the time-averaged values of vertical eddyviscosities and diffusivities at most depth levels are larger than the background values (e.g., Figure2.28); thus NEMO selects the values calculated from the GLS turbulence scheme. Reduced viscosityenables more shear between the upper layer and layers below and this is reflected in the modeledparticle trajectories (e.g., lower right panel, Figure 2.12), mainly by reducing the near-surface along-strait flows, which are driven by the barotropic tides. However, reduced viscosity has little impacton the near-surface salinity value of the Fraser River plume (Figure 2.10) as this parameter onlydirectly controls the vertical momentum transfer. Reducing the background vertical eddy diffusivityreduces vertical mixing, resulting in a fresher near-surface waters (Figure 2.10). However, this doesnot necessarily mean that the plume is fresher everywhere with lowered background vertical eddydiffusivity because plume advection is also changed by changing background vertical eddy diffusivity.As a result, the difference in salinity at some places is saltier and others fresher (e.g., Figure 2.13).However, if one looks at the averaged surface salinity in the plume region (2.1) and (2.2) over October1-31, 2014, a 0.5 practical salinity unit fresher plume is generated with reduced background vertical452.4. Discussioneddy diffusivity.Our values for vertical viscosity and diffusivity are typical. For example, 1× 10−5 m2s−1 wasused as background value of vertical eddy viscosity to simulate the Rhine River plume using theDelft3D-Flow model with a vertical grid resolution of 2 m. It was considered to generate the mostrealistic salinity profiles compared to vertical viscosity values of 1× 10−2, 1× 10−3, 1× 10−4 and 1×10−6 m2s−1 (Jacobs, 2004). Background vertical eddy viscosity and diffusivity of 5× 10−6 m2s−1was selected to model the Columbia River plume using Regional Ocean Modeling System with aterrain-following coordinate. Vertical resolution in the model varies and an average resolution is1.0 m in the upper 4 m in a depth of 100 m (MacCready et al., 2009; Liu et al., 2009).Although model results are in a better agreement with the observations by adding a longerand deeper river channel as well as reducing background vertical eddy viscosity and diffusivity,substantial discrepancies still exist. For example, the cross-strait velocities are still weak comparedto ferry-based observations (e.g., Figure 2.9); the near-surface plume is still saltier compared againstCTD and ferry observations (Figure 2.9, 2.10 and 2.13), which could be partially caused by incorrectplume position. In addition, relatively strong northward mean surface flows from June 22 to August25, 2016 are found north of the river mouth off the banks compared to HF radar data (Figure D.2),which might be related to the absence of the Steveston jetty in the model.2.4.3 Four Important Forcing Mechanisms Reproduced by NumericalSimulationsEffects of river flowsThe increase in amount of fresh water stratifies the water column (a, e and i of Figure 2.17, 2.18),increases the size of plume and expands the near-field region. These features are significant in May462.4. Discussionwith wind greater than 5 m s−1 (e and h of Figure 2.20) compared to October and January whenwind magnitude is greater than 5 m s−1. A larger near-field region when the wind is strongerthan 5 m s−1 in May compared to October and January is probably an artifact of a non-local tideimpact. Without tides, more freshwater was transported to the south to fulfill the stronger estuarinecirculation due to the lack of tidal mixing in Haro Strait. This southward advection, particularly ofthe May plume result could also be due to advection caused by winds from north-west. Winds fromnorthwest advect the plume more to the southeast compared to October, when prevailing windsfrom south-east advect the plume against the mainland coast. In either of these two cases, the sizeof modeled near-field region is not directly related to mixing, but a result of advection.During weak wind periods, the near-field region dominates the plume region in October andJanuary. However, the near-field region is smaller in May compared to October and January forweak winds (c of Figure 2.20).Halverson and Pawlowicz (2008) investigated Fraser River plume salinity affected by river dis-charge and tides based on four-years (2003-2006) of ferry observations. They suggested that theferry only samples the near-field plume during high river flow but not during low flow in winter whenthe near-field plume remains much closer to the river mouth. With a threshold of 0.1 for our Rmixbased near-field criteria, our results show that the near-field is constrained near the river mouthin January and October, which agrees with their interpretation of the ferry data. Furthermore,they interpreted the result using a conceptual model for the combined effects of tides and river flowon plume salinity. For example, in summer, the position of salt-wedge is more downstream thanin winter (Kostaschuk and Atwood, 1990; Halverson and Pawlowicz, 2008), suggesting mixing atthe river mouth would be stronger since intense mixing occurs around the salt-wedge. As a result,the near-field plume area is larger in summer than in winter (Halverson and Pawlowicz, 2008). A472.4. Discussionmajor difference between the study of Halverson and Pawlowicz (2011) and my study is that theyestimated local intense mixing. On the other hand, my study measures the total mixing based onthe water column stratification.Overall, the impact of river discharge on the location of the plume centre is small, with no morethan 3 km northward movement between low river flow (January, 2016) to high river flow (May,2015). This small displacement that is observed is possibly caused by more northward spreadingwith increasing river discharge.Effects of windsWind is the major factor causing mixing in the Fraser River plume under moderate and low riverflows, provided the wind is greater than 5 m s−1 (Figure 2.20). Thus the far-field plume, definedas the region dominated by wind mixing, is the largest part of the plume. Wind typically mixesthe surface plume (Horner-Devine et al., 2015), which can be seen by the smaller area enclosed bythe 25.3 practical salinity isohaline plume boundary (Figure 2.27). In general, both the amount ofthe freshwater transport across the northern transect and plume centre location correlate with thewind. The 3-day time-averaged plume centre location can move 12 km when pushed by a strongwind (Figure 2.27 i) and the averaged rate of freshwater transport across the northern transect(Figure 2.24) caused by wind can exceed the river discharge rate during the same period (Table 2.3,Figure 2.25), indicating the importance of winds in freshwater transport.Wind impacts the spatial freshwater thickness pattern of the plume. Even without wind forcing,the freshwater thickness distribution and plume centre location vary about 2-3 km (Figure 2.27),which could be caused by the combined effects of river discharge and spring/neap variations due tothe tides.482.4. DiscussionA plume bulge is an unsteady anticyclonic eddy circulation that forms offshore of the river mouthin the absence of external forcing, such as winds [Section 1.5, Horner-Devine et al. (2015)]. It is veryunlikely that a bulge forms in the Fraser River plume. According to the bulge circulation theory,momentum of the bulge flow in a steady state satisfies the gradient-wind balance (Horner-Devineet al., 2015; Horner-Devine, 2009), where the pressure gradient is balanced by both the centrifugalforce and the Coriolis force. Modeled 3-day mean surface currents, in October in the without windforcing case, have a bulge-like pattern with velocities of about 0.01 m s−1. These are much smallerthan an inertial flow, v = fr, velocity