You may notice some images loading slow across the Open Collections website. Thank you for your patience as we rebuild the cache to make images load faster.

Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modeling ²³⁰Th (and ²³¹Pa) : as an approach to study the intermediate and deep water circulation in the… Yu, Xiaoxin 2017

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_2017_may_yu_xiaoxin.pdf [ 8.47MB ]
JSON: 24-1.0343591.json
JSON-LD: 24-1.0343591-ld.json
RDF/XML (Pretty): 24-1.0343591-rdf.xml
RDF/JSON: 24-1.0343591-rdf.json
Turtle: 24-1.0343591-turtle.txt
N-Triples: 24-1.0343591-rdf-ntriples.txt
Original Record: 24-1.0343591-source.json
Full Text

Full Text

Modeling 230Th (and 231Pa):As an approach to study the intermediate and deepwater circulation in the Arctic OceanbyXiaoxin YuB.Sc., Sun Yat-Sen University, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Oceanography)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April 2017c© Xiaoxin Yu 2017AbstractRecently observed 230Th concentrations in 2007 and 2009 documented very high230Th values within the Atlantic layer in the Canada Basin of the Arctic Ocean.Similar levels of high 230Th had only been previously observed in the Alpha Ridgeregion, implying that the Alpha Ridge is the potential source of the high 230Thwaters. As the Alpha Ridge is downstream in the classic cyclonic circulation, thatcirculation is believed to have changed. Motivated by this, a three-dimensionalArctic 230Th model is configured for the first time to study such change.To simulate the tracer, I coupled a scavenging model, which describes the ex-change of 230Th (and 231Pa) between the dissolved and particulate phases, to anoffline NEMO model (the Nucleus for European Modelling of the Ocean) that pro-vides the advection and mixing processes that redistribute the tracers within theocean. As the scavenging rates of such tracer elements are strongly affected byoceanic particle concentrations, the scavenging rates are parameterized as a func-tion of ice concentration, which, to a great extent, influences the biological processesin the water.Model output produced an increase of 230Th concentration in the south CanadaBasin. Sensitivity experiments confirm such change is not caused by a change inthe particle field but a change in the intermediate circulation from cyclonic to an-ticyclonic throughout the Amerasian Basin. This shift in circulation is the reasonfor a subsequent transport of high 230Th concentration from the Alpha Ridge to thesouth Canada Basin.The model circulation and density fields suggest that the change in the flow iscaused by increased dense water flux into the Arctic Ocean, primarily through theBarents Sea route. This increase of dense water inflow alters the density distributionin the Arctic and results in a quick adjustment in the Atlantic layer (∼ 1 year)through propagation of boundary trapped internal Kelvin waves.iiPrefaceThis project was initiated by my supervisor Dr. Susan Allen. This thesis is anoriginal work of the author, X. Yu. The observations reported in this thesis havetwo sources: the previously published data and unpublished data from Dr. RogerFrancois, Dr. Michiel van der Loeff, Dr. Melanie Grenier and Ole Valk (Grenier etal., in prep.; Valk et al., in prep.). The physical model outputs from the ANHA4(Arctic and Northern Hemisphere Atlantic) configurations, which were applied toforce the tracer model and to investigate the physics behind the change in the Arcticintermediate circulation pattern, are provided by Dr. Paul Myers and Xianmin Hufrom University of Alberta.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Recent changes in the Arctic Ocean . . . . . . . . . . . . . . . . . . 11.2 230Th and 231Pa as tracers . . . . . . . . . . . . . . . . . . . . . . . 31.2.1 Geochemical behavior of 230Th (and 231Pa) . . . . . . . . . . 41.2.2 Modeling study of 230Th (and 231Pa) . . . . . . . . . . . . . 41.3 Thesis objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.1 Model structure and regional Arctic configuration . . . . . . . . . . 62.1.1 Physical model . . . . . . . . . . . . . . . . . . . . . . . . . . 72.1.2 Scavenging model . . . . . . . . . . . . . . . . . . . . . . . . 82.2 Tracer parameterization . . . . . . . . . . . . . . . . . . . . . . . . . 82.3 Boundary and initial conditions . . . . . . . . . . . . . . . . . . . . 162.3.1 Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . 172.3.3 Model Spin-up . . . . . . . . . . . . . . . . . . . . . . . . . . 172.4 Description of numerical experiments . . . . . . . . . . . . . . . . . 172.5 Methods for analyzing model results . . . . . . . . . . . . . . . . . . 182.5.1 Method of model evaluation . . . . . . . . . . . . . . . . . . 182.5.2 Method of sensitivity analysis . . . . . . . . . . . . . . . . . 182.5.3 Method of boundary current analysis . . . . . . . . . . . . . 192.5.4 Method of physical component calculations . . . . . . . . . . 20ivTable of Contents3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.1 Model evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.1.1 Dissolved 230Th . . . . . . . . . . . . . . . . . . . . . . . . . 213.1.2 Particulate 230Th . . . . . . . . . . . . . . . . . . . . . . . . 233.1.3 Particle field . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2 Model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.3 Sensitivity experiments . . . . . . . . . . . . . . . . . . . . . . . . . 293.3.1 A constant ice case . . . . . . . . . . . . . . . . . . . . . . . 303.3.2 A different flow scheme case . . . . . . . . . . . . . . . . . . 313.4 Hydrographic comparisons of two flow scenarios . . . . . . . . . . . 343.4.1 Topostrophy analysis . . . . . . . . . . . . . . . . . . . . . . 343.4.2 Density analysis . . . . . . . . . . . . . . . . . . . . . . . . . 363.4.3 The Arctic inflow analysis . . . . . . . . . . . . . . . . . . . 394 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.1 Cause of observed increase in 230Th in the Canada Basin . . . . . . 424.1.1 Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 424.1.2 Sensitivity to sea ice concentration . . . . . . . . . . . . . . 424.1.3 Sensitivity to circulation pattern . . . . . . . . . . . . . . . . 434.1.4 General circulation pattern revealed by 230Th . . . . . . . . 444.2 Cause of circulation pattern change . . . . . . . . . . . . . . . . . . 454.3 Model limitations and perspectives . . . . . . . . . . . . . . . . . . . 495 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53AppendicesA Choice of sea ice concentration . . . . . . . . . . . . . . . . . . . . . 58B Preliminary model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60B.1 1D Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60B.2 2D Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62C Physical model evaluation . . . . . . . . . . . . . . . . . . . . . . . . 65D 231Pa and 230Pa/230Th in the intermediate layer . . . . . . . . . . 68D.1 Dissolved 231Pa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68D.2 Particulate 231Pa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69D.3 Dissolved 231Pa/230Th . . . . . . . . . . . . . . . . . . . . . . . . . . 71D.4 Particulate 231Pa/230Th . . . . . . . . . . . . . . . . . . . . . . . . . 72vTable of ContentsE Tracer distribution in deep layer . . . . . . . . . . . . . . . . . . . . 75E.1 Dissolved 231Pa and 230Th . . . . . . . . . . . . . . . . . . . . . . . 75E.2 Particulate 231Pa and 230Th . . . . . . . . . . . . . . . . . . . . . . 77E.3 Ratio of 231Pa/230Th . . . . . . . . . . . . . . . . . . . . . . . . . . 77viList of Tables2.1 Summary of ANHA4 runs used in this thesis. More details can befound at ( 72.2 Scavenging parameters estimated through a weighted least squarestechnique. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.3 Summary of the numerical runs. Runs beyond 2014 repeat the oceancirculation and sea ice field in 2014. . . . . . . . . . . . . . . . . . . 182.4 Data sources for model-data comparisons. . . . . . . . . . . . . . . . 18viiList of Figures1.1 Bottom topography and schematic of the Atlantic Water circulationin the Arctic Basin (modified from Spall (2013)). . . . . . . . . . . . 21.2 Conceptual map: 230Th cycle in the ocean. . . . . . . . . . . . . . . 42.1 Model grids and boundary conditions of the ANHA4 configuration. . 62.2 230Th measurements over the Arctic region. . . . . . . . . . . . . . . 92.3 Linear regressions on the observed profiles in the Eurasian Basin andthe Alpha Ridge region. . . . . . . . . . . . . . . . . . . . . . . . . . 112.4 Choosing the desorption rate: sum of intercept errors ∆c versus kd. 122.5 Particle trajectories reproduced by Ariane, a computational tool tocalculate 3D streamlines from the output velocity field using a La-grangian analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.6 Sea ice concentration series over the Alpha Ridge and the Laptevshelf side of the Eurasian Basin from NASA National Snow and IceData Center (Meier et al., 2013; Peng et al., 2013). . . . . . . . . . . 152.7 Initial conditions for particulate and dissolved 230Th at 1000 m depth. 162.8 Sections involved in sensitivity analysis and physical calculations. . 193.1 Simulated dissolved 230Th concentration in phase (a) I (2007), (b) II(2009), (c) III (2015) and (d) IV (2020). . . . . . . . . . . . . . . . 223.2 Simulated particulate 230Th concentration in phase (a) I (2007), (b)II (2009), (c) III (2015) and (d) IV (2020). . . . . . . . . . . . . . . . 243.3 Distribution of pseudo-particle in phase (a) I (2007), (b) II (2009),(c) III (2015) and (d) IV (2020). . . . . . . . . . . . . . . . . . . . . 263.4 Tukey box plots showing modeled and observed time series of dis-solved 230Th concentration at between 500-1500 m depth. . . . . . . 273.5 Observed versus modeled vertical profiles of 230Th. . . . . . . . . . . 283.6 Simulated dissolved 230Th concentration at 1000 m depth versus timefor (a) reference run and sensitivity experiments (I (b) and II (c)). . 303.7 Simulated particulate 230Th concentration at 1000 m depth versustime for (a) reference run and sensitivity experiments (I (b) and II (c)). 323.8 Vertical cross-section of along shore velocity along CAA line (section(a) in Fig. 2.8) in (a) reference and (b) Exp. 2. . . . . . . . . . . . . 33viiiList of Figures3.9 Topostrophy along isopycnal σθ = 28.0 in the reference run and Exp.2, suggesting significant change in the flow pattern is seen in thereference run but not in Exp. 2. . . . . . . . . . . . . . . . . . . . . 353.10 Analysis of density variations at 1000 m depth for 2006, 2007 and2008 for the reference run and Exp. 2. . . . . . . . . . . . . . . . . 373.11 The vertical distribution of isopycnal σθ = 28.0 in the reference runalong the 1000 m isobath around the Arctic boundary. . . . . . . . 383.12 The amount of dense water inflows (σθ¿28.05, from the Barents Seaand the Fram Strait) (upper panel) and average topostrophy in theCanada Basin (lower panel) versus time in both the reference run andExp. 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.13 Time series of dense water inflow (below σθ = 28.05 from a sectionacross the Fram Strait and a section along the Barents Sea shelf (Fig.2.8)) in the reference run (upper panel) and the Exp. 2 run (lowerpanel). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.14 Time derivatives of averaged topostrophy over the Canada Basin ver-sus the time series of averaged wind curl over the Canada Basin.Upper panel: the reference run, lower panel: Exp. 2. The windaffects the topostrophy on an annual time scale. . . . . . . . . . . . . 414.1 Sketch of general circulation pattern (a) before 2007, (b) after 2007at the intermediate layer (between σθ = 27.8 and σθ = 28.05) inferredfrom simulated 230Th. . . . . . . . . . . . . . . . . . . . . . . . . . . 444.2 Sea ice formation events from observations and model in the periodof 2002-2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48A.1 Data-model comparison on average sea ice pattern from 2002-2014 . 58B.1 Numerical solutions of one-dimensional 230Th model initialed withzero concentration, without background velocity. . . . . . . . . . . . 61B.2 Numerical solutions of one-dimensional 230Th model without back-ground velocity and with a steady state 230Th concentration fromlow ice region. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62B.3 Numerical solutions of two-dimensional 230Th model, with a constanteastward background velocity imposed for ten model years. . . . . . 63C.1 Locations of the data-model comparison (shaded regions): the bound-ary of the Canada and Makarov basins, the internal Canada Basin,the internal of central Arctic, the Kara sea opening and the BarentsSea opening. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65C.2 Physical model evaluations with respect to the observations from theWorld Ocean Database (WOD). . . . . . . . . . . . . . . . . . . . . . 66ixList of FiguresD.1 Horizontal distribution of simulated dissolved 231Pa concentration inthe intermediate layer in phase (a) I (2007), (b) II (2009), (c) III(2015) and (d) IV (2020). . . . . . . . . . . . . . . . . . . . . . . . . 69D.2 Horizontal distribution of simulated particulate 231Pa concentrationin phase (a) I (2007), (b) II (2009), (c) III (2015) and (d) IV (2020)in the intermediate layer. . . . . . . . . . . . . . . . . . . . . . . . . 70D.3 Horizontal distribution of simulated particulate 231Pa/230Th concen-tration in the intermediate layer in phase (a) I (2007), (b) II (2009),(c) III (2015) and (d) IV (2020). . . . . . . . . . . . . . . . . . . . . 72D.4 Simulated particulate 231Pa/230Th concentration in phase (a) I (2007),(b) II (2009), (c) III (2015) and (d) IV (2020). . . . . . . . . . . . . 73E.1 Simulated dissolve 231Pa and 230Th concentration in year 2002 (a)(b),2007 (c)(d) and 2015 (e)(f) . . . . . . . . . . . . . . . . . . . . . . . 76E.2 Simulated particulate 231Pa and 230Th concentration in year 2002(a)(b), 2007 (c)(d) and 2015 (e)(f) . . . . . . . . . . . . . . . . . . . 78E.3 Simulated dissolve and particulate 231Pa/230Th concentration in year2002 (a)(b), 2007 (c)(d) and 2015 (e)(f) . . . . . . . . . . . . . . . . 79xAcknowledgmentsFirst and most importantly, I would like to thank my advisor Dr. Susan Allen, forall the support throughout this work. Thanks for your guidance, encouragementand invaluable suggestions along the way. It has truly been a pleasure to be one ofyour students.I would also like to thank my committee members, Dr. Roger Francois andMelanie Grenier, for all insightful discussions and helpful feedback.Special thanks to everyone in the Waterhole with whom I have shared endlessdiscussions and to Doug Latornell for all technical help on Jasper, Python andMercurial.Last but not least, I would also like to acknowledge my family, especially myparents, and Ryan. I couldn’t possibly express all the thanks I have for you. Becausethe list is endless.xiChapter 1Introduction1.1 Recent changes in the Arctic OceanThe Arctic Ocean is recognized as an important component of the global circulation.It takes an Atlantic input, densifies it and returns it back to the Atlantic. The mod-ified water mass is an important component of meridional overturning circulation(Rudels and Friedrich, 2000). As the overturning circulation plays a crucial role inthe global varying climate through transporting heat around the world (Delworthet al., 2008), the Arctic ocean is not only crucial to local people there, but to all ofus.The Arctic Ocean is traditionally studied as a three-layer system: the surfacelayer, the intermediate layer (also known as Atlantic layer) and the deep layer, whichare separated by isopycnal σθ = 27.8 at around 400 m and ispycnal σθ = 28.0 ataround 1500 m, respectively (Lique et al., 2010). The warm, saline Atlantic layerserves as an important source of heat and salt transport to the Arctic Ocean. Anychange in this layer significantly impacts the Arctic surface layer, as well as thereturn flow back to the Atlantic Ocean.Important as the Atlantic layer is, efforts have been made in understanding itscirculation in the Arctic. The classic intermediate circulation pattern includes threecyclonic gyres inferred from the analysis of hydrographic observations (Rudels et al.,1994) . The first cyclonic gyre is in the Eurasian Basin: it receives inflows from FramStrait and the Barents Sea, transports this water along the Lomonosov Ridge andreturns it to Fram Strait. A fraction of the inflows (mainly from the Barents Seabranch), however, crosses the Lomonosov Ridge (Rudels et al., 1994) and developsinto two cyclonic gyres in the Canada Basin and the Makarov Basin. These flowseventually exit the Arctic Ocean through Fram Strait (Fig 1.1).However, this classic intermediate circulation is not stable. A recent study ongeo-tracer 129I revealed the circulation pattern in the Canada Basin changed from cy-clonic to anticyclonic, as implied by the arrival in 2007 of a large 129I spike (Karcheret al., 2012). Similarly, another geo-tracer study, 230Th, also reported that, in con-trast to the previous 230Th profile measured in the Canada Basin that displayeda linear increase in concentration with depth (Edmonds et al., 1998), an observed230Th profile in 2007 surprisingly documented very large deviations from linearitywithin the Atlantic Water. These deviations imply that recent changes occurred inthe Atlantic layer of the Canada Basin (Francois and Soon, 2010).11.1. Recent changes in the Arctic OceanHowever, while the observations point to pronounce changes, the forcing mech-anism in the Atlantic layer remains unclear. Originally, it was suggested, by anidealized barotropic model, that the Atlantic flow pattern was mainly driven by thelateral transport of positive potential vorticity (PV) (Yang, 2005). Later, a studyproposed that vertical PV flux, closely related to wind field changes, was a more im-portant factor (Karcher et al., 2007). They found that half the variance of AtlanticWater (AW) layer of the Amerasian Basin could be explained by the wind forcing.Zhang and Steele (2007) also emphasized the importance of the vertical mixing pro-cess in simulating the direction and intensity of the AW circulation. However, recentstudies suggested that changes in the vertical mixing only significantly altered thedepth of the halocline but not the AW circulation (Lique et al., 2015; Spall, 2013).Spall (2013) further showed that the ultimate driving force for the Atlantic Waterin the model was the salinity contrast between the Atlantic water and the Arcticshelves.Motivated by the observed changes in the tracer profiles and the unsolved puzzlein the forcing mechanism, this thesis aims at exploring the ongoing changes in theintermediate layer of the Arctic Ocean from a geo-tracer perspective.Figure 1.1: Bottom topography and schematic of the Atlantic Water circulation inthe Arctic Basin (modified from Spall (2013)).21.2. 