UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Ice-ocean interactions in Milne Fiord Hamilton, Andrew Kent 2016

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

Item Metadata


24-ubc_2016_november_hamilton_andrew.pdf [ 29.28MB ]
JSON: 24-1.0314106.json
JSON-LD: 24-1.0314106-ld.json
RDF/XML (Pretty): 24-1.0314106-rdf.xml
RDF/JSON: 24-1.0314106-rdf.json
Turtle: 24-1.0314106-turtle.txt
N-Triples: 24-1.0314106-rdf-ntriples.txt
Original Record: 24-1.0314106-source.json
Full Text

Full Text

Ice-ocean interactions in Milne FiordbyAndrew Kent HamiltonB. Sc., University of Alberta, 1996M. Sc., The University of British Columbia, 2006A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Civil Engineering)The University of British Columbia(Vancouver)August 2016c Andrew Kent Hamilton, 2016AbstractWidespread break up of ice shelves, glacier tongues, and the loss of ice-dammed epishelf lakes onthe northern coast of Ellesmere Island has motivated a need to understand fjord dynamics, yet theoceanography of fjords here is not well studied. Here, we present ocean profiling and mooring datacollected from 2011 to 2015 in Milne Fiord, the last ice shelf-epishelf lake-glacier tongue fjord inthe Arctic.The data reveal that seasonal and interannual variations of fjord water properties and circulationare strongly impacted by the presence of the Milne Ice Shelf. The ice shelf forms a floating damthat traps surface runoff resulting in strong stratification and an elevated fjord heat content. Waterexchange below the ice shelf is restricted to a narrow basal channel, prolonging the export of fjord-modified water, including subglacial runoff, by several months. In contrast, intermediate watersthat penetrate to the Milne Glacier grounding line are freely exchanged and respond to offshorevariations, with implications for submarine melting.Unexpectedly, the depth of the halocline between the epishelf lake and seawater, often used toinfer the thickness of the ice shelf, varied by several meters each year. This variability resulted fromrapid inflow of surface runoff during summer followed by slow drainage under the ice shelf overwinter, which is well modelled as hydraulically-controlled flow through a channel. A mixing eventalso abruptly changed the depth of the halocline by 1.5 m in less than 24 hours, indicating cautionmust be used when inferring ice shelf mass balance from halocline depth.Submarine melt rates, estimated by two independent approaches, are strongly dependent on thevertical distribution of heat in the fjord. Spatial variation of ice thickness resulted in a heterogeneousdistribution of melt. The highest estimated melt rate (4 m a1) occurred where the glacier was incontact with warm Atlantic Water at the grounding line, and enhanced near-surface melting is drivenby the elevated heat content of the upper water column. Estimated melt rates are limited by weakcurrents (⇠1 cm s1) in the fjord imposed by the presence of the ice shelf.iiPrefaceThis thesis presents the original research of the author, conducted under the supervision of B. Lavaland D. Mueller. The thesis is presented as three, self-contained manuscripts (Chapters 2, 3, and 4)for which I designed the research program, performed the data collection and analysis, and wrotethe manuscripts under the guidance of my supervisors. The contribution of my co-authors and col-leagues to each manuscript are noted below. Members of my supervisory committee, G. Lawrenceand H. Melling, provided comments on drafts of Chapters 2, 3, and 4. Chapter 1 provides the overallcontext for the work and places the manuscripts within that context. Chapter 5 presents the overallconclusions of the thesis and suggests opportunities for future work.A version of Chapter 2 is being prepared for submission to a peer-reviewed journal as ”Modifi-cation of ocean properties and circulation by an ice shelf and glacier tongue in Milne Fiord” by A.Hamilton, D. Mueller, B. Laval, A. White, C. Mortimer, and L. Copland. I was the lead investigatorfor the work presented in this chapter. A. White, C. Mortimer, and L. Copland collected and pro-cessed the ice-penetrating radar data from 2008, 2009, and 2011. D. Mueller and N. Wilson assistedwith the collection of and processed the ice-penetrating radar acquired in 2012 and 2013. Additionalfield data was acquired through collaboration with the SwitchYard Project (P.I. M. Steele, Univer-sity of Washington) and various publicly available remote sensing data products were utilized. Iwrote the manuscript and all co-authors provided comments.A version of Chapter 3 is being prepared for submission to a peer-reviewed journal as ”Dynamicresponse of the last remaining Arctic epishelf lake to seasonal and