of nearly 0.8 m s−1 given that the radius r of the eddy-likefeature is about 7 km (Figure 2.27 e). The modeled mean surface flow is also smaller than thegeostrophic buoyant flow estimated by thermal wind relationship:∂v∂z=gfρ0∂ρ∂x(2.16)Scaling analysis is performed to estimate V velocity:V ≈ gf∆zρ0∆ρ∆x≈ 0.4ms−1 (2.17)where ∆z is about 10 m for plume depth, f is approximately 1 × 10−4 s−1, ∆x is the length scaleover the bulge-like pattern (Figure 2.27 e), around 7 km, ∆ρ of this region is about 3 kg m−3 giventhe salinity change of 4. This 0.4 m s−1 velocity is much larger than modeled mean along-strait Vvelocity in the bulge-like area, which is only around 0.01 m s−1.492.4. DiscussionUsing the gradient-wind relationship, the expected flow should exceed flow under geostrophyalone because the centrifugal and pressure gradient terms are in the same direction. Then thedifference between the modeled speed and the expected speed further increases. Therefore, themean flow is not explained by the gradient-wind balance.On the other hand, the tidal excursion at the centre location of bulge is 7 km, which is comparableto the Rossby deformation radius in this region (Rd ∼7 km). The fluctuation caused by strong tidesmay inhibit the bulge formation in the SoG.Weak amplitude wind (0-5 m s−1) has little effect on the centre plume location and the amountof freshwater transport compared to moderate or strong wind conditions (Table 2.3). The winddirections during weak wind periods are usually variable. As a result, the net influence of weakwinds on the plume location is small.Effects of tidesThe region dominated by tide-induced mixing (near-field) is tightly constrained near the FraserRiver mouth during low and moderate river flow (Figure 2.20). Tides cause more mixing than thewinds inside the river channel and the salt-wedge propagation is more limited with just tides thanwith just winds (Figure 2.21). Mixing in the river channel is less intense in May, which is likelydue to the strong river outflow that causes stronger stratification. However, even under strong riverflow, the tides are still important in vertical mixing inside the river channel compared to the winds(Figure 2.21).The distance of salt-wedge propagation S′ (unit: m) is calculated from the empirical equation502.4. Discussion(Kostaschuk and Atwood, 1990):S′ = 82419m− 9835m× loge Q + 22458m× log10 T (2.18)where Q is the river discharge (unit: m3s−1) at Hope, T is the tidal height (unit: m) at PointAtkinson. The distance of salt-wedge (S′) is measured relative to Sand Heads in study of Kostaschukand Atwood (1990). It is converted to refer to Steveston by:S′s = S′ − 8300m (2.19)Modeled position of salt-wedge propagation is defined as the point of intersection of the 10practical salinity isohaline with the estuary bed (Kostaschuk and Atwood, 1990).The model results for the position of the tip of the salt-wedge when it is high tide at PointAtkinson are compared to the predictive regression model results based on observations using 2.18from Kostaschuk and Atwood (1990) (Figure 2.29). Model results with only river and winds showvery small variations in the salt-wedge position since there is no fortnightly tidal cycle. The salt-wedge location is similar between the tides only case and the combine all case and the position issubject to fortnightly tidal modulation. The striking feature of this set of comparisons is that themodeled salt-wedge intrusion is closer to the river mouth at Steveston by about 5 km compared tothe predictions, indicating too much mixing in the model. However, Kostaschuk and Atwood (1990)pointed out a couple of sources of errors in predicting this salt-wedge position, such as the time lagin discharge between Hope and the estuary, which results in potential errors in the salinity-dischargerelationship since variation in discharge at Hope may not be felt in the estuary for one or two days.512.4. DiscussionMore importantly, this prediction is based on a limited dataset (19 observations), which may havelimitations due to the quantity.Freshwater flux in the upper 10 m with tides shows large differences compared to that with riveronly (Figure 2.24). The amount of freshwater carried southward in the river only case, without tidesis 3-5 times greater than that with the tides. This large difference of freshwater amount is mostlikely due to the role of tidal mixing in Haro Strait (Figure 2.1). Tidal mixing significantly reducesestuarine exchange (Nagai and Hibiya, 2011). However, the reduced freshwater transport due to thelack of tides cannot last forever, as the freshwater input is the same for all the different forcing runs.Nonetheless, this result for a short period simulation is completely different compared to results ofLi and Rong (2012) for the Changjiang River plume located on the shelf, where they argued thatby adding tides, freshwater transport was enhanced. They interpreted this phenomenon as tidalcurrents forcing the plume water to move in the direction of the tidal currents, thus freshwatertransport downstream increased. Without tides, most of the Changjiang River plume was trappednear the river mouth and formed a bulge, which accumulated a large portion of freshwater, andtherefore led to less freshwater transported downstream. Comparison of Changjiang River plumeand Fraser River plume undoubtedly illustrates different roles of tides on plume transport betweenon the shelf and in a semi-enclosed ocean basin. On the shelf, tidal currents mainly advect theplume in the direction of the tidal currents, which impacts freshwater transport. However, thefreshwater transport is insensitive to the strength of turbulent mixing (Li and Rong, 2012). Onthe other hand, in a semi-enclosed ocean basin, such as in the SoG, tides contribute to mixing thatdecreases the freshwater transport by estuarine circulation and this reduction of freshwater flux isnot caused by advection due to the tides.522.4. DiscussionEffects of Coriolis forceThe Coriolis force is important in weakening the mixing of the plume, due to its role in producinga thicker plume because of the northward deflection of the plume along the mainland coast, andthis thicker plume is not as easily mixed by the winds. The Coriolis force results in fresher waterin English Bay. This phenomenon is robust and independent of river flow conditions (Figure 2.22).My study shows the Coriolis force generally reduces the Fraser River plume mixing in the far-fieldplume, which is in agreement with previous studies (Garvine, 1999).Generally, much more freshwater is carried northward across the northern transect (Figure 2.24)with the Coriolis force (Table 2.3), except during strong winds. Winds appear to impact the plumemore without the Coriolis force (Figure 2.25). Specifically, a moderate wind to the north results inmore northward flux. This is because the Coriolis force constrains the plume along the mainlandcoast and the depth of plume is thicker to conserve the plume buoyancy. A deeper plume is moredifficult to move by the winds.2.4.4 ConclusionsSurface flows and salinity in the Fraser River plume in a three-dimensional baroclinic model wereevaluated against various types of observations. The major problems in the model are too weakcross-strait flows, too strong along-strait flows and too salty a plume. By employing a longer anddeeper river channel (bathymetry #6) for the Fraser River, cross-strait flows improve substantiallycompared to both ferry-based salinity data and drifter data.Plume movements and properties are sensitive to values of vertical eddy viscosity and diffusivity.By reducing the background vertical eddy viscosity and diffusivity to 1× 10−5 and 1×10−6 m2s−1,respectively, both surface flows and salinity in the plume improved. From a single drifter-particle532.4. Discussioncomparison, the along-strait flows were