Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characterizing the lowest oxygen waters on the southern continental shelf off Vancouver Island Sahu, Saurav 2019

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


24-ubc_2020_may_sahu_saurav.pdf [ 12.74MB ]
JSON: 24-1.0385530.json
JSON-LD: 24-1.0385530-ld.json
RDF/XML (Pretty): 24-1.0385530-rdf.xml
RDF/JSON: 24-1.0385530-rdf.json
Turtle: 24-1.0385530-turtle.txt
N-Triples: 24-1.0385530-rdf-ntriples.txt
Original Record: 24-1.0385530-source.json
Full Text

Full Text

Characterizing the lowest oxygen waters on the southern continental shelf offVancouver IslandbySaurav SahuB.E. (Hons), Jadavpur University, 2013M.Tech, Indian Institute of Technology Kharagpur, 2016A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Oceanography)The University of British Columbia(Vancouver)November 2019c© Saurav Sahu, 2019The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoc-toral Studies for acceptance, the thesis entitled:Characterizing the lowest oxygen waters on the southern continental shelf off Vancouver Islandsubmitted by Saurav Sahu in partial fulfillment of the requirements for thedegree of MASTER OF SCIENCE in Oceanography.Examining Committee:Susan Allen, Earth, Ocean and Atmospheric Sciences, and Institute of Applied MathematicsSupervisorGonzalo S. Saldı´as, Earth, Ocean and Atmospheric SciencesSupervisory Committee MemberStephanie Waterman, Earth, Ocean and Atmospheric SciencesSupervisory Committee MemberPhilippe Tortell, Earth, Ocean and Atmospheric SciencesSupervisory Committee MemberCharles Hannah, Fisheries and Oceans CanadaSupervisory Committee MemberPhil Austin, Earth, Ocean and Atmospheric Sciences, and Institute of Applied MathematicsAdditional Supervisory Committee memberiiAbstractA shortage of dissolved oxygen in seawater can adversely impact marine life and ecosystems. Coastal wa-ters deeper than 100 m do not gain oxygen directly from the surface, and thus transient, seasonal, or permanentoxygen deficit conditions can occur in such deeper coastal waters. A low oxygen, dense pool of water is formedevery summer over the mid-shelf off southwest Vancouver Island in the Juan de Fuca Eddy region, known forits high primary productivity. In this thesis, the waters of the dense pool are traced back to their source usingLagrangian Particle tracking and a regional numerical ocean model, and upwelling hotspots of deep water lead-ing to the pool are discovered. The model accuracy in simulating the local circulation is evaluated based on astatistical skill score, root mean square error and bias upon comparing the model variables to the observations.The numerical model simulates the local circulation well, except it under-predicts the variation along isopyc-nals. Due to this lack of variation in the model, only the particles which agree well between the model resultsand the observations are selected using a K-Means clustering algorithm. Tracking these particles backwards intime showed that the dense pool under the Juan de Fuca Eddy is primarily composed of water from the Cali-fornia Undercurrent, water from Washington State shelf and offshore water. Signatures from mixing the sourcewaters in proportion closely approximate the final signatures of water inside the dense pool. The investigationof upwelling hotspots revealed that 1) the dense pool water primarily upwells through Spur Canyon and theconvoluted Juan de Fuca Canyon bathymetry near Swiftsure Bank, and 2) The southern side of Nitinat Canyonacts as a dominant upwelling site for water ending on the south-outer shelf of South Vancouver Island. Thisstudy helps in climate change predictions, as the pathways identified for different source water origins helpto determine the change in source composition over time and understand the potential low dissolved oxygenimplications over the Vancouver Island shelf, in Juan de Fuca Strait and in the Strait of Georgia.iiiLay SummaryA shortage of dissolved oxygen in seawater can adversely impact marine life and ecosystems. Coastalwaters deeper than 100 m do not gain oxygen directly from the atmosphere, and thus transient, seasonal, orpermanent oxygen deficit conditions can occur in such deeper coastal waters. This study answers the question:What are the sources and pathways of the low oxygen water that is observed every summer over the SouthVancouver Island shelf region? It was found that the poleward undercurrent flowing north along the continentalslope is the dominant contributor to the low oxygen deeper waters off South Vancouver Island shelf. Thisstudy helps in climate change predictions, as the pathways identified for different source water types helpto determine the change in source composition over time and understand the potential low dissolved oxygenimplications over the Vancouver Island shelf, in Juan de Fuca Strait and in the Strait of Georgia.ivPrefaceThis thesis is the original work of the author, Saurav Sahu. The research project presented in this thesiswas supervised by Susan Allen, who assisted with the model evaluation, interpretation of results, and thesismanuscript editing. The output of the numerical model NEP36, built using the Nucleus for European Mod-elling of the Ocean (NEMO) ocean modelling framework was provided to me by Li Zhai and Jean-PhilippePaquin (Fisheries and Oceans, Canada). This work is unpublished, but it is undergoing preparation for futurepublication.vContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vContents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Restriction on Cross-Shelf Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Upwelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3.1 Wind-Driven Coastal Ekman Vertical Transport . . . . . . . . . . . . . . . . . . . . . 31.3.2 Ekman Pumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3.3 Canyon-Driven Cross-Shelf Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3.4 Definition of Upwelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3.5 Other Mechanisms of Vertical Advection . . . . . . . . . . . . . . . . . . . . . . . . 41.3.6 Importance of Upwelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5vi1.4 West Coast of Vancouver Island . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4.1 Domain Description Emphasizing Local Bathymetry . . . . . . . . . . . . . . . . . . 61.4.2 Evidence and Importance of Upwelling on West Coast of Vancouver Island (WCVI) . . 71.4.3 Coastal Oceanic Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.4.4 Region Near the Juan de Fuca Eddy . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.5 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1 Observational Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.1 Scalar Observation Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.1.2 Vector Observation Data: Currents and Winds . . . . . . . . . . . . . . . . . . . . . . 122.2 Regional Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.2 Tracer Advection and Sub-Grid Scale Mixing . . . . . . . . . . . . . . . . . . . . . . 132.2.3 Model Forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.2.4 Initial and Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.5 Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3 Spice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.4 Particle Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.5 Classification of Source Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.1 Statistical Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173.2 Influence of the Local and Remote Winds on Undercurrent Strength Over the Slope . . . . . . 193.3 Model Prediction of the Dense Pool and Spice Front . . . . . . . . . . . . . . . . . . . . . . . 213.4 Choosing the Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.5 Relative Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.6 Source Water Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.7 Localised Regions of Upwelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30vii4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334.1 Tracing the Dense Pool Back to Its Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334.2 Dissolved Oxygen of Dense Pool From Source Water Signatures . . . . . . . . . . . . . . . . 364.3 Role of Canyons in Transporting Water Over the WCVI Shelf . . . . . . . . . . . . . . . . . . 384.4 Model Limitations and Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.4.1 Resolution of NEP36 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.4.2 Low Model Spice Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46A NEP36 Model Evaluations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55A.1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55A.2 Evaluation of Velocity Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56A.2.1 ADCP Data From Mooring A1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56A.2.2 Current Meter Data off Washington Coast . . . . . . . . . . . . . . . . . . . . . . . . 61A.3 Evaluation of Scalar Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67A.3.1 CTD Data From Pathways Cruise, 2013 . . . . . . . . . . . . . . . . . . . . . . . . . 67A.3.2 MVP Data from Pathways Cruise, 2013 . . . . . . . . . . . . . . . . . . . . . . . . . 76B Water Mass Characteristics for South Outer Shelf Water . . . . . . . . . . . . . . . . . . . . . 92C Water Mass Properties Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94viiiList of TablesTable 4.1 Dissolved oxygen concentration from source waters . . . . . . . . . . . . . . . . . . . . . 38Table A.1 Data used to evaluate the NEP36 model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Table C.1 Water mass properties table of source water masses contributing to the dense pool approxi-mated in terms of mean ± std . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94ixList of FiguresFigure 1.1 Bathymetry of the southern West Coast of Vancouver Island and northern Washington . . 6Figure 2.1 Classifying the sources of water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Figure 3.1 Statistical Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Figure 3.2 Wind Influence on undercurrent strength over the slope . . . . . . . . . . . . . . . . . . . 20Figure 3.3 Picture of pool and spice front . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Figure 3.4 Choosing the clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 3.5 Relative Contribution of source waters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Figure 3.6 Source water characteristics for “Dense Pool” from Particle Tracking . . . . . . . . . . . 28Figure 3.7 T-S diagram of source waters leading to “Dense Pool” . . . . . . . . . . . . . . . . . . . 30Figure 3.8 Localised regions of upwelling for “Dense Pool” . . . . . . . . . . . . . . . . . . . . . . 31Figure 3.9 Localised regions of upwelling for “South Outer Shelf Water” . . . . . . . . . . . . . . . 32Figure 4.1 Identification of the timing of upwelling of source waters . . . . . . . . . . . . . . . . . . 36Figure 4.2 Dissolved oxygen concentration from source waters . . . . . . . . . . . . . . . . . . . . . 37Figure B.1 Source water characteristics for “Souter outer shelf water” from particle tracking . . . . . 93xGlossaryCUC California UndercurrentCCS California Current SystemVICC Vancouver Island Coastal CurrentWCVI West Coast of Vancouver IslandxiAcknowledgmentsI want to express a deep sense of gratitude for my advisor Dr. Susan Allen for supporting me throughoutthis work. Susan has been incredibly supportive and offered not only her scientific knowledge and meticulousscrutiny to this thesis, but she has also helped me to settle in Canada and helped me in making significant careerdecisions. I would also like to thank my committee members, Dr. Gonzalo Saldias, Dr. Stephanie Waterman,Dr. Philippe Tortell, and Dr. Charles Hannah, for all the insightful discussions and helpful feedback.I want to thank everyone in the Waterhole, especially the Submarine Canyons-Arctic group working withSusan (Karina, Idalia, Birgit, and Gonzalo) for your suggestions during the weekly Friday meetings and theFriday lunch and beers. I would like to take this opportunity to thank Doug, who has solved many of mysoftware hiccups and helped me countless number of times. Also, thank you Sam, my work neighbour, for allthe wonderful discussions on football and numerical techniques.I am indebted to Ankita for her unwavering confidence in me and for supporting me each day during mytime at UBC. Thanks to you and Dhruv, I have so many happy memories and excursions of exploring the localmountains and the ocean. I look forward to making a whole lot more memories in the coming years. Dhruv,thank you for spending the summer of 2018 with me, a slightly turbulent time work-wise did not overly affectme personally because of your fun presence. This work and my life would be incomplete without mentioningthese people in my life, so here goes in no particular order: Brato, Spandan, Subhro, Pratyaya, and Sumit andmy school friend - Banhirup.I cannot describe in words, the appreciation towards my mom (Krishna) and dad (Ajit) for being the pillarsof support each day of my life. I want to dedicate this thesis to my parents for helping me reach here in life.You two have made the impossible possible.xiiChapter 1Introduction1.1 MotivationMarine life needs oxygen to perform aerobic metabolism. In waters low in dissolved oxygen, organismscan suffer from reduced growth rates, lower reproductive success, and higher mortality. The swimming per-formance of organisms also suffers, and this leads to, for example, a reduction in a fish’s ability to captureprey, escape mobile fishing gear, and avoid predators (Chabot and Claireaux, 2008; Ekau et al., 2010). Oceanicfield observations conducted in the California Current System (44.65◦ N to 44.00◦ N) during July-September2002 found a large number of dead fish and invertebrates in mid shelf waters near Oregon (Grantham et al.,2004). The death toll was due to the intensification of existing summer low oxygen oceanic conditions and fishtrying to seek refuge in shallow, atypical habitats (Grantham et al., 2004). The summer of 2006 was similar to2002, with the lowest oxygen conditions since 1950 existing over the Washington shelf (Connolly et al., 2010).Widespread mortality of benthic animals during the summers of 2002 (Grantham et al., 2004) and 2006 (Chanet al., 2008) due to severe low dissolved oxygen concentrations led to the shelf regions of the Northeast Pacificbeing termed “dead zones” (Bianucci et al., 2011).Conditions in the ocean are referred to as hypoxic conditions if the dissolved oxygen concentration lies be-low a certain threshold (∼1.4 mL L−1; (Gray et al., 2002)) or anoxic, if it is absent entirely. Hypoxic conditionsin shelf waters occur every summer over the Washington shelf region, but the coastal shelf regions of Oregonand Washington experienced the deadliest impact of low oxygen concentrations during the years of extremeoceanic conditions as in the summers of 2002 and 2006 (Chan et al., 2008; Connolly et al., 2010; Granthamet al., 2004). Concomitantly unprecedented anomalous changes in physical and biogeochemical properties of1the ocean were observed during the summer of those years (Connolly et al., 2010). Studies investigating hy-poxia in relatively shallow waters of the Oregon and Washington shelves (Chan et al., 2008; Connolly et al.,2010; Grantham et al., 2004) assign the persistently strong summertime winds, favouring vertical transport ofwater from deeper to shallower ocean depths and the local oxidation of organic matter as two prime causesfor hypoxia (Adams et al., 2013; Barth et al., 2007). Physical processes can initiate hypoxia over the shelf bytransporting oxygen depleted deeper waters upwards, but can also prevent the onset of anoxic conditions viathe advection and mixing of water masses having higher oxygen concentration (Adams et al., 2013). Thus anaccurate understanding of the coastal physical processes and knowledge of different water masses which findtheir way on the shelf is crucial for predicting the occurrence of summertime hypoxia in shelf waters of theNortheast Pacific.The shelf off southern Vancouver Island on the west coast of Canada is slightly different from the U.S.west coast in two aspects: 1) the equatorward summer winds are weaker and stronger episodic wind reversalsoccur during summer, and 2) the Vancouver Island Coastal Current (VICC), flowing north, nearshore alongthe Vancouver Island coast prevents denser hypoxic waters from approaching the coast (Bianucci et al., 2011).Although near-surface summertime hypoxia does not occur on the shelf region off the coast of southern Van-couver Island; this is not true for deeper shelf waters below the surface mixed layer. Historically waters withlow oxygen concentration (less than 1.5 mL L−1) are observed below the Juan de Fuca Eddy [Fig:1.1], foundover the Vancouver Island and Northern Washington shelf region (Freeland and Denman, 1982). The oxygenconcentration over the shelf region was lower than offshore (and deeper) locations on the same isopycnal andthe lowest along-isopycnal oxygen concentration (for the σθ ≈ 26.4 isopycnal) was recorded near the oceanbottom at LB08 [Fig:1.1], a sampling station situated in∼150 m depth and maintained by Fisheries and Oceans,Canada (Crawford and Pen˜a, 2013).Seasonal hypoxia in coastal ecosystems around the globe causes the ecological dynamics to be perturbedand has severe implications on sustaining commercial fisheries (Diaz, 2001; Naqvi et al., 2000; Rabalais andTurner, 2001). This thesis aims to improve the existing knowledge about the bottom waters near station LB08by identifying the sources of water which lead to this pool of low oxygen.In the following sections, I lead the reader through brief introductions of cross-shelf and vertical transportmechanisms, which have been widely studied in other regions. Subsequently, the reader is also introduced tothe prevailing coastal oceanic conditions off the West Coast of Vancouver Island (WCVI) that exist in summer.This chapter is concluded by defining the research questions that are answered in this study.21.2 Restriction on Cross-Shelf ExchangeThe coastal ocean separates the landmasses from the deep seas of the Earth. The shelf (in general, depthsless than 200 m), the shelf break (region of sharp seaward change in topographic gradient which is commonlyfound at about 200 m depth off WCVI and Northern Washington), and the slope (rapidly decreasing bottomdepth after the shelf break in the offshore direction) are common bathymetric constituents of the coastal oceanbottom. The coastal ocean is forced both locally and remotely by oceanic, atmospheric, bottom and terrestrialinteractions. Its response to the variety of forcing results in phenomena over a broad range of scales includingwaves, tides, vertical velocities, horizontal currents and jets, meanders, eddies, filaments, plumes, stratification,water