long-term forcing” by A. Hamil-ton, B. Laval, D. Mueller and W. Vincent. I was the lead investigator for the work presented in thischapter. W. Vincent collected and contributed conductivity-temperature-depth profiles from 2004to 2011, and D. Mueller collected profiles in 2009. D. Mueller performed the digitization of aerialand satellite imagery. I wrote the manuscript and all co-authors provided comments.A version of Chapter 4 is being prepared for submission to a peer-reviewed journal as ”Depth-dependent submarine melt rates of a glacier tongue and ice shelf in a High Arctic fjord” by A.Hamilton, D. Mueller, B. Laval, and W. van Wychen. I was the lead investigator for the work pre-sented in this chapter. W. van Wychen provided the ice surface velocity data. I wrote the manuscriptand all co-authors provided comments.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Review of relevant literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2.1 Northern Ellesmere Island . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2.2 Ice shelves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.2.3 Epishelf lakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.2.4 Fjord oceanography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.2.5 Glacial fjords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.2.6 Submarine melting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111.3 Thesis objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181.4 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.5 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 Modification of ocean properties and circulation by an ice shelf and glacier tonguein Milne Fiord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22iv2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.2.1 Field Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.2.2 Bathymetry and ice thickness mapping . . . . . . . . . . . . . . . . . . . 262.2.3 Hydrography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.2.4 Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.3.1 Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.3.2 Hydrography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.3.3 Circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372.3.4 Mooring timeseries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392.4.1 Geophysical setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392.4.2 Freshwater inflow and outflow . . . . . . . . . . . . . . . . . . . . . . . . 412.4.3 Submarine melt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422.4.4 External forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432.4.5 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462.6 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483 Dynamic response of the last remaining Arctic epishelf lake to seasonal and long-term forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643.2.1 Study site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643.2.2 Area and volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653.2.3 Hydrography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653.2.4 Current velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673.2.5 Tidal height . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673.2.6 Meteorological time series . . . . . . . . . . . . . . . . . . . . . . . . . . 673.2.7 Mooring time series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693.3.1 Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693.3.2 Stratification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693.3.3 Current velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 703.3.4 Lake depth: interannual variation . . . . . . . . . . . . . . . . . . . . . . 713.3.5 Lake depth: seasonal variation . . . . . . . . . . . . . . . . . . . . . . . . 713.3.6 Tides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71v3.3.7 Lake depth: internal waves . . . . . . . . . . . . . . . . . . . . . . . . . . 723.3.8 Spatial extent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733.3.9 Lake volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733.3.10 Meteorological time series . . . . . . . . . . . . . . . . . . . . . . . . . . 743.3.11 Salinity time series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743.3.12 Temperature time series . . . . . . . . . . . . . . . . . . . . . . . . . . . 753.3.13 January 2012 event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753.3.14 Long-term lake depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773.4.1 Area expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773.4.2 Depth changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773.4.3 Observational error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783.4.4 January 2012 mixing event . . . . . . . . . . . . . . . . . . . . . . . . . . 