reduced in magnitude by reducing the background verticaleddy viscosity. However, of the nine drifters, only two showed reduced along-strait velocities, twoshowed increased along-strait velocities and five showed no significant change. The near-surfacesalinity values were generally reduced by reducing the background vertical eddy diffusivity.Using the longer and deeper river channel (bathymetry #6), lowered background vertical eddyviscosity and diffusivity, mixing and transport processes within the Fraser River plume were stud-ied. The relative strength of momentum and internal wave speed was investigated. Instantaneousmomentum is generally weaker during peak floods than during peak ebbs. Internal wave speed isoverall higher during high river flow period than low and moderate river flow periods.The impact of river discharge, tides, winds and the Coriolis force on mixing and freshwater fluxwere investigated and the principle findings are:1. The amount of runoff impacts mixing by stratifying the water column. The size of the plumeand tide-induced mixing region expands with increasing river discharge. However, the plume centrelocation does not change much.2. The region dominated by tide-induced mixing is tightly constrained near the river mouthduring low and moderate river flow conditions. The tide-induced mixing in the river itself is strongerthan that due to the wind. The amount of freshwater flux across the northern transect in runs withtides decrease to 20-33% compared to runs without tides due to tidal mixing in Haro Strait thatsignificantly reduces the freshwater transport.3. Wind is the dominant factor in mixing in the Fraser River plume during low and moderateriver flow periods, provided the magnitude of wind is greater than 5 m s−1. For most of theplume, wind mixing controls the total mixing. Wind contributes to considerable change in both thefreshwater flux (2.14) and plume centre location.542.4. Discussion4. The Coriolis force constrains the plume along the mainland coast and the plume freshensEnglish Bay. The Coriolis force contributes to much more northward freshwater transport alongthe mainland coast when the wind is not strong. The plume is deeper and less easily affected bywinds.552.4. DiscussionTable 2.1: List of simulations for evaluation. νb and κb are background vertical eddy viscosity anddiffusivity, respectively.Run Scenario Bathymetryversionνb (m2s−1) κb (m2s−1) Wind Timeperiod1 a Baseline 2 1 × 10−4 1 × 10−5 Yes June 11 b Baseline 2 1 × 10−4 1 × 10−5 Yes October 22 Extended riverchannel5 1× 10−4 1× 10−5 Yes June 13 a Extended anddeepened riverchannel6 1× 10−4 1×10−5 Yes June 13 b Extended anddeepened riverchannel6 1× 10−4 1×10−5 Yes October 24 Lower viscosity 6 1× 10−5 1× 10−5 Yes October 25 Lower viscosityand diffusivity6 1× 10−5 1× 10−6 Yes October 26 Baselinebathymetrywithout wind2 1× 10−5 1× 10−6 No October 27 Extended anddeepened riverchannel withoutwind6 1×10−5 1×10−6 No October 28 Extended anddeepened riverchannel with jetty10 1×10−5 1×10−6 Yes October 29 Extended anddeepened riverchannel6 1×10−5 1×10−6 Yes October 31 June 15-29, 2015, Initialized June 14, no spin-up2 October 1-31, 2014, Initialized September 25, 5 days spin-up3 October 1-31, 2015, Initialized September 18, 12 days spin-up562.4. DiscussionTable 2.2: List of plume physics simulationsScenario Averaged river discharge(m3 s−1)Initialization Time periodLow river flow 1316 Dec 19, 2015 Jan 1-31, 2016Moderate riverflow2051 Sep 25, 2014 Oct 1-31, 2014High river flow 4970 Apr 18, 2015 May 1-31, 2015572.4.DiscussionTable 2.3: Mixing and transport parameters impacted by winds and the Coriolis force for various windevent periods during three river flow periods. Flux is across the northern transect (Figure 2.24). Changeof plume centre location is calculated with (2.15). Along-strait and cross-strait distance in the near-fieldregion starts from the river mouth.River flow Dates Wind condition Near-field (km) Coriolisimpacton flux(with -with-out)(m3s−1)Wind impact (with - without)Magnitude Direction(to)along-straitcross-straitFlux(m3s−1)Changeofplumecen-tre(km)Low1 Jan 27-28 Strong Northwest8 2.2 -2034 534 9.3Jan 17-19 Moderate Northwest8 2.4 -926 969 8.4Jan 8-10 Weak – – – 2066 374 3.1Moderate2 Oct 21-23 Strong Northwest2.6 2 278 3855 12.5Oct 18-20 Moderate Northwest6.5 2.2 -580 2687 12.1Oct 3-6 Weak – – – 1089 439 1.7High3 May 2-3 Strong Southeast 30 16 7957 8814 11.8May19-21Moderate Southeast 41 23 4829 6564 2.5May 9-11 Weak – – – 4573 4856 2.61 January 1-31, 20162 October 1-31, 20143 May 1-31, 2015582.4. DiscussionFigure 2.1: Domain and observation sites. Left panel is model domain (colored rectangular area)including bathymetry and rivers (green circles). Right panel displays the locations of observations.Two ferry routes (black lines) with terminal locations. Location of the CTD casts (stars) in theFraser River plume region and NS station (green star). Location of Sand Heads station (red circle),VENUS Central station (black circle). Subdomain that will be used to analyze the plume (boxbounded by four green solid lines). The northwest corner of this subdomain (box bounded by fourred lines), used to calculate a reference salinity. Half the subdomain is used to calculate plumecentre location (west half of green box as split by the green dashed line).592.4. DiscussionFigure 2.2: From left to right are bathymetry #2, #5 and #6, respectively. Centre panel showsstation locations. Right panel shows the along river transect (in green).602.4. DiscussionFigure 2.3: Thalweg of bathymetry#6 in the model. Distance starts from the 67 m isobath, andincreases upriver. 35 km is at New Westminster.612.4. DiscussionFigure 2.4: Model forcing for June 15-29, 2015. Top panel shows daily Fraser River discharge(black) at Hope and averaged river discharge during this time period (blue). Second panel from topis hourly modeled sea surface height at Point Atkinson. Third panel from top shows wind vectorswith north upward at Sand Heads. Arrows represent the to direction of wind. Bottom panel showsthe hourly model wind speed at Sand Heads.622.4. DiscussionFigure 2.5: Model forcing for October 1-31, 2014. Top panel shows daily Fraser River discharge(black) at Hope and averaged river discharge during this time period (blue). Second panel from topis hourly modeled sea surface height at Point Atkinson. Third panel from top shows wind vectorswith north upward at Sand Heads. Arrows represent the to direction of wind. Bottom panel showsthe hourly model wind speed at Sand Heads. Green, blue and red columns in each panel representthe weak, moderate and strong wind event periods, respectively.632.4. DiscussionFigure 2.6: Model forcing for January 1-31, 2016. Top panel shows daily Fraser River discharge(black) at Hope and averaged river discharge during this time period (blue). Second panel from topis hourly modeled sea surface height at Point Atkinson. Third panel from top shows wind vectorswith north upward at Sand Heads. Arrows represent the to direction of wind. Bottom panel showsthe hourly model wind speed at Sand Heads. Green, blue and red columns in each panel representthe weak, moderate and strong wind event periods, respectively.642.4. DiscussionFigure 2.7: Model forcing for May 1-31, 2015. Top panel shows daily Fraser River discharge (black)at Hope and averaged river discharge during this time period (blue). Second panel from top ishourly modeled sea surface height at Point Atkinson. hird panel from top shows wind vectors withnorth upward at Sand Heads. Arrows represent the to direction of wind. Bottom panel shows thehourly model wind speed at Sand Heads. Green, blue and red columns in each panel represent theweak, moderate and strong wind event periods, respectively.652.4. DiscussionFigure 2.8: Number of occurrence of salinity of 10.5 m depth at NS station (Figure 