mass and ice formation/transformation, internal dynamical instabilities, turbulence and mixing (Brinkand Robinson, 1998). The low frequency (sub-tidal) along-shelf currents over the continental shelf are usuallyan order of magnitude larger than the cross-shelf currents. This asymmetry is due to the tendency of along-shelf flow to occur along isobaths, roughly aligned along the coastline with larger scales in the along-shelfdirection as compared to the cross-shelf direction (Brink and Robinson, 1998). Dynamically this asymmetrycan be explained by the conservation of potential vorticity in a rotating fluid (Boyer et al., 2000). The Taylor-Proudman theorem applied to homogenous, geostrophic flow states that velocities can not vary with depth, thusconstraining the flow to occur along isobaths. This along isobath flow constraint, when extended to realisticstratified flows, holds true when the along-shelf length scale is much larger than the cross-shelf length scaleand mass conservation (or continuity) causes the cross-shelf velocity to be much smaller than the along-shelfvelocity (Brink and Robinson, 1998).1.3 Upwelling1.3.1 Wind-Driven Coastal Ekman Vertical TransportAlong-shelf wind stress drives a cross-shelf Ekman transport in the upper ocean (Csanady, 1982, 1981).During persistent southward wind conditions, with coast to the left in the Northern Hemisphere, a surfacedivergence occurs at the coast and to compensate for the loss of water near the surface, deeper subsurface waterrises near the coast. Thus the along-shelf wind induces vertical exchange. Owing to the associated drop incoastal sea level, partially counteracted by denser waters close to the coast, a cross-shelf pressure gradient isset up over a cross-shelf distance, determined by the Rossby radius of deformation. The Coriolis force balancesthis pressure gradient and this results in a coastal jet such as the summertime equatorward jet in the Northern3California Current System (CCS) (Checkley Jr and Barth, 2009).1.3.2 Ekman PumpingSpatial gradients in both along-shelf and cross-shelf wind stress (referred to as wind stress curl), may alsocause vertical transport by forcing divergence (or convergence) of horizontal Ekman transport. The subsequentvertical transport is called Ekman pumping (Sverdrup et al., 1942). Ekman pumping occurs further offshore ascompared to the coastal Ekman vertical exchange (Kraus and Businger, 1994) and is directly proportional tothe wind stress curl.1.3.3 Canyon-Driven Cross-Shelf ExchangeReal isobaths are rarely aligned parallel to the coast. When along-shelf flow encounters an abrupt topo-graphic change (short along-shelf length scale) such as a submarine canyon, the dynamic constraint of thetendency of flow to follow isobaths breaks down. The abrupt topographical change breaks down the constraintof geostrophy and the cross-shelf pressure gradient which was previously balanced by the Coriolis force, forcesthe flow along the axis of the canyon (Allen and Durrieu de Madron, 2009). Thus transport of water massescan occur to and from the open ocean along the axis of the canyon. I choose to call this type of exchange ascross-shelf break exchange, to differentiate it from a vertical exchange or vertical transport of water which hasbeen discussed previously.1.3.4 Definition of UpwellingTowards the purpose of ensuring clarity in defining upwelling for this work, I choose to define upwellingas the transport of water where a water parcel performs both cross-shelf break motion and vertical transportduring its trajectory. In other words, a water parcel has to be elevated from a certain depth in the water columnand concomitantly laterally cross the shelf break isobath for it to be defined as an upwelled water parcel.1.3.5 Other Mechanisms of Vertical AdvectionOther than wind-driven coastal Ekman vertical exchange and Ekman Pumping, there are other mechanismscausing vertical advection in the ocean such as tidally induced vertical advection (Thompson and Golding,1981), dynamic uplift for western boundary current systems (Tomczak and Godfrey, 2003) and transient upliftcaused by the passage of coastal trapped waves (Hormazabal et al., 2004; Shaffer et al., 1997). However thesewill not be further discussed in this work.41.3.6 Importance of UpwellingThe source of upwelled water is of significant interest, both dynamically and for its input to biogeochemicalcoastal budgets (Checkley Jr and Barth, 2009). If the source is at mid-depth away from the bottom boundarylayer as in case of strong stratified coastal systems (Huyer, 1976; Lentz and Chapman, 2004; Smith, 1981) theonshore flow carries the properties of subsurface pycnocline offshore waters; if the source of upwelled wateris in the bottom boundary layer, then it carries nutrient and trace metals due to its proximity to the seafloor(Johnson et al., 1999).When a cross-shelf flow is initiated by the presence of a shelf break discontinuity such as a submarinecanyon, these bathymetric features can act as conduits for transporting the nutrient-rich subsurface waters ontothe shelf; this water can then be subsequently uplifted by the wind-driven Ekman transport. Localized uplift ofnutrient-rich water into the euphotic zone provides an explanation for enhanced zooplankton biomass (Hickey,1995) and phytoplankton biomass (Allen and Durrieu de Madron, 2009) in the vicinity of some canyons. Thusupwelling can move nutrient-rich seawater from the slope region across the shelf break, onto the continentalshelf. This upwelled water, upon reaching the euphotic zone, accelerates the biological productivity of thatregion.The process of transporting water and trace gases like oxygen and carbon dioxide from the surface mixedlayer into the ocean interior is defined as ocean ventilation (Khatiwala et al., 2012). The dissolved oxygen at anymarine location depends primarily on age since ventilation and temperature of the water (Bograd et al., 2008;Chan et al., 2008; Whitney et al., 2007). Significant utilization often renders the oxygen concentration to be aminimum at mid-depth levels (∼1200 m) (Keeling et al., 2010; Sverdrup, 1938; Wyrtki, 1962). On the shelf,waters at lesser depth usually have higher concentrations because of more recent ventilation whereas deeperwaters (at ∼150 m depths) have lower concentration because of higher oxygen utilization. Local organismsmay suffer from stress or mortality if low oxygen conditions persist. Fish and crustacea are deemed to be mostsensitive to low oxygen conditions of all the marine taxa (Vaquer-Sunyer and Duarte, 2008).Based on the low dissolved oxygen concentration findings of Crawford and Pen˜a (2013) on the south Van-couver Island shelf, I lead the reader to explore the relevant local coastal oceanography of this region. Thefollowing section provides a brief description of how WCVI is influenced by cross-shelf as well as vertical ex-change of waters and description of the crucial aspects of its summertime coastal circulation. The subsectionends with a focus on influences of the coastal circulation on the Juan de Fuca Eddy region.51.4 West Coast of Vancouver Island1.4.1 Domain Description Emphasizing Local BathymetryThe western continental margin of Vancouver Island comprises a continental shelf, a steep slope followed bya broad continental rise. The shelf break, separating the slope from the shelf, is delineated by the 200 m contour.The region lies at the northern end of the CCS, a well researched eastern boundary current system (Hickey,1979, 1998). The shelf width narrows as one moves north from the Juan de Fuca Strait region [Fig:1.1]. Nearsouthern Vancouver Island, the bathymetry is more convoluted in shape with numerous shallow banks separatedby a series of roughly 100 m wide basins and the 400 m wide and 10 km long Juan de Fuca Canyon (Foremanet al., 2000b). North of Juan de Fuca Canyon are three shorter canyons, respectively, Nitinat, Barkley andClayoquot Canyons [Fig:1.1].Figure 1.1: Bathymetry of the southern West Coast of Vancouver Island coastal region and the coastal regionof northern Washington; the 200 m depth contour line approximates the shelf break. The shelf breakcanyons are labeled. The yellow circles identify the locations of the moving vessel profiler (MVP)data collected during the Pathways Cruise (2013) and the stars identify the location of mooring A1 andstation LB08. The Juan de Fuca Eddy region is shown as a blue shaded circular region over the shelf.Jdf denotes Juan de Fuca and C denotes canyons.61.4.2 Evidence and Importance of Upwelling on WCVISubmarine canyons incise the shelf break one after the other and modify the regional coastal circulationin this biologically productive region (Hickey, 1995). Current meter observations during summer (Huggettet al., 1987; Thomson et al., 1984) have shown significant onshore flow in the ocean interior (Freeland et al.,1984) off the WCVI. The upward vertical transport of water and onshore flow in the ocean interior duringsummer is caused by offshore Ekman transport (Freeland et al., 1984; Thomson and Gower, 1985) and thepresence of submarine canyons (Allen et al., 2001; Denman et al., 1981). Vertical transport can also be drivenby northward propagating shelf waves in the northern CCS (Battisti and Hickey, 1984). Southward flow over theouter continental shelf is associated with large vertical excursions of deep water near canyon heads (Allen et al.,2001). Moreover the flux of nitrate to the South Vancouver Island-Washington shelf through Barkley, Juan deFuca and Quinault Canyons is estimated to be ∼30-60 % of the local wind-driven nitrate flux contribution tothe euphotic zone (Connolly and Hickey, 2014).1.4.3 Coastal Oceanic CirculationThe two large Pacific Ocean atmospheric pressure systems are the Aleutian Low and the North PacificHigh (Thomson et al., 1989), located offshore near the Aleutian Island chain (centred around 60◦N, 160◦W)(Hood and Zimmerman, 1987) and northeast of Hawaii and west of California (centered around 35◦N, 140◦W)(Yokoi et al., 2009), respectively. The zone of separation between them migrates north and south dependingupon the relative strength of the two pressure systems. The separation zone lies near 50◦N during summer(northern Vancouver Island) and near 40◦N during winter (San Francisco) (Thomson and Ware, 1996). Theseasonal transition from one pattern to another is abrupt and causes spring and fall oceanic transitions (Strubet al., 1987; Thomson and Ware, 1996). The two large pressure systems drive two basin-scale Pacific gyrescalled the subtropical anticyclonic gyre and the sub-polar cyclonic gyre (Munk, 1950). The West Wind Driftor the Sub-Arctic current flows at the transition zone between these two large gyres and it divides into thepoleward flowing Alaska current and the equatorward flowing California Current near the continental marginof the eastern Pacific (Bograd et al., 1999; Tabata, 1975; Thomson, 1981). Offshore of south Vancouver Islandand northern Washington, the summer California current is replaced by the Alaska current in winter (Freelandand Cummins, 2005).In summer, over the shelf, the surface intensified, seasonal, wind-driven, baroclinic coastal jet called theshelf break current flows equatorward, and in winter, the nearly barotropic Davidson Current flows poleward7over the outer shelf (Thomson and Gower, 1998). The year-round California Undercurrent (CUC) flows pole-ward over the slope and shelf break (Connolly et al., 2014; Thomson and Krassovski, 2010). The CUC is char-acterized by high salinity, low dissolved oxygen and high nutrient content (Hickey, 1979; Huyer et al., 1989).For a 10-40 day period the temporal variability of CUC (at mooring A1 on the slope (∼500m depth shown inFig:1.1)) is coherent with winds further south (near Oregon) at a lag of 2±3 days (Thomson and Krassovski,2015). This relationship is consistent with the speed of poleward propagating coastal trapped waves (Thomsonand Krassovski, 2015). The outflow from Juan de Fuca Strait called the VICC (Hickey et al., 1991b; Thomsonet al., 1989) flows northward, close to the shore; the VICC is separated from the equator-bound shelf breakcurrent by a transition zone which narrows as one moves from the broader continental shelf of Barkley Soundnorth towards Estevan Point (Thomson et al., 1989).1.4.4 Region Near the Juan de Fuca EddyThe Juan de Fuca Eddy region is located offshore near southern Vancouver Island [Fig:1.1]. It is known tobe a site of high phytoplankton biomass (Trainer et al., 2002), high primary productivity (Marchetti et al., 2004)and enhanced higher trophic level biomass (Mackas and Robinson, 1997). Despite weaker upwelling favourablewind intensity in this region, as compared to the coast further south (Hickey, 1979), the highest productivityis recorded off the coasts of northern Washington and southern British Columbia (Hickey and Banas, 2003;Ware and Thomson, 2005). Ware and Thomson (2005) credited the high productivity in this region to (1) theyear-round freshwater inputs from the Columbia and Fraser Rivers making the upper water column stable, and(2) the supply of land derived nutrients. The Fraser River discharge drives the estuarine circulation in Juande Fuca Strait providing the freshwater source. The freshwater gets tidally mixed in shallow regions, withinthe strait, with the deep, nutrient-rich water flowing into the strait. The deepwater, entrained into the baseof the freshwater layer, contributes to the plankton production (Mackas et al., 1980). According to Crawfordand Dewey (1989) tidal mixing over the shelf contributes only a small fraction (less than 10%) of the flux ofnutrients supplied by the estuarine outflow from the Juan de Fuca Strait, which is the largest source of nutrients.The Juan de Fuca Eddy region is also known to be an initiation site for Pseudo-nitzschia blooms. Its neurotoxin,domoic acid harms benthic fisheries (MacFadyen et al., 2008; Trainer et al., 2002).The VICC is frequently described as a inner shelf current flowing north from the mouth of Juan de Fuca Strait(Hickey et al., 1991b; Thomson et al., 1989); however water mass analysis studies by Mackas et al. (1987) andresearch on coastal circulation carried out by Crawford (1988) and MacFadyen et al. (2005) showed cross-shelf8transport of the outflow from the strait due to the presence of the cyclonic Juan de Fuca Eddy near the mouth ofthe strait. The Juan de Fuca Eddy is a topographically arrested feature that develops in spring, recedes duringfall, and is not observed in winter (Freeland and Denman, 1982). A fraction of the estuarine outflow from theJuan de Fuca Strait gets advected around the eddy margin and the eddy grows through summer [∼15 km eddyradius in July to ∼35 km eddy radius in September] (MacFadyen et al., 2008). The eddy surface circulationis retentive (up to 32 days) during weak upwelling-favourable winds and frequent wind reversals, whereas it isleaky at its southern perimeter during typical southbound summer winds (MacFadyen et al., 2008). Upwellingnear Cape Flattery plays a role in the formation of the eddy after which it moves to the location, slightlynorthward of Spur Canyon, where it has been observed historically (Foreman et al., 2008). The influenceof CUC in near-surface eddy waters increases throughout the summer from June to September; this upwelledCUC and the cross-shelf advection of fresh Juan de Fuca Strait outflow make the northern CCS nutrient-rich(MacFadyen et al., 2008).A low oxygen pool of water is observed in August occupying ∼50 km of the shelf region below the surfacewaters of the Juan de Fuca Eddy (below ∼70 m depth) (Freeland and Denman, 1982). The dissolved oxygencontent of the pool ranged from 1.06 mL L−1 in the bottom waters (∼150 m) to 2 mL L−1 at 75 m depth(Freeland and Denman, 1982) and can be easily identified from the closed 2 mL L−1 dissolved oxygen isoplethand the closed 26.4 σt contour at 75 m depth (Freeland, 1983). This low oxygen, dense water found below theJuan de Fuca Eddy has led to postulates of upwelling of low oxygen CUC flowing over the continental slopeonto the shelf region (Foreman et al., 2008; Freeland and Denman, 1982; MacFadyen et al., 2008).The presence of well mixed, dense shelf water, near the head of Spur Canyon was also observed duringobservations conducted in August, 2013 (Klymak et al., 2013). This well mixed, dense shelf water was clearlydelineated in spice (Klymak et al., 2013). Seawater density depends on both the temperature and amount ofsalt in the water. Thus two water masses having different temperatures can have the same density if they havedifferent salinity. We call warm salty water “spicy” and cold fresh water “minty.” The well mixed, dense shelfwater below the Juan de Fuca Eddy was low in spice and separated from the high spice outer shelf water bya sharp front along the 26.4 potential density (σθ ) surface (Klymak et al., 2013). A water mass analysis studyconducted by Mackas et al. (1987), also found presence of CUC in deeper waters of the southern WCVI shelfalong with contributions from offshore water.91.5 Research QuestionsCUC has been observed in deeper waters of southern WCVI shelf and the upwelling of CUC onto the shelfthrough canyons has been previously postulated (Foreman et al., 2008; Freeland and Denman, 1982; Mac-Fadyen et al., 2008). At station LB08, bottom oxygen content matched the oxygen concentration at 450 mdepth on the adjacent slope region and the bottom water was similar in density to the water at the shelf break(∼200 m depth) further offshore (Crawford and Pen˜a, 2013). This leads us to think that upwelling could havea role in the formation of dense water near LB08. So this leaves a knowledge gap about the sources of densewater and their pathways onto the shelf, allowing us to ask the following questions:1. What are the sources of dense water near LB08?2. Are the canyons involved in upwelling that leads to the presence of dense waters near LB08? If the answeris yes, we choose to ask further: Which are the most important conduits, in particular, the pathways ontothe shelf and the upwelling hotspots?3. Why do the near-bottom waters at LB08, located in the region of the Juan de Fuca Eddy, have such lowdissolved oxygen concentration?Surveys conducted during the 2013 Pathways cruise (Klymak et al., 2013) observed the dense pool of waterbelow the Juan de Fuca Eddy, which was previously observed (Freeland and Denman, 1982) and simulated inmodeling studies of the eddy (Foreman et al., 2008; MacFadyen and Hickey, 2010). In this study, I track thepotential sources of this dense water observed on the shelf during summer and investigate the pathways of deepwater upwelled onto the shelf. In addition, this study aims to quantify the relative percentage of slope waterthat chooses a certain path in arriving at specific positions on the shelf off WCVI.I lead the reader to the following chapter describing the numerical