230Th and 231Pa as tracers1.2 230Th and 231Pa as tracersRecent warming of the climate opens up the possibility of taking measurementsin the Arctic. However, as the existing velocity data in the Arctic does not havegood spatial and temporal coverage, this study does not directly apply velocitymeasurements to study the recent change in the Atlantic layer, but uses a tracermethod.Traditional tracers for ocean circulation are temperature and salinity. However,in the intermediate Arctic, variations in temperature and salinity are generally small.As such, a long-lived, highly particle-reactive isotope, 230Th, is better suited forArctic circulation studies because of its large spatial variations caused by the largecontrast in productivity between the seasonally ice-free shelves and the perennialice-covered deep basins. This isotope is commonly studied with 231Pa and theirratio is a promising proxy of oceanic processes in both the present and past marineenvironment (Marchal et al., 2000; Yu et al., 2001).Geochemical cycles for 231Pa and 230Th are very similar, however, their useful-ness as tracers of ocean circulation is mainly a result of the difference in particlereactivity between 231Pa and 230Th in the water column - 231Pa has a higher affinityfor biogenic opal and authigenic manganese oxides compared with the other majorsedimentary components, such as calcium carbonate and clay/lithogenic material(smaller in size), while 230Th is scavenged more equally among different particletypes (Bacon and Anderson, 1982). This difference in particle affinities results in asignificant difference in their oceanic residence time (20-40 years for 230Th, 100-200for 231Pa (Yu et al., 2001)).The 231Pa/230Th ratio is most extensively applied in understanding the AtlanticMeridional overturning circulation because 231Pa can be advected further south than230Th, which leads to variations in the 231Pa/230Th ratio. Consequently, the ratiorecorded in sediments preserves information on ventilation rate and geometry of theoverturning circulation in the past, while that in the water column helps to constrainthe long term mean, present circulation.The second most popular 231Pa and 230Th application is in the Arctic. In the1990s, chemical oceanographers used observed distributions of 230Th and 231Pa inthe Arctic to explain the basin-scale ventilation rates (as well as the pronounceddifferential scavenging) (Scholten et al., 1995). However, studies of 231Pa and 230Thdiscontinued in late 1990s. It was not until the 2007 Arctic explorations of theWilfried Laurier and the Polarstern that the observations of these isotopes restarted.While 231Pa/230Th has been widely used in the ocean, an interesting questionto ask is if the information provided by 231Pa/230Th can be acquired from only oneof these elements. If we take a step back and reconsider why 231Pa/230Th can beused as a circulation proxy, we can find the long travel distance of 231Pa is the mainreason. Compared to the large scale overturning circulation, the spatial scale ofthe Arctic circulation is much smaller, to such a degree that the travel distance for31.2. 230Th and 231Pa as tracers230Th is sufficient to indicate the Arctic water pathways. An argument to supportthis is that the residence time of 230Th in the water column, approximately 22 yearsin both the Canada and Eurasian basins and 39-45 years over the Alpha Ridgeand in the Makarov Basin (Scholten et al., 1995; Trimble et al., 2004), is of similarmagnitude to the travel time (∼30 years) of the intermediate water in the Arctic(Lique et al., 2010).Consequently, 230Th alone is sufficient to act as a circulation proxy in the ArcticOcean. In my master’s research, the distribution of 231Pa and 230Th are both simu-lated. However, for the reason mentioned above, this thesis presents only simulated230Th results. The results for 231Pa are shown in Appendix D.1.2.1 Geochemical behavior of 230Th (and 231Pa)230Th (and 231Pa) is added to seawater by the decay of Uranium isotopes with aconstant production rate (Fig. 1.2). The dissolved phases are removed from the wa-ter column by attaching to marine particles, through a process known as scavenging,and sink with marine particles in the particulate form. Since the exchange betweenthe dissolved and particulate phase is reversible, the particulate phase will also des-orb from particles and return to water column as they sink. Such exchange occurscontinuously until the marine particles eventually reach the seafloor. Meanwhile,currents play an important role in redistributing the tracers over the ocean.Figure 1.2: Conceptual map: 230Th cycle in the ocean.1.2.2 Modeling study of 230Th (and 231Pa)With more 230Th observations available, many recent efforts have been made tomodel the 230Th (and 231Pa) distribution in the ocean. The typical way to simulatethe tracer is to couple a chemical scavenging model to a physical flow field, withscavenging coefficients estimated from observations (Dutay et al., 2009; Lerner et al.,2016; Luo and Lippold, 2015; Marchal et al., 2007, 2000; Roy-Barman, 2009).However, most of the models tend to underestimate the potential complexityof the system by using a vertically uniform particle sinking velocity and constantscavenging coefficients. The former simplification implicitly assumes no variation41.3. Thesis objectivesin the particulate tracer field which is not supported by observations. The latter isalso not appropriate given rapidly changing biology fields.Lately, improvements were made (Dutay et al., 2009; Lerner et al., 2016) to simu-late 231Pa and 230Th distribution by generating the scavenging coefficients based onsimulated particle fields. Although some large deviations between the observationsand their simulations remain, the allowance of vertical and spatial variations in thescavenging coefficients lead to a significantly better fit to the data. This methodwas used as inspiration for the modeling of tracers in this study.1.3 Thesis objectivesIn this study, a three-dimensional, Arctic 230Th (and 231Pa) model is configured forthe first time.I couple a scavenging model, which describes the exchange of 230Th (and 231Pa)between the dissolved and particulate phases, to an offline ocean dynamics model(the Nucleus for European Modelling of the Ocean - NEMO) that provides theadvection and mixing processes that redistribute the tracers within the ocean tosimulate the tracer distributions in the Arctic ocean. In this attempt, the scavengingcoefficients are not treated as constants but vary spatially and vertically based onparticle concentrations.This study addresses the following research questions:1. Can the model produce the observed change of 230Th? What does the ob-served change imply about the Arctic circulation?2. What is the reason for the circulation changes in the Atlantic water layer inthe Arctic?This thesis is organized as follows. Chapter 2 gives a detailed description of themodel construction. Chapter 3 presents the tracer model results and analyzes thephysical flow fields. Further implications of the results will be discussed in Chapter4. A summary section is provided in Chapter 5.5Chapter 2Methods2.1 Model structure and regional Arctic configurationTo model the tracer distribution, the Nucleus for European Modelling of the Ocean(NEMO v3.4) is used with two sub-models nested inside: first, a scavenging modelwhich simulates the exchange process between the dissolved and particulate form;second, an ocean dynamics model which transports and mixes tracers within theocean. More detailed descriptions can be found in Section 2.1.1 and 2.1.2.The regional Arctic configuration used in this study is the Arctic NorthernHemisphere Atlantic configuration (ANHA4) and was initially configured by Uni-versity of Alberta for the VITALS (Ventilation, Interactions and Transports Acrossthe Labrador Sea) project based on the NEMO framework. It includes a three-dimensional, free surface, hydrostatic, primitive-equation ocean general circulationmodel and a dynamic-thermodynamic sea-ice model (Louvain-la-Neuve: LIM2).Note that there is no tidal forcing implementation in current configuration.Figure 2.1: Model grids for the ANHA4 configuration. The grid is plotted every10 grid points. The contours are the bathymetry. The black circle at latitude 70◦delimits the tracer boundary.The ANHA4 configuration is contained within open boundaries at 20◦S latitude62.1. Model structure and regional Arctic configurationin the Atlantic and across the Bering Strait. It has a quarter-degree horizontalresolution (Fig. 2.1), with the highest horizontal resolution located in Dease Strait(near the artificial pole over northern Canada), and the lowest resolution at theequator. In the vertical, it uses 50 levels with layer thickness smoothly increasingfrom 1.05 m at the surface to 453 m at the deepest level. To ensure model stability,the time step is set to be 1080 s; NEMO uses a Leapfrog algorithm for time-stepping.Tracer transportation is simulated by the total variation diminishing scheme(TVD). For tracer lateral diffusion, an isopycnal Laplacian operator with a horizon-tal eddy diffusivity of 300 m2s−1 is applied. Vertical mixing uses a turbulent kineticenergy dependent vertical diffusion (TKE) scheme that was parameterized usinga second-order turbulent closure scheme. The model bathymetry is derived from1-minute gridded elevations/bathymetry for the world (ETOPO1) data. Partialz-steps are utilized to better resolve the sea floor.The ANHA4 configuration has been successfully applied to study several fresh-water processes in the Arctic regions, as well as circulation and deep convection ofthe Labrador sea (Holdsworth and Myers, 2015). In this study, it is applied to boththe physical fields and the scavenging model.2.1.1 Physical modelFive-day model outputs from University of Alberta are used as our ocean dynamicinputs. Among available ANHA4 runs, I chose two velocity fields, EXH001 andEXH005, to conduct this study (Table 2.1). These two simulations were initializedwith Global Ocean Reanalyses and Simulations (GLORYS) data, ran with riverrunoff taken from the 12-month climatological data of Dai and Trenberth (2002)and forced by interannual atmospheric data from CMC GDPS reforecasts (CGRF)(Smith et al., 2014).It is particularly interesting to conduct studies on these two runs because theyare forced by the same input files, except for the precipitation field, but the mod-eled velocity fields turn out to be surprisingly different. To be specific, the EXH001boundary currents in the Arctic intermediate layer shifted from cyclonic to anti-cyclonic during the run, while those in EXH005 remained cyclonic over the entiremodel period.Simulation Integration Initial condition Climatology forcingEXH001 2002-2014 GLORYS2v3 CGRF(less precipitation)EXH005 2002-2014 GLORYS2v3 CGRF(more precipitation)Table 2.1: Summary of ANHA4 runs used in this thesis. More details can be foundat ( contrast between these two runs in terms of circulation pattern will be usedto understand how the tracers respond to currents. Meanwhile, the temperature,72.2. Tracer parameterizationsalinity and velocity fields from these runs will be applied to investigate what con-tributes to the change in circulation pattern.2.1.2 Scavenging modelThe second component needed in this study is a reversible scavenging model (Baconand Anderson, 1982). The tracer scavenging processes are controlled by absorption,desorption and sinking. Provided tracer concentrations are in dissolved form xd andparticulate form xp, they are described by the following equations:∂xd∂t= q − kaxd + kdxp + advection+mixing (2.1)∂xp∂t= kaxd − kdxp − s∂xp∂z+ advection+mixing (2.2)where q (dpm m−3y−1) is the production rate of 230Th, ka (y−1) and kd (y−1) arethe exchange rate constants of 230Th by adsorption and desorption respectively, s(m y−1) is the sinking rate of particles and z (m) is vertical depth, positive upward.These equations neglect the radioactive decay of 230Th itself (230Th half-life: 75000years).Production rate, q, sinking rate, s, scavenging coefficients ka and kd togethercontrol the tracer vertical distributions. Among all these parameters, the productionrate is the best known (0.0267 dpm m−3y−1 for 230Th), while the rest are left tobe determined. (Same situation for 231Pa. Details and results can be found inAppendix D. )2.2 Tracer parameterizationThe main challenge of this work is constructing a spatial distribution of scavengingintensity within the model domain based on existing 230Th data. The following sec-tion first introduces data used in the parameterization, then describes methods todetermine scavenging intensity at each sample location and finally presents an ap-proximation method to interpolate the scavenging coefficients from isolated samplelocations to the entire Arctic domain.ObservationsI use water column particulate and dissolved data obtained from a dataset compiledby Henderson and Wilkinson (2012) to constrain the unknown scavenging rates inthis thesis. The data mainly come from cruises in 1983 and 1991 (Fig. 2.2).82.2. Tracer parameterizationFigure 2.2: 230Th measurements over the Arctic regions. Data are available for theCanada Basin, Alpha Ridge, Makarov Basin and Eurasian Basin. Data from cruises1983 and 1991 are used in the parameterzation (para.) and initial conditions (IC),while those measured in 2000s are used in model evaluation (section 3.2).Determination of scavenging rates at sample locationsAssuming that the tracer profile is at steady state and that advection and mixingprocesses are negligible, the solutions of (2.1) and (2.2) are:xp =qsz + c (2.3)xd =qka+kdka(qsz + c)(2.4)where c is an intercept from integrating (2.1). If the effect of the mixed layer isneglected, particulate 230Th produced at z = 0 tends towards 0, which gives thesurface boundary condition, c = 0.The linearity of the tracer analytic solutions (2.3) and (2.4) enables us to esti-mate the scavenging coefficients through a least square regression. To ensure thathigh-quality data influences the regression more than low-quality data, a weightedleast square regression technique is applied; data is weighted by measurement errorsreported in the original publications.Equation (2.3) suggests that steady state particulate element concentrations areproportional to the inverse of its sinking rate. Given that the slope of particulatephase determined by linear regression is αp±αp , the sinking rate s can be estimated.92.2. Tracer parameterizations =qαp(2.5)The corresponding error is estimated by error propagation rules (Ku, 1966):s = s√√√√(αpαp)2(2.6)Similarly, (2.4) indicates that the dissolved element has an intercept that isinversely related to the adsorption rate and a slope that is controlled by the sinkingvelocity and the ratio of desorption rate and adsorption rate. Provided the linearregression slope for dissolved phase is αd ± αd , the ratio of adsorption rate overdesorption rate can be estimated:kakd=qαds=αpαd(2.7)The corresponding error is: kakd=kakd√√√√(αpαp)2+(αdαd)2(2.8)However, (2.7) only provides a relationship between absorption and desorptionrates. Inspired by an early study (Bacon and Anderson, 1982), which found a strongpositive correlation between the adsorption rate coefficient and suspended matterconcentration but did not find such dependence in desorption rate, I treated 230Thdesorption rates as a constant over the Arctic domain. Therefore, the absorptionrate can be calculated:ka = kdαpαd(2.9)This simplification on kd decreases the degrees of freedom such that the ana-lytical solution fits the slope but not the intercept and thus creates a uniform indepth discrepancy between the observation and the analytical solution (Fig. 2.3).This discrepancy can be quantified by the difference in the intercept between theobservation and the analytical solution, ∆c.102.2. Tracer parameterization∆c = βd − qka(2.10)in which βd is the linear regression intercept from the observed profiles; q/ka is theintercept from the analytical solution.As such, with an iteration on (2.7) for all stations (or groups if one groups somestations together), the sum of ∆c (∑∆c) from all stations (or all groups) can beused to find the best ka by searching for a kd that minimizes the total discrepancybetween all observations and analytical solutions.Selecting the dataHowever, one obvious disadvantage of this method is that it tends to generate highuncertainty in scavenging intensity when the advection effect is strong compared toparticle sinking (Venchiarutti et al., 2008). Stations close to Fram Strait receivestrong lateral advection from the Atlantic Ocean (profiles are not linear deeper thanapproximately 1000 m depth). For this reason, I did not apply data from thoselocations in the paramerization (marked as triangles in Fig. 2.2).Figure 2.3: Linear regressions of the observed 230Th profiles in the Eurasian Basinand the Alpha Ridge region. The left panel shows a map with observation locations.The middle panel shows linear regressions (dashed lines) on observed particulate230Th (points). The solid lines are the analytical solutions recalculated from esti-mated scavenging coefficients; Right panel: same but for the dissolved 230Th profile.Tracer profiles show that 230Th concentrations over the Alpha Ridge and in theMakarov Basin (green) increase with depth faster than those in the Eurasian Basin(blue).The parameterization method established above was performed based on threeEurasian Basin stations (‘EB group’) and two central Arctic stations (‘AR group’),112.2. Tracer parameterizationone over the Alpha Ridge and one in the Makarov Basin (Fig. 2.3). Since particlesin each basin have similar sources, stations that are