793.4.5 Freshwater budget . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 813.4.6 Outflow hydraulics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 823.4.7 Implications for MIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 853.6 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 873.7 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 884 Depth-dependent submarine melt rates of a glacier tongue and ice shelf in a HighArctic fjord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 994.2.1 Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 994.2.2 Bulk mass balance of MGT . . . . . . . . . . . . . . . . . . . . . . . . . 1004.2.3 Ice thickness, motion and terminus position . . . . . . . . . . . . . . . . . 1004.2.4 Surface mass balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014.2.5 Divergence of ice flux along MGT . . . . . . . . . . . . . . . . . . . . . . 1024.2.6 Ocean thermodynamic melt model . . . . . . . . . . . . . . . . . . . . . . 1024.2.7 Ocean properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054.2.8 Ocean circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054.2.9 Temperature timeseries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1064.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1064.3.1 Milne Glacier grounding line . . . . . . . . . . . . . . . . . . . . . . . . . 1064.3.2 Glacier surface velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . 1074.3.3 Ocean properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107vi4.3.4 Ocean model sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . 1084.3.5 Submarine melt rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1084.3.6 Area-averaged melt rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 1094.3.7 Spatial distribution of melt rates . . . . . . . . . . . . . . . . . . . . . . . 1094.3.8 Thermal driving depth-dependence . . . . . . . . . . . . . . . . . . . . . 1114.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1114.4.1 Grounding line melt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1114.4.2 Near-surface melt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1124.4.3 Circulation in Milne Fiord . . . . . . . . . . . . . . . . . . . . . . . . . . 1144.4.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1164.5 Summary and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1184.6 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1204.7 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1225 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325.2 Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1355.3 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139viiList of TablesTable 3.1 Milne Fiord epishelf lake depth, area, volume and related observations. . . . . . 87Table 4.1 Annual horizontal displacement of ablation stakes on the Milne Glacier. . . . . 120Table 4.2 Values of physical constants and parameterizations used in ocean thermody-namic ice shelf ablation model. . . . . . . . . . . . . . . . . . . . . . . . . . . 121Table 4.3 Submarine melt rates (m a1) in Milne Fiord from 2011 to 2015. . . . . . . . . 121viiiList of FiguresFigure 1.1 Northern Ellesmere Island . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Figure 1.2 Milne Fiord schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21Figure 1.3 Ice-ocean interface schematic . . . . . . . . . . . . . . . . . . . . . . . . . . 21Figure 2.1 Overview map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48Figure 2.2 Milne Fiord bathymetry and ice thickness . . . . . . . . . . . . . . . . . . . . 49Figure 2.3 Hydrographic profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50Figure 2.4 Full depth temperature salinity plot . . . . . . . . . . . . . . . . . . . . . . . 51Figure 2.5 Deep water renewal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Figure 2.6 Temperature salinity plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Figure 2.7 2012 hydrographic sections . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Figure 2.8 2013 hydrographic sections . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Figure 2.9 Subglacial runoff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56Figure 2.10 Current velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57Figure 2.11 Mooring timeseries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58Figure 2.12 Mooring temperature line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Figure 2.13 Fjord heat content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60Figure 2.14 Offshore hydrographic section . . . . . . . . . . . . . . . . . . . . . . . . . . 