model, particle tracking algorithm andthe analysis methods used for answering the research questions. The subsequent chapter details the findings ofthis study. The implications of the results are discussed in Chapter 4. The final chapter concludes this study,answering each of the research questions.10Chapter 2MethodsIn order to determine the source waters of the low oxygen pool under the Juan de Fuca Eddy, water particleswere tracked back in time based on the circulation from a numerical ocean model. The Eulerian circulationmodel and the Lagrangian tracking algorithm is described below, but first, the observations used to evaluatethe numerical model are introduced. This section also defines the variable “spice” used for analysis, and theapproach used for classifying different sources of water is described.2.1 Observational Data2.1.1 Scalar Observation DataThe Moving Vessel Profiler (MVP) survey during the Pathways 2013 Cruise (Klymak et al., 2013) wasconducted from 21 August 2013 to 30 August 2013. This was focused mainly on the western edge of theJuan de Fuca Eddy region (tracks are shown in Fig:1.1) between Swiftsure bank (48.55◦ N, 125.0◦ W) and LaPerouse Bank (48.67◦ N, 125.83◦W). The MVP is a towed profiler instrumented with a Seabird 39 conductivity-temperature-depth (CTD) sensor, a fluorometer, an optical backscatter sensor, and a dissolved oxygen sensor.Observational data presented in this study are primarily from the MVP surveys, although CTD data from aship-lowered Seabird 911-plus CTD were also used for evaluating the results of the numerical model. TheMVP data were processed as in Klymak et al. (2015), and the processed CTD cast data are maintained by theInstitute of Ocean Sciences Data Archive (, which are publicly available upon request to them.112.1.2 Vector Observation Data: Currents and WindsMooring A1 (48.53◦ N, 126.2◦ W), situated on the continental slope (∼500m depth), is located approxi-mately 60 km off southern Vancouver Island [Fig:1.1]. It is maintained by Richard E. Thomson of Fisheriesand Oceans, Canada. An Acoustic Doppler Current Profiler (ADCP) measures water velocities at A1, andthe data from 7 May 2013 to 31 August 2013 were used to compare the model simulated currents againstobserved velocities. The processed data were downloaded from the Institute of Ocean Sciences Data Archive( The observed and the model current velocities at this location were correlatedwith local and remote winds to understand how observed and modeled sub-surface currents at A1 respond to thewind forcing. Two periods were used: during the time of the cruise in August 2013 and from 28 February 2013till 31 August 2013. Both the winds and the current velocities were rotated to a principal axis system to obtainthe alongshore component at the respective stations, following Kosro (1987). The local winds, from the buoyat La Perouse Bank (station 46206) maintained by Environment and Climate Change, Canada, were obtainedfrom the Canadian wave data archive ( The remote wind data were obtained from the buoy (station 46050) at Stonewall bank (44.68◦ N& 124.51◦ W, west of Newport, Oregon) maintained by National Data Buoy Center (NDBC). The data weredownloaded from the NOAA-NDBC website ( Historical current meter data offthe coast of Washington from the Vancouver Island Coastal Current experiment (June to November 1984) wasalso compared with the model velocities during the summer months of 2013 to confirm the models ability toreplicate the coastal circulation in the region of study off the coast of Washington. The current meter data areavailable from the Institute of Ocean Sciences Data Archive ( Regional Model2.2.1 DescriptionThe numerical model output analyzed in this study is a regional and temporal subset (covering the WCVIregion shown in Fig:1.1 for the summer of 2013) of an updated version of the NEP36 model, previously devel-oped by Lu et al. (2017). The updated NEP36 version used in this study uses the same model bathymetry as Luet al. (2017) but with updated surface forcing and changes to the mixing parameterization. This hindcast wasrun from 2007 to 2016. Daily values of model output fields of temperature, salinity, horizontal velocities (Uand V) were utilized for the sliced WCVI domain from 28 February 2013 to 10 October 2013.12The horizontal model grids follow the tri-polar ORCA configuration (Barnier et al., 2007) with 1/36◦resolution in latitude/longitude (∼1.7 km average grid spacing). The sliced subset of the NEP36 model outputincludes 32 vertical levels from the surface to 541 m water depth with higher vertical resolution (∼1 m) close tothe surface to ∼87 m at 541 m depth. The model uses z-levels with partial cells near the bottom with thicknessvarying from a minimum of 6 m or 20% of the full size of the cell. Following Roullet and Madec (2000), themodel uses VVL (variable volume level) which allows the thickness of the vertical level to vary based on thechanges in sea surface elevation.The NEP36 model grid is closely aligned along geographical latitude and longitude. The regional domainin latitude and longitude [Fig:1.1] is well represented by the orthogonal model grid in x and y, respectively[Fig:2.1]. The subsequent figures in this study will be referenced relative to this model grid.2.2.2 Tracer Advection and Sub-Grid Scale MixingThe Total Variance Dissipation scheme (TVD) is used by the model for tracer advection as described byZalesak (1979). The vertical eddy diffusivity for tracers (temperature and salinity) and viscosity for momentumare computed using the General Length Scale (GLS) scheme (Umlauf and Burchard, 2003). This has beenupdated from the Lu et al. (2017) version of the model, which used the Mellor & Yamada 1.5 turbulence closurescheme (Axell, 2002; Blanke and Delecluse, 1993; Gaspar et al., 1990; Mellor and Blumberg, 2004; Mellor andYamada, 1982). The bottom friction is parameterized using a quadratic law with a fixed drag coefficient of5×10−3. The horizontal mixing of tracers and momentum is parameterized by a Laplacian scheme alongisopycnal and horizontal levels, respectively. Following Smagorinsky (1963), the horizontal viscosity is relatedto the horizontal velocity shear with 0.7 as the constant parameter. The background horizontal eddy diffusivityand viscosity are set to 20 m2s- Model ForcingThe surface of the NEP36 model is forced by hourly wind, sea level pressure, air temperature, specifichumidity, short and long wave radiation and freshwater fluxes taken from Climate Forecast System Version 2(CFSv2) (Saha et al., 2014) which differs from the Climate Forecast System Reanalysis (CFSR, Saha et al.(2010)) used by the previous version (Lu et al., 2017). Both CFSv2 and CFSR are products of the US NationalCentre for Environmental Prediction (NCEP), and CFSv2 can be considered a seamless extension of CFSR(Saha et al., 2014). Monthly climatology of Morrison et al. (2012) provides the river runoff at the topmost level13as surface precipitation with enhanced vertical mixing near the river mouths. There are six rivers included inthe WCVI sliced sub-domain namely the Alberni, San Juan, Hoko, Quinault, Bedwell and Gold Rivers.2.2.4 Initial and Boundary ConditionsThe model is initialized on 1 Jan 2007 using the daily mean temperature (T) and salinity (S), horizontalvelocities (U and V) and sea surface height (SSH) from PSY4, an operational global analysis product producedby Mercator Ocean (Lellouche et al., 2013). PSY4 has a resolution of 1/12◦ in latitude and longitude.PSY4 also provides daily averaged temperature, salinity, non tidal SSH and velocities at the model’s lateralopen boundaries. Tidal SSH and depth-averaged currents for 8 tidal components (K1, O1, P1, Q1, M2, S2,N2 and K2) are obtained from Webtide ( with a solution for the North East Pacific Ocean based on Foreman et al. (2000a). The lateral openboundaries use the radiation condition of Flather (1976) based on the PSY4 barotropic velocity normal to theboundary and the difference between PSY4 and modeled SSH. Baroclinic velocities, T and S are relaxed to theprescribed values from PSY4 over a width of 10 grid points. The relaxation parameter decreases from unity atthe lateral open boundary to zero at the tenth grid cell inside of the boundary.2.2.5 Model EvaluationThe model performance is evaluated based on a quantitative model skill, proposed by Willmott (1981), givenby WS = 1−MSE/(〈(|m−〈o〉|+ |o−〈o〉|〉)2), where the highest values WS = 1 implies perfect agreementbetween model and observation and a zero value means strong disagreement. The Willmott skill score takesinto account the mean bias between model simulated and observed variable values, besides considering thevariability about that mean bias. In the above equation, m and o are the model and the observed variablesrespectively and 〈〉 denotes a mean; MSE denotes the mean squared error given by (〈(m−o)2〉), RMSE denotesthe root mean squared error given by (〈(m−o)2〉)1/2 and the normalised bias is calculated as ((m−o)−〈 m−o〉)2/(o−〈 o〉)2. The model fields of temperature, salinity, density and spice are evaluated against the MVPand CTD observations collected during the Pathways 2013 cruise (Klymak et al., 2013). The model horizontalvelocities are evaluated against the observed currents at station A1. The metrics of model performance namely,the bias, RMSE and Willmott skill scores are calculated comparing the model variables against the observations.142.3 SpiceMunk (1981) referred to the difference between along isopycnal (same density) points in a property-property plot (T-S diagram) as “spice”; the varying heat content (balanced by altered salinity) was responsiblefor this difference. Defined by Flament (2002), the spice field is orthogonal to isopycnal surfaces, and thusspice can be utilized to convert the various combinations of potential temperature and salinity into a singleconserved quantity on a potential density surface. A more recent spiciness formulation (Roquet et al., 2015) ofspice given by McDougall and Krzysik (2015), has been used for this study. This definition of spice has beenused for investigating the water masses of the North-East Pacific (Davis et al., 2008; Pelland et al., 2013).2.4 Particle TrackingThe sources of water were determined by backward integration of particle trajectories using ARIANE, anoffline mass-preserving particle tracking algorithm (Blanke and Raynaud, 1997; Do¨o¨s, 1995)( grima/Ariane/), corrected for use with variable volume (VVL). Using the back-ground NEP36 model velocity fields, ARIANE calculates the trajectories analytically within each grid cellbased on 3-D linearly interpolated velocities. For each time step, the salinity, temperature, and density of thewater parcel are also computed based on the Eulerian NEP36 tracer variables. The change in T and S along aparticle trajectory includes the changes through mixing from the parent model. Selected particles are releasedon 26 August 2013 at their respective grid locations and traced back to their sources six months before theseeding date. The individual particle trajectories are sensitive to change of the date of seeding of the particlesand the tracks are observed to change when they are released a day earlier or later but percentage contributionsfrom individual source waters remains within 1%.2.5 Classification of Source WaterThe origins of source water are classified based on the geographical origin and known source water names.For this classification, a region is selected inside of the model domain that surrounds the low oxygen densepool. Both the south water and CUC water enter the region from the southern edge, and the difference in theirclassification is made based on their starting source depths. If the particles entering the south edge originatedfrom a depth deeper than 200 m, they are classified as CUC whereas if they originated from a depth shallowerthan 200 m they are called south water. If the particles are still inside the grid box after 6-month of particleback tracking, they are classified as local water. The particles which originated in the Juan de Fuca Strait are15called the strait outflow water, and similarly, the north and offshore water are so named [Fig:2.1].Figure 2.1: Study area showing the box used to track the different sources of water. The labelled arrowsindicate the entry of a source of water across the relevant edge of the regional box in blue; The southwater and CUC cross the same boundary but they are classified differently based on their source depthbeing less than or deeper than 200 m, respectively.16Chapter 3Results3.1 Statistical Model EvaluationThe NEP36 model performance is evaluated by comparing the model daily averaged temperature and salin-ity fields to the observation data, from the Pathways MVP [Fig:3.1], at corresponding dates and locations. Thecomparison provides high Willmott skill scores of 0.94 and 0.93, low root mean squared error (RMSE) of1.09 ◦C and 0.38 PSU, as well as low bias values of 0.19 ◦C and 0.33 PSU for temperature and salinity, re-spectively. When compared to the Willmott skill scores for different regional models evaluated for temperatureand salinity, for northern CCS (Giddings et al., 2014; Stone et al., 2018), the similarly high Wilmott skill scoresfrom NEP36 evaluations [Fig:3.1 a and b] indicate a reasonably accurate daily estimate of the fields. The modelis slightly warmer and fresher close to the surface compared to the observations as evidenced by the major-ity of the near-surface points lying above and below the one-to-one line, respectively [Fig:3.1 a and b]. Thisnear-surface bias is not expected to impact our analysis, as this study attempts to track the dense pool (found atdepths > 70 m) back to its sources, and it will be revealed in section 3.6, that most of the sources contributingto it originate from even deeper depths (> 100 m). On a T-S diagram [Fig:3.1 c], the model is observed tosimulate the mixing line, seen along the σθ ≈ 26.4 contour, similar to the MVP observations. However, thevariation in spice [seen in the spread of points along σθ contour in fig:3.1 c] in the model is much smaller thanthe MVP observations along 26.4 σθ , indicating a much larger along-isopycnal spice variability in the observa-tions as compared to the numerical model. The along-isopycnal model and observed spice are further analyzedin section 3.4.17Figure 3.1: Scatter plot of a) model temperature against observed temperature and b) model salinity againstobserved salinity, during the dates of the Pathways Cruise, August 2013. The scatter points are coloredby their depth. c) The same points plotted on a T-S diagram, with the model points in blue and theMVP observations in magenta. The solid and dotted contour lines in c) indicate contours of constantpotential density (σθ ) and spice, respectively. The black contour lines with labels in a) and b) representthe density of the scatter points. The contour values represent the fraction of points that lie outside thecontour.The model performance was also evaluated against the velocity observations from mooring A1 [AppendixA] resulting in a good Willmott skill score (0.71) and a low root mean squared error (0.08 ms−1) for thealongshore velocity, at 150 m depth. The 0.71 value of Willmott skill score indicated a reasonable match of themodel alongshore velocity with the observations, when compared to the similar skill scores from other modelsevaluated at location A1 (Connolly et al., 2014). The magnitude of observed velocities are much larger than themodel velocities at depths deeper than 200 m at A1. This discrepancy is because poleward flow over the slope18(CUC) in the model is slightly shifted offshore of A1 (approximately 2 model grid points offshore ≈3.4 km) atthese depths. As a result the model velocities (at depths >=200 m) at ≈3.4 km offshore of A1 compare muchbetter with the observed velocities at A1, at those depths [Appendix A]. The model successfully reproducedthe depth (∼120 m) at which summertime currents switch from equatorward to poleward at slope station A1.The model was also successful in replicating the historical summertime current directions observed over theWashington shelf from current meter data [Appendix A]. As we will see in section 3.5, much of the water inthe dense pool comes from the south of the Vancouver Island shelf, including the Washington shelf. The modelstratification was similar to the observed stratification confirmed from high Willmott skill score (0.95), low bias(0.22 σθ ) and low RMSE (0.43 σθ ) between model density and density observed by the Pathways MVP, andalso from comparing the model temperature, salinity and density depth-profiles with CTD data collected duringthe Pathways cruise [Appendix A]. The reader is referred to Appendix A for a more rigorous and detailedevaluation of the NEP36 model for our regional domain.3.2 Influence of the Local and Remote Winds on Undercurrent Strength Overthe SlopeTo evaluate the response of model currents to local and remote wind forcing, as compared to the observa-tions, the winds and currents (model and observed) were studied from 28 February 2013 to 31 August 2013.The local and remote winds correlate with both the simulated currents and observed currents at A1. From June2013 to August 2013, the alongshore flow (model and observed) at A1 is mostly southward, down to a depthof ∼100 m. At depths deeper than 100 m, the direction of currents switches to predominantly northward withmaximum northward current speeds observed at depths of ∼250 m [Appendix A]. Current vectors and windswere rotated to the direction of their principal axis, assumed to represent the alongshore direction. The un-dercurrent transport, defined as the strength of the poleward current below the southward flow, was calculatedasCUCs(t) =∫ Ddv(z, t)dz (3.1)where CUCs(t) is the California Undercurrent transport (m2s-1) at time t, v(z, t) is the poleward velocity (m s-1)at depth z (m) and time t, d is the depth at which the current turns poleward from the equatorward surfacecurrents, and D is the bottom depth at A1 (∼500 m). Alongshore wind stress was calculated from the alongshorewinds at La Perouse (station 46206) and Newport (station 46050) using the drag formulation of Large and19Pond (1981). The local (La Perouse) and remote wind (Newport) wind stresses are observed to have a visualcoherence with both the observed and modeled undercurrent transport at A1 [Fig:3.2 a and b]. The modeltransport is in phase with the observed transport but underestimates the observed transport [Fig:3.2 a and b].The model transport is lower than the observed transport at A1, because the model poleward velocities (atdepths deeper than 200 m) are slightly shifted offshore [Appendix A].Figure 3.2: Time series of alongshore wind stress and undercurrent transport at station A1 a) for the monthof August 2013 and b) for a period of 6 months from March to September 2013; the time series of thewind stress are given in