geographically close are groupedtogether for a better least squares regression performance.Fig. 2.4 indicates the intercept discrepancy between observations and analyticalsolutions is smallest when kd = 0.8 y−1. Therefore, the 230Th desorption rate is setto 0.8 y−1 in this study. Estimated magnitudes of ka are consistent with previousresearch (Table 2.2). In this study, ka is 0.24 y−1 in the Eurasian Basin, which fallswithin the range of 0.12 to 0.55 y−1, shown by Trimble et al. (2004). Estimatedka of the AR group is 0.1 y−1, which is, again, comparable to previously reportedvalues of 0.16 y−1 (Bacon et al., 1989).Although estimated 230Th absorption rates are in good agreement with previouswork, the desorption rate, kd, estimated in this study is much lower than previousestimates which are 1.5 - 4.0 y−1 (Bacon et al., 1989). This discrepancy means moreparticulate phase will stay in the water column as compared to that from Baconet al. (1989), because less desorption will occur in the current parameterizationscheme. However, because adsorption of 230Th onto particles is balanced by lossesdue to desorption and sinking processes, the lower desorption rate is compensated bya higher particle sinking rate in my parameterization. In other words, even thoughthere is a difference in desorption rate between previous work and this study, theratio between dissolved and particulate phases is similar due to a more efficientsinking process.Figure 2.4: Choosing the desorption rate: sum of intercept errors ∆c versus kd. Theminimum, kd= 0.8 y−1, was chosen.Region ka (y−1) s (my−1) kdEB group 0.24 ±0.04 1040± 80 0.8AR group 0.10 ± 0.03 520± 100 0.8Table 2.2: Scavenging parameters estimated through weighted least squares tech-nique. ka and kd are absorption rate and desorption rate, respectively. s is sinkingrate. Values are higher for the Eurasian Basin (EB) than for the Alpha Ridge (AR).122.2. Tracer parameterizationParamerization of scavenging rates over the ArcticConsidering that the distribution of 230Th is closely related to particle concentrationdistribution (Bacon and Anderson, 1982), and that primary production in the Arcticis strongly affected by sea ice as a result of its control on the light availability(Brown and Arrigo, 2012) (especially by the summer minimum ice, because themajor portion of biologic productivity occurs during the summer (Kim et al., 2015)),I conjecture that high biological productivity around the ice free margins of theArctic Ocean creates a region of enhanced removal of trace elements by scavengingonto particles, while the low biological productivity area that sits under permanentice produces a region of weak scavenging.The value of this conjecture lies in the relationship that it implies between thescavenging coefficients and sea ice concentration. The application of this conjec-ture allows an Arctic-wide scavenging intensity approximation from the horizontalvariations of the summer minimum sea ice concentration.Ice-particle modelHowever, the influence from the summer minimum sea ice concentration is not im-mediately reflected in tracer profiles at all depths. Shortly after the time-varyingparticle fields originate from different sea ice concentrations at the surface, they sinkand are advected. As such, to complete the conjecture, it is important to take intoconsideration this sinking, advection and mixing motion (2.11):∂x∂t= s∂x∂z+ ~u · ∇x+mixing (2.11)where x is the ice fraction that was at the surface when the particle formed. Becauseparticles that scavenge the tracer are assumed to move at same speed as tracer’sparticulate form, the sinking speed, s, of particles produced at the surface undervarious sea ice cover is assumed to be identical to the sinking rate of the correspond-ing particulate tracer. The output from this model is called ‘pseudo-particle field’ asit tracks the ice fraction that has an implicit relation to the particle concentration.This scheme is expected to capture the variation of scavenging in a changing Arctic.Scavenging-ice relationAs mentioned above, it takes some time (more precisely, a few years) for the oceanicparticles to fall from upper ocean to deeper water, so the surface ice concentrationtakes years to affect scavenging rates in the deep sea. Therefore, an important stepis to determine where particles were initially formed. To investigate this problem,I performed a backward integration of trajectories on NEMO model outputs usingAriane, a particle tracking tool based on an algorithm given by Do¨o¨s (1995) andBlanke and Raynaud (1997) ( (Fig. 2.5).132.2. Tracer parameterizationInitial positions were distributed randomly within the sample locations area.Particles were released at different depths (1000 m, 2000 m and at the seafloor). Be-cause the time coverage of EXH001 and EXH005 is 2002-2014 and the velocity fieldfrom the late 1990s to the early 2000s does not have significant change, trajectorycalculations are based on the first three years velocity field from EXH001 (2002-2004), with 1.5x velocity speed (the reason for increasing the velocity explained inSection 4.2). Particles are integrated backward in time until they reach the surface.I added an appropriate sinking term of 500 m y−1 into Ariane to approximate thesinking process of the marine particles.Figure 2.5: Particle trajectories reproduced by Ariane, a computational tool tocalculate 3D streamlines from the output velocity field using a Lagrangian analysis.Black points show the initial locations and the lines are the particle trajectories. Aand B areas refer to the areas where the particles are initially form in surface (i.e.origin of the seafloor particles of each EB and AR group). Particles in the deepEurasian Basin are laterally transported from the Laptev side of the Eurasian Basinas they sink, whereas those over the Alpha Ridge move very little laterally.Judging by the long pathways of particles released in the Eurasian Basin (Fig.2.5), the sampling locations in the Eurasian Basin are considered to receive stronglateral transportation from the Laptev shelf side of the Eurasian Basin by boundarycurrents. Since particles released in the central Arctic only moved a few kilometers,the sampling locations from the AR group are treated as places that directly receive142.2. Tracer parameterizationvertical particles from above. This result is consistent with earlier research thatshows water masses in the Eurasian Basin interior are carried by slope currents(Pnyushkov et al., 2013), while those over the Alpha Ridge remain in the interiordue to weak ventilation (Aagaard, 1989).Based on this analysis, scavenging coefficients from the Eurasian Basin group(Table 2.2) are related to five-year-averaged sea ice condition from the Laptev shelfside of the Eurasian Basin (location B in Fig. 2.5), while those from the AlphaRidge group are related to five-year-averaged sea ice condition from the centralArctic (location A in Fig. 2.5).Information of historical sea ice concentration is taken from NOAA/NSIDC Cli-mate Data Record dataset (Meier et al., 2013; Peng et al., 2013). Satellite-derivedsea ice was used instead of the sea ice time series from the model due to bias in themodel ice concentration (Appendix A). The averaged sea ice concentration for theEurasian Basin group and the Alpha Ridge group are 0.62 and 0.97, respectively(Fig. 2.6).Figure 2.6: Sea ice concentration series over the Alpha Ridge and the Laptev shelfside of the Eurasian Basin from NASA National Snow and Ice Data Center (Meieret al., 2013; Peng et al., 2013). The grey shading shows the potion of the sea ice timeseries used to estimate the sea ice condition where the particles originally formed.With these numbers, the linear relation inferred by the conjecture can easily befound. Values for sinking and adsorption coefficients used in this paper are:s = (−1387× i+ 1874) my−1 (2.12)ka = (−0.38× i+ 0.46) y−1 (2.13)where i is ice fraction from the ice particle model.152.3. Boundary and initial conditions2.3 Boundary and initial conditions2.3.1 Initial conditionsInitializing from zero concentration, the time scale for 230Th to reach equilibriumis on an order of decades. (Take 230Th at 1000 m depth as an example. Theaverage 230Th concentration at this depth is approximately 0.4-0.5 dpm m−3 in theArctic region. With a production rate at 0.0267 dpm m−3 y−1 and assuming thereis no exchange between different phases and no lateral mixing between differentwater masses, it already takes 230Th at least 15-19 years to accumulate to 0.4-0.5dpm m−3 (Appendix B.1). With the mixing and exchange processes, it takes evenlonger.) Given the high resolution in ANHA4, obtaining initial conditions by thismethod is computationally expensive. There are two other options to generate theinitial fields: first, initializing the tracer fields from analytical solutions; second,initializing from existing observations.Figure 2.7: Initial conditions for particulate and dissolved 230Th at 1000 m depth.The colored points are observational data. The spatial pattern of dissolved 230Thshows high values in the central Arctic and lower values in the Canada and Eurasianbasins.Generating the initial fields from analytical solutions is quick, however, becausethe analytical solution assumes the advection is negligible, the corresponding initialfield purely reflects scavenging effects. On the contrary, initializing the tracer fieldwith observed data gives an initial 230Th field that will not only reflect particlescavenging, but also the impact of advection.162.4. Description of numerical experimentsGiven a relatively uniform horizontal distribution of observed 230Th, I generatedthe initial fields from all available observations measured before 2000 (Fig. 2.7).Total 230Th data are included in initializing tracer distributions in order tocompensate for a lack of spatial coverage of particulate and dissolved data in theCanada and Makarov Basins. To extract dissolved and particulate information fromthe total 230Th data, I use 90% of the total as the dissolved phase and 10% as theparticulate phase since the particulate and dissolved data in the Arctic suggests thatthe particulate fraction accounts in average for 10-20% of the total 230Th, with theremaining 80-90% in dissolved form.Vertical linear interpolation was performed to obtain tracer values at the modeldepths first then a horizontal interpolation on the 14× 14◦model grid was implemented(Fig. 2.7). The interpolation was performed using PyKrige 0.1.2 Python package( Boundary conditionsThe lateral boundaries for the tracer model is 70◦ N (Fig. 2.1). Tracer values atthe boundary are set equal to the initial condition.Surface boundary condition for the particulate phase is set to 0; for the dissolvedphase is set to q/ka.The bottom boundary uses the Neumann boundary conditions to constrain thevalue at the deepest level - the first derivative in the deepest layer concentration isthe same as that directly above.2.3.3 Model Spin-upThe spin-up period is 12 years. There are two reasons for this spin-up length. First,most of the data used in initialization is from 1991. Therefore, spinning up themodel for at least a decade enables the tracer to evolve from the 1991 conditions toa 2002 state. Second, there is no significant change in 230Th concentration reportedin the Arctic during the late 1990s and the time derivative of the modeled tracer isalmost constant after running for this length of time. As such, this spin-up lengthis considered to be appropriate.The temporal coverage of EXH001 and EXH005, 2002-2014, does not include1991-2001. As the circulation pattern does not have significant changes throughoutthis period, the model is spun up with 2002-2004 circulation repeated four times togive the 12 years.2.4 Description of numerical experimentsThe sensitivity experiments are designed to test contributions from potential affect-ing factors (Table 2.3). The reference run is forced by flow fields from EXH001; seaice concentration used for the ice particle model varies based on satellite-derived sea172.5. Methods for analyzing model resultsice data. The model is run from 2002-2020. Since the physical fields from EXH001(as well as EXH005) only cover the period of 2002-2014, runs later than 2014 usethe 2014 ocean circulation result and sea ice field as dynamic inputs.The first experiment is designed to investigate tracer sensitivity to ice concen-tration by holding it constant at 2002 values. It is forced by identical dynamics asthat in the reference run. The second experiment is forced by a different flow field(EXH005) to test the tracer sensitivity to circulation. Its sea ice concentration isidentical to that in the reference run.Simulation Integration Forcing field Sea iceReference run 2002-2020 EXH001 annualExp. 1 2002-2020 EXH001 constantExp. 2 2002-2020 EXH005 annualTable 2.3: Summary of the numerical runs. Runs beyond 2014 repeat the oceancirculation and sea ice field in 2014.2.5 Methods for analyzing model results2.5.1 Method of model evaluationTo assess the accuracy of the tracer model, point-to-point comparison between sim-ulated results and existing data are made. Data were collected during four cruises:Polarstern 2007, Wilfried Laurier 2007, GEOTRACES cruises 2009 and 2015 (Table2.4).Reference Year Location diss.230Th tot.230ThGrenier et al. (in prep.) 2007 Canada Basin YValk et al. (in prep.) 2007 Eurasian Basin Y YGrenier et al. (in prep.) 2009 Canada Basin YGrenier et al. (in prep.) 2015 Canada Basin YValk et al. (in prep.) 2015 Eurasian Basin YTable 2.4: Data sources for model-data comparisons.2.5.2 Method of sensitivity analysisThe sensitivity analysis is evaluated on section (a) in Fig. 2.8 to explore uniqueproperties of each run. This section is selected because it connects the CanadaBasin to the Alpha Ridge and is a key pathway for high 230Th in the model.182.5. Methods for analyzing model resultsFigure 2.8: Sections involved in sensitivity analysis and physical calculations. Theshaded area is the Canada Basin region used for average topostrophy (2.14) andwind stress curl. Section (a) the Alpha Ridge to the Canada Basin is used for traceranalysis; Section (b) Kara and Barents Sea opening and (c) Fram Strait are usedfor the dense water inflow calculation.Tracers at approximately 1000 m depth are chosen as representative of the tracerevolution in the intermediate layer. Results are visualized as Hovmoller diagramsto highlight the tracer distribution evolution.2.5.3 Method of boundary current analysisIn order to visualize model vector flow fields, a scalar field, topostrophy, is employedin diagnosing circulation (Holloway et al., 2007). In the northern hemisphere, posi-tive topostrophy corresponds to flows that tend to keep shallow water to the right.Therefore, positive topostrophy characterizes cyclonic boundary currents in the Arc-tic, while negative topostrophy represents anticyclonic boundary flows.The definition of topostrophy (T ) is:T = ~U ×∇H zˆ (2.14)where ~U is yearly mean horizontal velocity, ∇H is the gradient of total depth andzˆ stands for the unit vertical vector. In order to characterize the change of flow192.5. Methods for analyzing model resultspattern in the Canada Basin, a sum of topostrophy over Canada Basin is used (Fig.2.8).2.5.4 Method of physical component calculationsCalculations of dense water influxInflux calculation focuses on two main passages: the Barents Sea and Fram Strait(Fig. 2.8). In this study, the dense water influx is the total flux of water denserthan σθ = 28.05, which represents the separation of the Atlantic layer and the deeplayer (Lique et al., 2010).20Chapter 3ResultsThe first half of this section presents model results and model evaluation. The secondhalf displays comparison analysis between two different dynamics fields (salinity andtemperature fields) from EXH001 and EXH005, to understand potential factors thatcontribute to the different behaviors in the two flow fields.3.1 Model evolution3.1.1 Dissolved 230ThThe tracer analysis presented here is at a depth of 1000 m. I chose this depth fortwo reasons: first, the section of the 2009 230Th observations that deviates fromlinearity was centered within the Atlantic Water (Francois and Soon, 2010), fromapproximately 500-1500 m; second, as mentioned in Section 2.5, the Atlantic layeris defined as a layer between isopycnal σθ = 27.8 and σθ = 28.05, with the shallowlimit at 200-400 m depth and the deeper limit at approximately 1500 m depth (Liqueet al., 2010). Therefore, I assume isopycnal σθ = 28.0 is a reasonable central locationfor the core Atlantic layer, which is roughly at depth 1000 m.Over the time series, the simulated 230Th pattern in the Arctic Ocean expe-rienced a substantial change. The evolution of simulated dissolved 230Th can beseparated into four phases: phase I - before the change (2007), phase II - during thechange (2009), phase III - shortly after the change (2015) and phase IV - long afterthe change (2020). Note that the characteristic year for each phase was chosen tomatch a cruise date for which we have 230Th data: 2007, 2009, 2015.In phase I, simulated dissolved 230Th showed highest concentrations over theAlpha Ridge and showed lower values in both the Canada and the Eurasian basins(Fig. 3.1 (a)). This spatial pattern is in good agreement with the previous obser-vation patterns, which showed that scavenging intensity was lower in the centralArctic and higher in the Canada Basin and the Eurasian Basin (Edmonds et al.,1998; Scholten et al., 1995). One noticeable feature in this phase is that a low230Th tongue originating from the low sea ice Canada Basin displayed a northwardextension along the Canada Basin side of the Alpha Ridge.213.1. Model evolutionFigure 3.1: Simulated dissolved 230Th concentration in phase (a) I (2007), (b) II(2009), (c) III (2015) and (d) IV (2020). Tracer concentrations shown here areannually averaged model results. In phase I, a low 230Th concentration tongue isseen extending from the Canada Basin (red arrow in (a)). In phase III, a similartongue but of high 230Th extended from the Alpha Ridge (red arrow in (c)). Highconcentrations seen in phase I (2007) and phase II (2009) moved southward alongthe CAA in phase III (2015) and dissipated in phase IV (2020).In phase II, the overall spatial variation of dissolved 230Th concentrations