61Figure 3.1 Map of Milne Fiord epishelf lake . . . . . . . . . . . . . . . . . . . . . . . . . 88Figure 3.2 Epishelf lake water column profiles . . . . . . . . . . . . . . . . . . . . . . . 89Figure 3.3 Seasonal deepening of the MEL halocline . . . . . . . . . . . . . . . . . . . . 90Figure 3.4 Cumulative PDDs versus depth . . . . . . . . . . . . . . . . . . . . . . . . . . 91Figure 3.5 Drainage pathway of MEL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Figure 3.6 Epishelf lake mooring timeseries . . . . . . . . . . . . . . . . . . . . . . . . . 92Figure 3.7 January 2012 mixing event . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Figure 3.8 Milne epishelf lake depth changes . . . . . . . . . . . . . . . . . . . . . . . . 94Figure 3.9 Epishelf lake outflow schematic . . . . . . . . . . . . . . . . . . . . . . . . . 95Figure 3.10 Epishelf lake outflow model . . . . . . . . . . . . . . . . . . . . . . . . . . . 96ixFigure 4.1 Milne Fiord map and ice draft . . . . . . . . . . . . . . . . . . . . . . . . . . 122Figure 4.2 Milne Glacier grounding line . . . . . . . . . . . . . . . . . . . . . . . . . . . 123Figure 4.3 Ice surface velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124Figure 4.4 Hydrographic profiles for melt model . . . . . . . . . . . . . . . . . . . . . . 125Figure 4.5 Current speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125Figure 4.6 Ocean model sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126Figure 4.7 MGT basal melt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Figure 4.8 Submarine melt rates in Milne Fiord from mooring profile . . . . . . . . . . . 128Figure 4.9 Submarine melt rates in Milne Fiord from offshore profile . . . . . . . . . . . 129Figure 4.10 Difference in submarine melt rates in Milne Fiord . . . . . . . . . . . . . . . . 130Figure 4.11 Thermal driving timeseries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131Figure 5.1 Milne Fiord present day schematic . . . . . . . . . . . . . . . . . . . . . . . . 137Figure 5.2 Milne Fiord future schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . 138xList of SymbolsAL Area of the epishelf lakeab Basal ablation rate at ice-ocean interfaceas f Surface ablation rate at ice-air interfaceb Effective width of the basal channelCd Drag coefficient for ice-ocean interfaceCe Discharge coefficient for weir equationch Constant for weir equationci Specific heat capacity of icecw Specific heat capacity of seawaterDEL Depth of epishelf lakeE Energyg Gravitational accelerationg0 Reduced gravitational accelerationH Ice thicknessh Water column heightL Distance dimensionLi Latent heat of fusion of iceN2 Brunt-Vaisala frequency squaredPb Pressure at ice-ocean interfacePE Potential energyQTb Latent heat flux at ice-ocean interfaceQTw Sensible heat flux across ice-ocean boundary layerQTi Sensible heat flux into iceQTw Sensible heat flux across ice-ocean boundary layerQSb Salt or ”freshwater” flux at ice-ocean interfaceQSi Salt flux into iceQSw Salt flux across ice-ocean boundary layerxiQc f Glacier calving volumetric flowQgl Glacier grounding line volumetric flowQfw Total freshwater fluxQsg Subglacial freshwater fluxQs f Surface freshwater fluxQb Submarine melt freshwater fluxU Velocity of ambient waterUi Velocity of iceu⇤ Friction velocityu Depth-averaged glacier velocityRi gradient Richardson numberSb Salinity at ice-ocean interfaceSw Salinity of ambient waterSi Salinity of iceS1 Salinity of outflowing layer (Knudsen relation)S2 Salinity of inflowing layer (Knudsen relation)Ti Internal ice shelf temperatureTb Ice-ocean boundary temperatureTw Ambient water temperatureT fb Freezing temperature at ice-ocean interfaceT fw Freezing temperature of ambient waterTwT fw Thermal drivingV Fjord volumeZint Interpolated depth of glacier bed below sea levelZc Depth of glacier centreline below sea levelZm Depth of glacier margin below sea levelz Depthz(N2max) Depth of maximum Brunt-Vaisala frequency squarede Ratio of iceberg width to heightGT Turbulent transfer coefficient for heatGS Turbulent transfer coefficient for saltG(TS) Turbulent transfer coefficient for heat and saltgT Turbulent exchange velocity for heatgS Turbulent exchange velocity for saltki Thermal diffusivity of icel1,2,3 Empirical constantsxiiY Composite water propertyrw Density of ambient waterri Density of icerw Density of seawaterri Density of icero Reference densityco Conservative water propertyxiiiAcknowledgmentsThe work presented in this thesis has been made possible through the support of a number of orga-nizations and numerous individuals. I would first like to thank the funding agencies that allowed meto pursue my studies, including the Natural Sciences and Engineering Research