shades of brown while the time series of transport are shown in blue; the twovertical red lines mark the start and end dates for collecting MVP observations during the PathwaysCruise 2013. Note the visual coherence between winds and the undercurrent strength at A1.During August 2013, a lag-lead correlation analysis conducted on the model current strength versus theNewport and La Perouse wind stresses resulted in 0.69 and 0.59 Pearson correlation coefficient, respectively20with a lag of 1 day. The model undercurrent strength at A1 thus correlates better with the remote winds atNewport than the local winds at La Perouse. However, the transport calculated from the ADCP correlates betterand equally with both the remote and local winds, evidenced by the observed transport having 0.76 and 0.75correlation with the same lag of 1 day with the remote and local winds, respectively. The local winds werethemselves correlated with the remote winds with a 0.6 correlation. This correlation between the winds (localand remote) with the transport (model and observed) at A1 suggests that both the remote and local winds forcethe undercurrent strength on the WCVI slope for the month of August, 2013.The same lag-lead correlation analysis was also performed for 6 months from 28 February to 31 August2013 for the model transport versus wind stress at remote and local locations. This resulted in 0.52 and 0.44Pearson correlation coefficient with a lag of 1 day for the Newport and La Perouse wind stresses, respectively.On the other hand, the undercurrent transport calculated from the ADCP gave 0.48 and 0.4 correlation with thesame lag of 1 day from 5 May to August 2013 (the ADCP data was not recorded for the months of April, Marchand February in 2013). Similar to the analysis done for the month of August 2013, the model undercurrenttransport correlates better with the remote winds at Newport than the local winds at La Perouse; however, theADCP currents do not have as high a correlation as observed for the month of August when the longer timeseries is considered. The ADCP correlates similarly with the local and remote winds as the model undercurrentstrength does for the extended time series.3.3 Model Prediction of the Dense Pool and Spice FrontThe model’s ability to simulate the observed well mixed, dense water near LB08, delineated in along-isopycnal spice (Klymak et al., 2013) is analyzed. The model successfully identifies the pool of dense waternear station LB08 [Fig:1.1]. The structure of the eddy, near the surface, is revealed by closed fresher isopycnalsnear the surface [Fig:3.3 a] in a vertical cross-section chosen through LB08 (green line in Fig:3.3 b). Movingtowards the shelf from further offshore, the spice transition at around 70 m depth (below the depths at which theJuan de Fuca Eddy is observed (Foreman et al., 2008; Freeland and Denman, 1982)) is coincident with a domeof isopycnals (σθ >= 26.3 in Fig:3.3 a) that indicates the pool of dense water, observed during the PathwaysCruise (Klymak et al., 2013). The model also successfully identifies the spatial front in spice observed alongthe 26.4 isopycnal (Klymak et al., 2013), near the region of the shelf break [Fig:3.3 b]. The water offshorealong this isopycnal is spicier than the water in the dense pool over the shelf near LB08 [Fig:3.3 b]. Additionalcomparisons of the model predicted spice pool and front to the MVP observations from the cruise are included21in Appendix A.Figure 3.3: a) Model calculated spice (in colour) and potential density (as black contour lines) averaged forthe month of August, 2013. b) Model spice on the 26.4 isopycnal is represented in color for the monthof August 2013 with depth contours in black. The green dashed line through LB08 in b) marks theregion of the vertical cross-section in a)3.4 Choosing the ParticlesThe model successfully simulates the dense pool of water [Fig:3.3 a] and the along-isopycnal spice front[Fig:3.3 b]. However, the along-isopycnal spice variation in the model results is an order of magnitude lessthan what is observed in the MVP observations for each of the three selected isopycnals 26.3, 26.4 and 26.5[Fig:3.4 a, b and c ]. These isopycnals are chosen, because the model simulates the dense pool as the domeof isopycnals given by σθ >= 26.3 with very little change in density beneath the σθ ≈ 26.5 isopycnal overthe shelf [Fig:3.3 a], and also because the model successfully identifies the dense pool as a shelf region of lowalong-isopycnal spice on the σθ ≈ 26.4 isopycnal [Fig:3.3 b] similar to Klymak et al. (2013). The particlesselected for tracking must represent the dense pool, in both the model results and observations to identify thesources of dense pool water correctly. In other words, given the noted differences in along-isopycnal spicein the model versus observations, the particles of dense pool selected for tracking are the subset that show aconsistent low along-isopycnal spice value in both the predicted model results and in the observations. Towardsthis intent, the model versus observed spice space values were clustered using a simple KMeans clusteringalgorithm (MacQueen et al., 1967), which assigns each data point to the closest corresponding centroid using22standard Euclidean distance [Fig:3.4 d, e and f]. Thus it was possible to extract the true positive clusters. Atrue positive cluster in this context conveys the same characteristic information (along-isopycnal spice) in bothmodel results and observations. In simpler words, one true positive cluster is particles that have both low spicepredicted by the model and low spice observed (blue cluster in Fig:3.4), and the other is particles that have highspice predicted by the model and high spice observed (orange cluster in Fig:3.4). This clustering was separatelyperformed for each of the three selected isopycnals.23Figure 3.4: The columns (a,d,g), (b,e,h) and (c,f,i) represent the analysis for each of the isopycnals 26.3,26.4 and 26.5, respectively. The plots in the top row (a,b,c) represent the model vs. observed spicespace for each of the three isopycnals and how the points are distributed in the bi-variate space with theupper and right distributions representing the uni-variate tendencies for the model and observed spiceindividually. The middle row (d,e,f) shows the clustering of the points based on K-Means clusteringalgorithm with the line representing a linear fit of the points in model-observed spice space. The bottomrow (g,h,i) reveals the location of the clusters in the model domain. The blue and orange clusters showa consistent relative along-isopycnal spice value in the model results and the observations and are thusselected as the particles for tracking.24Location of the Clusters:The location of the clusters in our model domain is determined, and the low-low spice cluster lies in theLB08 region for all three isopycnals. We chose to call the low-low spice cluster “dense pool” water as theseparticles identify the dense pool of water below the Juan de Fuca Eddy, both in the model and the observations.The other true positive cluster was seen to mostly occur on the south outer shelf region. For the south outer shelfregion [Fig:3.4 g,h and i ], only those locations located over the shelf (particle x index between 590 and 615)were chosen to be called “south outer shelf water” and the points which lay offshore beyond the shelf break werenot considered further. Particles from the dense pool and particles from the south outer shelf water were trackedbackward in time using ARIANE to trace the location of their sources. Particles were traced backward from 26August 2013 (end of the cruise) to their positions six months before (as of 28 February 2013). The particles forthe dense pool (or the south outer shelf water) were chosen to belong to the blue (or orange) locations on eachof the three isopycnals 26.3, 26.4 and 26.5. The location of the true positive clusters is slightly different foreach of the three isopycnals [Fig:3.4]. In addition to selecting the particles belonging to the three isopycnals,select points in between these three isopycnals are chosen. The points in between isopycnals are chosen only ifa model vertical grid cell is present between two of the isopycnals. That particle would be seeded for trackingif it lay vertically below the true positive cluster of the isopycnal above (blue for dense pool or orange for southouter shelf water). To help explain this let us consider the example, if we consider the 26.3 isopycnal to be120 m deep and the 26.4 isopycnal to be 130 m deep at the same location and, that at the same spatial locationin the domain, there is a model vertical grid point at 125 m. Then this particle would be selected for trackingif it was found to lie vertically below a true positive cluster location for the 26.3 isopycnal. In addition the truepositive particle belonging to the 26.3 isopycnal would be tracked.3.5 Relative ContributionsA classification of source waters using the geographical boundaries in the model domain [Fig::2.1] is usedin quantifying the relative contribution of the source waters leading to the “dense pool” and south outer shelfwater. The dominant source water contributions towards both the dense pool and the south outer shelf watercomes from the CUC, south and offshore waters. The CUC is the highest contributor to the dense pool, and thesouth water is the dominant contributor to the south outer shelf water [Fig:3.5]. The strait outflow, north waterand the local water contributions to the dense pool and south outer shelf water are smaller compared to the otherthree.25Figure 3.5: Relative contribution of origins of source water, towards each of the two types of final shelf waterbased on the particle tracking. Waters from the CUC, south and offshore are the dominant contributorstowards both the dense pool and the south outer shelf water.3.6 Source Water CharacteristicsFor each of the Lagrangian parcels, the source signatures of temperature, salinity, density, spice, sourcedepth, and trajectory time to the final location are investigated. It is important to know the time taken by aparticle in reaching its final location over the shelf as a longer trajectory time could imply more modification dueto mixing during its journey. A kernel density estimate (KDE) is computed for each of the finite data samplesof source water properties of temperature, salinity, density, spice, starting depths, and travel time in days. Theprobability density of the individual source waters is weighted based on their percentage contributions such thatthe combined area of all the source waters for each property is 1. The much widely used rule of thumb “Scott”bandwidth is used to select the bandwidth for the Gaussian kernel (Scott, 2015). The kernel density estimatehelps in estimating the characteristics of the source properties by determining a probability density functionfrom a set of finite data samples akin to a continuous variable. The final signature of the “dense pool” in termsof temperature, salinity, and density is closely approximated, within a standard deviation, by mixing the sourcewater signatures utilizing the relative fractional contributions of each source water type following Mamayev(2010). The spread (standard deviation) of the approximated final distribution of signatures is calculated based26on the formula of standard deviation propagation (Rouaud, 2013).S f =n∑i=1mi.Si (3.2)σ2f =n∑i=1mi.(σi)2 (3.3)where S f is the mean of the final signature calculated based on the fraction of the individual source waters (mi)and their mean signatures (Si). Similarly, the standard deviation of the final signatures σ f is computed usingthe standard deviation of the individual source signatures σi, and n = 6 is the total number of source waters.27Figure 3.6: Source water characteristics from particle tracking: Panels (d), (e), (j), (k), (f) and (l) representthe shaded distribution from the kde estimates of individual source water characteristics of temperature,salinity, density, spice, source depths and trajectory time, respectively leading to the final dense pool,shaded in grey in panels (a), (b), (c), (g) and (h). The distribution in red lines in each of (a), (b), (g)and (h) represents the final distribution of the dense pool, estimated from kde estimates of source watersignatures using equation:3.2 and equation:3.3. The vertical red (dotted) line represents the estimatedmean of the final dense pool signature and the vertical gray (dotted) line represents the sample mean.28The CUC is colder and more saline than the other source waters and possesses a distinct signature in allthe properties of temperature, salinity, density and spice [Fig:3.6 (d), (e), (j), (k) and Fig:3.7]. It also has themaximum density and high spice among the source waters, but the majority of the south water (shallower thanthe CUC) has higher spice than the CUC. The strait outflow exits the Juan de Fuca Strait at depths shallowerthan 50 m and is significantly warmer and fresher than the other source waters [Fig:3.6 and Fig:3.7]. Thenorth water sources are at ∼100 m and the offshore source depths are mostly ∼150 m. The distinct peaksin distributions of trajectory time (in days) could imply the tendency to choose specific paths, during theirtravel, to arrive at the shelf in batches at specific times of the year; this suggests the possible existence oflocalized regions of upwelling, which is discussed in the next section. The distinct peaks in the distributionsof trajectory time [Fig:3.6 l] exactly coincide to generally upwelling winds being interrupted by northerly windreversals and relaxations [Fig:3.2 b]. The spring oceanic transition in WCVI, which marks the transition from adownwelling favorable wind regime to an upwelling favorable wind regime occurred in early April for the year2013 (Hourston and Thomson, 2018). Almost half of the offshore water took more than 150 days to arrive atthe shelf locations [Fig:3.6 l]; this fraction of offshore water left before the spring oceanic transition (Thomsonand Ware, 1996). These waters, which entered our box before the spring transition, were inside the boundingbox [Fig:2.1] during summer and would have been local water as per our classification of source waters if theparticles were only tracked for the summer months. The wide variation of spice of the offshore water [Fig:3.6k] leads to the wider spread in the estimated spice distribution [Fig:3.6 h] calculated from mixing the sourcewaters. The reader is referred to Appendix B for the distributions of the source water properties for the southouter shelf water which are very similar to those of the dense pool. One of the important differences betweenthe distributions is that most of the offshore water contributing to the south outer shelf water takes over 120days to arrive at the final shelf locations and the distribution of south water contributing to the south outer shelfwater possesses a much wider spread in salinity and thus in density and spice as compared to the south watercontributing to the dense pool.29Figure 3.7: Temperature and Salinity of the individual source waters represented on a T-S diagram. Theellipses are the 95% confidence covariance error ellipses representing a characteristic source water.The CUC water is cold, salty and well mixed, since the CUC error ellipse almost forms a straight line inthe T-S diagram. The water from the Strait Outflow is warmer and fresher than the other source waters.3.7 Localised Regions of UpwellingBased on our definition of upwelling, for a Lagrangian water parcel to be considered an upwelled waterparcel it must rise from a depth deeper than 200 m and laterally cross the 200 m isobath. This definition ensuresthat upwelling combines vertical transport with cross-shelf motion of the water parcel. The percentages of par-ticles that were upwelled were calculated to be 56% and 36% for the dense pool and the south outer shelf water,respectively out of the total number of particles that were tracked. The locations where the upwelled waterscrossed the shelf break isobath (200 m) and the locations of their depth change from deeper than 200 m to shal-lower than 200 m are shown in (a) and (b), respectively in each of Fig:3.8 and Fig:3.9. Shelf break irregularitiessuch as the presence of Spur, Juan de Fuca, Barkley & Nitinat Canyons act as hot spots of upwelling for theparticles. The upwelled dense pool predominantly chooses Spur Canyon and the north rim of the Juan de FucaCanyon (near Swiftsure Bank) as a site for final crossing of the 200 m isobath [Fig:3.8 a]. The significant depth30change occurs inside each of Barkley, Nitinat, Juan de Fuca and Spur Canyons prior to crossing the shelf breakisobath [Fig:3.8 b].Figure 3.8: Upwelling regions relevant to the pathways of upwelled dense pool water identified by twocriteria a) shelf break crossing b) depth change of particles from deeper to shallower than 200 m,expressed in terms of percentage of total upwelled particles leading to the dense pool.The southern side of Nitinat Canyon acts as the primary region of shelf break crossing for the south outershelf water and both Barkley and Nitinat Canyons act equally as regions of depth change for the upwelledsouth outer shelf water [Fig:3.9]. The localization of shelf break crossing and depth change of the upwelledwater parcels near canyons indicates the dominant role played by the submarine canyons of the West Coast ofVancouver Island towards the formation of deep shelf water off southern WCVI.31Figure 3.9: Upwelling regions relevant to the pathways of upwelled south outer shelf water identified bytwo criteria a) shelf break crossing b) depth change of particles from deeper to shallower than 200 m,expressed in terms of percentage of total upwelled particles leading to south outer shelf water.32Chapter 4DiscussionThis study aims to describe the sources of water that contribute to the dense pool observed on the WCVI shelfnear LB08 during summer and the path taken by the source waters to reach this portion of the shelf. This sectionprovides a review of the source waters obtained by tracing the shelf water backward in time and compares thesignatures of the source waters with previous regional observational, modeling and water mass analysis studies.The implications of the different sources of water on the expected dissolved oxygen concentration in the densepool, in the context of possible local hypoxia are also discussed. Finally, this section provides context on theregional impact of local canyons, causing upwelling of deeper waters onto the shelf region.4.1 Tracing the Dense Pool Back to Its SourceThe joint source water and transit-time distributions contribute crucial information about the transport pro-cesses in the flow (Haine and Hall, 2002). Thus this study outlines the importance of Lagrangian particletracking in characterizing water types and quantifying the source waters leading to the dense pool near