didnot change dramatically compared to phase I. However, there was a small change in223.1. Model evolutionthe south Canada Basin - the northward-moving, low 230Th tongue seen in phase Idisappeared. Instead, this low 230Th concentration water mass accumulated in thesouth the Canada Basin.In phase III, changes in the spatial pattern of dissolved 230Th became moreobservable. First of all, there was a high 230Th concentration tongue developedwith a very pronounced southward progression (Fig. 3.1 (c)). Also, the simulatedhigh 230Th concentration zone was displaced further south toward the Canada Basin,compared to its previous position which had been maintained through phase I andII. Meanwhile, the low 230Th concentration that accumulated in the south CanadaBasin in phase II now moved eastward to the Chukchi Plateau side of the CanadaBasin.A notable feature revealed by phase IV (Fig. 3.1 (d)) is that the previouslyhigh 230Th concentration water over the Alpha Ridge was completely flushed awayand replaced by new, relatively lower 230Th from adjacent water masses. The high230Th concentration water mass moved further south to the Canada Basin andprogressively dissipated during its journey to the south. By phase IV, a transitionto a new, remarkably different spatial pattern had been accomplished.Observations (Francois and Soon, 2010) indicate similar changes - high 230Thvalues began to invade the Canada Basin in 2007 and were more wide spread in thebasin in 2009. However, the modeled tracer evolution in the Canada Basin has adifferent timing - the emergence of each phase is slightly delayed compared to theobservations. Note that although timing does not match well, the model capturedthe main trends of 230Th evolution. This indicates the model has a similar behaviorwith observations but it does not reproduce the change at the right time.3.1.2 Particulate 230ThCompared to the distribution of dissolved 230Th, the particulate distribution hasmore spatial and temporal variation (Fig. 3.2).In phase I, the spatial pattern of particulate 230Th was similar to the dissolved230Th pattern. There was a low particulate 230Th tongue extending from the southCanada Basin, along the edge of the Canada Basin, to the Alpha Ridge, where thehighest particulate 230Th was found (Fig. 3.2 (a)). Unlike the dissolved phase, therewere some intermediate concentrations of particulate 230Th at the seasonally ice freeedge of the Canada Basin and the Eurasian Basin.233.1. Model evolutionFigure 3.2: Simulated particulate 230Th concentration in phase (a) I (2007), (b)II (2009), (c) III (2015) and (d) IV (2020). Tracer concentrations shown here areannually averaged model results. High concentration seen over the Alpha Ridge inphase I (2007) and II (2009) moved southward along edge of the Canada Basin inphase III (2015) and IV (2020).Phase II exhibited a domain-wide increase in modeled particulate 230Th (Fig.3.2 (b)). The largest increases occurred in the Canada Basin and the MakarovBasin. This is associated with the dramatically low sea ice coverage event observedin 2007, when the sea ice coverage was at its 2nd lowest on record. The two-year243.1. Model evolution(2007-2009) delay in responding to the sea ice is due to the sinking process.In phase III, as the sea ice concentration slightly rebounded after 2007, theArctic-wide high particulate 230Th pattern disappeared - the simulated 230Th droppedback to its previous level (Fig. 3.2 (c)). However, the high particulate 230Th con-centration over the Alpha Ridge was not maintained. It developed into a narrowtongue that propagated along the northern edge of the Canadian Arctic Archipelago,similar with that seen in the dissolved 230Th.This high particulate 230Th tongue arrived at the Chukchi plateau side of theCanada Basin in phase IV, with the rest of the Amerasian Basin filled with lowparticulate 230Th concentration water and the Eurasian Basin occupied with inter-mediate particulate 230Th concentration water. (Fig. 3.2 (d)).Unlike the high dissolved 230Th concentration tongue which dissipated along itspathway, the particulate 230Th tongue increased its value during its way southward.Unfortunately, there is not particulate 230Th data available to evaluate the simu-lated particulate 230Th evolution over the modeling period. However, judging by thereasonable behavior of simulated dissolved 230Th, one can conclude that althoughany analysis for simulated particulate 230Th should be undertaken with care, thisevolution is reasonable.3.1.3 Particle fieldThe particle field is generated by the ice-particle model (2.11) that simulates themovement of advected and sinking particles. Note that the ‘particle’ field here is nottracking the particle concentration but an ice fraction which has an implicit relationto the particle concentration; this field will be referred to as the ‘pseudo-particlefield’. Keep in mind that the sea ice fraction and the marine particle concentrationare inversely correlated; low/high values of the pseudo-particle field are associatedwith high/low particle concentrations, respectively.In phase I, high pseudo-particle values accumulated over the central Arctic region(Fig. 3.3 (a)). Much lower pseudo-particle values were found around the east edge ofthe Eurasian Basin and the south boundary of the Canada Basin. The low pseudo-particle values in the Canada Basin exhibited a weak northward tongue along thebasin boundary.In phase II, the tongue disappeared along with a significant shrinking in the highpseudo-particle pool (Fig. 3.3 (b)). This Arctic-wide decrease in the pseudo-particlevalues is coincident with the appearance of Arctic-wide increase in particulate 230Thconcentrations. During this shrinking event, there is a newly-formed low pseudo-particle area developed around the Chukchi plateau; this location also documents atransient occurrence of high particulate 230Th zone at the same time (Fig. 3.2).In phase III, the high pseudo-particle pool went back to its normal size. However,different from previous years, the high pseudo-particle region extended from theAlpha Ridge to the south Canada Basin with the low pseudo-particle water mass253.1. Model evolutionmoving toward the Chukchi Plateau side. This spatial pattern remained in phaseIV.Despite the temporal change, there is a robust spatial pattern to the pseudo-particle field - the highest pseudo-particle value (suggesting the lowest particle con-centration) always appears north of the Canadian Archipelago, while lowest pseudo-particle value (suggesting the highest particle concentration) is located along thesouth edge of the Canada Basin and the east edge of the Eurasian Basin. It is alsonotable that the low pseudo-particle locations have high similarity with the highparticulate 230Th zones.Figure 3.3: Distribution of pseudo-particle in phase (a) I (2007), (b) II (2009),(c) III (2015) and (d) IV (2020). Note that low/high pseudo-particle values areassociated with high/low particle concentrations, respectively. The low pseudo-particle locations have high similarity with the high particulate 230Th zones.263.2. Model evaluation3.2 Model evaluationCanada BasinTo assess model accuracy, a model and data comparison was conducted. The mainfocus was on the Canada Basin where the temporal coverage of 230Th data wasgreatest.As indicated previously, the tracer model has a delay in the change of 230Th con-centration in the Canada Basin. To study the lag between the observation and thesimulation, time series of 230Th concentration in the south Canada Basin (shadedarea in Fig. 3.4) at 1000 m depth are presented. In the model, an increase inthedissolved 230Th concentration started in 2009, reached a maximum in 2015 andthe maximum 230Th values decreased slightly since then (Fig. 3.4). In the obser-vations, however, the dramatic increase of 230Th appeared in 2009 instead of 2015.It suggests the modeled change of 230Th concentration in the Canada Basin is sixyears delayed.Therefore, in order to have a fair data-model comparison, the model compari-son conducted here is evaluating the model behavior at different ‘evolution status’,rather than directly comparing data with model results in the same sample year.Suggested by the lag, the observed 2009 data was compared with simulated resultin phase III (in model year 2015, when the highest simulated 230Th concentrationoccurred in the Canada Basin). Data from the GEOTRACES cruise in 2015 iscompared with phase IV (2020, which is 5 years after phase III).Figure 3.4: Tukey box plots showing modeled and observed time series of dissolved230Th concentration at between 500-1500 m depth. Each box shows data rangefrom first to third quartile, and median (central horizontal line); black dots showextremes (Tukey, 1977). The green boxes (and black points) represent the simulated230Th concentrations in the Canada Basin (the shaded area). The blue boxes showthe observed data.273.2. Model evaluationThe 2007 and 2015 observed 230Th profiles in the Canada Basin are close to butnot exactly, straight lines (Fig. 3.5 (a-i)). This shape is captured by the simulatedprofiles in the Canada Basin, which show an increase with depth. Also, the simulated230Th magnitudes appear realistic compared to the observations in 2007 and 2015(Figure 3.5 (a-c, g-i)). ) The major deviation from the observations is found inthe intermediate layer in 2009, where the model tends to underestimate the 230Thconcentration. However, the model predicts reasonable values in the surface and thedeep layer.Figure 3.5: Observed versus modeled 230Th. Panels (a-u) show observed 230Th pro-files (points) from the Canada Basin, Makarov Basin, Alpha Ridge and the EurasianBasin versus model results (curves) (*: total 230Th; without *: dissolved 230Th).Sample locations are mapped with different colors. Panel (v) is an linear regressionof observations versus simulations. Red dashed line is a reference line showing y=x.283.3. Sensitivity experimentsOther basinsIn the Eurasian Basin, simulated dissolved 230Th is consistent with observations inboth magnitude and vertical structure (Fig. 3.5 (l-u)).As for the central Arctic (areas above the Alpha Ridge and in the MakarovBasin) that are permanently covered with ice, the observation profiles are not aslinear as those in the Canada Basin and Eurasian Basin. More specifically, thevalues of observed 230Th in the intermediate layer always exceeds (or equals to) thevalues in the deep layer. As such, while the simulated 230Th in this region appearsto be realistic in the surface and deep layers, the simulated 230Th in the intermediatelayer is underestimated (Fig. 3.5 (j,k)). For example, at a depth of 1000 m, theaverage 230Th concentration of the observed data is approximately 0.6 dmp m−3but the model only produces a range of 0.3-0.5 dmp m−3. This reveals that theestimated scavenging intensity in this specific layer is still not low enough in themodel.In order to provide a more quantitative evaluation of the simulation, a linearregression between model and 230Th observations is performed (Fig. 3.5). Theresult shows large scatter between the model and the observations when the valueof observed 230Th is greater than 0.8 dmp m−3 and lower than 0.2 dmp m−3. Thisindicates the model performance is less satisfactory in simulating very high and verylow 230Th values.However, the values of the slope of the linear regression, an indicator to assessthe ability of the model in simulating the tracer distribution, gives 0.91 ± 0.03,a value that is very close to 1. The discrepancy between the reference and theregression is in acceptable range. It reveals that despite the uncertainty in producingvery high/low 230Th values, the model, in general, is in good agreement with theobservations. This can be supported by the reasonable spatial contrast between thehigh ice region in the central Arctic and the low sea ice zone in both the Canadaand Eurasian basins, and the well simulated vertical structures that are broadlyconsistent with observations.3.3 Sensitivity experimentsAs the 230Th concentrations are controlled by the scavenging process and the phys-ical process such as advection/diffusion, I conducted two sensitivity experiments(see section 2.4) to test the contribution of each process to understand the changein 230Th concentrations.293.3. Sensitivity experimentsFigure 3.6: Simulated dissolved 230Th concentration at 1000 m depth versus timefor (a) reference run and sensitivity experiments (I (b) and II (c)). The left end ofthis section is to the south in the Canada Basin and the right end extends north tothe Alpha Ridge. Run descriptions can be found in section 2.4. Results show thatthe reference run is similar with Exp. 1 but very different from Exp. A constant ice caseDissolved 230ThThe only difference between sensitivity Exp. 1 (Fig. 3.6 (b)) and the reference run(Fig. 3.6 (a)) is the sea ice concentration, which varies annually in the referencerun but is held constant at 2002 values in Exp. 1. Experiment results show a highsimilarity between Exp. 1 and the reference run throughout the model period.Focusing on a vertical section along the north edge of the CAA (section (a) inFig. 2.8), we see, in both runs, two distinct water masses. Between 2002 and 2015,the Canada Basin side was occupied with relatively low dissolved 230Th concentra-tion water, while the Alpha Ridge side had much higher 230Th concentrations. Inbetween, water masses from both sides took turns to dominate. From 2002 to 2007,the low 230Th water from the Canada Basin dominated the mid-section. However,starting from approximately 2007, this situation changed - the low 230Th water masswas slowly pushed back toward the Canada Basin by high dissolved 230Th from theAlpha Ridge. By 2015, the whole section was filled with the high 230Th water.Although these two runs are very similar, a slight difference exists: tracer con-centrations from the constant ice case (Exp. 1), were always higher than that of303.3. Sensitivity experimentsthe reference run. This difference was less obvious in the early part of the run butbecame more noticeable in time.Particulate 230ThFor the particulate phase, comparing results for variable ice versus constant ice alsoshows high similarity in the tracer evolution (Fig. 3.7 (a)(b)).As in section 3.3.1, we focus on the vertical section north of CAA, where wesee it is dominated by a low particulate 230Th water mass from the low ice CanadaBasin and a high particulate 230Th water mass from the Alpha Ridge. In both thereference and Exp. 2 runs, the high 230Th labeled water dominated the sectionshortly after 2007. The similarity in the qualitative behavior between these runsreveals that the ice field has limited effect on the tracer distributions.It should be noted that, in the reference run with variable ice (Fig. 3.7 (a)), thereis a short-term, section-wide increase in particulate 230Th recorded from approxi-mately 2014 - 2015. This occurred 2-3 years after 2012, when sea ice concentrationwas at record low. The length of this time coincides with that required for theparticles to sink from surface to the current depth.This fluctuation in particulate 230Th, however, is not found in Exp.1, in whichthe sea ice concentration was the same in each year. These findings indicate that,although the persistent increase in particulate 230Th field that began in 2007 wasnot caused by the ice, the short-time scale fluctuations were caused by it. Thisexplanation can equally apply to the temporary Arctic-wide increase in particulate230Th seen in phase II that showed up 2-3 years after 2007 (Fig. 3.2).3.3.2 A different flow scheme caseDissolved 230ThWhile Exp. 1 tests the tracer sensitivity to changes in sea ice concentration, Exp.2 tests how sensitive the tracer is in response to changes in the circulation.Exp. 2 (Fig. 3.6 (c)) produced significantly different results as compared tothose of the reference run. Between 2002 and 2007, the dissolved 230Th distributionis similar in the two runs. Low 230Th water occurred on the Canada Basin sideand high 230Th labeled water occurred on the Alpha Ridge side. However, after2007, high 230Th water in the reference run spread southward but the change didnot occur in Exp. 2. This experiment suggests that the 230Th distribution is verysensitive to the circulation.Next, I investigated the velocity fields in the reference run and in Exp. 2 (Fig.3.8).In the period 2002 - 2006, a strong northward velocity dominated in the se-lected section in both runs. This flow transported low 230Th concentrations fromthe Canada Basin to the Alpha Ridge. A significant change took place in 2007 - ve-313.3. Sensitivity experimentsFigure 3.7: Simulated particulate 230Th concentration at 1000 m depth versus timefor (a) reference run and sensitivity experiments (I (b) and II (c)). The left end ofthis section is at south the Canada Basin and the right end extends north to theAlpha Ridge. Runs descriptions can be found in section 2.4. Results show that thereference run is more similar to Exp. 1 than to Exp. 2.locity direction totally reversed in the reference run; however, the velocity directionin Exp. 2 remained northward. This velocity shift is coincident with the southwardpropagation of high 230Th labeled water in the reference run. Almost immediatelyafter the circulation pattern switched, the tracer concentrations increased progres-sively. The high 230Th concentration eventually arrived in the Canada Basin in2015.The distance between the Alpha Ridge and the south Canada Basin is approxi-mately 1500 kilometers. With an average southward velocity of 0.006 m s−1 in thereference run, the time for high dissolved 230Th arrive at the south Canada Basinis approximately 8 years. This is consistent with the tracer evolution we see in thereference run.Particulate 230ThAs with dissolved 230Th, results from reference run and Exp. 2 show that thedifferences between these runs increased after 2007. After 2007, the mid-section ofline CAA in the reference run was filled with high particulate 230Th concentrationfrom the Alpha Ridge; however, in Exp. 2 this region had low particulate 230Th from323.3. Sensitivity experimentsFigure 3.8: Vertical cross-section of along shore velocity along CAA line (section(a) in Fig. 2.8) in (a) reference and (b) Exp. 2. Results show the velocity changeddirection in the reference run but not in Exp. 2.the south Canada Basin. This pattern, similar with that in dissolved 230Th results,suggests particulate 230Th has a rapid response to the circulation (Fig. 3.8)).Section summaryExp. 