2.1) in January2016, October 2014 and May 2015 from left to right, respectively. Dashed green line labels thesalinity threshold (2.1). Overall, no more than 1% of the time does the plume reach beyond 10 mdepth during these three months according to the salinity criteria.662.4. DiscussionFigure 2.9: Comparison of the 1.5 m depth model results for Runs 1b, 3b, 4 and 5, and theferry-based salinity along diagonal ferry route (left panel) and south ferry route (right panel) onOctober 8, 2014, from 3:15 to 5:15 (UTC), and 2:15 to 3:40 (UTC), respectively. On the rightpanel, results of Run 4 and Run 5 are underneath that of Run 3b. Run 5 (Lower both viscosity anddiffusivity) produces the closest minimum salinity value compared to observations on the diagonalroute. Location of minimum salinity value moves more than half way towards the observed minimumin Run 3b, 4 and 5 for diagonal route. Run 3b, 4 and 5 show improvement in reducing discrepancyof minimum salinity location compared to Run 1b on the south route.672.4. DiscussionFigure 2.10: Statistics of bias (model - data) of eighty-six (upper two panels) and twenty-three (lowertwo panels) ferry crossing comparisons of minimum salinity value and location between modeledresults of run 1a, run 2, run 3a (upper two panels) and and ferry-based salinity along the diagonaltrack, and between modeled results of run 1b, run 3b, run 4, run 5 (lower two panels) and ferry-based data along the diagonal track. Errorbars show the standard error of the bias for each run,respectively. Run 3a improves plume location by generating stronger cross-strait velocities, butdoes not reduce the salinity value (upper two panels). Run 5 decreases the salinity values.682.4. DiscussionFigure 2.11: Daily averaged fresh water fraction along the diagonal ferry route of June 16-29, 2015.Even after extending and deepening the river channel (Run #3a), the fresh water fraction is stilltoo low compared to the observations.692.4. DiscussionFigure 2.12: A drifter comparison with modeled surface particles. Upper left panel: observeddrifter. Upper right panel: Run 1b (blue dots) and run 3b (magenta dots). Lower right panel: run4 (teal dots) and run 5 (red dots). Lower left panel shows the sea surface height at Vancouver,BC, Canada, with the blue and red vertical lines representing the release and recovery time for thisdrifter at 16:10 (UTC) on October 8, 2014 and at 20:55 (UTC) on October 9, 2014, respectively.Model cross-strait velocities are slightly stronger with run 3b than run 1b. Along-strait flows arereduced substantially with run 4 and 5.702.4. DiscussionFigure 2.13: Scatter diagram of the observed CTD versus modeled results and cast locations onthe map. Errorbars show the minimum and maximum hourly-average salinity values and symbolsshow the model daily averaged salinity. Cast numbers are labeled. In the panels on the right, thered color indicates modeled salinity is saltier than the observation at the shallowest available depthof the observed cast, blue indicates the modeled salinity is fresher and white shows the modeledsalinity value is within 1 compared to observations. This depth ranges between 1.5 and 3.5 m.Upper panel: observation versus results of run #3b. Middle panel: observation versus results ofrun #4. Lower panel: observation versus results of run #5. Variation of viscosity and diffusivityhas little impact on the salinity profiles.712.4. DiscussionFigure 2.14: Transect locations, instantaneous volume flux, daily integrated volume flux and saltflux. Locations of transects (Upper panel). Panel b and f: hourly volume flux integrated (2.4) overfull depth for run #6 (red), integrated over full depth for run #7 (green) and integrated over onlythe top 4 m for run #7 (blue) across the transect shown in Panel a and e. Hours since 00:30 (utc)on October 1, 2014. Panel c and g: as Panel b and f but for daily volume flux. Panel d and h: asPanel b and f but for the salinity integral (2.5). With a longer and deeper river channel (bathymetry#6) (green), larger amplitude fluctuations in the hourly volume flux and much saltier water areobtained.722.4. DiscussionFigure 2.15: Statistics of averaged distance between the observed drifters and various modeledparticles after one hour. Errorbars show the standard error of the averaged distance of each modelrun (Run #1b, 3b, 4 and 5), respectively. There is no statistically significant improvement with thelonger and deeper river channel (Run #3b), lower viscosity (Run #4) and lower both viscosity anddiffusivity cases (Run #5).732.4. DiscussionFigure 2.16: Comparison of modeled particle tracks. Upper left panel: run #1b baseline case.Upper right panel: run #3b extended and deepened river channel. Lower left panel: run #4 lowerviscosity. Lower right panel: run #5 lower both viscosity and diffusivity case. In each case, inaddition to particles released at the drifter release time, tracks of particles released half an hourearlier and later are shown. For each released time, in each run case, particles are released at eightother grid points that surround the grid point closest to the drifter. The blob of black diamondsshow the released particle points. The observed drifter track is shown in Figure 2.12. Particlesreleased half an hour late overall give the best results.742.4. DiscussionFigure 2.17: Salinity time series at NS station (Figure 2.1) under moderate wind conditions (5-10 m s−1) during low, moderate and high river discharge periods from left to right, respectively,and for only river forcing, river and tides, river and winds, and combine all forcings from top tobottom, respectively. White color contour indicates the plume boundary for each period (2.1).Mixing patterns with winds show more similarity to all forcings case during low and moderate riverflow periods while tide-generated mixing controls the mixing structure in the high river flow periodin May.752.4. DiscussionFigure 2.18: Salinity time series at NS station (Figure 2.1) as for Figure 2.17 but under weak windconditions (0-5 m s−1). White color contour indicates the plume boundary for each period (2.1).River and tides case shows similar mixing as the combine all forcings case during all three timeperiods.762.4. DiscussionFigure 2.19: Potential energy relative to completely unmixed state (2.11) at NS station (Figure 2.1)for river only (blue), river and tides (red), river and winds (green), and combine all forcings case(purple) in low, moderate and high river flow periods, from top to bottom, respectively. Green, blueand red columns represent the weak, moderate and strong wind event periods, respectively. Potentialenergy generated by river flow only is non-negligible and on average winds create considerablepotential energy.772.4. DiscussionFigure 2.20: Spatial distribution of Pr (2.7) for low, moderate and high river flow periods from topto bottom during weak, moderate and strong wind events from left to right, respectively. Teal colorcontour indicates the plume boundary for each period (2.1). Moderate and strong wind dominatesmixing in the plume region during low and moderate river flow periods. High river flow expandsboth the plume region and the near-field region defined as where the Pr is less than 0.1.782.4. DiscussionFigure 2.21: Instantaneous salinity along the river transect (Figure 2.2) when it is high tide atPoint Atkinson in January, October and May for combine all forcing case from left to right, andfor river and winds forcing, river and tides forcing, and combine all forcing from top to bottom,respectively. Distance reference is the start point of this transect at the river mouth. Highlightedwhite contour line labels the 10 salinity value. Much more mixing is induced by tides in the riverchannel compared to winds during all three river flow periods.792.4. DiscussionFigure 2.22: The surface salinity with