Council of Canada(NSERC), the Association of Canadian Universities for Northern Studies (ACUNS), and The Uni-versity of British Columbia (UBC).My Ph.D. co-supervisors, Dr. Bernard Laval and Dr. Derek Mueller, have my gratitude fortheir valuable guidance and encouragement over the course of this journey. Bernard was extremelygenerous in allowing me the freedom to forge my own research direction and having confidence inall my endeavours, from AUVs to icebergs. Derek’s intellectual curiosity and shared passion for theNorth inspired the origins of this thesis and for that I will always be grateful.This thesis has benefited from the knowledge of members of my supervisory committee, Dr.Greg Lawrence and Dr. Humfrey Melling, who both provided helpful insight into various aspectsof the research. Humfrey’s expertise in polar oceanography and his dedication to scientific rigourwere invaluable to the thesis and my growth as a researcher. I must also mention Dr. Eddy Carmackwho sent me off on my first journey to the icy seas many years ago, and Dr. Warwick Vincent whowas instrumental in inspiring an interest in Ellesmere Island.In my time at UBC I have enjoyed the intellectual company of many great people, too numerousto name them all. Current and past EFM group members, including Alex Forrest, Mona Rahmani,Anirban Guha, Kelly Graves, and Ted Tedford, to name a few, have made time inside and outside theRusty Hut enjoyable. Christian Schoof and the Glaciology group welcomed an outsider into theirranks and taught me a few things about ice. The Physical Oceanography group provided stimulatingseminars and allowed me to keep one foot in salty water at all times.The research in this thesis was only possible with the tireless efforts of all my field colleagues.First and foremost, I must thank Dr. Derek Mueller and Dr. Luke Copland for inviting me on thenorthern Ellesmere Ranger patrol in 2008, an incredible experience that first introduced me to thefascinating fjord that would later become the focus of my Ph.D. Two years later they again wel-comed a polar hitch-hiker into their camp at Purple Valley and have been generous with equipment,field experience, enthusiasm, and good nature since that time. Derek was a seemingly tireless andxivconstantly enjoyable field companion and it was a pleasure to trek across the ice with him. Otherswho I have had the great fortune to toil and laugh with in the field, and who have assisted in collect-ing much of the data presented in this thesis, include Adrienne White, Nat Wilson, Jill Rajewicz,Adam Garbo, and Kevin Xu. Field work on Ellesmere Island would not be possible without thelogistical support of all the individuals at PCSP in Resolute and the pilots who fly to the edge ofthe map. Luke Copland, Adrienne White, Colleen Mortimer, Wesley van Wychen, and WarwickVincent have also been extremely generous with the sharing of hard-won data.I must extend a special thank you to my family, who have encouraged and supported my aca-demic and life adventures from the beginning. To all the Hamiltons and Finlays, thank you. Ofspecial distinction are my parents Bill and Linda, who have always supported, encouraged, andinspired me, I truly thank you for that.Finally, to my wife Karran, and my sons Feynman and Sebastian. It has been quite a journey.Thank you for everything.xvChapter 1IntroductionThe widespread break up of ice shelves, glacier tongues, and the loss of ice-dammed fjord lakes(epishelf lakes) along the northern coast of Ellesmere Island, Canada, has motivated a need to un-derstand the dynamics of fjords here. However, an understanding of how these changes may beinfluenced by, or influence, ocean properties and circulation in fjords along this coast is limited by ascarcity of oceanographic observations. This thesis aims to: 1) describe the physical oceanographyand ice-ocean interactions of a northern Ellesmere Island fjord before the collapse of ice; 2) inves-tigate factors influencing the seasonal and interannual dynamics of an epishelf lake; and 3) estimatethe magnitude and spatial distribution of submarine melting of an ice shelf and glacier tongue. Toaddress these goals the thesis presents the results of an extensive oceanographic and glaciologicalfield study undertaken between 2011 and 2015 in Milne Fiord, the last ice shelf-epishelf lake-glaciertongue fjord system in the Arctic.1.1 BackgroundSince 2000, the northern coast of Ellesmere Island has undergone accelerated loss of ice shelves(Mueller et al., 2003; Mueller et al., 2008; White et al., 2015a) and the floating termini of glaciers(Pope et al., 2012; Copland et al., 2015). Ice shelves along this coast differ from those surroundingAntarctica, which are the seaward extensions of