LB08.While the surface structure of the eddy and its impact on the Washington shelf was studied using drifters anda diagnostic finite element circulation model by MacFadyen et al. (2008, 2005), our study found the source ofthe dense pool of water that exists below the eddy. The dense pool of water was first identified by Freeland andDenman (1982) as a pool of low oxygen appearing on the shelf and the authors hypothesized that presence ofthe eddy near the head of Spur Canyon plays a role in drawing deeper waters, which are low in oxygen and highin nutrients, up onto the shelf. Our efforts in quantifying the relative percentages of source water contributionshow that besides CUC water being a dominant contributor to the dense waters of the pool, the pool also hasrelatively significant contributions from south and offshore waters [Fig:3.5]. The contribution of the south wa-33ter was not considered in the previous water mass analysis of Mackas et al. (1987) in their examination of thetracer concentrations from the dominant water masses affecting the coastal WCVI. The temperature and salinityvalues of CUC waters observed by Mackas et al. (1987) matches the source signatures of CUC waters in thisstudy [Fig:3.6 and Appendix C]. The south water originates south of the geographical boundary considered forour analysis [Fig:2.1] and has source depths shallower than 200 m as compared to the CUC, which starts fromdepths deeper than 200 m. This study finds a high proportion of south water (22.5 %) in the dense pool. Thiswater is warmer and fresher than the CUC [Fig:3.6 d,e] and possesses a higher dissolved oxygen concentrationas well (which we see later). The south water arrives at WCVI from the Washington shelf region during similartimes as the CUC [Fig:3.6 l].Although the cross-shelf advection of the VICC has been historically observed to contribute to the sur-face structure of the eddy (Crawford, 1988; MacFadyen et al., 2008, 2005), the outflow from the strait onlycontributes 2 % towards the dense pool of water found below the eddy. The water from the north VancouverIsland shelf contributes 5 % towards the dense water of the pool and also towards the south outer shelf water[Fig:3.5]. The surface intensified summertime shelf break jet (Hickey, 1998; Ikeda and Emery, 1984; Thomsonand Gower, 1998), flowing southward, does not contribute significantly towards deeper shelf waters off south-ern WCVI. This study finds that Spur Canyon and the sharp bathymetric curvature in Juan de Fuca Canyon nearSwiftsure Bank are the major shelf break crossings for upwelled water in the dense pool [Fig:3.8a]. The shelfbreak canyons such as Barkley and Nitinat Canyon also act as conduits for vertical transport of the upwelledwater parcels leading to the dense pool [Fig:3.8b].Deep Water Inflow Into the Juan De Fuca Strait:In a seasonal water mass analysis performed for the Juan de Fuca Strait and Strait of Georgia, Masson (2006)referred to inflowing dense water at the mouth of Juan de Fuca Strait as deep inflow from the shelf. Its tem-perature and salinity signatures of 6.4 ◦C and 33.9 PSU, respectively match the source signatures of CUC water[Fig:3.7 (d) and (e)] that end up in the dense pool. Thus, the identical signatures of the two waters show thatthe deep shelf water inflow into the Strait of Georgia as part of an estuarine exchange in the strait is also CUCthus showing the importance of the slope undercurrent not only on WCVI dynamics but also in the deepwaterentrainment into Juan de Fuca Strait.34Timing of Upwelling Onto the Shelf:The spring oceanic transition marks the time of the year when wind-driven currents switch from wintertimepoleward to summertime equatorward directions. This change in local circulation plays an important role in thecoastal ocean variability of WCVI (Thomson and Ware, 1996). Tracking the particles on the shelf, both from thedense waters of the pool and the south outer shelf water, backward for up to 6 months provided insights about thetime taken for the individual source waters to reach the shelf. Particle tracking was also performed for 4 monthsof the summer dating back to 29th April 2013 and starting from 26th August 2013 and the results showed thata significant proportion (35 %) of water was still inside our bounding box [Fig:2.1], which we call the localwater. However, the particle tracking conducted for 6 months, dating back to 28 February 2013 only shows6.5 % of local water; which means almost 30 % of the water belonging to the dense pool entered our regionaldomain (bounding box in Fig:2.1) between 28th February 2013 to 29th April 2013, a time of the year in 2013which included the spring oceanic transition season in WCVI (Hourston and Thomson, 2018). Based on the lowpercentages of particles still remaining inside the region [Fig:2.1] with 6 % and 3.4 % contributions to the densepool and south outer shelf water, respectively, it was decided not to track the particles further back in time than28 February 2013. Almost half of the offshore water took more than 150 days to arrive at the shelf locations[Fig:3.6 l], this time precedes the spring oceanic transition in 2013. This half of the offshore water started itsjourney towards the WCVI shelf before April 2013 but only reached the WCVI shelf during August; a travel timeof more than 150 days. Cumulative wind stress has been used to identify the persistence of upwelling favorablewinds during a period of time (Barth et al., 2007; Pierce et al., 2006). Persistent upwelling favorable southwardwinds lead to a steeply decreasing cumulative wind stress [after mid July in Fig:4.1], whereas the periods ofvariable winds produce no change in cumulative wind stress with time [during mid June and during the start ofMay in Fig:4.1]. The CUC (and south water) arrives at the WCVI shelf predominantly at these two distinct times[Fig:3.6 l], one at the start of May 2013 (spending a trajectory lifetime of ∼115 days) and then another in midJune 2013. During these times, the winds (both local and remote) are variable with frequent interruptions ofupwelling favourable southward winds by northward wind reversals and relaxation of southward winds [Fig:3.2b) and Fig:4.1]. The other half of the offshore water also arrives at the boundary during the start of May. Thisstudy thus emphasizes the role of the transient oceanic flow regimes forced by variable winds, during earlysummer of 2013 in drawing these offshore waters, the CUC and south water into proximity of WCVI shelfregion. A combination of summertime coastal dynamics and the local coastal circulation must then cause thewaters to reach the shelf region.35Figure 4.1: Time series of cumulative alongshore wind stress for the oceanic summer of 2013 (spring transi-tion to fall transition); Two periods of variable wind conditions identified by regions of almost constant(horizontal) cumulative wind stress (both local and remote) during 1) mid June 2013 and 2) the start ofMay 2013. The CUC, south and almost half of the offshore source waters reach the WCVI shelf duringthese two times in 2013.4.2 Dissolved Oxygen of Dense Pool From Source Water SignaturesThe mean dissolved oxygen obtained from CTD and bottle data (collected during Pathways Cruise) corre-sponding to each of the source waters [Fig:4.2] was calculated [Table:4.1] and considered as the characteristicdissolved oxygen concentration for each source water. The straight line in T-S plot for the final shelf particles[Fig:3.1] indicates that the dense pool over the shelf is well mixed. The characteristic oxygen value of the densepool was calculated by considering the mixing of the source waters utilizing their respective characteristic dis-solved oxygen concentration value and relative fractions [using equation:3.2]. The resultant dissolved oxygenconcentration (2.18 mL L−1) obtained as a characteristic value for the dense pool is higher than that observedby Freeland and Denman (1982). This difference is likely because the calculation employed here to determinethe properties of the mixed water does not take into consideration the oxygen utilized due to biology. Freelandand Denman (1982) found a value ranging from 1.06 to 2 mL L−1 of the oxygen isopleths representing the poolas one moves from bottom depths in the pool (∼150 m) to depths of around 100 m, respectively. The mean dis-solved oxygen concentration (in August) measured by the Pathways cruise was 1.45 mL L−1 for the isopycnallevels at which we seeded the particles for tracking the source waters. Thus the dissolved oxygen value of the36dense pool measured by the Pathways Cruise (Klymak et al., 2013) is close to the value observed by Freelandand Denman (1982). The measured dissolved oxygen content of the pool is 1.45 mL L−1, and the mixing ofsource waters providing a value of 2.18 mL L−1 for the waters of the dense pool, implies an oxygen utilizationrate of 2.30 mL L−1 per year or 0.102 mmol per cubic m per year using a mean lifetime of 116 days for thecombined source waters. The southern shelf waters of WCVI get advected into the deep waters of BarkleySound during summer (Pawlowicz, 2017). The implied oxygen utilization rate value (2.30 mL L−1 per year)from this study is close to the 2.4 ± 0.5 mL L−1 per year value of oxygen utilization rate observed in deepwaters near Barkley Sound by Pawlowicz (2017). This value is higher than oxygen utilization rates usuallyobserved at similar depths in different shelf regions (Hartnett and Devol, 2003; Siedlecki et al., 2015).Figure 4.2: T-S plot from Pathways CTD and bottle data with the dissolved oxygen mL L−1 (in color). Thecovariance ellipses are the same as in Fig:3.7. The covariance ellipse for local water is not shown herefor clarity. The dissolved oxygen of a source water is considered to be the mean dissolved oxygenconcentration of the points lying within the ellipse belonging to a source particular source water. Thesevalues are reported in Table:4.1.37Type of Source Water Percentage (%) Dissolved Oxygen(mL L−1)CUC Water 42 1.43South Water 22.5 3.0Offshore Water 21.4 2.45North Water 5.2 2.38Strait Outflow Water 2.4 4.87Local Water 6.5 2.16Dense Pool 100 2.18Table 4.1: Dense pool dissolved oxygen concentration from source water fraction and characteristic dis-solved oxygen using Equation:3.24.3 Role of Canyons in Transporting Water Over the WCVI ShelfEquatorward flow forced by an alongshore wind or generated by poleward propagating shelf waves is as-sociated with observations of flow up the canyon axis (Allen et al., 2001; Allen and Hickey, 2010; Hickey,1997). The canyons present off the Washington coast and WCVI, subjected to the equatorward summertimeflow cause enhanced vertical transport (Allen and Hickey, 2010; Freeland and Denman, 1982; Hickey, 1997)and contribute to cross-shelf break exchange and to the nutrient flux over the shelf (Hickey and Banas, 2008).The role of shelf break canyons such as Barkley and Nitinat Canyon has been well understood in terms ofhow individually each of these canyons can modify shelf flow depending on the currents and canyon geometry(Allen and Hickey, 2010; Mirshak and Allen, 2005). A long canyon such as the Juan de Fuca Canyon or Mon-terey Canyon features a canyon head that extends far into the coastal region and does not reach the continentalshelf depth until close to the coast; such long canyons have been studied to a much lesser extent and havebeen observed to cause steady upwelling even with weak equatorward winds compared to shelf break canyons(Allen, 2000; Waterhouse et al., 2009). The submarine canyon flow response to alongshelf flow has mostlybeen studied by treating them as individual features in an idealized coastal ocean (Allen, 2000; Waterhouseet al., 2009), but in the realistic shelf region such as WCVI, the canyons incise the shelf break one after theother. This study is performed in a realistic regional scale modeling domain where all the canyons are presentand flow in one canyon is not free from being modified by another canyon sharing the shelf break. Only three38other studies have inspected the cumulative impact of multiple canyons on a regional scale, namely Jordi et al.(2006) for the northwest Mediterranean, Ka¨mpf (2010) for the southern coast of Australia, and Connolly andHickey (2014) for northern CCS. Connolly and Hickey (2014) studied the cumulative effect of the multiplecanyon system in northern CCS and concluded that the canyon system caused particles to upwell from greatersource depths (from the slope) than a single long canyon (Juan de Fuca Canyon) system. This study was ableto specify the crucial shelf break pathways through each canyon that contribute to the upwelled water particlesover the WCVI shelf. An example is that the downstream (southward) side of Nitinat Canyon is the primaryshelf break crossing site for particles ending up over the south outer shelf region in WCVI. Also, the head ofthe Spur Canyon, together with the convoluted bathymetry in Juan de Fuca Canyon near Swiftsure Bank act assignificant upwelling regions responsible for creating the dense pool of water studied here.4.4 Model Limitations and Perspectives4.4.1 Resolution of NEP36 ModelThe NEP36 numerical model results for WCVI analyzed in this study have 1/36◦ horizontal resolution,which may fall short of delineating all the flow features inside a submarine canyon. Too low resolution wouldlead to a reduction in the magnitude of cross-shelf flux upwelled within a canyon. Assuming a bottom Ekmanlayer thickness in the coastal WCVI to be roughly 10 to 15 m (Crawford and Thomson, 1991), and consideringa NEP36 vertical grid resolution of 87 m at 541 m depth, it can be inferred that the model may have underesti-mated the onshore transport by ∼20 % (Dawe and Allen, 2010). A vertical grid thickness less than 20 % of thebottom Ekman layer thickness is estimated to be needed to accurately model the canyon upwelled flux within1 % of the actual flux (Dawe and Allen, 2010). Thus, improving the NEP36 model resolution can be expectedto lead to an improved representation of canyon upwelling.4.4.2 Low Model Spice VariabilityThe spice simulated in the dense pool by the NEP36 numerical model is significantly less variable than theobserved spice at the same isopycnal levels. The reason for the lack of along-isopycnal spice variation wasnarrowed down to the lack of model spice variation in a region near the south boundary of our sub-domain,from comparing the NEP36 model results with glider data near the Washington shelf. An along isopycnalcomparison of model versus observed spice, for the glider locations, revealed that along the 26.4 isopycnal, the39observed spice ranging from 0 to -0.25 was only simulated by the model as a 0 to -0.08 variation in spice. Aninaccurate spice representation in the southern boundary conditions of the original full domain NEP36 modelis anticipated to have caused this lack of spice variability.40Chapter 5ConclusionsUnderstanding the origin of waters that form a feature of interest, including fields such as temperature andsalinity and also bio-geochemical fields and pollutant concentrations requires tracking of water to its sourcelocations (Fischer et al., 2013; Grant et al., 2005). Larval dispersion by coastal advection, which has orderone influence on fish stock dynamics (Cowen et al., 2006; Jackson and Strathmann, 1981; Roughgarden et al.,1988), can also be determined by particle tracking. The concept of Lagrangian water parcel displacement anddispersion has been traditionally measured by tracking water parcels using drifting buoys to characterize path-ways of water and compute dispersal statistics (Dever et al., 1998; LaCasce, 2008; Poulain and Niiler, 1989;Swenson and Niiler, 1996). Although drifting buoys may provide a fairly good idea of circulation and disper-sal patterns, major limitations exist in terms of spatiotemporal sampling, important in an upwelling favorablecoastal region such as WCVI. This study is aimed at extending the existing knowledge about the sources andpathways of water contributing to the dense pool under the biologically productive Juan de Fuca Eddy on thesouthern WCVI shelf during summer. The following research questions are addressed in this thesis.1. What are the sources of dense water near LB08?Station LB08 is located in the Juan de Fuca Eddy region at around 150 m depth over the WCVI shelf. Alow oxygen dense pool of water is observed at subsurface depths in this region [Fig:3.3]. In this study,waters of the dense pool are identified by clustering the model values of along-isopycnal spice with theobserved along isopycnal spice. Subsequently, the cluster, agreed by the model simulations and observa-tions to have low along isopycnal spice values, is chosen to represent the waters of the dense pool. TheCalifornia Undercurrent (CUC) water flowing over the slope region (deeper than 200 m) off the Washing-41ton shelf is the dominant contributor to the dense pool and supplies 45 % of the dense pool water[Fig:3.5].The south water originates from the same Washington shelf geographical boundary as the CUC [Fig:2.1]but with source depths shallower than 200 m. The south water and offshore water also contribute sig-nificantly to the dense pool, each contributing ≈20 % of the total [Fig:3.5]. The source depths of thesouth waters are slightly deeper than the offshore waters, both being less than 200 m [Fig:3.6]. Thisstudy thus highlights contributions of the CUC along with the south and offshore water, rather than flowfrom the north, as being the dominant sources of deep water on the southern WCVI shelf. CUC watersare usually associated with low oxygen and high nutrient concentration. The low oxygen waters of CUCcan influence the Juan de Fuca Eddy region, potentially having severe implications such as hypoxia whenaided by oxidation of organic matter. The south water and offshore waters have higher dissolved oxygenconcentrations [Table:4.1] and thus larger percentages of these waters would ameliorate the low oxygenimpact of CUC.2. Are the canyons involved in upwelling that leads to the presence of dense waters near LB08? If the an-swer is yes, we choose to ask further: Which are the most important conduits,in particular, the pathwaysonto the shelf and the upwelling hotspots ?The CUC water finds its way to the WCVI shelf. The canyons act as upwelling regions for the deep watersthat end up on the WCVI shelf. The canyons off WCVI play a prominent role in enabling the cross-shelfexchange and vertical transport of waters deeper than 200 m onto the shelf region. The canyons namely,Barkley, Nitinat, Juan de Fuca and the Spur Canyon are the dominant topographical conduits of shelfbreak crossing and vertical transport for waters leading to the dense pool. The Spur Canyon and thenorthern rim of the Juan de Fuca Canyon near Swiftsure Bank are the dominant regions for shelf breakcrossing and Barkley and Nitinat Canyons are the dominant locations of vertical transport as the wa-ters