the Coriolis force, without the Coriolis force and their differ-ence (With Coriolis - without Coriolis) at instantaneous peak ebbs when velocity at the river mouthreaches a maximum. From left to right are with the Coriolis force, without the Coriolis force andtheir difference. Black color contour indicates the plume boundary for each period (2.1). From topto bottom are January 29, 2016, October 19, 2015 and May 15, 2015, respectively. The effect ofthe Coriolis force is to constrain the plume along the mainland coast, resulting in fresher water inEnglish Bay.802.4. DiscussionFigure 2.23: Froude Number, momentum and internal wave speed at instantaneous peak ebb inthree river flow periods. From left to right are Froude Number, momentum and internal wavespeed (2.12) spatial distribution at instantaneous peak ebb. Teal color contour indicates the plumeboundary for each period (2.1). From top to bottom are January 9, 2016, October 9, 2014 and May31, 2015, respectively. Momentum peaks toward the southern boundary of the subdomain due tostronger tides in this region. Internal wave speed is generally higher in May, a high river dischargeperiod.812.4. DiscussionFigure 2.24: Timeseries of cumulative freshwater transport over the upper 10 m across the transectshown in the inset map. The runs are river only (OnlyRiver), river and tides (nowind), withoutthe Coriolis force but with other forcings (nof) and combine all forcings (combineall) cases. Upperpanel: each run case in January 2016. Middle panel: each run case in October 2014. Lower panel:each run case in May 2015. Positive values represent northward flux. The southward freshwaterflux is the largest in the river only (OnlyRiver) case, followed by the case without the Coriolis force(nof).822.4. DiscussionFigure 2.25: Daily integral of freshwater flux for without the Coriolis force (nof), combine all forcings(all) and river and tide (nowind) in October, 2014. Corresponding model wind information at SandHeads station (Figure 2.1) with arrows represent direction of wind going to (lower panel). Green,blue and red columns represent the weak, moderate and strong wind event periods, respectively.For combine all forcings and without the Coriolis force cases, the flux generally matches the winds,with the flux more sensitive to winds in the without the Coriolis force case.832.4. DiscussionFigure 2.26: Freshwater thickness (2.15) spatial distribution and plume centre location (2.15, star)for January, October and May period from left to right, respectively for river only case. Red linecuts off the Fraser River channel and Squamish river source. Change of plume centre location issmall due to variations in river discharge.842.4. DiscussionFigure 2.27: Plume centre location (2.15, star) as well as freshwater thickness (2.15) spatial distribu-tion for weak, moderate and strong wind events in October, 2014 from top to bottom, respectivelyand with wind forcing, without wind forcing, and their difference (With wind - without wind) fromleft to right, respectively. Change of plume centre location under weak wind conditions is small.Large changes of location occur during moderate and strong wind events.852.4. DiscussionFigure 2.28: Background values and modeled values of vertical eddy diffusivity and viscosity aver-aged over October 8 to 10, 2014 (Run # 3 and 5, Table 2.1) at VENUS Central station (Figure 2.1).Vertical dash lines represent background values of vertical eddy viscosity and diffusivity, dot curvesare values calculated by the NEMO model as the maximum between the GLS turbulence schemeand the background values. LowerBoth refers to run #5 (Table 2.1) and Higherboth denotes run#3 (Table 2.1). Background value for LowerBoth (blue) and the HighBoth (green). With lowerbackground values, more depths use the values that GLS turbulence closure predicts.862.4. DiscussionFigure 2.29: Comparison of the position of the salt-wedge tip in the estuary between predictiveresults based on 19 observations and model results of river plus winds, river plus tides and allforcing, for low, moderate and high river flow from top to bottom. The salt-wedge position issimilar between the river plus tides and river plus tides plus winds case, although most of themodeled salt-wedge intrusions are closer to the river mouth than the predictions.87Chapter 3Conclusions3.1 Research QuestionsThe conclusions here are summarized to answer each of the research questions that are proposed inChapter 1.1. How well do the modeled flows and salinity compare to the observations?Ferry-based salinity data, drifter data, CTD profiles and water level data are used to evaluatethe model plume results. Ferry salinity data is available throughout the study period with smallgaps and can provide near-surface salinity comparisons along the ferry tracks. The drifter data Iam using is concentrated in October, 2014. CTD casts provide salinity profile comparisons, which Imostly focused in the central SoG where the Strait is directly influenced by the Fraser River outflow.Water level data are used to evaluate the tidal heights in the river channel produced by the modelby different sensitivity experiments. Overall, near-surface salinity values of the baseline run are ina similar range compared with ferry salinity data. The problems discovered in drifter, ferry-basedsalinity and CTD comparisons are a saltier plume and the plume position being too close to themainland BC coast. In addition, the model predicts much stronger along-strait flows relative to the883.1. Research Questionsdrifter data.2. What geometry of the Fraser River should be used in the NEMO model and how sensitiveare salinity and surface currents in the plume to the geometry of the Fraser River estuary and theregion around its mouth?By referring to the Canadian Hydrographic chart, a longer river channel (bathymetry #5) wascreated. Salinity values are reduced by employing this longer river channel, but the modeled plumestill lacks strong enough cross-strait flows. By further deepening the river mouth region (bathymetry#6), cross-strait velocities are improved considerably and tidal amplitudes inside the river channelcorrespond with observed water level data.3. How do vertical eddy viscosity and diffusivity affect the plume properties?Hindcast comparisons are made to compare with drifter data as well as ferry-based salinity. Al-though lowered background vertical eddy viscosity significantly reduced the along-strait flows intwo out of nine drifter-particle comparisons, the overall capability of the model to track the sur-face currents was not improved statistically. Lowered background vertical diffusivity reduces modelsalinity, which decreases the discrepancy between ferry observations and the model results.4. What is the importance of the tides, winds and the Coriolis force on the mixing of the plumeand how does the importance vary with different river flows?893.1. Research QuestionsBy conducting fifteen simulations in total, which covered three river flow levels (low, moderateand high), for river only, river plus tide, river plus winds, all forcing, all forcing except the Coriolisforce for each of the flow periods, the relative importance of tides, winds, the Coriolis force andriver discharge on mixing has been investigated. Tides are found to control mixing in the riverchannel during all river flow periods compared to the winds. The near-field region is tightly con-strained near the mouth during low and moderate river flows. During high river flow periods, thetide-dominated mixing region expands. Winds are the major forcing of mixing over a large areaof plume during low and moderate river flow periods when the wind magnitude is greater than5 m s−1. The Coriolis force results in fresher water being constrained along the mainland coastand freshening English Bay, mainly due to the more northward