continental ice sheets. Here, ice shelves generallyform in situ at the mouths of fjords from accretion of sea ice below and snow accumulation above,with additional mass acquired from tributary glaciers and low-elevation ice caps (Jeffries, 2002).The largely marine origin of ice shelves here also distinguishes them from glacier tongues, thefloating extensions of outlet glaciers at the head of fjords (Jeffries, 2002). From 2000 to 2011 fiveof the six remaining major ice shelves on northern Ellesmere Island broke up or calved completelyfrom the coast, reducing ice shelf area from 1043 km2 to 546 km2 (Mueller et al., 2008; White et al.,2015a). Over the same period, there has been a widespread loss of multi-year land-fast sea ice (Popeet al., 2012) and the collapse of glacier tongues (Pope et al., 2012; Copland et al., 2015).1The breakup of ice shelves on northern Ellesmere Island has resulted in the loss of associatedice shelf-dammed lakes. Where an ice shelf spans the mouth of a fjord, it can form a floating damthat can trap surface meltwater runoff inside the fjord down to the draft of the ice shelf, resultingin a perennial layer of freshwater directly overlying seawater. The freshwater layer Surface ice-cover at this high-latitude (>80N) and low tidal action (www.tides.gc.ca) limits mixing of thefresh surface layer with seawater below and as a result the freshwater layer can last for decades orlonger (Veillette et al., 2008), creating what is known as an epishelf lake. The halocline betweenfreshwater and seawater can be incredibly sharp, with observed salinity gradients of >10 g kg1m1 (Veillette et al., 2008), and its depth is thought to be determined by the draft of the impoundingice shelf (Vincent et al., 2001). The existence of the epishelf lake is entirely dependent on theintegrity of the ice shelf dam, and the breakup of ice shelves along this coast has resulted in the lossof several epishelf lakes (Veillette et al., 2008), including the catastrophic drainage of the DisraeliFiord epishelf lake caused by the fracturing of the Ward Hunt Ice Shelf between 2000 and 2002(Mueller et al., 2003). The breakup of ice shelves can therefore cause dramatic changes to fjordstratification.The collapse of coastal ice along northern Ellesmere Island has been linked to recent atmo-spheric warming (Copland et al., 2007; Copland et al., 2015). Since 2000, annual mean air temper-atures in the Canadian Arctic Archipelago (CAA) rose by 1-2C (Sharp et al., 2011) resulting in asharp increase in mass loss from the ⇠146 x 103 km2 glaciated area in the region (Gardner et al.,2011; Sharp et al., 2011; Lenaerts et al., 2013). The increased meltwater runoff meant the CAA wasthe single largest contributor to sea level rise outside of Greenland and Antarctica from 2007-2008(Gardner et al., 2011). This meltwater runoff is funnelled through fjords where it is transformed byfjord processes before being exported to the open ocean.Growing evidence from other regions suggests that ocean forcing can also be an important factorin the loss of coastal ice. Submarine melting driven by the ocean accounts for three quarters of themass loss via ablation from Antarctic ice shelves (Rignot et al., 2013). In Greenland, up to 80% ofthe mass loss of glacier tongues has been attributed to ocean driven melting (Rignot, 1998; Rignotand Steffen, 2008), and submarine melting of outlet glaciers has been identified as a plausible mech-anism contributing to the both the retreat and increased flow rate of tidewater glaciers and collapseof glacier tongues (Holland et al., 2008b; Motyka et al., 2011; Joughin et al., 2012; Straneo et al.,2012). A regional rise in air temperatures over the Greenland Ice Sheet has increased surface runoff,much of which enters the ocean at the bed of tidewater glaciers, leading to enhanced convectivelydriven circulation and melting at the ice-ocean interface (Straneo and Cenedese, 2015). The rate ofsubmarine melt and the release of freshwater, from both surface runoff and submarine discharge,to the coastal ocean is dependent on the dynamics of fjords, which are generally poorly understoodeven in the few fairly well-studied Greenland fjords (Straneo and Cenedese, 2015). It is very likelythat many of the same ocean-driven processes that have influenced changes in Greenlandic fjords,2and below Antarctic ice shelves, could be driving the changes occurring along the northern coast ofEllesmere Island.Several characteristics set the fjords of northern Ellesmere Island apart from fjords in otherregions and justify focused study. First, very little oceanographic data has been collected in fjordsalong this coast, so further observations are a valuable contribution to the literature. Second, fjordwater