change their depth from deeper than 200 m to shallower than 200 m inside these canyons. Morethan 50 % of upwelled waters forming the south outer shelf water cross the downstream side of NitinatCanyon. This study outlines the pathways of shelf break exchange through each of the four canyons onWCVI - Barkley, Nitinat, Juan de Fuca and its arm, the Spur Canyon playing a crucial role in transportingdeep water onto different regions of the WCVI shelf.423. Why do the near-bottom waters at LB08 located in the region of the Juan de Fuca Eddy, have such lowdissolved oxygen concentration?The dominant contribution of CUC waters mixed with other source waters (mainly the south water andoffshore waters) combined with the significant oxygen utilization in the Juan de Fuca Eddy region pro-duces the low oxygen in the dense pool of water. CUC water is known for its low oxygen content and highnutrient concentration. The mixing of the source waters in their relative proportions provides an oxygenconcentration of 2.18 mL L−1. The observed oxygen at those depths in the dense pool was 1.45 mL L−1and thus the implied oxygen utilization rate was calculated to be a 2.5 mL L−1 per year, which is rel-atively high. The high oxygen utilization is attributed to the high biological productivity in this regionduring summer.The study made the following key contributions to regional coastal oceanography research. The firstcontribution is in the application of a high resolution, fully realistic regional ocean model (NEP36), andthe use of model results to track water back to its sources. Although analytical and idealised numericalmodels are excellent tools for understanding the physical mechanisms of oceanographic processes suchas upwelling, a statistically evaluated realistic model eliminates most of the assumptions of an idealisedmodel (such as constant winds or no tides for example) and thus provides an improved portrayal of a real-world oceanographic event or process. A machine learning approach (K-Means Clustering) was adoptedto select the particles for tracking backwards in time to ensure the model particles truly represented thewater of the dense pool. Thus, the geographical sources of the low oxygen, dense pool of water wererevealed, and the reason for it having a low oxygen content was also explained in this study. While pastresearch using both drifter based studies and modeling efforts has focused on the surface structure ofthe Juan de Fuca Eddy in this biologically productive region, this study advances our understanding ofthe sources of water lying below the eddy in the form of a dense pool that appears “well-mixed” fromobservations aboard the 2013 Pathways cruise (Klymak et al., 2013).The second contribution was characterizing the source waters leading to the dense pool. Utilization of adata-driven statistical approach to transition from finite data samples to continuous distribution functionsof source water signals (in terms of temperature, salinity, density, spice, trajectory time and source depths)enabled us to understand the signatures of the source waters better. Mixing of the source waters closelyapproximated the final signatures of water inside the dense pool. The knowledge of final mixed shelf43water was inferred by uncertainty propagation using the source water signatures.The final and most important contribution was identifying the regional submarine canyons on the WCVIshelf break that facilitate upwelling of deep waters onto the shelf. Interaction of submarine canyons ina multiple canyon system is not understood well and this study, conducted in a realistic multiple canyonmodel domain, outlines the relative contribution of each canyon in upwelling waters deeper than 200 monto the shelf. The canyons being the conduits of upwelling implies the upwelled water comes from amuch deeper source than waters from traditional shelf break upwelling (Allen and Durrieu de Madron,2009; Allen, 2004; Allen and Hickey, 2010; Howatt and Allen, 2013). CUC waters have been foundover the WCVI shelf by previous water mass studies (Mackas et al., 1987), but the specific pathwaysthrough which the undercurrent ends up on the shelf were previously unknown. The impact of upwellingof slope waters of CUC through these canyons onto the WCVI shelf has crucial implications to WCVIregional dynamics and biological productivity because of the cold, salty signatures of CUC and also itslow oxygen content and high nutrient concentrations. The dominant hotspots of upwelling of waterleading to the dense pool, such as the Spur Canyon and the northern rim of Juan de Fuca Canyon, revealthe contribution of small scale topography in creating a dense pool of low oxygen water on the shelf.Following the conclusions of this study, future work is proposed as follows. A significant proportion ofthe offshore waters arrives on the WCVI shelf region before the spring oceanic transition. Thus, followingthe work of Thomson and Ware (1996), which identified the time of spring transition on WCVI, theimpact of this spring transition on the WCVI needs to be studied more thoroughly. The response of thecanyon-driven flow during such a dramatic time of year, involving switching of current directions, isalso less understood as submarine canyon research often focuses on the upwelling season (summer) ordownwelling season (winter). Although the local or remote wind forcing did not seem to influence thedepth versus time trends of the individual source waters leading to the waters over the shelf, the time seriesof each of the geographically classified source waters needs to be closely investigated with focus being tounearth representative clusters and outliers, based on their preference to take a particular trajectory duringtravel. This analysis would help us understand why a representative cluster of source water chooses aparticular path of travel over another trajectory, or what causes it to upwell. Does the upwelling occur asa response of the multiple submarine canyon system to the mean flow, or as a single canyon response tofluctuations in the mean flow or both? The impact of propagating coastal trapped waves in this region on44the trajectory of each representative cluster of source water needs to be investigated to fully understandthe dynamics of the WCVI coastal region. This analysis could potentially reveal a wealth of informationon the interaction of multiple canyons and their cumulative impact on the upwelling of deep waters onthe shelf as well as the response of such a system to the external forcing.45BibliographyAdams, K. A., Barth, J. A., and Chan, F. (2013). Temporal variability of near-bottom dissolved oxygen duringupwelling off Central Oregon. J. Geophys. Res. Oceans, 118(10):4839–4854. → page 2Allen, S. and Durrieu de Madron, X. (2009). A review of the role of submarine canyons in deep-oceanexchange with the shelf. Ocean Sci., 5(4):607–620. → pages 4, 5, 44Allen, S., Vindeirinho, C., Thomson, R., Foreman, M. G., and Mackas, D. (2001). Physical and biologicalprocesses over a submarine canyon during an upwelling event. Can. J. Fish. Aquat. Sci., 58(4):671–684. →pages 7, 38Allen, S. E. (2000). On subinertial flow in submarine canyons: Effect of geometry. J. Geophys. Res. Oceans,105(C1):1285–1297. → page 38Allen, S. E. (2004). Restrictions on deep flow across the shelf-break and the role of submarine canyons infacilitating such flow. Surv. Geophys, 25(3-4):221–247. → page 44Allen, S. E. and Hickey, B. (2010). Dynamics of advection-driven upwelling over a shelf break submarinecanyon. J. Geophys. Res. Oceans, 115(C8). → pages 38, 44Axell, L. B. (2002). Wind-driven internal waves and Langmuir circulations in a numerical ocean model of thesouthern Baltic Sea. J. Geophys. Res. Oceans, 107(C11):25–1. → page 13Barnier, B., Brodeau, L., Le Sommer, J., Molines, J.-M., Penduff, T., Theetten, S., Treguier, A.-M., Madec, G.,Biastoch, A., Bo¨ning, C. W., et al. (2007). Eddy-permitting ocean circulation hindcasts of past decades.Clivar Exchanges, 42(12):8–10. → page 13Barth, J. A., Menge, B. A., Lubchenco, J., Chan, F., Bane, J. M., Kirincich, A. R., McManus, M. A., Nielsen,K. J., Pierce, S. D., and Washburn, L. (2007). Delayed upwelling alters nearshore coastal ocean ecosystemsin the northern California current. Proc. Natl. Acad. Sci., 104(10):3719–3724. → pages 2, 35Battisti, D. S. and Hickey, B. M. (1984). Application of remote wind-forced coastal trapped wave theory to theOregon and Washington coasts. J. Phys. Oceanogr., 14(5):887–903. → page 7Bianucci, L., Denman, K., and Ianson, D. (2011). Low oxygen and high inorganic carbon on the VancouverIsland Shelf. J. Geophys. Res. Oceans, 116(C7). → pages 1, 2Blanke, B. and Delecluse, P. (1993). Variability of the tropical Atlantic Ocean simulated by a generalcirculation model with two different mixed-layer physics. J. Phys. Oceanogr., 23(7):1363–1388. → page 13Blanke, B. and Raynaud, S. (1997). Kinematics of the Pacific Equatorial Undercurrent: An Eulerian andLagrangian approach from GCM results. J. Phys. Oceanogr., 27(6):1038–1053. → page 1546Bograd, S. J., Castro, C. G., Di Lorenzo, E., Palacios, D. M., Bailey, H., Gilly, W., and Chavez, F. P. (2008).Oxygen declines and the shoaling of the hypoxic boundary in the California Current. Geophys. Res. Lett.,35(12). → page 5Bograd, S. J., Thomson, R. E., Rabinovich, A. B., and LeBlond, P. H. (1999). Near-surface circulation of thenortheast Pacific Ocean derived from WOCE-SVP satellite-tracked drifters. Deep Sea Res. Part II,46(11-12):2371–2403. → page 7Boyer, D. L., Zhang, X., and Pe´renne, N. (2000). Laboratory observations of rotating, stratified flow in thevicinity of a submarine canyon. Dyn. Atmos. Oceans, 31(1-4):47–72. → page 3Brink, K. H. and Robinson, A. R. (1998). The Global Coastal Ocean-Processes and Methods, volume 10. TheSea, John Wiley, Hoboken, N. J., 604 pp. → page 3Chabot, D. and Claireaux, G. (2008). Environmental hypoxia as a metabolic constraint on fish: the case ofAtlantic cod, Gadus morhua. Marine Pollution Bulletin, 57(6-12):287–294. → page 1Chan, F., Barth, J., Lubchenco, J., Kirincich, A., Weeks, H., Peterson, W. T., and Menge, B. (2008).Emergence of anoxia in the California current large marine ecosystem. Science, 319(5865):920–920. →pages 1, 2, 5Checkley Jr, D. M. and Barth, J. A. (2009). Patterns and processes in the California Current System. Prog.Oceanogr., 83(1-4):49–64. → pages 4, 5Connolly, T., Hickey, B., Geier, S., and Cochlan, W. (2010). Processes influencing seasonal hypoxia in thenorthern California Current System. J. Geophys. Res. Oceans, 115(C3). → pages 1, 2Connolly, T. P. and Hickey, B. M. (2014). Regional impact of submarine canyons during seasonal upwelling.J. Geophys. Res. Oceans, 119(2):953–975. → pages 7, 39Connolly, T. P., Hickey, B. M., Shulman, I., and Thomson, R. E. (2014). Coastal trapped waves, alongshorepressure gradients, and the California Undercurrent. J. Phys. Oceanogr., 44(1):319–342. → pages 8, 18Cowen, R. K., Paris, C. B., and Srinivasan, A. (2006). Scaling of connectivity in marine populations. Science,311(5760):522–527. → page 41Crawford, W. R. (1988). The use of Loran-C drifters to locate eddies on the continental shelf. J. Atmos.Oceanic Technol., 5(5):671–676. → pages 8, 34Crawford, W. R. and Dewey, R. K. (1989). Turbulence and mixing: Sources of nutrients on the VancouverIsland continental shelf. Atmos.–Ocean, 27(2):428–442. → page 8Crawford, W. R. and Pen˜a, M. A. (2013). Declining oxygen on the British Columbia continental shelf.Atmos.–Ocean, 51(1):88–103. → pages 2, 5, 10Crawford, W. R. and Thomson, R. E. (1991). Physical oceanography of the western Canadian continentalshelf. Cont. Shelf Res., 11(8-10):669–683. → page 39Csanady, G. (1982). On the structure of transient upwelling events. J. Phys. Oceanogr., 12(1):84–96. → page3Csanady, G. T. (1981). Circulation in the coastal ocean. In Adv. Geophys., volume 23, pages 101–183.Elsevier. → page 347Davis, R. E., Ohman, M. D., Rudnick, D. L., and Sherman, J. T. (2008). Glider surveillance of physics andbiology in the southern California Current System. Limnol. Oceanogr., 53:2151–2168. → page 15Dawe, J. T. and Allen, S. E. (2010). Solution convergence of flow over steep topography in a numerical modelof canyon upwelling. J. Geophys. Res. Oceans, 115(C5). → page 39Denman, K., Mackas, D., Freeland, H., Austin, M., and Hill, S. (1981). Persistent upwelling and mesoscalezones of high productivity off the west coast of Vancouver Island, Canada. Coastal upwelling, 1:514–521.→ page 7Dever, E. P., Hendershott, M., and Winant, C. (1998). Statistical aspects of surface drifter observations ofcirculation in the Santa Barbara Channel. J. Geophys. Res. Oceans, 103(C11):24781–24797. → page 41Diaz, R. J. (2001). Overview of hypoxia around the world. J. Environ. Qual., 30(2):275–281. → page 2Do¨o¨s, K. (1995). Interocean exchange of water masses. J. Geophys. Res. Oceans, 100(C7):13499–13514. →page 15Ekau, W., Auel, H., Po¨rtner, H.-O., and Gilbert, D. (2010). Impacts of hypoxia on the structure and processesin pelagic communities (zooplankton, macro-invertebrates and fish). Biogeosciences, 7(5):1669–1699. →page 1Fischer, H. B., List, J. E., Koh, C. R., Imberger, J., and Brooks, N. H. (2013). Mixing in inland and coastalwaters. Elsevier, 478pp. → page 41Flament, P. (2002). A state variable for characterizing water masses and their diffusive stability: spiciness.Prog. Oceanogr., 54(1-4):493–501. → page 15Flather, R. (1976). A tidal model of the north-west European continental shelf. Mem. Soc. R. Sci. Liege.,10:141–164. → page 14Foreman, M., Callendar, W., MacFadyen, A., Hickey, B., Thomson, R., and Di Lorenzo, E. (2008). Modelingthe generation of the Juan de Fuca Eddy. J. Geophys. Res. Oceans, 113(C3). → pages 9, 10, 21Foreman, M., Crawford, W., Cherniawsky, J., Henry, R., and Tarbotton, M. (2000a). A high-resolutionassimilating tidal model for the northeast Pacific Ocean. J. Geophys. Res. Oceans, 105(C12):28629–28651.→ page 14Foreman, M., Thomson, R., and Smith, C. (2000b). Seasonal current simulations for the western continentalmargin of Vancouver Island. J. Geophys. Res. Oceans, 105(C8):19665–19698. → page 6Freeland, H., Crawford, W., and Thomson, R. (1984). Currents along the Pacific coast of Canada.Atmos.–Ocean, 22(2):151–172. → page 7Freeland, H. J. (1983). A seasonal upwelling event observed off the west coast of british columbia, canada. InCoastal Oceanography, pages 217–223. Springer. → page 9Freeland, H. J. and Cummins, P. F. (2005). Argo: A new tool for environmental monitoring and assessment ofthe worlds oceans, an example from the NE Pacific. Prog. Oceanogr., 64(1):31–44. → page 7Freeland, H. J. and Denman, K. L. (1982). A topographically controlled upwelling center off southernVancouver Island. J. Mar. Res., 40:1069–1093. → pages 2, 9, 10, 21, 33, 36, 37, 3848Gaspar, P., Gre´goris, Y., and Lefevre, J.-M. (1990). A simple eddy kinetic energy model for simulations of theoceanic vertical mixing: Tests at station Papa and Long-Term Upper Ocean Study site. J. Geophys. Res.Oceans, 95(C9):16179–16193. → page 13Giddings, S., MacCready, P., Hickey, B., Banas, N., Davis, K., Siedlecki, S., Trainer, V. L., Kudela, R.,Pelland, N., and Connolly, T. P. (2014). Hindcasts of potential harmful algal bloom transport pathways onthe Pacific Northwest coast. J. Geophys. Res. Oceans, 119(4):2439–2461. → page 17Grant, S. B., Kim, J. H., Jones, B. H., Jenkins, S. A., Wasyl, J., and Cudaback, C. (2005). Surf zoneentrainment, along-shore transport, and human health implications of pollution from tidal outlets. J.Geophys. Res. Oceans, 110(C10). → page 41Grantham, B. A., Chan, F., Nielsen, K. J., Fox, D. S., Barth, J. A., Huyer, A., Lubchenco, J., and Menge, B. A.(2004). Upwelling-driven nearshore hypoxia signals ecosystem and oceanographic changes in the NortheastPacific. Nature, 429(6993):749. → pages 1, 2Gray, J. S., Wu, R. S.-s., and Or, Y. Y. (2002). Effects of hypoxia and organic enrichment on the coastalmarine environment. Mar. Ecol. Prog. Ser., 238:249–279. → page 1Haine, T. W. and Hall, T. M. (2002). A generalized transport theory: Water-mass composition and age. J.Phys. Oceanogr., 32(6):1932–1946. → page 33Hartnett, H. E. and Devol, A. H. (2003). Role of a strong oxygen-deficient zone in the preservation anddegradation of organic matter: A carbon budget for the continental margins of northwest Mexico andWashington State. Geochimica et Cosmochimica Acta, 67(2):247–264. → page 37Hickey, B., Thomson, R., Yih, H., and LeBlond, P. (1991a). Velocity and temperature fluctuations in abuoyancy-driven current off Vancouver Island. J. Geophys. Res. Oceans, 96(C6):10507–10538. → pages61, 62Hickey, B., Thomson, R., Yih, H., and LeBlond, P. (1991b). Wind and buoyancy-driven fluctuations in theVancouver Island Coastal Current. J. Geophys. Res. Oceans, 96:C6. → page 8Hickey, B. M. (1979). The California current system-hypotheses and facts. Prog. Oceanogr., 8(4):191–279.