1 indicates the ice-particle field has a more rapid impact on the particulate230Th than the dissolved 230Th. More importantly, it demonstrates that tracerspatial patterns persist even though the sea ice conditions are different. On theother hand, Exp. 2 demonstrates that the circulation field is crucial in creating thehigh 230Th zone in the Canada Basin.As such, the increase in observed and modeled 230Th concentrations in the southCanada Basin suggests an establishment of southward directed currents along theCAA line. As the water transport in the Arctic mainly occurs along the boundaries,the velocity direction along the CAA line implies the general circulation pattern inthe Canada Basin. Thus the change of transport direction along the CAA line (fromnorth to south) indicates a general change in the Canada Basin (from cyclonic toanticyclonic).That engenders a follow-up question: what factor leads to the change in interme-diate circulation pattern in the reference run and why did the anticyclonic patternnot show up in Exp. 2?333.4. Hydrographic comparisons of two flow scenarios3.4 Hydrographic comparisons of two flow scenariosThis section provides further analysis on the two different circulation runs. I firstinvestigated the topostrophy evolution during the model period, then explored hy-drography differences in the runs along the boundary of the Arctic Ocean.The years chosen to display here are different from those in the tracer analysis.As revealed by the simulated 230Th, the flow pattern in the reference run startedto change during 2007. As a result, I selected 2006, 2007, 2008 to investigate thetransition of the flow field. For a comparison purpose, analysis in Exp. 2 alsocontains these years, even though the change is not seen in this run.3.4.1 Topostrophy analysisAs mentioned before, isopycnal σθ = 27.8 and σθ = 28.05 are the upper and lowerbounds for the Atlantic layer. Therefore, isopycnal σθ = 28.0, centered in betweenisopycnal σθ = 27.8 and σθ = 28.05, is considered as a representative depth foranalyzing the flow pattern in the Atlantic layer. I conducted a topostrophy analysisfor this layer as a means of comparing the model circulations.The results show that, in 2006 (before the change), the topostrophy fields alongthe isopycnal σθ = 28.0 are similar in the reference run and Exp. 2 (Fig. 3.9). Inthis year, positive topostrophy dominates over the whole Arctic.As mentioned in Chapter 2.5.3, positive and negative topostrophy characterizescyclonic and anticyclonic boundary currents, respectively. This positive topostrophythus shows that, after the Atlantic water enters the Arctic system through the eastside of Fram Strait, it flows in a cyclonic direction in the Eurasian Basin. Part of thewater splits from the main loop shortly after it passes the Laptev shelf and entersthe Amerasian Basin. After that, this branch flows into the Canada Basin, via theMakarov Basin and the slope of the Chukchi Sea. This pattern can be revealed bya cyclonic transport of low 230Th (in both dissolved and particulate form) along theArctic boundary in phase I .During 2007 in the reference run, most of the topostrophy along the boundaryof the Amerasian Basin changed from positive to negative. This reversal in thesigns of topostrophy field shows the water masses in the Amerasian Basin were intransition from a cyclonic rotation to an anticyclonic rotation. This flow pattern isin contrast to the dominant pattern in the early years and it leads to the transportof high 230Th concentration water from the Alpha Ridge to the south Canada Basinin the reference run, with arrival in 2015.343.4. Hydrographic comparisons of two flow scenariosFigure 3.9: Topostrophy along isopycnal σθ = 28.0 in the reference run and Exp.2, suggesting significant change in the flow pattern is seen in the reference run butnot in Exp. 2.353.4. Hydrographic comparisons of two flow scenariosHowever, while negative topostrophy occupied most of the Amerasian Basin inthe reference run, only a small area with negative topostrophy was found in Exp.2, around the Chukchi Plateau. It indicates the development of anticyclonic flowin Exp. 2 preliminarily started from the east side of the Amerasian Basin. Thisnegative area in Exp. 2 is coincident with the region with the strongest negativetopostrophy signal in the reference run.The reference run shows that shortly after 2007, the positive topostrophy in theCanada Basin completely disappeared. The negative topostrophy extended farthertowards the Alpha Ridge, which indicates the weak anticyclonic flow in the previousyear further developed into an Amerasian-basin-wide anticyclonic flow. Also, thisanticyclonic flow was intensified compared to 2007, as revealed by more negativetopostrophy values in this region. A notable feature is that the absolute values ofnegative topostropy decreased along the Amerasian boundary, from the east to thewest.However, while the anticyclonic flow was fully developed in the Amerasian Basinin the reference run, the situation in Exp. 2 was quite different. The results showa stronger positive topostrophy signal in the Amerasian Basin, which means theboundary current not only remained cyclonic but was enhanced. Meanwhile, the2007 small negative signal close to the Chukchi plateau vanished.3.4.2 Density analysisIn order to understand the hydrographic distributions within the model domain,density variations in 2006, 2007 and 2008 at 1000 m depth, the same depth as usedfor the tracer analysis, are investigated.Result shows that the density spatial distribution are similar in both runs in2006 (Fig. 3.10). Less dense water occurred on the Russia side of the Eurasian andMakarov Basin, with much denser water existing in the central Arctic and aroundthe edge of the Barents Sea.In 2007, when the topostrophy in the Amerasian Basin reversed sign in thereference run, the boundary current along the Barents and Kara Sea exhibited higherdensity in both runs. However, compared to the reference run, the increase of densitywas less obvious in the Exp. 2 - its high density terminated at the Laptev side ofthe Eurasian Basin.In contrast, the increasing density signal in the reference run spread cyclonicallyall the way from the Eurasian Basin to the CAA side of the Canada Basin, althoughthe signal dissipated along its pathway.In 2008, despite the enhanced anticyclonic flow revealed by the negative topostro-phy from the reference run (Fig. 3.9), this high isopycnal signal propagated cycloni-cally around the Arctic. An noticeable feature is that this high density signal wastrapped by the shelf slope and always kept the shelf to its right. Also note that itspropagation speed was, clearly, much faster than the simulated 230Th signal. Thesefeatures point to an existence of an internal Kelvin wave.363.4. Hydrographic comparisons of two flow scenariosFigure 3.10: Analysis of density variations at 1000 m depth for 2006, 2007 and 2008for the reference run and Exp. 2. The upper panels show density fields for 2006.The lower panels show the density differences in 2007 and 2008, as compared to2006, indicating a denser boundary current in the Amerasian Basin in the referencerun but not in Exp. 2. 373.4. Hydrographic comparisons of two flow scenariosTo further study of the propagation of the dense water, an investigation of thecore Atlantic layer (isopycnals σθ = 28.0 ) in the Amerasian Basin was conductedon the reference run.Figure 3.11: The vertical distribution of isopycnal σθ = 28.0 in the reference runalong the 1000 m isobath around the Arctic boundary. The left panel maps thelocation of the 1000 m isobath. The right panel shows time evolution of isopycnalσθ = 28.0 in the Amerasian Basin along the Arctic boundary (displayed the map). Asthe bathymetry around the Chukchi Plateau is complicated to contour, the verticalposition of isopycnal σθ = 28.0 in this region is dotted. Result shows the isopycnalrose in 2007 in the Makarov Basin. The signal reached all the way to the AlphaRidge, although in diminished amplitude.Following the isopycnal σθ = 28.0 (the core layer of Atlantic water), we can seethat the depth of this isopycnal σθ at the Makarov Basin changed dramatically from1300 m to 500 m during 2006 - 2007. The density contrast created a wedge-shapeisopycnal in the Makarov Basin.The isopycnal σθ = 28.0 in the rest of the basin responded accordingly - it waslifted by approximately 50 m at the west side of the Amerasian Basin (the CanadaBasin plus the Alpha Ridge) in no more than one year. The timing of this event iscoincident with the time when topostrophy reversed signs in the Amerasian Basin(Fig. 3.9).As flow at this time scale is geostrophic, it is not a surprise that the changein Atlantic isopycnal shape is synchronous with the change in the circulation. Thelift of the isopycnals implies the density increased at the continental slope, whichimplies vertical shear of the boundary current by the thermal wind relationship. Thisvertical shear implies a more anticyclonic flow in the intermediate layer comparedto the deep layer.In 2008, the isopycnal σθ = 28.0 in the Makarov Basin was pushed furtherup. The wedge-shape isopycnal remained in the Makarov Basin. In this year, theisopycnal σθ = 28.0 in the rest of the basin was lifted another 50 m. Note thatprevious density analysis indicates while the density along the Arctic boundary383.4. Hydrographic comparisons of two flow scenariosincreased after 2007, that in the interior basin decreased (Fig. 3.10). Therefore,the further lifting of isopycnal σθ = 28.0 means that density contrast between theboundary and the internal ocean were further increased, which created an enhancedanticyclonic flow as implied by the topostrophy field.The density and the isopycnal analysis suggests that the major difference be-tween the reference run and Exp. 2 is whether the dense water succeeds in pene-trating into the deep Arctic and reshaping the isopycnals. When a dramatic lift ofisopycnal slope developed along the Arctic boundary (the reference run), one canexpect an anticyclonic flow; however, if the dramatic lift of isopycnal slope is notdeveloped (Exp. 2), the anticyclonic flow will not be generated.3.4.3 The Arctic inflow analysisFigure 3.12: The amount of dense water inflows (σθ > 28.05, from the BarentsSea and the Fram Strait) (upper panel) and average topostrophy in the CanadaBasin (lower panel) versus time in both the reference run and Exp. 2, indicating anincrease dense water inflow is coincident with a change of topostrophy.In order to change the shape of the Atlantic layer, there must be a sufficient inflowof dense water. I integrated all inflows where σθ > 28.05 from a section across theFram Strait and a section along the Barents Sea shelf (Fig. 2.8; sections b and c,respectively) to further complete the puzzle.In the reference run, the dense inflow started to increase in 2005 (Fig. 3.12).This moderate increase was followed by a dramatic increase starting in 2006. Thehighest influx occurred in 2011 with a slightly decrease in the following years.During this period, topostrophy in the Canada Basin (Fig. 3.12 (a)) persistently393.4. Hydrographic comparisons of two flow scenariosdecreased in the reference run. It changed from positive to negative in 2007, whichindicated the change from the cyclonic to the anticyclonic circulation pattern.On the contrary, Exp. 2 maintained a positive topostrophy in the Canada Basinduring the model period. Note that its dense water inflow did not dramaticallyincrease as that in the reference run.While the influence from dense water inflow has been highlighted, its source isleft to be determined. There are two branches of inflow that can bring dense waterinto the Arctic, one is the Fram Strait branch and the other is the Barents Seabranch.Figure 3.13: Time series of dense water inflow (below σθ = 28.05 from a sectionacross the Fram Strait and a section along the Barents Sea shelf (Fig. 2.8)) in thereference run (upper panel) and the Exp. 2 run (lower panel). Results suggest thedense water influx that significantly increased after 2006 in the reference run mainlycomes from the Barents Sea. However, such change did not appear in Exp. 2.The increase in the dense water inflow in the reference run starting in 2006 wasmainly caused by a significant increase in the amount of Barents Sea dense water.Before that, the dense water was mainly from the Fram Strait. However, the densewater source in Exp. 2 continued to come mainly through the Fram Strait from theAtlantic Ocean (Fig. 3.13). It can be concluded that the difference between tworuns is attributed to the increase of dense water input from the Barents Sea.Finally, there is one last question left to be solved. Despite the differencesin the long term linear trend in topostrophy, one may notice that the short termtopostrophy variations in both runs are similar. Why is that? Given the fact thatboth runs are forced by the same wind field, I further investigated the correlationbetween the wind field and the change of topostrophy.The wind pattern over the Canada Basin (Fig. 3.14) is anticyclonic due to the403.4. Hydrographic comparisons of two flow scenariosFigure 3.14: Time derivatives of averaged topostrophy over the Canada Basin versusthe time series of averaged wind curl over the Canada Basin. Upper panel: thereference run, lower panel: Exp. 2. The wind affects the topostrophy on an annualtime scale.Beaufort high. The time derivative of topostrophy is well correlated with the windseries. The strongest anticyclonic wind was seen in 2007, when topostrophy hadlarge drops in both runs. The second largest decrease occurred almost at the end ofruns, 2013, when topostrophy was observed to have a similar downward trend. Thetopostrophy responds to the wind curl in less than one year.41Chapter 4Discussion4.1 Cause of observed increase in 230Th in the CanadaBasin4.1.1 Model EvaluationThe data-model comparison supports the results of the tracer simulation in threeways. First, the general vertical structure simulated by the model is in good agree-ment with observations. Second, the contrast between the low ice areas and high iceareas is reasonably reproduced. Third, similar to the observations, the simulated230Th experienced an increase in the Canada Basin in the reference case.However, there are two main discrepancies between model and data.First, there is a 6 year difference in the arrival time of high 230Th concentration inthe Canada Basin between the observation and simulation. This delay in the changeof the simulated tracer is caused by a late change of the velocity pattern in thereference run, suggested by the sensitivity experiments. One can expect the arrivaltime of high 230Th concentration will get closer to the observation if the changeof circulation pattern occurs earlier. Due to the difficulty in accurately modelingall features of the Arctic circulation, I cannot expect to generate an accurate mapof 230Th in this study: rather this attempt is a means to answer what circulationchanges could have altered the 230Th distribution in the real world.Second, there is a major dissolved 230Th deviation from the observations foundover the Alpha Ridge and in the Makarov Basin region. The model consistentlyunderestimated the tracer concentration in this region, which suggest an overesti-mation of scavenging intensity over the Alpha Ridge (more discussion in Section4.3). In this model, the high 230Th concentration over the Alpha Ridge is the maincontributor to the high 230Th observed in the Canada Basin. Therefore, the un-derestimation of scavenging coefficients over the Alpha Ridge directly causes theunderestimation of the increase of simulated 230Th concentration in the CanadaBasin.4.1.2 Sensitivity to sea ice concentrationExp. 