flow that Coriolis force generates.With increasing of river discharge, both the plume size and the tide-induced mixing region increases.5. What is the importance of the tides, winds and the Coriolis force on the transport of theplume and how does the importance vary with different river flows?Freshwater flux across the northern transect in simulations with tides decreases considerably com-pared to that with only the river. This decrease is likely caused by the tidal mixing in Haro Straitthat greatly reduces the freshwater transport. In general, the amount of freshwater flux and plumecentre location correspond with the winds. Under strong winds from north-west, the plume centrecan move 11 km to the south-east. The Coriolis force causes much more northward freshwatertransport compared with no Coriolis force when the wind is not strong. I found that without theCoriolis force, winds appear to impact the flux more than with the Coriolis force because windswill affect a thinner plume layer more easily and the plume is more spread out and thinner without903.2. Implications and Future Workthe Coriolis force. Although increasing river discharge expands the plume region, the plume centrelocation changes little.3.2 Implications and Future WorkThe NEMO model has shown promising skill in predicting the Fraser River plume properties andposition with a longer and deeper river channel (bathymetry #6), reduced background vertical eddyviscosity and diffusivity. These results and predictions have great practical implications for oil spillevents, search and rescue, biological productivity and marine pollutant transport.Investigation of mixing and transport processes within the plume strengthens the understandingof the transformation of river discharge and its evolution into the ambient ocean water. A detailedunderstanding of different forcing mechanisms responsible for plume mixing and transport is espe-cially useful. This study is the first comprehensive investigation of the impact of river discharge,tides, winds and the Coriolis force on the Fraser River plume. It also provides a realistic wind andtide dominant example for comparison to future plume theories that include both these processes.A number of problems still exist in the model. In addition to what was mentioned in theDiscussion 2.4.2, more recent drifter trajectories imply there might be a bulge-like recirculationpattern further away from the river mouth (Figure H.1). However, it is unclear if the model resultscapture this feature with the improved configuration. Surface mean currents are generally weakerthan those are observed by HF radar (Figure D.1). The surface currents are influenced by internaltides as well (Thomson, 1975). Allen (2016) identified that one of the potential origins of internaltides is at the shelf-break at the tidal banks. Interaction of internal tides and the Fraser Riverplume would provide be another interesting topic to investigate the surface currents in the FraserRiver plume.913.2. Implications and Future WorkAs mentioned in the Discussion 2.4.3, the freshwater thickness pattern in the absence of windsis associated with increasing river discharge and residual currents induced by tides. Further inves-tigation of the evolution of plume spatial distribution without wind forcing or under weak windcondition would be helpful to understand plume advection.In addition, separation of Fraser River and other rivers and inlets (e.g., Squamish river, BurrardInlet) in the simulations will provide a comprehensive way to measure the mixing and transport inthe Fraser River plume although this current study generally shows that many forcing mechanismscontribute to plume dynamics.92BibliographyAllen, S. (2016). Accurate modelling of surface currents and internal tides in a semi-enclosed coastalsea. https://agu.confex.com/agu/os16/preliminaryview.cgi/Paper89180.html. presentedat 2016 Ocean Sciences Meeting, AGU/ASLO/TOS, New Orleans, LA, 21-27 Feb. Accessed:2016-12-28.Blanke, B. and Raynaud, S. (1997). Kinematics of the Pacific Equatorial Undercurrent: an Eulerianand Lagrangian approach from GCM results. Journal of Physical Oceanography, 27(6):1038–1053.Chang, K. T. (2006). Introduction to Geographic Information Systems. pp. 117–122. McGraw-HillHigher Education Boston.Crean, P., Murty, T. S., and Stronach, J. (1988). Mathematical Modelling of Tides and EstuarineCirculation: The Coastal Seas of Southern British Columbia and Washington, Lect. Notes onCoastal and Estuarine Stud., vol. 30. Springer, New York.Fong, D. A. and Geyer, W. R. (2002). The alongshore transport of freshwater in a surface-trappedriver plume. Journal of Physical Oceanography, 32(3):957–972.Foreman, M., Walters, R., Henry, R., Keller, C., and Dolling, A. (1995). A tidal model for Juan-de-Fuca Strait and the Southern Strait of Georgia. Journal of Geophysical Research, 100(C1):721–740.93BibliographyGarvine, R. W. (1999). Penetration of buoyant coastal discharge onto the continental shelf: Anumerical model experiment. Journal of Physical Oceanography, 29(8):1892–1909.Garvine, R. W. (2001). The impact of model configuration in studies of buoyant coastal discharge.Journal of Marine Research, 59(2):193–225.Geyer, W. R. and Farmer, D. M. (1989). Tide-induced variation of the dynamics of a salt wedgeestuary. Journal of Physical Oceanography, 19(8):1060–1072.Halverson, M. and Pawlowicz, R. (2011). Entrainment and flushing time in the Fraser River es-tuary and plume from a steady salt balance analysis. Journal of Geophysical Research: Oceans,116:C08023.Halverson, M. and Pawlowicz, R. (2016). Tide, wind, and river forcing of the surface currents inthe Fraser River plume. Atmosphere-Ocean, 54(2):131–152.Halverson, M. J. (2009). Multi-timescale analysis of the salinity and algal biomass of the FraserRiver plume from repeated ferry transects. PhD thesis, The University Of British Columbia(Vancouver). https://open.library.ubc.ca/cIRcle/collections/ubctheses/24/items/1.0053481. Accessed: 2016-12-28.Halverson, M. J. and Pawlowicz, R. (2008). Estuarine forcing of a river plume by river flow andtides. Journal of Geophysical Research: Oceans, 113:C09033.Hetland, R. D. (2005). Relating river plume structure to vertical mixing. Journal of PhysicalOceanography, 35(9):1667–1688.Horner-Devine, A. R. (2009). The bulge circulation in the Columbia River plume. Continental ShelfResearch, 29(1):234–251.94BibliographyHorner-Devine, A. R., Fong, D. A., Monismith, S. G., and Maxworthy, T. (2006). Laboratoryexperiments simulating a coastal river inflow. Journal of Fluid Mechanics, 555:203–232.Horner-Devine, A. R., Hetland, R. D., and MacDonald, D. G. (2015). Mixing and transport incoastal river plumes. Annual Review of Fluid Mechanics, 47:569–594.Houghton, R., Chant, R., Rice, A., and Tilburg, C. (2009). Salt flux into coastal river plumes: Dyestudies in the Delaware and Hudson River outflows. Journal of Marine Research, 67(6):731–756.Jacobs, W. (2004). Modelling the Rhine River plume. PhD thesis, TU Delft,Delft University of Technology. http://repository.tudelft.nl/islandora/object/uuid:cf8e752d-7ba7-4394-9a94-2b73e14f9949?collection=education. Accessed: 2016-12-28.Kostaschuk, R. and Atwood, L. (1990). River discharge and tidal controls on salt-wedge positionand implications for channel shoaling: Fraser River, British Columbia. Canadian Journal of CivilEngineering, 17(3):452–459.Kourafalou, V. H., Oey, L.