→ pages 6, 8Hickey, B. M. (1995). Coastal submarine canyons. In Topographic effects in the ocean, pages 95–110. ’AhaHulika’a, proceedings of the Hawaiian Winter Workshop, University of Hawaii. → pages 5, 7Hickey, B. M. (1997). The response of a steep-sided, narrow canyon to time-variable wind forcing. J. Phys.Oceanogr., 27(5):697–726. → page 38Hickey, B. M. (1998). Coastal oceanography of western North America from the tip of Baja California toVancouver Island. The Sea, 11:345–393. → pages 6, 34Hickey, B. M. and Banas, N. S. (2003). Oceanography of the US Pacific Northwest coastal ocean and estuarieswith application to coastal ecology. Estuaries, 26(4):1010–1031. → page 8Hickey, B. M. and Banas, N. S. (2008). Why is the northern end of the California Current System soproductive? Oceanography, 21(4):90–107. → page 38Hood, D. W. and Zimmerman, S. T. (1987). Gulf of Alaska: Physical environment and biological resources.Technical report, National Ocean Service, Anchorage, AK (USA). Ocean Assessments Div. → page 749Hormazabal, S., Shaffer, G., and Leth, O. (2004). Coastal transition zone off Chile. J. Geophys. Res. Oceans,109(C01021). → page 4Hourston, R. A. and Thomson, R. (2018). Wind-driven upwelling/downwelling along the Northwest coast ofNorth America: Timing and Magnitude, P.C., King, S.A., and Boldt, J. (Eds.). State of the physical,biological and selected fishery resources of Pacific Canadian marine ecosystems in 2017. Can. J. Fish.Aquat. Sci., 3266:164–169. → pages 29, 35Howatt, T. and Allen, S. (2013). Impact of the continental shelf slope on upwelling through submarinecanyons. J. Geophys. Res. Oceans, 118(10):5814–5828. → page 44Huggett, W., Crawford, W., Thomson, R., and Woodward, M. (1987). Coastal Ocean Dynamics Experiment(CODE), Part 1-5. Data of current observations, 19. → page 7Huyer, A. (1976). Comparison of upwelling events in 2 locations-Oregon and Northwest Africa. J. Mar. Res.,34(4):531–546. → page 5Huyer, A., Kosro, P. M., Lentz, S. J., and Beardsley, R. C. (1989). Poleward flow in the California Currentsystem. In Poleward flows along eastern ocean boundaries, pages 142–156. Springer. → page 8Ikeda, M. and Emery, W. (1984). A continental shelf upwelling event off Vancouver Island as revealed bysatellite infrared imagery. J. Mar. Res., 42(2):303–317. → page 34Jackson, G. A. and Strathmann, R. R. (1981). Larval mortality from offshore mixing as a link betweenprecompetent and competent periods of development. The American Naturalist, 118(1):16–26. → page 41Johnson, K. S., Chavez, F. P., and Friederich, G. E. (1999). Continental-shelf sediment as a primary source ofiron for coastal phytoplankton. Nature, 398(6729):697. → page 5Jordi, A., Basterretxea, G., Orfila, A., and Tintore´, J. (2006). Analysis of the circulation and shelf-slopeexchanges in the continental margin of the northwestern Mediterranean. Ocean Sci., 2(2):173–181. → page39Ka¨mpf, J. (2010). On preconditioning of coastal upwelling in the eastern Great Australian Bight. J. Geophys.Res. Oceans, 115(C12071). → page 39Keeling, R. F., Ko¨rtzinger, A., and Gruber, N. (2010). Ocean deoxygenation in a warming world. Ann. Rev.Mar. Sci., 2:199–229. → page 5Khatiwala, S., Primeau, F., and Holzer, M. (2012). Ventilation of the deep ocean constrained with tracerobservations and implications for radiocarbon estimates of ideal mean age. Earth and Planetary ScienceLetters, 325:116–125. → page 5Klymak, J., Dewey, R., Allen, S. E., and Waterman, S. (2013). Open ocean to inner sea-RV Falkor, a jointexpedition. Ocean Networks Canada and Schmidt Ocean Institute Cruise Rep., 224 pp., cruise report final skj.pdf. → pages9, 10, 11, 14, 21, 22, 37, 43Klymak, J. M., Crawford, W., Alford, M. H., MacKinnon, J. A., and Pinkel, R. (2015). Along-isopycnalvariability of spice in the North Pacific. J. Geophys. Res. Oceans, 120(3):2287–2307. → page 11Kosro, P. M. (1987). Structure of the coastal current field off northern California during the Coastal OceanDynamics Experiment. J. Geophys. Res. Oceans, 92(C2):1637–1654. → page 1250Kraus, E. B. and Businger, J. A. (1994). Atmosphere-ocean interaction, volume 27. Oxford University Press.→ page 4LaCasce, J. (2008). Statistics from Lagrangian observations. Prog. Oceanogr., 77(1):1–29. → page 41Large, W. and Pond, S. (1981). Open ocean momentum flux measurements in moderate to strong winds. J.Phys. Oceanogr., 11(3):324–336. → page 19Lellouche, J.-M., Le Galloudec, O., Dre´villon, M., Re´gnier, C., Greiner, E., Garric, G., Ferry, N., Desportes,C., Testut, C.-E., Bricaud, C., et al. (2013). Evaluation of global monitoring and forecasting systems atMercator Oce´an. Ocean Sci., 9(1):57. → page 14Lentz, S. J. and Chapman, D. C. (2004). The importance of nonlinear cross-shelf momentum flux duringwind-driven coastal upwelling. J. Phys. Oceanogr., 34(11):2444–2457. → page 5Lu, Y., Li, J., Lei, J., et al. (2017). Impacts of model resolution on simulation of meso-scale eddies in theNortheast Pacific Ocean. Satellite Oceanography and Meteorology, 2(2):328. → pages 12, 13MacFadyen, A., Hickey, B., and Cochlan, W. (2008). Influences of the Juan de Fuca Eddy on circulation,nutrients, and phytoplankton production in the northern California Current System. J. Geophys. Res.Oceans, 113(C8). → pages 8, 9, 10, 33, 34MacFadyen, A. and Hickey, B. M. (2010). Generation and evolution of a topographically linked, mesoscaleeddy under steady and variable wind-forcing. Cont. Shelf Res., 30(13):1387–1402. → page 10MacFadyen, A., Hickey, B. M., and Foreman, M. G. (2005). Transport of surface waters from the Juan deFuca Eddy region to the Washington coast. Cont. Shelf Res., 25(16):2008–2021. → pages 8, 33, 34Mackas, D. L., Denman, K. L., and Bennett, A. F. (1987). Least squares multiple tracer analysis of water masscomposition. J. Geophys. Res. Oceans, 92(C3):2907–2918. → pages 8, 9, 34, 44Mackas, D. L., Louttit, G. C., and Austin, M. J. (1980). Spatial distribution of zooplankton and phytoplanktonin British Columbian coastal waters. Can. J. Fish. Aquat. Sci., 37(10):1476–1487. → page 8Mackas, D. L. and Robinson, C. L. (1997). Physical, biological and fisheries oceanography of a largeecosystem (west coast of Vancouver Island) and implications for management. Oceanologica Acta,20(1):191–200. → page 8MacQueen, J. et al. (1967). Some methods for classification and analysis of multivariate observations. InProceedings of the fifth Berkeley symposium on mathematical statistics and probability, volume 1, pages281–297. University of California Press. → page 22Mamayev, O. I. (2010). Temperature-salinity analysis of world ocean waters, volume 11. Elsevier, 374pp. →page 26Marchetti, A., Trainer, V. L., and Harrison, P. J. (2004). Environmental conditions and phytoplanktondynamics associated with Pseudo-nitzschia abundance and domoic acid in the Juan de Fuca Eddy. Mar.Ecol. Prog. Ser., 281:1–12. → page 8Masson, D. (2006). Seasonal water mass analysis for the Straits of Juan de Fuca and Georgia. Atmos.–Ocean,44(1):1–15. → page 34McDougall, T. J. and Krzysik, O. A. (2015). Spiciness. J. Mar. Res., 73(5):141–152. → page 1551Mellor, G. and Blumberg, A. (2004). Wave breaking and ocean surface layer thermal response. J. Phys.Oceanogr., 34(3):693–698. → page 13Mellor, G. L. and Yamada, T. (1982). Development of a turbulence closure model for geophysical fluidproblems. Rev. Geophys., 20(4):851–875. → page 13Mirshak, R. and Allen, S. E. (2005). Spin-up and the effects of a submarine canyon: Applications to upwellingin Astoria Canyon. J. Geophys. Res. Oceans, 110(C2). → page 38Morrison, J., Foreman, M., and Masson, D. (2012). A method for estimating monthly freshwater dischargeaffecting British Columbia coastal waters. Atmos.–Ocean, 50(1):1–8. → page 13Munk, W. (1981). Internal waves and small-scale processes. Evolution of Physical Oceanography, edited byB. Warren, and C. Wunsch :pp: 264–291, MIT Press, Cambridge, Mass. → page 15Munk, W. H. (1950). On the wind-driven ocean circulation. Journal of Meteorology, 7(2):80–93. → page 7Naqvi, S., Jayakumar, D., Narvekar, P., Naik, H., Sarma, V., D’souza, W., Joseph, S., and George, M. (2000).Increased marine production of N2O due to intensifying anoxia on the Indian continental shelf. Nature,408(6810):346. → page 2Pawlowicz, R. (2017). Seasonal Cycles, Hypoxia, and Renewal in a Coastal Fjord (Barkley Sound, BritishColumbia). Atmos.–Ocean, 55(4-5):264–283. → page 37Pelland, N. A., Eriksen, C. C., and Lee, C. M. (2013). Subthermocline eddies over the Washington continentalslope as observed by Seagliders, 2003–09. J. Phys. Oceanogr., 43(10):2025–2053. → page 15Pierce, S. D., Barth, J. A., Thomas, R. E., and Fleischer, G. W. (2006). Anomalously warm July 2005 in thenorthern California Current: Historical context and the significance of cumulative wind stress. Geophys.Res. Lett., 33(22). → page 35Poulain, P.-M. and Niiler, P. P. (1989). Statistical analysis of the surface circulation in the California CurrentSystem using satellite-tracked drifters. J. Phys. Oceanogr., 19(10):1588–1603. → page 41Rabalais, N. N. and Turner, R. E. (2001). Coastal hypoxia: consequences for living resources and ecosystems.Coastal and estuarine studies 58, American Geophyical Union, Washington,DC. → page 2Roquet, F., Madec, G., McDougall, T. J., and Barker, P. M. (2015). Accurate polynomial expressions for thedensity and specific volume of seawater using the TEOS-10 standard. Ocean Modelling, 90:29–43. → page15Rouaud, M. (2013). Probability, Statistics and Estimation: Propagation of Uncertainties in ExperimentalMeasurement. Lulu Press, Morrisville, NC, USA, 252pp. → page 27Roughgarden, J., Gaines, S., and Possingham, H. (1988). Recruitment dynamics in complex life cycles.Science, 241(4872):1460–1466. → page 41Roullet, G. and Madec, G. (2000). Salt conservation, free surface, and varying levels: a new formulation forocean general circulation models. J. Geophys. Res. Oceans, 105(C10):23927–23942. → page 13Saha, S., Moorthi, S., Pan, H.-L., Wu, X., Wang, J., Nadiga, S., Tripp, P., Kistler, R., Woollen, J., Behringer,D., et al. (2010). NCEP climate forecast system reanalysis (CFSR) 6-hourly products, January 1979 toDecember 2010. Research Data Archive at the National Center for Atmospheric Research, Computationaland Information Systems Laboratory, Boulder, CO, → page 1352Saha, S., Moorthi, S., Wu, X., Wang, J., Nadiga, S., Tripp, P., Behringer, D., Hou, Y.-T., Chuang, H.-y., Iredell,M., et al. (2014). The NCEP climate forecast system version 2. Journal of Climate, 27(6):2185–2208. →page 13Scott, D. W. (2015). Multivariate density estimation: theory, practice, and visualization. John Wiley & Sons,377pp. → page 26Shaffer, G., Pizarro, O., Djurfeldt, L., Salinas, S., and Rutllant, J. (1997). Circulation and low-frequencyvariability near the Chilean coast: Remotely forced fluctuations during the 1991–92 El Nino. J. Phys.Oceanogr., 27(2):217–235. → page 4Siedlecki, S. A., Banas, N. S., Davis, K. A., Giddings, S., Hickey, B. M., MacCready, P., Connolly, T., andGeier, S. (2015). Seasonal and interannual oxygen variability on the Washington and Oregon continentalshelves. J. Geophys. Res. Oceans, 120(2):608–633. → page 37Smagorinsky, J. (1963). General circulation experiments with the primitive equations: I. The basicexperiment. Monthly Weather Review, 91(3):99–164. → page 13Smith, R. L. (1981). A comparison of the structure and variability of the flow field in three coastal upwellingregions: Oregon, Northwest Africa, and Peru. Coastal Upwelling, 1:107–118. → page 5Stone, H. B., Banas, N. S., and MacCready, P. (2018). The effect of alongcoast advection on Pacific Northwestshelf and slope water properties in relation to upwelling variability. J. Geophys. Res. Oceans,123(1):265–286. → page 17Strub, P. T., Allen, J. S., Huyer, A., and Smith, R. L. (1987). Large-scale structure of the spring transition inthe coastal ocean off western North America. J. Geophys. Res. Oceans, 92(C2):1527–1544. → page 7Sverdrup, H. U. (1938). On the process of upwelling. J. Mar. Res., 1(2):155–164. → page 5Sverdrup, H. U., Johnson, M. W., Fleming, R. H., et al. (1942). The Oceans: Their physics, chemistry, andgeneral biology. Prentice-Hall New York, 511pp. → page 4Swenson, M. S. and Niiler, P. P. (1996). Statistical analysis of the surface circulation of the California Current.J. Geophys. Res. Oceans, 101(C10):22631–22645. → page 41Tabata, S. (1975). The general circulation of the Pacific Ocean and a brief account of the oceanographicstructure of the North Pacific Ocean Part I-circulation and volume transports. Atmosphere, 13(4):133–168.→ page 7Thompson, R. O. and Golding, T. (1981). Tidally induced upwelling by the Great Barrier Reef. J. Geophys.Res. Oceans, 86(C7):6517–6521. → page 4Thomson, R. E. (1981). Oceanography of the British Columbia Coast. Can. Spec. Publ. Fish. Aquat. Sci.,50:1–291. → page 7Thomson, R. E., Crawford, W. R., and Huggett, W. (1984). Water property observations off the west coast ofVancouver Island during CODE: May 1979 to September 1980. Department of Fisheries and Oceans,Institute of Ocean Sciences. → page 7Thomson, R. E. and Gower, J. F. (1985). A wind-induced mesoscale eddy over the Vancouver Islandcontinental slope. J. Geophys. Res. Oceans, 90(C5):8981–8993. → page 753Thomson, R. E. and Gower, J. F. (1998). A basin-scale oceanic instability event in the Gulf of Alaska. J.Geophys. Res. Oceans, 103(C2):3033–3040. → pages 8, 34Thomson, R. E., Hickey, B., and LeBlond, P. (1989). The Vancouver Island coastal current: fisheries barrierand conduit. Can. Spec. Publ. Fish. Aquat. Sci., 108:265–296. → pages 7, 8, 61, 62Thomson, R. E. and Krassovski, M. V. (2010). Poleward reach of the California Undercurrent extension. J.Geophys. Res. Oceans, 115(C09027). → page 8Thomson, R. E. and Krassovski, M. V. (2015). Remote alongshore winds drive variability of the CaliforniaUndercurrent off the British Columbia-Washington coast. J. Geophys. Res. Oceans, 120(12):8151–8176. →page 8Thomson, R. E. and Ware, D. M. (1996). A current velocity index of ocean variability. J. Geophys. Res.Oceans, 101(C6):14297–14310. → pages 7, 29, 35, 44Tomczak, M. and Godfrey, J. (2003). Regional oceanography: An introduction, 390 pp. Daya, New Delhi. →page 4Trainer, V. L., Hickey, B. M., and Horner, R. A. (2002). Biological and physical dynamics of domoic acidproduction off the Washington coast. Limnol. Oceanogr., 47(5):1438–1446. → page 8Umlauf, L. and Burchard, H. (2003). A generic length-scale equation for geophysical turbulence models. J.Mar. Res., 61(2):235–265. → page 13Vaquer-Sunyer, R. and Duarte, C. M. (2008). Thresholds of hypoxia for marine biodiversity. Proc. Natl. Acad.Sci., 105(40):15452–15457. → page 5Ware, D. M. and Thomson, R. E. (2005). Bottom-up ecosystem trophic dynamics determine fish production inthe Northeast Pacific. Science, 308(5726):1280–1284. → page 8Waterhouse, A. F., Allen, S. E., and Bowie, A. W. (2009). Upwelling flow dynamics in long canyons at lowRossby number. J. Geophys. Res. Oceans, 114(C5). → page 38Whitney, F. A., Freeland, H. J., and Robert, M. (2007). Persistently declining oxygen levels in the interiorwaters of the eastern subarctic Pacific. Prog. Oceanogr., 75(2):179–199. → page 5Willmott, C. J. (1981). On the validation of models. Physical Geography, 2(2):184–194. → page 14Wyrtki, K. (1962). The oxygen minima in relation to ocean circulation. Deep Sea Res. and OceanographicAbstracts, 9(1):11–23. → page 5Yokoi, S., Takayabu, Y. N., and Chan, J. C. (2009). Tropical cyclone genesis frequency over the western NorthPacific simulated in medium-resolution coupled general circulation models. Climate Dynamics,33(5):665–683. → page 7Zalesak, S. T. (1979). Fully multidimensional flux-corrected transport algorithms for fluids. J. Comp. Phys.,31(3):335–362. → page 1354Appendix ANEP36 Model EvaluationsA.1 Data DescriptionData Site Data Type Latitude, Longitude Dates Temporal ResolutionA1 ADCP 48.53◦ N, 126.2◦ W 2013-07-07 to 2014-05-24 30 minW01 Current Meter 48.31◦ N, 124.82◦ W 1984-06-20 to 1984-11-26 20 minW02 Current Meter 48.04 N, 124.89 W 1984-06-20 to 1984-11-26 20 minFalkor Locations CTD refer Figure: A.1 2013-08-21 to 2013-08-30 NAPathways Surveys MVP Along tracks 2013-08-21 to 2013-08-30 NATable A.1: Data used to evaluate the NEP36 modelMooring A1 (48.53◦ N, 126.2◦ W) situated on the continental slope (500 m depth) is located approximately60 km on the seaward side of the southwest portion of Vancouver Island coast [Figure:A.1]. It is maintained byFisheries and Oceans Canada on the continental slope. The shelf broadens as we move south from that region.The data used in the present work also includes the moored current meter data collected as part of the LaPerouse project led by the Local Dynamics and Processes Section at the Institute of Ocean Sciences (IOS) inSidney, BC, Canada. The MVP survey during Pathways 2013 Cruise was largely focused on the western edgeof the Juan de Fuca Eddy region between Swiftsure bank (48.55◦ N, 125.0◦ W) and La Perouse Bank (48.67◦N, 125.83◦ W); special emphasis was laid on the Tully Canyon in an attempt to cross the observed, strong andpersistent front between the well-mixed water in the eddy and the shelf water that was seen to exist south-east55from La-Peruse to “Allen Bank” (47.5240◦ N, 122.4693◦ W).Figure A.1: Map of NEP36 slice model bathymetry showing the locations of the Falkor CTD and the posi-tions of the data sites A1, W01 and W02.A.2 Evaluation of Velocity FieldsA.2.1 ADCP Data From Mooring A1The data (ADCP and NEP36 model) at A1 were plotted as quivers of horizontal velocities versus depth.From the Figure:A.2 it can be observed that the model velocity directions closely follows that of the observa-tions at their respective depths; southward equator-bound (from surface to 110 m) and polewards (from 150 mto deeper); also the model is seen to predict the depth of change of direction from southward flow to northwardflow (∼140 m) with impressive accuracy. In order to gain an improved understanding of the performance of themodel in predicting the dynamics of the region, the statistical skill scores were calculated from the scatter plotof the model against the observation data. To reduce the tidal influence, the ADCP data was averaged to dailyvalues.56Figure A.2: Quivers of daily averaged U and V velocity (m sec−1) ADCP data from mooring site A1 (indi-cated in figure:A.1). The arrow axis aspect ratio is 1 so that if U=V the orientation of the arrow on theplot is 45◦ counter-clockwise from the horizontal axis (positive to the right).A Willmott skill score of 0.71 and a root mean squared error of 0.08 at 150 m depth [Fig:A.3] indicatedthat the model agreed fairly well with the ADCP data gathered from the mooring site. The poleward flowingCUC shifts in the model, to around 2 model grid points or approximately 3.4 Km offshore of A1 [Fig:A.4] andthus the deeper (> 150 m depth) model alongshore velocities compared better to their ADCP counterparts atA1 [Fig:A.5].57Figure A.3: Statistical Scores of alongshore velocities in m sec−1 (model vs observations) at their respectivedepths for the months of July, August and September 2013Figure A.4: Model alongshore velocity m sec−1 as a) vertical section through A1 (shown as green dottedline) and, b) as horizontal shaded contour at 318 m depth, both averaged for the months of June, Julyand August. The figure shows how the poleward uncercurrent (CUC) shifts offshore in the model atdepths deeper than 200 m; thus a location offshore of A1 was chosen to compare the ADCP velocitiesfor further evaluation.58Figure A.5: Statistical Scores of alongshore velocities in m sec−1 (model at 3.4 Km offshore at A1 vs ob-servations at A1) at their respective depths for the months of July, August and September 2013As it was our intention to understand the prevailing real world oceanic conditions and corresponding modelpredicted scenarios during the time of the Pathways Cruise 2013, the model versus the ADCP alongshorevelocity data were plotted for specific days in August [Figure:A.6]. From figure:A.7 it is observed that thealongshore model velocity closely matched the ADCP alongshore velocity during early August but significantlydeviated later during the month (the duration of the cruise) as shown in Figure:A.8 when these two days wereplotted separately.59Figure A.6: Model Alongshore velocities in m sec−1 (solid lines) versus ADCP alongshore velocity (dashedlines) with depth for different days in AugustFigure A.7: Model velocities (m sec−1) versus ADCP with depth for 2 August 201360Figure A.8: Model velocities m sec−1 versus ADCP with depth for 31 August 2013From the Figures, it can be concluded that the model did agree with the ADCP data fairly convincingly,however it is to be noted that during late August 2013, strongly poleward conditions in ADCP were underesti-mated by the model which had a weaker northward flow at the same depths.A.2.2 Current Meter Data off Washington CoastThe Vancouver Island Coastal Current Experiment (June to November, 1984) was conducted with observa-tions recorded from 51 current meters and two Aanderaa wind recorders. The experiment attempted to monitorthe coastal counter current thus providing detailed information in near-surface and near shore regions. From thelocations W01 and W02 (indicated in Figure: A.1), the model simulated currents (for year 2013) replicates theobserved strong persistent northward flow at W01, also mentioned in Hickey et al. (1991a), believed to occuras a result of tidal rectification Thomson et al. (1989). At W02 (at deeper levels) the model estimated flow is abit different (more southward flow is observed) which agrees with the observations.61Figure A.9: Quivers of U and V velocity (m sec−1) from current meter data at W01 (indicated in figure A.1).The arrow axis aspect ratio is 1 so that if U=V the orientation of the arrow on the plot is 45 degreescounter-clockwise from the horizontal axis (positive to the right).Upon reducing the high frequency current meter data at W01 to daily averaged values, a persistent north-ward flow was revealed at the current meter depth (45m) also referenced in previous literature Hickey et al.