1 indicates that the temporal change of sea ice concentration will result ina subsequent change in 230Th concentration by altering the scavenging intensity.Between the two phases of 230Th, the particulate form responds to the sea ice424.1. Cause of observed increase in 230Th in the Canada Basinchange more rapidly. In both phases, the difference between the Exp. 1 and thereference run grows slowly in time, suggesting the scavenging adjustment occurson a relatively long time scale. However, the variation of sea ice accounts for aless than 0.05 dpm m−3 variation in simulated 230Th concentration at intermediatedepth. Since the overall change at the same depth is more than 0.3 dpm m−3, thisexperiment indicates that the change in sea ice concentration over the model periodis not sufficient to cause the observed change in 230Th concentration.However, one should keep in mind that this result does not mean sea ice concen-tration is not important. On the contrary, the spatial pattern of sea ice concentrationplays an extremely important role in controlling the distribution of 230Th - becausea concentration contrast between the Canada Basin and the Alpha Ridge is requiredto cause the observed change.4.1.3 Sensitivity to circulation patternComparison between runs with the same ice concentration but different circulationsshows that the change in observed 230Th concentration is directly caused by theshift in circulation. This result is consistent with findings that non-linear profiles of230Th in the Arctic are generally due to lateral transport (Roy-Barman, 2009).This change in circulation also leads to a decrease of 230Th concentration over theAlpha Ridge, in addition to the observed increase in the Canada Basin. However,it does not result in a significant change in the Eurasian Basin.The decreasing tracer concentration over the Alpha Ridge suggests that thebuild-up of high concentration over the Alpha Ridge requires both low scavengingrates and limited exchange. It leads to a new implication, if the intermediate patternremains anticyclonic in the Amerasian Basin, the high concentration will not occuragain, because the high concentration source cannot be rebuilt. However, if theflow becomes cyclonic again, a high concentration in the Canada Basin might occurnext time the anticyclonic flow re-establish, provided high 230Th concentration canrebuild in the high ice region over the Alpha Ridge (at a time scale of a few decadesaccording to a one-dimensional scavenging model - Appendix B.1).Another important implication from this tracer simulation is that the real-worldtransition from cyclonic to anticyclonic pattern is stronger than that shown in themodel. Assuming the intermediate circulation pattern changed in 2004, as suggestedby Karcher et al. (2012), the observed peak of high 230Th concentration arrived in theCanada Basin 5 years afterward. In this study, the circulation pattern shifted during2007 and the simulated high 230Th peak arrived 8 years later. In the simulation,the newly-formed anticyclonic circulation was weak from 2007 to 2010. It increasedprogressively and slowly gained speed over 2010. However, judging by the arrivaltime of observed high 230Th concentration, one can infer that the anticyclonic flowin the real world is different from the simulation - it could be 1.6x stronger, which isquite possible due to the grid resolution (>10 km) versus Rossby deformation radius(∼6 km) (more discussion in Section 4.2).434.1. Cause of observed increase in 230Th in the Canada BasinThe modeled arrival of simulated high 230Th is consistent with the modeledvelocity. The distance between the Canada Basin and the Alpha Ridge is approxi-mately 1500 kilometers, which, given 5 years, implies a mean real world anticyclonicvelocity is of around 0.01 ms−1, compared to the modeled velocity of 0.006 ms− General circulation pattern revealed by 230ThFigure 4.1: Sketch of general circulation pattern (a) before 2007, (b) after 2007 atthe intermediate layer (between σθ = 27.8 and σθ = 28.05) inferred from simulated230Th. The flow pattern in the Eurasian Basin remains cyclonic while that in theCanada Basin changed to anticyclonic.The evolution of 230Th revealed that there are two circulation patterns duringthe model period (Fig. 4.1). The first circulation pattern contains three cyclonicgyres: in the Eurasian, Makarov and the Canada basins. This is consistent withprevious knowledge of the Arctic circulation (Rudels et al., 1994). The secondcirculation pattern also has cyclonic flow in the Eurasian Basin, however, it hasa completely reversed circulation pattern in the Canada Basin. Previous researchsuggested a small anticyclonic recirculation cell in the Beaufort sea in an overallcyclonic intermediate circulation in the Amerasian Basin (McLaughlin et al., 2009).However, the simulated tracer distribution indicates the switch to an anticyclonicflow is more than a regional phenomenon in the Beaufort sea but, actually, an entirereversal in the Amerasian basin (Fig. 4.1). Otherwise, with just a small anticyclonicregion in the Canada Basin, it is impossible for the high concentration 230Th fromthe Alpha ridge to reach the Canada Basin. Note that this larger scale change notonly exists in the model, it is also supported by the real-world 230Th data.The topostrophy (Fig. 3.9) indicates two counter rotating flows in the MakarovBasin, with the outer flow cyclonic and the inner flow anticyclonic. Unfortunately,444.2. Cause of circulation pattern changethe simulated 230Th distribution in the Makarov Basin does not have enough contrastto reflect these flows.One feature of the simulation that can be further investigated is the persistenceof the anticyclonic flow. In this study, the run is not long enough to track whetherthe anticyclonic circulation is enhanced in time or weakens or disappears.4.2 Cause of circulation pattern changeWhile the tracer model and the tracer observations both confirm the change incirculation, the model uncertainty in timing is not sufficient to conclude when thechange happened. However, the combination of two physical flow fields provides aperspective on the mechanism of such change. This study proposes that the changein circulation regime during the mid 2000s was associated with a strengthening ofthe Arctic dense water inflow (mainly from the Barents Seas).Effect from windThis attribution of the change is different from a previous finding that suggested thatstrong Ekman pumping was the reason for anti-cyclonic intermediate flow (Karcheret al., 2012). Their conclusion came from a strong anticyclonic wind pattern in 2004that was coincident with the year when the Atlantic layer shifted flow direction. Inboth my physical runs (the reference run and Exp. 2), a strong anti-cyclonic windpattern dominated in model year 2007 (Fig. 3.14). Coincidentally, the change offlow pattern in the reference run also showed up at this year.However, disagreement arises when considering Exp. 2, which is forced by thesame wind field as the reference run and therefore also experienced strong anti-cyclonic winds in 2007, did not exhibit an anti-cyclonic circulation.This study recognizes that wind has an impact on the topostrophy variation inthe intermediate layer by compressing intermediate flow at the continental slopebut does not agree that the wind is the dominant reason for anticyclonic flow inthe intermediate layer. Judging by the response to the wind field in Exp. 2, theresponse happens at a relatively short temporal scale on the order of a year (Fig.3.14) and once the anti-cyclonic wind pattern weakens, a rebound in topostrophy isseen immediately.Effect from dense inflowAnalysis on different flow fields suggests that the change in topostrophy is affectedby both a short term effect from the wind patterns (Fig. 3.14) and a long term effectfrom the variation of dense Arctic inflow (Fig. 3.12). This result is consistent witha previous study that concluded that the intensity of remote inflow and intensityof the surface circulation were the main factors that affected the strength of thecirculation in the Atlantic water (Lique et al., 2015). However, since the change in454.2. Cause of circulation pattern changewind pattern that governs the surface flow is not sufficient to cause the reversal ofthe flow pattern, my work emphasizes the importance of ‘remote forcing’.In the reference run, the long term linearly decreasing trend in topostrophy iswell correlated to a large volume of dense water transported into the Arctic system.This correlation suggests that the dense water input is a necessary condition foranticyclonic movement in the intermediate layer in the Canada Basin. This findingis in agreement with an earlier finding that recognized the importance of dense shelfwater in terms of explaining the temperature and salinity characteristics observedin the Canadian Basin (Rudels et al., 1994).This study suggests that a possible mechanism responsible for the circulationchange is a high density wedge that develops and forms an internal Kelvin waveafter the dense water inflow alters the previous density distribution. This deductioncomes from two reasons: First, the propagation of high density values (or ‘signals’)is trapped along the continental slope; and despite the anticyclonic movement in theintermediate flow, which has the coastline on its left, the density signal propagatesconsistently with the coast on its right. Second, the propagation of the densitysignal is much faster than the model velocity. The average boundary velocity speedin the reference run at 1000 m depth is 0.006-0.008 m s−1. Ignoring the transport inthe Eurasian Basin, the distance along the boundary of the Amerasian Basin aloneis more than 3000 km. It would take more than ten years for the density signal toadvect. However, this signal propagated throughout the Arctic in no more than oneyear (Fig. 3.10 and Fig. 3.11). This is only possible for a wave, not advection.The internal Kelvin wave velocity for a two-layer fluid is (eg. Støylen and Fer(2014)):c =√ρ2 − ρ1ρ2gH1H2H1 +H2(4.1)where ρ2 and ρ1 are the density in the upper and lower layer; H1 and H2 are theupper and lower layer thicknesses, respectively. Assuming ρ1 and ρ2 are 1028.0 kgm−3 and 1028.1 kg m−3 , and H1 and H2 are 1000 m and 2000 m, the propagationof the internal Kelvin wave is 25000 km y−1. In that case, it is not surprising thatthe isopycnal signal can circle the Amerasian Basin in just one year. This resultconforms to previous study that suggests the circulation in the AW layers can adjustto any change of forcing through the propagation of boundary trapped waves (Liqueet al., 2015).The Rossby radius of deformation (R) is:R =cf(4.2)where c is the wave speed and f is the Coriolis frequency. With f = 2Ω sin(90◦)464.2. Cause of circulation pattern change(∼ 1.458× 10−4), the Rossby radius is approximately 6 km. As the grid resolution(>10 km) is approximately 1.6x larger than the Rossby radius, it is likely thatthe wave is only marginally resolved and thus explains why the model velocity wasslower than the realistic one that suggested by 230Th observations.Source of the dense water inflowThe dense water mainly comes from the Barents Sea (Fig. 3.13). The formation ofdense Barents Sea Water is facilitated by brine rejection due to ice formation. Dueto the dramatic changes in sea ice conditions, the salinity of Barents Sea water hasincreased over the last 30 years (Oziel et al., 2016; Skogseth et al., 2005).The highest-ever recorded bottom salinity, 35.83, was observed in 2002 in thedeepest pool of Storfjorden (north-west Barents Sea) (Skogseth et al., 2005). Tothe north of Severnaya Zemlya (between the Barents Sea and the Kara Sea), a verysaline and dense bottom layer was also observed in 2007 (Rudels et al., 2013). Theseobservations all imply that the inflow into the Arctic Ocean is becoming denser andthus it is possible that the dense water source has increased.This argument is also supported by a recent modeling study that finds a 1.5xincrease of salt flux under a warming climate, which leads to a threefold increase ofshelf-slope volume flux below the warm core of Atlantic water (Ivanov and Watan-abe, 2013).The timing of these increases in density agrees with the 230Th data which re-vealed an anticyclonic flow from at least 2007. However, our model does not re-produce the exact time when the circulation change. Therefore, combining theobservations in the Barents Sea with the previous dense water analysis, I proposethat, if the dense water recorded in 2002 entered the Arctic Ocean, it is possiblythis source that triggered the anticyclonic flow in the Amerasian Basin.Note that as release of the dense water plume requires multiple conditions, suchas offshore (easterly) winds and large ice production events (Skogseth et al., 2005),the anticyclonic circulation pattern in the real Arctic should reverse, back to cy-clonic, once the dense inflow decreases.Also, because the dense water can be produced by ice formation and heat lossin the real world, this study suggests that whenever these (or similar) processes sig-nificantly increase the amount of dense water inflow, there is a potential to generateanticyclonic flow in the Amerasian Basin.Finally, there is one question left to be answered: Exp. 2 corrected a precipi-tation error in the reference run. Why then is the circulation in Exp. 2 unable toreproduce the 230Th observation?As mentioned before, dense water on the shelf is mainly formed by the brinerejection during ice formation events, as in cold, polar regions, changes in salinityaffect ocean density more than changes in temperature. I compared the amount ofsea ice formation in both the observation and the simulation. The annual sea ice474.2. Cause of circulation pattern changeFigure 4.2: Sea ice formation events from observations and model in the period of2002-2013. The red and purple lines show the change of sea ice concentration de-rived from satellite data using Bootstrap and NASA Team Algorithms respectively.The blue and green lines represents the increase in sea ice concentration from thereference run and the Exp. 2, respectively. Results indicate that, the ice formationpredicted in the model is less than that from observations.formation is approximated by subtracting the minimum sea ice condition (previousSeptember) from maximum sea ice condition (March). Both sea ice concentrationsfrom the Bootstrap (bt) and NASA team (nt) algorithms were used (Singarayer andBamber, 2003). The sea ice formations computed from the nt and bt algorithms aresimilar and are much larger than those simulated by the model runs (Fig. 4.2). Assuch, we can conclude that there is less salt released in the modeled Barents Seashelf region than in the real world.For this reason, the amount of dense water in Exp. 2 is not sufficient to alterthe shape of the Atlantic water isopycnals and thus is not able to alter the circula-tion. On the contrary, the lack of salt released from sea ice in the reference run iscoincidently compensated by the effect of reduced precipitation.It should be noted that we can not exclude the possibility that the drift ofmodel salinity has an influence on the distribution of dense inflow, as the densewater inflow in the reference run is apparently too much. However, this possibilitydoes not weaken our conclusion that an increase in dense water is the main reasonfor the circulation change.484.3. Model limitations and perspectives4.3 Model limitations and perspectivesAn unavoidable source of uncertainty is poor temporal and spatial coverage of Arcticobservations. However, this factor is not within our control. During the constructionof this model, several assumptions and extrapolations were made in order to use thelimited data to establish an Arctic-wide model. This section will focus on discussinghow those assumptions bring uncertainty into the simulated results.The model uncertainty comes from the following steps:• Biases related to circulationBiases related to the physical field is the main reason why the model has a six-year delay in the increase of 230Th concentration. Therefore, the biases related tothe physical field can not be ignored. In order to predict a precise 230Th map inthe future, it will be essential to simulate an accurate flow field first. This could beachieved by improving the ice model as well as increasing the model resolution.• Uncertainty in parameterizing scavenging coefficientsUncertainty in parameterization is another source for the model uncertainty.There are three sources of uncertainty in parameterzing the scavenging coefficients.First, the observation profiles under the high ice zone (over the Alpha Ridgeand in the Makarov Basin) strongly deviate from linear. However, although I amaware that the non-linear shape of tracer profiles increases the uncertainty in theparameterization, limited data in the high ice region does not allow me to filter theunwanted stations.Second, as lateral transport is ignored, the 230Th based scavenging rates mightbe biased by assuming a one-dimensional model (Roy-Barman, 2009). For example,for a place that receives lateral transport from another place with higher 230Th,the scavenging intensity would be underestimated; similarly, those affected by lower230Th water would be overestimated. Although the weak advection in Arctic justifiesthe method of using a one-dimensional model to estimate scavenging coefficients,this problem still generates potential uncertainty in this work. Also, we shouldnote that in this one-dimensional model, we assume zero particulate concentrationat the surface. Thus, this model does not take external sources (from sea ice orriver or dust from atmosphere) into consideration. This has a larger influence onthe particulate form than the dissolved form because the magnitude of a potentialexternal source is small compared with the dissolved 230Th concentrations.Third, the uncertainty in the linear relationship between sea ice concentrationand scavenging coefficients based on only two groups of observations can not be ig-nored. This uncertainty is probably the largest of the three. We can not exclude thepossibility that the relationship between sea ice concentration and particle concen-tration might not always be linear. For example, it is very likely that the change in494.3. Model limitations and perspectivesscavenging coefficients might be flatter when the sea ice concentration is approaching0 % and 100% and steeper in the middle (like an S-Curve). Also, there is only oneparticle type considered in the model. It is fair to believe that with more particletypes included, approximations of the scavenging coefficients can be more accurate.