-Y., Wang, J. D., and Lee, T. N. (1996). The fate of river discharge onthe continental shelf: 1. Modeling the river plume and the inner shelf coastal current. Journal ofGeophysical Research: Oceans, 101(C2):3415–3434.LeBlond, P. H. (1983). The Strait of Georgia: Functional anatomy of a coastal sea. CanadianJournal of Fisheries and Aquatic Sciences, 40(7):1033–1063.Li, M. and Rong, Z. (2012). Effects of tides on freshwater and volume transports in the ChangjiangRiver plume. Journal of Geophysical Research: Oceans, 117:C06027.Liu, Y., MacCready, P., Hickey, B. M., Dever, E. P., Kosro, P. M., and Banas, N. S. (2009).95BibliographyEvaluation of a coastal ocean circulation model for the Columbia River plume in summer 2004.Journal of Geophysical Research: Oceans, 114:C00B04.MacCready, P., Banas, N. S., Hickey, B. M., Dever, E. P., and Liu, Y. (2009). A model studyof tide-and wind-induced mixing in the Columbia River estuary and plume. Continental ShelfResearch, 29(1):278–291.MacDonald, D. G. and Geyer, W. R. (2004). Turbulent energy production and entrainment at ahighly stratified estuarine front. Journal of Geophysical Research: Oceans, 109:C05004.Madec, G. (2012). Coauthors, 2012: NEMO ocean engine, version 3.4. Institut Pierre-Simon LaplaceNote du Pole de Modelisation, 27:357. www.nemo-ocean.eu/content/download/21612/97924/file/NEMO_book_3_4.pdf. Accessed: 2016-12-28.Marinone, S., Pond, S., and Fyfe, J. (1996). A three-dimensional model of tide and wind-inducedresidual currents in the central Strait of Georgia, Canada. Estuarine, Coastal and Shelf Science,43(2):157–182.Masson, D. and Cummins, P. F. (2004). Observations and modeling of seasonal variability in theStraits of Georgia and Juan de Fuca. Journal of Marine Research, 62(4):491–516.Masunaga, E., Fringer, O. B., and Yamazaki, H. (2016). An observational and numerical study ofriver plume dynamics in Otsuchi Bay, Japan. Journal of Oceanography, 72(1):3–21.Mikelbank, B. A. (2001). Quantitative Geography: perspectives on spatial data analysis. Geograph-ical Analysis, 33(4):370–370.Nagai, T. and Hibiya, T. (2011). The processes of semi-enclosed basin–ocean water exchange acrossa tidal mixing zone. Journal of Oceanography, 67(4):533–539.96BibliographyOey, L.-Y. and Mellor, G. (1993). Subtidal variability of estuarine outflow, plume, and coastalcurrent: A model study. Journal of Physical Oceanography, 23(1):164–171.Ohlmann, J. C., White, P. F., Sybrandy, A. L., and Niiler, P. P. (2005). GPS-cellular driftertechnology for coastal ocean observing systems. Journal of Atmospheric and Oceanic Technology,22(9):1381–1388.Pawlowicz, R., Riche, O., and Halverson, M. (2007). The circulation and residence time of theStrait of Georgia using a simple mixing-box approach. Atmosphere-Ocean, 45(4):173–193.Pond, S. and Pickard, G. (1983). Introductory dynamical oceanography 2nd Edition. 327pp. Perga-mon Press Oxford.Reffray, G., Bourdalle-Badie, R., and Calone, C. (2015). Modelling turbulent vertical mixing sen-sitivity using a 1-d version of NEMO. Geoscientific Model Development, 8(1):69–86.Riche, O. (2011). Time-dependent inverse box-model for the estuarine circulation and primary pro-ductivity in the Strait of Georgia. PhD thesis, University of British Columbia. https://open.library.ubc.ca/cIRcle/collections/ubctheses/24/items/1.0053142. Accessed: 2016-12-28.Royer, L. and Emery, W. (1982). Variations of the Fraser River plume and their relationship toforcing by tide, wind and discharge. Atmosphere-Ocean, 20(4):357–372.Royer, L. and Emery, W. (1985). Computer simulations of the Fraser River plume. Journal ofMarine Research, 43(2):289–306.Smith, G. C., Roy, F., Mann, P., Dupont, F., Brasnett, B., Lemieux, J.-F., Laroche, S., and Be´lair,S. (2014). A new atmospheric dataset for forcing ice–ocean models: Evaluation of reforecasts97Bibliographyusing the Canadian global deterministic prediction system. Quarterly Journal of the Royal Me-teorological Society, 140(680):881–894.Soontiens, N., Allen, S. E., Latornell, D., Le Soue¨f, K., Machuca, I., Paquin, J.-P., Lu, Y., Thomp-son, K., and Korabel, V. (2016). Storm surges in the Strait of Georgia simulated with a regionalmodel. Atmosphere-Ocean, 54(1):1–21.Stronach, J. (1981). The Fraser River plume, Strait of Georgia. Ocean Management, 6(2):201–221.Tabata, S. (1972). The movement of Fraser River-influenced surface water in the Strait of Georgiaas deduced from a series of aerial photographs, Pacific Marine Sci. Technical report, Report.Tedford, E., Carpenter, J., Pawlowicz, R., Pieters, R., and Lawrence, G. A. (2009). Observationand analysis of shear instability in the Fraser River estuary. Journal of Geophysical Research:Oceans, 114:C11006.Thomson, R. E. (1975). Longshore current generation by internal waves in the Strait of Georgia.Canadian Journal of Earth Sciences, 12(3):472–488.Thomson, R. E. (1981). Oceanography of the British Columbia coast, volume 56. Canadian Spe-cial Publications of Fisheries and Aquatic Sciences. http://publications.gc.ca/site/eng/9.816310/publication.html. Accessed: 2016-12-28.Umlauf, L. and Burchard, H. (2003). A generic length-scale equation for geophysical turbulencemodels. Journal of Marine Research, 61(2):235–265.Unesco (1981). Background Papers and Supporting Data on the Practical Salinity Scale 1978.Technical Report 37, UNESCO, Paris, France.98Wang, C. (2015). Oxygen budgets and productivity estimates in the Strait of Georgia from a contin-uous ferry-based monitoring system. PhD thesis, The University of British Columbia (Vancou-ver). https://open.library.ubc.ca/cIRcle/collections/ubctheses/24/items/1.0167147.Accessed: 2016-12-28.Wu, H., Zhu, J., Shen, J., and Wang, H. (2011). Tidal modulation on the Changjiang River plumein summer. Journal of Geophysical Research: Oceans, 116:C08017.Yin, K., Harrison, P. J., Pond, S., and Beamish, R. J. (1995a). Entrainment of nitrate in the FraserRiver estuary and its biological implications. i. Effects of the salt wedge. Estuarine, Coastal andShelf Science, 40(5):505–528.Yin, K., Harrison, P. J., Pond, S., and Beamish, R. J. (1995b). Entrainment of nitrate in theFraser River estuary and its biological implications. ii. Effects of spring vs. neap tides and riverdischarge. Estuarine, Coastal and Shelf Science, 40(5):529–544.Yin, K., Harrison, P. J., Pond, S., and Beamish, R. J. (1995c). Entrainment of nitrate in the FraserRiver estuary and its biological implications. iii. Effects of winds. Estuarine, Coastal and ShelfScience, 40(5):545–558.99Appendix AInverse Distance WeightingInterpolation Method for Ferry-basedSalinity comparisonThe Inverse Distance Weighting Interpolation method (Chang, 2006) is applied to interpolate modelgrid points onto the ferry routes and acquire model near-surface salinity values along the track(S mod track) by using the following equation.S mod track(x) =∑9i=1 ki(x)S mod(i)∑9i=1 ki(x)(A.1)where ki(x) = d(x, xi)−1, x refers to the interpolated point on the ferry track, xi denotes aninterpolating point of model grid point, d is the distance from NEMO model grid point to theinterpolated point on the ferry and is calculated on a sphere based on their longitudes and latitudes(Mikelbank, 2001), i ranges from 1 to 9 as I first find the nearest model grid point from interpolatedpoint, and further select the other eight model grid points, which surround it. S mod(i) denotesmodeled salinity value for each model grid point based on a total nine model points. If there are100Appendix A. Inverse Distance Weighting Interpolation Method for Ferry-based Salinity comparisonmore than two model grid points out of these nine model points on land, this position is consideredas a land point. If there are fewer than two grid points out of these nine model points are on land,then this position is still used by averaging the available water grid points.101Appendix BTidal Amplitude Comparison Insidethe Fraser River ChannelTable B.1: Tidal amplitude comparisons inside Fraser River between extended river (bathymetry#5), extended and deepened river (bathymetry #6) and observationsStationnamesDistancefrommouth(km)Observations Bathymetry #5 Bathymetry #6Max(m)Min(m)Mean(m)Max(m)Min(m)Mean(m)Max(m)Min(m)Mean(m)Steveston 0 3.49 2.10 2.88 3.67 2.04 2.89 3.71 2.03 2.90Deas IslandTunnel18 3.05 1.85 2.58 1.46 1.01 