(1991a) and Thomson et al. (1989). This northward flow was nicely captured by the model [Figure:A.11];however the magnitude of mean model velocity at 45m depth are much smaller than those of the current meters(0.12 m/sec as compared to 0.35 m/sec).62Figure A.10: Daily Averaged horizontal U and V velocities in m sec−1 (quivers) with depth at W01 (indi-cated in figure A.1). The arrow axis aspect ratio is 1 so that if U=V the orientation of the arrow onthe plot is 45 degrees counter-clockwise from the horizontal axis (positive to the right).63Figure A.11: Quivers of NEP36 Model horizontal (U and V) velocities in m sec−1 plotted at different modeldepths at a grid location closest to W01. The arrow axis aspect ratio is 1 so that if U=V the orientationof the arrow on the plot is 45 degrees counter-clockwise from the horizontal axis (positive to theright).64Figure A.12: Quivers of U and V velocity (m sec−1) from current meter data at W02 (indicated in figureA.1). The arrow axis aspect ratio is 1 so that if U=V the orientation of the arrow on the plot is 45degrees counter-clockwise from the horizontal axis (positive to the right).The high frequency current meter data at W02 was also averaged daily to observe any noticeable informationif gone unnoticed due to tidal influence in the mean flow. One important aspect of the flow at W02 is thesouthbound flow direction at a depth level (current meter) of 65m [Figure: A.13]. The model captures thatnicely [Figure:A.14] and has similar mean flow speed as the current meter data (close to 0.2 m/sec for bothmodel and current meters).65Figure A.13: Daily Averaged horizontal U and V velocities in m sec−1 (quivers) with depth at W02 (indi-cated in figure A.1). The arrow axis aspect ratio is 1 so that if U=V the orientation of the arrow onthe plot is 45 degrees counter-clockwise from the horizontal axis (positive to the right).66Figure A.14: Quivers of NEP36 Model horizontal (U and V) velocities in m sec−1plotted at different modeldepths at a grid location closest to W02. The arrow axis aspect ratio is 1 so that if U=V the orientationof the arrow on the plot is 45 degrees counter-clockwise from the horizontal axis (positive to theright).A.3 Evaluation of Scalar FieldsA.3.1 CTD Data From Pathways Cruise, 2013The CTD data gathered at different depths were plotted against the spatial contours of model results foreach of the scalar fields of temperature, salinity and spice, interpolated to the same depth. The daily modelresults were averaged for the last ten days of August during the time of CTD cast (21-30 August, 2013). Thefront in each of the scalar fields observed in the CTD casts as one moves away from the coast at each modeldepth (separating the water near the coast from the deep ocean) is well captured in the model at the 200m shelfbreak isobath. The cold dense water near the mouth of the strait at 109m depth is of special interest to us.One CTD station near the mouth of the strait was plotted as a comparison of model fields with observations,the model seemed to follow the observed profiles of temperature, salinity, spice and potential densities.67Figure A.15: Model vs observed profiles of temperature, salinity, spice and potential density near the en-trance of the strait of Juan de FucaThe model in general in seen to overestimate the temperature [Figure:A.16 and Figure:A.17] and this causesa expected mismatch in the spice for the same depths [Figure:A.20 and Figure:A.21]. The model salinity isreally close to the observed salinity of the casts as seen in figure:A.18 and figure:A.19.68Figure A.16: Observed Temperature (Falkor CTD) plotted as circular scatter points with a cyan outlineagainst the weekly averaged model output of Temperature on the same depth level for the sameweek in August 2013 as a background spatial contour. A closer match is indicated if the colour of thescatter points are found similar to the background colour of the spatial contour colour only delineatedby a cyan circle representing the boundary of the scatter point69Figure A.17: Observed Temperature (Falkor CTD) plotted as circular scatter points with a cyan outlineagainst the weekly averaged model output of Temperature on the same depth level for the sameweek in August 2013 as a background spatial contour. A closer match is indicated if the colour of thescatter points are found similar to the background colour of the spatial contour colour only delineatedby a cyan circle representing the boundary of the scatter point70Figure A.18: Observed Salinity (Falkor CTD) plotted as circular scatter points with a cyan outline againstthe weekly averaged model output of Salinity on the same depth level for the same week in August2013 as a background spatial contour. A closer match is indicated if the colour of the scatter pointsare found similar to the background colour of the spatial contour colour only delineated by a cyancircle representing the boundary of the scatter point71Figure A.19: Observed Salinity (Falkor CTD) plotted as circular scatter points with a cyan outline againstthe weekly averaged model output of Salinity on the same depth level for the same week in August2013 as a background spatial contour. A closer match is indicated if the colour of the scatter pointsare found similar to the background colour of the spatial contour colour only delineated by a cyancircle representing the boundary of the scatter point72Figure A.20: Spice calculated from observation data (Falkor CTD) plotted as circular scatter points with acyan outline against the weekly averaged model output of Spice on the same depth level for the sameweek in August 2013 as a background spatial contour. A closer match is indicated if the colour of thescatter points are found similar to the background colour of the spatial contour colour only delineatedby a cyan circle representing the boundary of the scatter point73Figure A.21: Spice calculated from observation data (Falkor CTD) plotted as circular scatter points with acyan outline against the weekly averaged model output of Spice on the same depth level for the sameweek in August 2013 as a background spatial contour. A closer match is indicated if the colour of thescatter points are found similar to the background colour of the spatial contour colour only delineatedby a cyan circle representing the boundary of the scatter point74The statistical scores were calculated for all the CTD casts vs the model variables for all the depths. Themodel seems to perform well from the respective high Willmott Skill scores, low RMSE and bias values whenall the data are considered [Figure:A.22]; however this is not true as revealed when the depths of only 60m to120m were considered and the statistical scores were recalculated. The overestimation of temperature (seenin the spatial plots previously) in the model became more evident at depths close to 100m and bias was reallylarge for the corresponding model temperatures[Figure:A.23].Figure A.22: Statistical Scores of model vs observations at their respective depths (given by colour of thescatter points)Figure A.23: Statistical Scores of model vs observations at their respective depths (shown by the colourscale in legend)75A.3.2 MVP Data from Pathways Cruise, 2013Figure A.24: Scatter plot of a) model density against observed density during the dates of the PathwaysCruise, August 2013. The scatter points are colored by their depth. The black contour lines withlabels represent the density of the scatter points. The contour values represent the fraction of pointsthat lie outside the contour.76Figure A.25: Survey Tracks for MVP during Pathways Cruise 2013 plotted on top of the NEP36 modelbathymetry mapThe spice at different model depths were interpolated to the potential density 26.4 isopycnal level usinglinear piecewise interpolation. The spice from the MVP surveys (conducted on the corresponding dates inAugust 2013 when the MVP experiments were performed) were plotted as scatter points on that along-isopycnalspice in an attempt to see how well the model matches the physics during the cruise in the small domain of ourinterest, similar to the CTD casts at different depths. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by a cyancircle representing the boundary of the scatter point. A common theme from all the surveys showed in thesubsequent figures is the overestimation of the spice closer to the coast by the model which is mainly due to theoverestimation of temperature in depths of 60m to 120m, which are the depths which mainly correspond to the26.4 isopycnal close to the coast.The model was successful in capturing the sharp front in spice as one moves north-west from the mouth of77Juan de Fuca strait and before encountering the shelf break; most evident in the Survey E which has significantcrossings of the front. The model results show that spicier water gets upwelled not through the Tully canyon butto its right [Figure:A.29]; the water to the right of the canyon has the same spice as water further southwardsinside the Juan de Fuca canyon; this prompted us to take a look at the model velocities on this isopycnal.The model simulated horizontal velocities (U and V) was interpolated to this isopycnal. The currents onthe same isopycnal directed northwards lays assertion to our finding of upwelling occuring to the right of theTully Canyon. Intense downwelling is also seen towards the left of the Tully canyon axis but the water insidethe canyon seems stationary on this isopycnal.Figure A.26: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of Spice on the same isopycnal (rho = 26.4) for the date inAugust 2013 as a background spatial contour. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by acyan circle representing the boundary of the scatter point78Figure A.27: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of horizontal velocities (U and V velocity quivers) on the sameisopycnal (rho= 26.4) for the date in August 2013. The location of the MVP positions and spice wereretained from the previous plot of the same survey to understand the location of upwelling around theTully canyon and why the Pathways Cruise missed it.79Figure A.28: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of Spice on the same isopycnal (rho = 26.4) for the date inAugust 2013 as a background spatial contour. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by acyan circle representing the boundary of the scatter point80Figure A.29: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of horizontal velocities (U and V velocity quivers) on the sameisopycnal (rho= 26.4) for the date in August 2013. The location of the MVP positions and spice wereretained from the previous plot of the same survey to understand the location of upwelling around theTully canyon and why the Pathways Cruise missed it.81Figure A.30: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of Spice on the same isopycnal (rho = 26.4) for the date inAugust 2013 as a background spatial contour. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by acyan circle representing the boundary of the scatter point82Figure A.31: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of horizontal velocities (U and V velocity quivers) on the sameisopycnal (rho= 26.4) for the date in August 2013. The location of the MVP positions and spice wereretained from the previous plot of the same survey to understand the location of upwelling around theTully canyon and why the Pathways Cruise missed it.83Figure A.32: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of Spice on the same isopycnal (rho = 26.4) for the date inAugust 2013 as a background spatial contour. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by acyan circle representing the boundary of the scatter point84Figure A.33: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of horizontal velocities (U and V velocity quivers) on the sameisopycnal (rho= 26.4) for the date in August 2013. The location of the MVP positions and spice wereretained from the previous plot of the same survey to understand the location of upwelling around theTully canyon and why the Pathways Cruise missed it.85Figure A.34: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of Spice on the same isopycnal (rho = 26.4) for the date inAugust 2013 as a background spatial contour. A closer match is indicated if the colour of the scatterpoints are found similar to the background colour of the spatial contour colour only delineated by acyan circle representing the boundary of the scatter point86Figure A.35: Spice calculated from observation data (MVP data) plotted as circular scatter points with acyan outline against the model output of horizontal velocities (U and V velocity quivers) on the sameisopycnal (rho= 26.4) for the date in August 2013. The location of the MVP positions and spice wereretained from the previous plot of the same survey to understand the location of upwelling around theTully canyon and why the Pathways Cruise missed it.Once the sharp front was identified from the MVP observations (and supported by model), it was necessaryto ask the question: What characteristics does the water mass on either side of the front look like and how welldoes the model predict its properties as compared to the observations made during the MVP ?A plot of both observation and model (for the model locations where the MVP survey B were conducted) inthe property-property (T-S) diagram showed a distinct straight line [Figure:A.36] which indicates mixing lineof “old” water. The figure:A.36 similar line in the model showed sharper rise in density for same increase ofsalinity or decrease of temperature. The two different water mass properties close to the surface layer are shownby forks in both model and MVP which mean the predominant presence of two different types of water masses87(one cold, less saline and the other warmer,more saline). Figure:A.36 shows a sharper slope in model T-S ascompared to MVP at larger depths (or denser waters).Figure A.36: T-S diagram for Survey A. The observations are plotted in purple as compared to the modelresults in redTo find the difference in water mass characteristics as one moves up the Tully Canyon can be seen inFigure:A.37 where the T-S plots are made for surveys B,D and G as these surveys are the ones around theTully canyon [Figure:A.25]. The mixing line is distinctly visible in the Figure:A.37 where all the three surveyswere plotted implying the existence of “old” mixed water during passage through the Tully Canyon, howeverthe MVP data is more spread-out at the 26.4 isopycnal line as compared to the model, this needs furtherinvestigation as the question remains:Does the model happen to isolate the different water mass properties at our chosen isopycnal level, or ratherhow well it is able to isolate the different water mass at the 26.4 isopycnal ?88Figure A.37: T-S diagram for Surveys B,D and G. The MVP data are plotted in the left subplot and themodel in the right subplotWhen ispycnals are not parallel to constant pressure (isobaric) surfaces, the density field can modify thepressure gradient such that it varies with depth. The baroclinic pressure gradient causes surfaces of constantdensity to be not parallel to surfaces of constant pressure. If a water is homogeneous with depth (with constantdensity) it will always be barotropic.Having compared the model alongshore velocity with ADCP measurements at mooring A1 earlier, it wouldbe nice to know how the model alongshore baroclinic velocity tendencies compared with the observation. Wecan predict this from calculating the cross-shore baroclinic pressure gradients for both the model and the MVPmeasurements.We selected 2 points twice namely points 1,2 ; and points 3,4 along cross-shore transits of the MVP surveyE with an intention to calculate the cross-shore baroclinic pressure gradient.89Figure A.38: 4 points chosen along Survey E MVP track to calculate two cross-shore baroclinic pressuregradient. The background map is the NEP36 model bathymetry showing the depth contoursFigure A.39: Baroclinic pressure gradient vsdepth between points 1,2Figure A.40: Baroclinic pressure gradient vsdepth between points 3,490Both figure:A.39 and figure:A.40 showed a higher pressure near the coast and thus the flow appeared tohave a baroclinic northward flow tendency which increased with depth. Both the model and the MVP agreed tothis in directional sense.91Appendix BWater Mass Characteristics for South OuterShelf Water92Figure B.1: Source water characteristics from particle tracking: Panels (d), (e), (j), (k), (f) and (l) representthe shaded distribution from the kde estimates of individual source water characteristics of temperature,salinity, density, spice, source depths and trajectory time, respectively leading to the final south outershelf water, shaded in grey in panels (a), (b), (c), (g) and (h). The distribution in red lines in eachof (a), (b), (g) and (h) represents the final distribution of the south outer shelf water, estimated fromkde estimates of source water signatures using equation:3.2 and equation:3.3. The vertical red (dotted)line represents the estimated mean of the final south outer shelf water signature and the vertical gray(dotted) line represents the sample mean.93Appendix CWater Mass Properties TableType of Source Water Temperature (◦C) Salinity (PSU) Density (σθ )CUC Water 6.7 ± 0.47 34 ± 0.04 26.6 ± 0.09South Water 7.7 ± 0.32 33.7 ± 0.22 26.3 ± 0.22Offshore Water 7.6 ± 0.7 33.3 ± 0.53 26.0 ± 0.5North Water 7.9 ± 1.01 33.1 ± 0.56 25.8 ± 0.58Strait Outflow Water 9.5 ± 1.12 31.4 ± 0.44 24.3 ± 0.42Local Water 7.5 ± 0.74 33.5 ± 0.75 26.1 ± 0.65Table C.1: Water mass properties table of source water masses contributing to the dense pool approximatedin terms of mean ± std94


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items