• Simplified in particle dynamicsThe particle dynamics is another source of uncertainty. I did not consider aggre-gation/disaggregation/remineralization processes in the ice-particle model (2.11).The incomplete particle dynamics cycle leads to an inaccuracy in particle-relatedbehavior and will be reflected in the tracer profile. This disadvantage will be ex-aggerated when it comes to the deeper ocean. For example, the break-down ofbiogenic particles in deep water, which may result in a noticeable decrease in thenumber of particles and thus a non-linear increase in the dissolved phase, will notbe reproduced in the model. A more realistic particle dynamics in the ocean wouldbe the largest area open for improvement in further modeling work.50Chapter 5SummaryThis work coupled a reversible scavenging model to an ocean flow field using NEMOv3.4 to model the water column distribution of dissolved and particulate 230Thover the Arctic region. The three-dimensional 230Th in Arctic is simulated forthe first time. Due to the lack of particle fields in the model, I established an ice-particle model to approximate the particle distribution. The model is run from 2002-2020 with a 12-year spin-up period. Two sensitivity experiments were conducted tounderstand the simulated 230Th behaviors, with one testing the impact from changesin circulation and the other testing the influence of changing particle fields causedby sea ice concentration variations.The objective of this model work is to address the following research questions:1. Can the model produce the observed change of 230Th? What does the observedchange imply about the Arctic circulation?The model produced a reasonable spatial pattern that agrees with early mea-surements. During the run, this spatial pattern changed when the flow field in thereference run switched from cyclonic to anticyclonic in the intermediate layer of theCanada Basin. It resulted in a subsequent increase in simulated 230Th concentrationin the south Canada Basin. This change in simulated 230Th concentration is con-sistent with the observed change. Thus it is fair to say that the model is successfulin capturing the main feature of the 230Th distribution.Suggested by the tracer model, the observed change in 230Th profile indicatesthe flow pattern in the Arctic Ocean has experienced a transition. Before 2007,intermediate circulation in the Canada Basin was cyclonic, which transported low230Th concentration towards the Alpha Ridge. In the late 2000s, the intermediateflow changed to an anticyclonic pattern. This change results in a reversal in 230Thtransport - high 230Th concentrations from the Alpha Ridge was transported into thesouth Canada Basin. Consequently, measurements in 2009 documented an unusual,non-linear 230Th profile.The tracer model did not reproduce the exact year when the 230Th increased andtherefore it is difficult to derive conclusions on when the anticyclonic flow started.Judging by the 230Th data only, the anticyclonic flow existed before 2007.2. What’s the reason for circulation changes in the Arctic?There are two processes that contribute to circulation changes: local wind and51Chapter 5. Summaryremote inflow.The influence of wind occurs on a relatively short time scale (in less than a year)but the persistently decreasing topostrophy adjustment occurred over a much longertime scale. As such, the reversal of topostrophy sign is a result of an increase denseinflow from Fram Strait and, mainly, from the Barents Sea. The increase of densewater inflow alters the density distribution in the Arctic. A lift of isopycnals gener-ates an internal Kelvin wave which carries an anticyclonic signal to the AmerasianBasin. The source of the dense water is located in the Barents Sea. The increase ofdense water source is probably caused by enhanced ice formation over the BarentsSea.There was a large ice formation event, together with the highest-ever recordedbottom salinity, observed in 2002 in the Barents Sea region (Skogseth et al., 2005).It is possible that the change in intermediate circulation was triggered by this event.As this three-dimensional Arctic 230Th model is established for the first time,this study also highlighted some limitations/difficulties in accurately simulating thetracer distribution. Future improvement can focus on two aspects: first, improvingthe scavenging model by finding a more realistic particle distribution through a morecomprehensive relation between sea ice and biology productivity and by developingof a more complete particle cycle; second, improving the physical field by betterparameterizing the ice model and increasing the model resolution. Regarding theshift in the intermediate circulation, this work was not able to resolve the sensitiv-ity of the intermediate circulation to the increasing amount of dense water inflow.More simulation experiments that explore the relationship between the amount ofdense water inflow and the dynamics of the intermediate layer are required to fullyunderstand the response mechanism.Since the intermediate layer is crucial to the overturning circulation and this the-sis has proven that the flow pattern in this layer has significantly changed, one canspeculate that the outflow from the Arctic Ocean would be saltier and stronger asa way to balance the intensified dense water inflow. If this is the case, the overturn-ing circulation would be driven by a higher pole-to-tropics density contrast, withthe condition that the density field in the tropical region experiences no significantchange. Thus, instead of slowing down due to a decreasing contrast in the tem-perature field under a warming scenario, the future overturning circulation couldpossibly be strengthened as a result of the larger density contrast. Moreover, theshift in the overall circulation pattern of the intermediate layer is expected to pro-duce changes in the heat distribution in the Arctic Ocean. Further exploration ofthe heat balance in the context of a new established intermediate circulation schemewould greatly contribute to the understanding of Arctic Ocean dynamics.52BibliographyAagaard, K. (1989). A synthesis of the Arctic ocean circulation. InternationalCouncil for the Exploration of the Sea (ICES).Bacon, M. P. and Anderson, R. F. (1982). Distribution of thorium isotopes betweendissolved and particulate forms in the deep sea. Journal of Geophysical Research:Oceans, 87(C3):2045–2056.Bacon, M. P., Huh, C.-A., and Moore, R. M. (1989). Vertical profiles of some naturalradionuclides over the Alpha Ridge, Arctic Ocean. Earth and Planetary ScienceLetters, 95(1-2):15–22.Blanke, B. and Raynaud, S. (1997). Kinematics of the Pacific equatorial under-current: An Eulerian and Lagrangian approach from GCM results. Journal ofPhysical Oceanography, 27(6):1038–1053.Brown, Z. W. and Arrigo, K. R. (2012). Contrasting trends in sea ice and primaryproduction in the Bering Sea and Arctic Ocean. ICES Journal of Marine Science:Journal du Conseil, 69(7):1180–1193.Dai, A. and Trenberth, K. E. (2002). Estimates of freshwater discharge from conti-nents: Latitudinal and seasonal variations. Journal of hydrometeorology, 3(6):660–687.Delworth, T., Clark, P., Holland, M., Johns, T., Kuhlbrodt, T., Lynch-Stieglitz, C.,Seager, R., Weaver, A., Zhang, R., et al. (2008). The potential for abrupt changein the Atlantic Meridional Overturning Circulation.Do¨o¨s, K. (1995). Interocean exchange of water masses. Journal of GeophysicalResearch: Oceans, 100(C7):13499–13514.Dutay, J.-C., Lacan, F., Roy-Barman, M., and Bopp, L. (2009). Influence of particlesize and type on 231Pa and 230Th simulation with a global coupled biogeochemical-ocean general circulation model: A first approach. Geochemistry, Geophysics,Geosystems, 10(1).Edmonds, H. N., Moran, S. B., Hoff, J. A., Smith, J. N., and Edwards, R. L. (1998).Protactinium-231 and thorium-230 abundances and high scavenging rates in thewestern Arctic Ocean. Science, 280(5362):405–407.53BibliographyFrancois, R. and Soon, M. (2010). Dramatic changes in the dissolved 230Th concen-tration of seawater in Canada Basin between 1995 and 2009: a transient Arcticcirculation signal? In EGU General Assembly Conference Abstracts, volume 12,page 2989.Henderson, G. and Wilkinson, J. (2012). Global database of literature and unpub-lished water column Pa and Th data. Retrieved from compilations/water-column th and pa dataset notes and references.Holdsworth, A. M. and Myers, P. G. (2015). The Influence of high-frequency at-mospheric forcing on the circulation and deep convection of the Labrador Sea.Journal of Climate, 28(12):4980–4996.Holloway, G., Dupont, F., Golubeva, E., Ha¨kkinen, S., Hunke, E., Jin, M., Karcher,M., Kauker, F., Maltrud, M., Maqueda, M., et al. (2007). Water properties andcirculation in Arctic Ocean models. Journal of Geophysical Research: Oceans,112(C4).Ivanov, V. and Watanabe, E. (2013). Does Arctic sea ice reduction foster shelfbasinexchange? Ecological Applications, 23(8):1765–1777.Karcher, M., Kauker, F., Gerdes, R., Hunke, E., and Zhang, J. (2007). On the dy-namics of Atlantic Water circulation in the Arctic Ocean. Journal of GeophysicalResearch: Oceans, 112(C4).Karcher, M., Smith, J. N., Kauker, F., Gerdes, R., and Smethie, W. M. (2012).Recent changes in Arctic Ocean circulation revealed by iodine-129 observationsand modeling. Journal of Geophysical Research: Oceans, 117(C8).Kim, M., Hwang, J., Kim, H. J., Kim, D., Yang, E. J., Ducklow, H. W., La Hyoung,S., Lee, S. H., Park, J., and Lee, S. (2015). Sinking particle flux in the sea icezone of the Amundsen shelf, Antarctica. Deep Sea Research Part I: OceanographicResearch Papers, 101:110–117.Ku, H. (1966). Notes on the use of propagation of error formulas. Journal ofResearch of the National Bureau of Standards, 70(4).Lerner, P., Marchal, O., Lam, P. J., Anderson, R. F., Buesseler, K., Charette,M. A., Edwards, R. L., Hayes, C. T., Huang, K.-F., Lu, Y., et al. (2016). Testingmodels of thorium and particle cycling in the ocean using data from station GT11-22 of the US GEOTRACES North Atlantic section. Deep Sea Research Part I:Oceanographic Research Papers, 113:57–79.Lique, C., Johnson, H. L., and Davis, P. E. (2015). On the interplay between thecirculation in the surface and the intermediate layers of the Arctic Ocean. Journalof Physical Oceanography, 45(5):1393–1409.54BibliographyLique, C., Treguier, A.-M., Blanke, B., and Grima, N. (2010). On the originsof water masses exported along both sides of Greenland: A Lagrangian modelanalysis. Journal of Geophysical Research: Oceans, 115(C5).Luo, Y. and Lippold, J. (2015). Controls on 231Pa and 230Th in the Arctic Ocean.Geophysical Research Letters, 42(14):5942–5949.Marchal, O., Franc¸ois, R., and Scholten, J. (2007). Contribution of 230Th measure-ments to the estimation of the abyssal circulation. Deep Sea Research Part I:Oceanographic Research Papers, 54(4):557–585.Marchal, O., Franc¸ois, R., Stocker, T. F., and Joos, F. (2000). Ocean thermohalinecirculation and sedimentary 231Pa/230Th ratio. Paleoceanography, 15(6):625–641.McLaughlin, F. A., Carmack, E. C., Williams, W. J., Zimmermann, S., Shimada,K., and Itoh, M. (2009). Joint effects of boundary currents and thermohalineintrusions on the warming of Atlantic water in the Canada Basin, 1993–2007.Journal of Geophysical Research: Oceans, 114(C1).Meier, W., Fetterer, F., Savoie, M., Mallory, S., Duerr, R., and Stroeve, J. (2013).NOAA/NSIDC Climate Data Record of Passive Microwave Sea Ice Concentration,Version 2. ID, Boulder, Colorado, USA.Oziel, L., Sirven, J., and Gascard, J.-C. (2016). The Barents Sea frontal zones andwater masses variability (1980–2011). Ocean Science, 12(1):169–184.Peng, G., Meier, W., Scott, D., and Savoie, M. (2013). A long-term and repro-ducible passive microwave sea ice concentration data record for climate studiesand monitoring. Earth System Science Data, 5(2):311–318.Pnyushkov, A. V., Polyakov, I. V., Ivanov, V. V., and Kikuchi, T. (2013). Structureof the Fram Strait branch of the boundary current in the Eurasian Basin of theArctic Ocean. Polar Science, 7(2):53–71.Roy-Barman, M. (2009). Modelling the effect of boundary scavenging on Thoriumand Protactinium profiles in the ocean. Biogeosciences, 6(12):3091–3107.Rudels, B. and Friedrich, H. J. (2000). The transformations of Atlantic water in theArctic Ocean and their significance for the freshwater budget. In The freshwaterbudget of the Arctic Ocean, pages 503–532. Springer.Rudels, B., Jones, E., Anderson, L., and Kattner, G. (1994). On the intermediatedepth waters of the Arctic Ocean. The polar oceans and their role in shaping theglobal environment, pages 33–46.55BibliographyRudels, B., Schauer, U., Bjo¨rk, G., Korhonen, M., Pisarev, S., Rabe, B., andWisotzki, A. (2013). Observations of water masses and circulation in the EurasianBasin of the Arctic Ocean from the 1990s to the late 2000s. OS Special Issue:Ice-Atmosphere-Ocean interactions in the Arctic Ocean during IPY: the Damoclesproject, 9(1):147–169.Scholten, J., Van Der Loeff, M. R., and Michel, A. (1995). Distribution of 230Th and231Pa in the water column in relation to the ventilation of the deep Arctic basins.Deep Sea Research Part II: Topical Studies in Oceanography, 42(6):1519–1531.Singarayer, J. S. and Bamber, J. L. (2003). EOF analysis of three records of sea-iceconcentration spanning the last 30 years. Geophysical Research Letters, 30(5).Skogseth, R., Fer, I., and Haugan, P. M. (2005). Dense-water production and over-flow from an Arctic coastal polynya in Storfjorden. The Nordic Seas: An Inte-grated Perspective, pages 73–88.Smith, G. C., Roy, F., Mann, P., Dupont, F., Brasnett, B., Lemieux, J.-F., Laroche,S., and Be´lair, S. (2014). A new atmospheric dataset for forcing ice–ocean mod-els: Evaluation of reforecasts using the Canadian global deterministic predictionsystem. Quarterly Journal of the Royal Meteorological Society, 140(680):881–894.Spall, M. A. (2013). On the circulation of Atlantic Water in the Arctic Ocean.Journal of Physical Oceanography, 43(11):2352–2371.Støylen, E. and Fer, I. (2014). Tidally induced internal motion in an Arctic fjord.Nonlinear Processes in Geophysics, 21(1):87–100.Trimble, S., Baskaran, M., and Porcelli, D. (2004). Scavenging of thorium isotopesin the Canada Basin of the Arctic Ocean. Earth and Planetary Science Letters,222(3):915–932.Tukey, J. W. (1977). Exploratory data analysis. First Edition, Reading, Mass:Addison-Wesley Publishing Co.Venchiarutti, C., Jeandel, C., and Roy-Barman, M. (2008). Particle dynamics studyin the wake of Kerguelen island using thorium isotopes. Deep Sea Research PartI: Oceanographic Research Papers, 55(10):1343–1363.Yang, J. (2005). The Arctic and subarctic ocean flux of potential vorticity and theArctic Ocean circulation. Journal of Physical Oceanography, 35(12):2387–2407.Yu, E.-F., Francois, R., Bacon, M., and Fleer, A. (2001). Fluxes of 230Th and231Pa to the deep sea: Implications for the interpretation of excess 230Thand 231Pa/230Th profiles in sediments. Earth and Planetary Science Letters,191(3):219–230.56Zhang, J. and Steele, M. (2007). Effect of vertical mixing on the Atlantic Waterlayer circulation in the Arctic ocean. Journal of Geophysical Research: Oceans,112(C4).57Appendix AChoice of sea ice concentrationThere are two options to embed sea ice concentration into the tracer model. Thefirst choice is to directly apply simulated sea ice results from the existing physicalruns (Table 2.1). The alternative one is to use satellite-derived sea ice data fromNOAA/NSIDC Climate Data Record dataset (Meier et al., 2013; Peng et al., 2013).Holdsworth and Myers (2015) suggests that the model generally does a good jobof representing sea ice in the Labrador Sea. However, while the simulated sea iceis reasonable in the Labrador Sea, it is less than satisfactory in the Arctic region.A data-model comparison shows that while the observed data only has high seaice concentrations on the CAA side of the Canada Basin, the model has high seaice concentration in the entire Canada Basin (Fig. A.1). Also, while the observedsea ice is thin in the shelf region, the model displays thick sea ice there. Thisresult indicates that the spatial variation of the modeled sea ice deviates from theobservations.In this study, the variation of sea ice concentration is crucial in setting thescavenging intensity in the Arctic region, therefore, I chose observed sea ice data asmodel inputs.Figure A.1: Data-model comparison on average sea ice pattern from 2002-2014.From left to right: observations from (a) the NASA Team (NT), (b) Bootstrap(BT) algorithm and (c) the reference run. Result shows that the sea ice spatialpattern from the model deviates from the observations.However, the observed sea ice dataset contains ice concentration products from58Appendix A. Choice of sea ice concentrationtwo well established algorithms: the NASA Team (NT) algorithm and the Boot-strap (BT) algorithm. Given that there are considerable differences in total icecovered area between these two products due to the difference in processing passivemicrowave radiation data, the following paragraphs discuss the choice between theseproducts.Previous work analyzes the ice products from these two algorithms using Em-pirical Orthogonal Function (EOF). Result indicates that, in general, the NASAteam data produce lower ice concentrations than the Bootstrap. However, the ma-jor modes of variability derived from the EOF analysis from both products wereessentially the same (Singarayer and Bamber, 2003).To model the tracer distribution, variability in sea ice concentration is moreimportant than absolute magnitude. As such, the high similarity of the most signif-icant modes of variability makes the choice of dataset less critical. Considering thatlarge areas of artificial-saturation observed in the BT records, I selected productsfrom NT as sea ice inputs in this study.Note that not all satellites pass close enough to the North Pole for their sensorsto collect data there. This missing data around the North Pole is filled in by nearest-neighbor interpolation using Scipy.interpolate Python package.59Appendix BPreliminary modelB.1 1D ModelThe preliminary 1D model solves a system of coupled, linear equations. It is setup on a 1D domain with a zero background velocity field imposed. Steady statenumerical solutions are compared with analytical solutions for consistency.Initialized with zero concentrationWith a hypothetical state of no advection, there is no lateral transport of 230Thconcentration. As such, for the steady state, the numerical solutions for 230Th (inboth forms) linearly increase in magnitude with depth (Fig. B.1: left and midpanels). The time required for tracers to reach steady state increases from surfaceto seafloor. The magnitudes of the numerical solutions are consistent with theanalytical results.Initialized with zero concentrations, the time for 230Th at 1000 m depth toachieve steady state is approximately 18 years under the ice-free and and approxi-mately 125 years under the permanent ice area. Both forms of tracer increase fast inconcentrations for the first few years and their values slowly converge to the steadystates (Fig. B.1: right panels).Initialized with steady state 230Th concentrations from low iceregionAn interesting question to ask will be: if the 230Th profiles were initialized with asteady state low 230Th concentration from a low ice region, instead of zero concen-tration, how long it takes to build up high 230Th values.In this part, the initial 230Th concentrations are generated from a ice-free, steadystate 230Th profile. Result shows that 230Th at 1000 m depth spends approxi-mately 125 years to reach a new steady state (Fig. B.2). This length of time is,surprisingly, not significantly shorter than that initialized with zero concentration(125∼130 years). This quick experiment supports that the significant increase inobserved 230Th in such a short time scale (in less than 5 years) is not caused by achange in scavenging intensities.60B.1. 1D Model(a) under low ice condition(b) under high ice conditionFigure B.1: Numerical solutions of one-dimensional 230Th model without back-ground velocity. The left and middle panels in (a) and (b) show the evolutions of230Th vertical profiles (dissolved and particulate, respectively) under different sea iceconditions. The right panels focus specifically on the evolutions at 1000 m depth,with gray lines indicating the steady state values. Results show that the steadystate solutions in both phases propagate from surface to seafloor and that the timeto reach steady state would be longer for profiles under a high ice condition.61B.2. 