1.27 3.55 2.04 2.80New West-minster36 2.28 1.43 1.91 0.84 0.57 0.71 3.02 1.78 2.42Mission 52 0.37 0.24 0.31 0.84 0.57 0.70 3.00 1.71 2.37102Appendix CExtended, Deepened River ChannelWith JettyThe real jetty is about 8.8 km long and 10-20 m wide. Bathymetry #10 was created by addingthe Steveston Jetty based on bathymetry #6 (Figure C.1). About 5% of the model grid points inthe deepened river mouth region are removed by adding the jetty. Compared to bathymetry #6,bathymetry #10 might impact tidal propagation around the river mouth. In the model, the widthof this jetty is about 500 m and length is 7 km in total. This jetty geometry does not seem toimprove the surface flows (Figure C.1, run #8, Table 2.1) compared to the longer and deeper riverchannel cases (Run #3b, Table 2.1). The reason could be the unrealistic width of this jetty in themodel rather than a few metres in the real world. The wide jetty blocks too much plume watergoing to the north.103Appendix C. Extended, Deepened River Channel With JettyFigure C.1: Bathymetry #10 with the Steveston jetty, longer and deeper bathymetry #6 as wellas comparison between the observed drifter and modeled particles with these bathymetries. Leftpanel: the same drifter track as Figure 2.12 shows. Middle two panels: bathymetry #10 with theSteveston Jetty based on bathymetry #6 in the upper panel and bathymetry #6 in the lower panel.Right two panel: modeled particles for run #3b (blue dots) in the upper panel and for run #8 (tealdots) in the lower panel, respectively. The black contours are bathymetry of 5 m and 10 m and thewhite contour is bathymetry of 20 m. Much more southward along-strait flow is generated by thejetty bathymetry.104Appendix DMean Surface CurrentsFigure D.1: Mean surface flows of HF radar data (left) and model mean surface currents (right,run #9, Table 2.1) during the same time period in October 1-31, 2015. Mesh color indicates thestrength of mean surface currents. Tendency of flows from the data is northward beween 49◦N and49.10◦N whereas model results show southward flows. Model flows are much weaker.105Appendix D. Mean Surface CurrentsFigure D.2: Mean surface currents of HF radar data (green) and model results with improvedconfigurations (red) from June 22-August 25, 2016. Direction of mean flows between observationand model in the area of 48.9-49.1◦N and -123.5-123.2◦E is basically the same. Strong modeledmean currents in the region north of 49.1◦N and west of the banks are seen. Figure credit: NancySoontiens and Mark Halverson.106Appendix D. Mean Surface CurrentsMean surface flows in the SoG are assumed to be caused by the mean outflow of fresh water,assuming tides and winds contribute very little response over periods of weeks and longer (Halversonand Pawlowicz, 2016). To isolate the mean surface currents, an average over a long analysis period isemployed (Halverson and Pawlowicz, 2016). The modeled mean surface flow of October 1-31, 2015is evaluated against HF radar data during the same time period. The results show almost oppositeflow direction between the modeled and observed currents where modeled mean surface currents gosouthward from 49.10◦N south but the observational mean currents indicate an anti-cyclonic eddymotion at around 49.10-49.15◦N (Figure D.1). Southward flow in the model is most likely due to theestuarine circulation in the SoG with freshwater flowing out of the Strait towards Haro Strait. It ismore difficult to understand the flow pattern of the observations although Halverson and Pawlowicz(2016) reluctantly attribute this pattern to bulge behaviour.Mean modeled surface currents with an improved configuration in summer shows more consis-tency in direction with HF radar data (Figure D.2), especially in the Fraser River plume outflowregion within 48.9-49.1◦N and -123.5-123.2◦E. The magnitude of modeled mean currents is similarwith HF radar data in the region within 48.9-49.1◦N and -123.5-123.2◦E although strong modeledmean currents in the region northern than 49.1◦N west of the banks are found.107Appendix EFroude Number at Peak FloodsThe internal wave speed (2.12) does not change much between peak flood and ebb in a given day(g, h and i of Figure 2.23 and Figure E.1). Internal wave speed is higher to the north of the rivermouth during high river flow periods (Figure 2.23 i, Figure E.1 i), decreasing down to 0.2-0.6 m s−1during the low and moderate river flow periods (Figure 2.23 g and h, Figure E.1 g and h). Overall,flow speed is weaker over the entire subdomain at peak flood compared to peak ebb.108Appendix E. Froude Number at Peak FloodsFigure E.1: Froude Number, momentum and internal wave speed at instantaneous peak flood inthree river flow periods. From left to right are Froude Number, momentum and internal wave speedspatial distribution at instantaneous peak flood. Teal color contour indicates the plume boundaryfor each period (2.1). From top to bottom are January 9, 2016, October 9, 2014 and May 31, 2015,respectively. Momentum is less significant at peak floods compared to peak ebbs. Internal wavespeed does not change much compared to peak ebb.109Appendix FCoriolis Impact on Surface Salinity inthe English BayIt is robust that the Coriolis force results in a fresher surface plume in English Bay and maximumsalinity difference can reach 15 in May (Figure F.1).110Appendix F. Coriolis Impact on Surface Salinity in the English BayFigure F.1: Time series of averaged surface salinity in the English Bay. Left panel: map of partof the SoG and the red box is where surface salinity is averaged over in English Bay. Right panel:time series of averaged surface salinity within the red box in English Bay at instantaneous peakebbs with the Coriolis force (all, green) and without the Coriolis force (nof, blue) during low (Jan,2016), moderate (October, 2014) and high (May, 2015) river flow periods.111Appendix GCentre Plume Location for the OtherTwo River Flow PeriodsWind impacts on the plume centre location during low and high river flow periods are shown (FigureG.1 and G.2). Similar with Figure 2.27, the movement of the centre of the plume location followsthe winds and is significant when wind magnitude is stronger than 5 m s−1.112Appendix G. Centre Plume Location for the Other Two River Flow PeriodsFigure G.1: Plume centre location with wind forcing and without wind forcing of wind events inJanuary, 2016. Plume centre location (star) as well as freshwater thickness (2.15) spatial distributionfor weak, moderate and strong wind events in January from top to bottom, and with wind forcing,without wind forcing, and their difference (With wind - without wind) from left to right. Thechange of the centre of the plume under weak wind conditions is small. Large changes occur duringmoderate and strong wind events.113Appendix G. Centre Plume Location for the Other Two River Flow PeriodsFigure G.2: Plume centre location with wind forcing and without wind forcing of wind events inMay, 2015. Plume centre location (star) as well as freshwater thickness (2.15) spatial distributionfor weak, moderate and strong wind events in May from top to bottom, and with wind forcing,without wind forcing, and their difference (With wind - without wind) from left to right. Thechange of the centre of the plume under weak wind conditions is small. Large changes occur duringmoderate and strong wind events.114Appendix HSurface Drifter TracksFigure H.1: 47 surface drifter tracks after 2.2 days released at Sand Heads. Red circle denotes theend position after 2.2 days duration. The drifters are released at random time during a tidal cycle.The tendency for most of the drifters is northward and the tracks pattern resemble a bulge-likecirculation. Figure credit: Romain Di Costanzo and Rich Pawlowicz.115

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0343600/manifest

Comment

Related Items