2D ModelFigure B.2: Numerical solutions of one-dimensional 230Th model without back-ground velocity and with a steady state 230Th concentration from low ice region.Panels structure is the same as Fig. B.1. The amount of time to reach steady stateis found to be just slightly shorter than that initialized with zero concentration (<5years).B.2 2D ModelThe 2D model not only adds an extra dimension to the 1D model, but also includesan addition process: the advection process. In this test, the model is forced by anidealized eastward constant flow. The model domain is 5000 m by 1000 km, with a20 km horizontal resolution and a 100 m vertical resolution. The model is separatedinto two regions with different characteristics: the first 400 km in x direction isassumed to be covered by permanent ice; the rest (400 - 1000 km in x direction),however, is ice-free. Their scavenging rates are estimated with 100% and 0% iceconcentration, respectively. This model is established to examine how the tracerdistribution evolves with the advection process imposed.With advection, tracer concentrations from the high ice region was advected tothe low ice covered zone. With a constant 100 km y−1 velocity, the front of the high230Th concentration water mass (originally at x = 400 km) reached the right edgeof the model boundary (at x = 1000 km) during the sixth year. The flow pathwayis reflected by the movement of the high 230Th.Note that the high dissolved 230Th concentration dissipated after it was trans-ported to the ice free zone which has higher scavenging intensities. On the contrary,the particulate 230Th gained large amounts of concentration from the huge drop indissolved 230Th concentration. In order to balance the decrease in dissolved 230Th,the maximum particulate 230Th concentration even exceeded that originally devel-oped under the high ice region.62B.2. 2D Model(a) dissolved 230Th(b) particulate 230ThFigure B.3: Numerical solutions of two-dimensional 230Th model, with a constanteastward background velocity imposed for ten model years. Consistent with modeledvelocity speed (100 km y−1), the front of the high 230Th concentration water mass(originally at x = 400 km) reached the right edge of the model boundary (at x =1000 km) during the sixth year. However, all panels similarly display significantincreases in particulate 230Th concentrations while the high 230Th water masses aretransported eastward.63B.2. 2D ModelBack to the high ice region, the scavenging rates are low enough to build up high230Th (in both phases). However, with continuous transport of low 230Th water massinto the high ice zone ( 0-400 km in x direction), the high 230Th concentration wasnot able to re-establish.64Appendix CPhysical model evaluationThis section conducted a quick evaluation on the performances of the physicalmodel’s salinity and temperature fields. The observed data used in the comparisonis from the World Ocean Database (WOD). The evaluation focused on six Arcticsub-regions: the boundary of the Canada and Makarov basins, the internal CanadaBasin, the interior of the central Arctic, the Kara sea opening and the Barents Seaopening (Fig. C.1). This comparison focuses on the intermediate layer: the datafrom 1000 ± 50 m were chosen; model results all from the 1000 m depth.Figure C.1: Locations of the data-model comparison: the boundary of the Canadaand Makarov basins, the internal Canada Basin, the internal of central Arctic, theKara sea opening and the Barents Sea opening. Points indicate the sample locations.The magnitude of simulated salinity compares very well with the observed salin-ity fields in both EXH001 and EXH005 (used in the reference run and Exp. 2,respectively). Observations suggest that there are large variations in the MakarovBasin and the Kara Sea opening. This feature is captured in the model. For the restof the sub-domains, the modeled salinity also compares favorably to the observedsalinity, which is in the range of 34.88-34.94.For the temperature field, the largest variations are also seen in the Makarov65Appendix C. Physical model evaluation(I) salinity(II) temperatureFigure C.2: Physical model evaluations with respect to the observations from theWorld Ocean Database (WOD). The model simulations of the intermediate depthin six sub-domains are generally in line with observations.66Appendix C. Physical model evaluationBasin and in front of the Kara Sea and the Barents Sea openings. The results showthat the magnitude of simulated temperature in both runs generally corresponds wellto the observations with exceptions in the Kara Sea and the Barents Sea openingarea. In this region, the simulated temperature is approximately 0.2-0.3 ◦C higherthan that from observations. However, as the density/stratification in the Arctic,which includes the Kara Sea opening, is salinity-dominant, the influence caused bytemperature deviations from the observations are limited. As such, it is fair toassume that the overall Arctic dynamics is not strongly affected.There is a hardly noticeable feature revealed by this evaluation: the modeledsalinity field in EXH001 is found to have higher values than that in EXH005. Thisdifference mainly occurred in the Barents Sea and the Kara Sea openings. From thelimited observations in the Barents Sea opening, slightly higher values of salinityfield are found compared to those from the model. This result supports the previousargument that insufficient salt is released in the modeled Barents Sea region inEXH005.Summarizing, the model simulations of the intermediate depth in six sub-domainsare in line with observations and the model is fit for the purpose of the study.67Appendix D231Pa and 230Pa/230Th in theintermediate layerIn order to maintain consistency with the earlier 230Th analysis (Chapter 3), the231Pa and 230Pa/230Th concentrations are shown in year 2007, 2009, 2015, 2020 at1000 m depth for the reference run.D.1 Dissolved 231PaThe model behavior of dissolved 231Pa is almost the same as dissolved 230Th duringthe model period (Fig. D.1). Due to this similarity, the 231Pa results are put in thisappendix for compactness and clarity of the thesis.In phase I (2007), the highest concentration of dissolved 231Pa was located inthe Alpha Ridge area, with some low 231Pa water masses from the Canada Basinextending cyclonicly along the basin boundary. In phase II (2009), the spatialpattern did not change significantly. A slight difference to 2007 is that the low231Pa water mass retreated back to the Canada Basin, rather than extending to theAlpha Ridge.In phase III (2009) and IV (2015), it is clear that the high dissolved 231Pawater masses moved from the Alpha Ridge, via the south Canada Basin, toward theChukchi Plateau and the low 231Pa concentration water moved toward the MakarovBasin and the central Arctic.The high similarity between the 231Pa and 230Th evolution suggests that bothof the tracers are similarly affected by lateral transports and scavenging.The change in the highest 231Pa location, as well as the change in the low 231Pawater propagation direction, suggests the flow pattern reversed from cyclonic toanticyclonic during the model period. This result is consistent with that revealedby the 230Th simulation.Unfortunately, 231Pa observations in the Canada Basin are only available in2007 and 2015, but not in 2009 when the high 230Th concentrations were detected.However, given the similarity in the 230Th and 231Pa cycles, one can expect that theincrease in 231Pa concentration found in phase III and IV is a real feature.68D.2. Particulate 231PaFigure D.1: Horizontal distribution of simulated dissolved 231Pa concentration inthe intermediate layer. Results are displayed in phase (a) I (2007), (b) II (2009),(c) III (2015) and (d) IV (2020). Tracer concentrations shown here are annuallyaveraged model results. High concentrations seen in phase I (2007) and phase II(2009) moved southward along the CAA in phase III (2015) and dissipated in phaseIV (2020).D.2 Particulate 231PaThe evolution of particulate 231Pa exhibited a similar pattern to that of particulate230Th. In phase I (2007) and II (2009), the highest particulate 231Pa was foundunder the seasonally ice-free area in the Canada Basin. Intermediate particulate231Pa concentration water was located over the Alpha Ridge. The Eurasian Basinand the south Makarov Basin had low particulate 231Pa water masses.69D.2. Particulate 231PaFigure D.2: Horizontal distribution of simulated particulate 231Pa concentration inthe intermediate layer. Results are displayed in phase (a) I (2007), (b) II (2009), (c)III (2015) and (d) IV (2020). There was an domain-wide increase in the particulate231Pa concentration, with the high particulate 231Pa constrained to the ChukchiPlateau side of the Canada Basin in phase III and IV.In phase III (2015) and IV (2020), the particulate 231Pa experienced an Arctic-wide increase. The previous low 231Pa in the Eurasian Basin and the south MakarovBasin increased to an intermediate level. Meanwhile, high particulate 231Pa concen-trations were more constrained to the Chukchi Plateau side of the Canada Basin.This spatial pattern is different from the ice-type particle field, which suggests anadvection signal. Unfortunately, I do not have particulate 231Pa data to evaluatethis change at this stage.70D.3. Dissolved 231Pa/230ThD.3 Dissolved 231Pa/230ThThe distribution of dissolved 231Pa/230Th is, again, similar to the spatial pat-tern in dissolved 230Th, but reversed. Where the 230Th concentration is high, the231Pa/230Th ratio turns out to be low (as 230Th is the denominator). In phase I, ahigh 231Pa/230Th pool occurred in the Canada Basin, with a noticable north-westtrend towards the Alpha Ridge. The rest of the Arctic mainly consisted of low231Pa/230Th water because of the relatively low scavenging intensities which werecaused by high sea ice condition.The spatial pattern in phase II did not change significantly as compared to phaseI. However, the north-west spread of the high 231Pa/230Th terminated. The high231Pa/230Th poll retreated back to the south Canada Basin. Meanwhile, due to asharp decrease in sea ice concentration at the east Arctic Ocean, the 231Pa/230Thin the south Makarov Basin and the Russia side of the Eurasian Basin increased.Phase III depicts a significant change in the 231Pa/230Th field. The high 231Pa/230Thwater now moved east from the south Canada Basin, through the Chukchi Plateau ,to the Makarov Basin. The low 231Pa/230Th water over the Alpha Ridge was movedsouthward to fill the south Canada Basin. Meanwhile, an intermediate 231Pa/230Thwater developed in the Eurasian Basin.The spatial pattern in phase IV further evolved and became less similar to thatin dissolved 230Th. In this phase, low 231Pa/230Th occurred in the Canada Basinand over the Alpha Ridge, while high 231Pa/230Th occupied the Makarov Basin andmedium 231Pa/230Th dominated the Eurasian Basin. This pattern reflected threemain gyres in the intermediate circulation.The difference in spatial pattern between dissolved 230Th and 231Pa/230Th sup-ports the idea that 231Pa/230Th is a useful tool for larger scale circulation. However,because the residence time of the Arctic Ocean intermediate water is relatively shortas compared to the tracers’, the usefulness of 231Pa/230Th is limited as its spatialvariation can be very quickly homogenized (eg. Fig. D.3).71D.4. Particulate 231Pa/230ThFigure D.3: Horizontal distribution of simulated particulate 231Pa/230Th concen-tration in the intermediate layer. Results are displayed in phase (a) I (2007), (b)II (2009), (c) III (2015) and (d) IV (2020). Tracer concentrations shown here areannually averaged model results. High concentrations seen in phase I (2007) andphase II (2009) moved anticyclonic to the Chukchi Plateau side of the Canada Basinin phase III (2015) and dissipated in Phase IV (2020).D.4 Particulate 231Pa/230ThEvolution of the particulate 231Pa/230Th is consistent with that seen in the dis-solved 231Pa/230Th (Fig. D.4). Similar to dissolved 231Pa/230Th, a spread of high231Pa/230Th was seen in the Canada Basin in phase I (2007). It extended north-westinto the Alpha Ridge region. Meanwhile, a transport of low 231Pa/230Th water fromthe Eurasian Basin to the Amerasian Basin was observed in the simulation.72D.4. Particulate 231Pa/230ThFigure D.4: Simulated particulate 231Pa/230Th concentration in phase (a) I (2007),(b) II (2009), (c) III (2015) and (d) IV (2020). The high particulate 231Pa/230Thwater originally propagated northward toward the Alpha Ridge but finally accumu-lated in the Makarov Basin.In phase II (2009), the north-west spread of high 231Pa/230Th water in theCanada Basin discontinued. A significant amount of low 231Pa/230Th water fromthe Eurasian Basin was transported further into the central Arctic.Phase III (2015) shows that the low 231Pa/230Th concentration pool traveledalong the Canada side of the Alpha Ridge and arrived in the south Canada Basin.The change in spatial pattern was also witnessed by the shift in the high 231Pa/230Thpool location, which was seen in both the south Canada Basin and in the MakarovBasin.Phase IV (2020) indicates an increase in overall 231Pa/230Th values due to thin-73D.4. Particulate 231Pa/230Thner sea ice condition. However, if we focus on the spatial pattern instead of theabsolute value, the reader can find that the particulate 231Pa/230Th was relativelylow in the Canada Basin; very high in the Makarov Basin and intermediate in theEurasian Basin. These three different particulate 231Pa/230Th water masses coin-cided with three intermediate depth circulation loops. This result is similar withthat inferred from the dissolved 231Pa/230Th.74Appendix ETracer distribution in deep layerE.1 Dissolved 231Pa and 230ThWater beneath the intermediate layer is considered as the deep layer. Tracer con-centrations from a depth of 3000 m are chosen to present model results for thislayer.Among the limited observations in deep water, there is no significant changesobserved (Fig. 3.5). Here, the year 2002, 2007 and 2015 are chosen to represent thetracer pattern before, during and after the intermediate flow pattern changes.The dissolved 231Pa and 230Th distributions in 2002 (before the shift in theintermediate layer occurred) showed the Canada Basin, the Makarov Basin and theEurasian Basin contained the lowest, the highest and intermediate values in both230Th and 231Pa concentrations, respectively (Fig. E.1).The spatial pattern in 2007, when the shift in the intermediate circulation oc-curred, did not significantly change as compared to 2002. The main differences wereseen in the low 230Th and 231Pa concentration tongues, which extended further northtoward the Alpha Ridge.In 2015 after the intermediate change happened, the low concentration tonguestill existed along the boundary of the Canada Basin; however, instead of movingnorth-west toward the Alpha Ridge side, it extended all the way to the ChukchiPlateau. Meanwhile, the high 230Th and 231Pa at this layer decreased dramaticallyin the Makarov and Eurasian basins. However, unlike tracers in the intermediatelayer, there were not high tracer concentration tongues developed at this depth.Since the velocity is slow (approximately 0.002m s−1) in the deep layer, theadvection effect is small as compared to the upper ocean. It is likely that the 2015pattern is caused by the sinking from the intermediate layer. The scavenging processdominates the deep ocean tracer distribution.75E.1. Dissolved 231Pa and 230ThFigure E.1: Simulated dissolve 231Pa and 230Th concentration in year (a,b) 2002,(c,d) 2007 and (e,f) 2015. The low tracer concentrations in the south Canada Basinfound in 2002 and 2007 extended to the Chukchi Plateau side of the Canada Basinin 2015.76E.2. Particulate 231Pa and 230ThE.2 Particulate 231Pa and 230ThResult shows that, in 2002, the high concentrations of both particulate 230Th and231Pa were constrained to the Chukchi Plateau side of the seasonally ice-free CanadaBasin (Fig. E.2). In 2007 and 2015, these zones dissipated but relatively high 230Thand 231Pa concentration could still be seen on the Chukchi Sea side of the CanadaBasin. Again, there were no high particulate tracer tongues seen at this depth.E.3 Ratio of 231Pa/230ThSurprisingly, while the dissolved and particulate 231Pa and 230Th distributions onlyprovide obscure circulation patterns, their ratios indicated a clear cyclonic movementof the 231Pa/230Th ratio in the Canada Basin in 2007, from the south Canada Basinto the west basin along the boundary of the CAA; and an anticyclonic movement in2015, revealed by the east-moving high 231Pa/230Th tongues from the south CanadaBasin to the Chukchi Plateau side of the Canada Basin. Such change suggests thatthe Arctic, to be specific, the Canada Basin, experienced a significant change incirculation pattern.However, as the deep layer receives plenty of sinking particles from the upperocean, it is hard to quantify how much of the change in tracer concentration wascaused by the intermediate layer. It therefore appears difficult to derive some defini-tive conclusions upon the behavior of the trace elements from these results.As for the Eurasian Basin, the high 231Pa/230Th moved in a cyclonic patternover the model period. As a result, one can expect there is no significant change inflow field in this (and the intermediate) layer.77E.3. Ratio of 231Pa/230ThFigure E.2: Simulated particulate 231Pa and 230Th concentration in year 2002 (a)(b),2007 (c)(d) and 2015 (e)(f). High tracer concentrations zones over the south edge ofthe Canada Basin found in 2002 spread into the central Canada Basin in the 2007and 2015.78E.3. Ratio of 231Pa/230ThFigure E.3: Simulated dissolve and particulate 231Pa/230Th concentration in year2002 (a)(b), 2007 (c)(d) and 2015 (e)(f). In the Canada Basin, a high ratio zonedeveloped along the CAA side in 2007 then it flowed along the south edge of theCanada Basin to the Chukchi Plateau side in 2015.79


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