properties here appear to be strongly linked to the presence of ice, yet the interactions betweenthe ocean and ice are not well understood. How the loss of ice affects fjord properties, or the roleof the ocean in driving these changes, is hampered by the lack of observations from systems priorto the loss of ice. Third, the ongoing dramatic loss of ice means the opportunity to study ice-ocean interactions is limited. Fourth, source waters for fjords here come directly from the ArcticOcean, which may alter fjord processes compared to other ocean basins with different water masscharacteristics and circulation. Finally, the geophysical features of the region, particularly the oncewidespread ice shelves and associated epishelf lakes, are, to the best of the author’s knowledge,unique in all the Arctic. This may be the last opportunity to study such a system before its collapse.This thesis investigates the ocean properties of Milne Fiord (8235’N; 81W), the site of the lastremaining ice shelf-epishelf lake system in the Arctic (Veillette et al., 2011a). The Milne Ice Shelf(205 km2) spans the mouth of the fjord and is the last intact ice shelf along the northern coast ofEllesmere Island. The ice shelf dams an epishelf lake that occupied an area of approximately 52.5km2 as of 2009 (Mortimer, 2011). At the head of Milne Fiord the Milne Glacier terminates in a16 km long glacier tongue. Previous work in the fjord has included glaciological studies of the iceshelf and glacier tongue (Jeffries, 1985; Jeffries, 1986b; Narod et al., 1988; Mortimer et al., 2012),remote sensing analysis of the extent of the fresh epishelf lake ice (Mortimer et al., 2012; Veilletteet al., 2008), sampling of the freshwater ecosystem of the epishelf lake (Veillette et al., 2011a), andshallow hydrographic profiles focusing on the depth of the epishelf lake (Jeffries, 1985; Veilletteet al., 2008; Veillette et al., 2011a). The oceanography of the fjord has not previously been studied.In the following section relevant literature is reviewed to provide background on the topicsaddressed in the thesis, and to place the work in the context of the current state of knowledge.1.2 Review of relevant literature1.2.1 Northern Ellesmere IslandThe coast of northern Ellesmere Island is characterized by numerous fjords (narrow, deep inlets,carved by glaciers, often with a sill at the mouth) that open to the Arctic Ocean (Fig. 1.1). Althoughthe fjords of northern Ellesmere Island vary substantially in geometry (the largest being the NansenSound fjord system extending 340 km inland), a typical fjord along the northern coast is 30-50 kmlong and 4-8 km wide. The bathymetry in this region is in general very poorly constrained, butexisting observations suggest fjords here are typically a few hundred meters deep with sill depths3between 50-350 m (Ford and Hattersley-Smith, 1965). Bathymetric soundings collected along thecoast show the continental shelf is relatively shallow (⇠250 m) and the shelf break lies ⇠100 kmoffshore where it slopes steeply to depths over 2000 m (Jakobsson et al., 2012).Due to its high-latitude, the northern coast of Ellesmere Island receives no solar radiation dur-ing the polar night from mid-October through February, and continual solar radiation from April toSeptember. This results in low annual average air temperatures (⇠ -18C at Alert between 1971-2000; www.ec.gc.ca) but with large amplitude seasonal temperature cycle (varying up to 50C be-tween winter and summer). Surface temperatures over ice masses exceed the freezing point duringonly two to three months of the year. The Canadian High Arctic receives low amounts of precipita-tion (<0.5 m w.e. a1; Braithwaite, 2005), so regional surface mass balance of ice caps and glaciersis largely governed by the high variability in melt production which is strongly correlated withsummer surface air temperatures, which in turn, are highly dependent on local synoptic conditions(Gardner et al., 2011 and references therein).Long-term glacier mass balance records from the Queen Elizabeth Islands (QEI: Ellesmere,Devon, and Axel Hieberg Islands) indicate that prior to the 1980s the ice masses in this regionwere largely in balance (Sharp et al., 2011; Van Wychen et al., 2014), however since 2005 changingsummer atmospheric circulation patterns have led to increased surface melt and longer melt seasons(Gardner et al., 2011; Sharp et al., 2011; Lenaerts et al., 2013). Consequently, surface mass loss hasincreased sharply in recent years (averaging -7± 18 Gt a1 between 2004 and 2006, and -61± 18 Gta1 between 2007 and 2009; Gardner et al., 2011). Flow speeds of glaciers in the QEI are relativelyslow, with maximum velocities of tidewater glaciers of 30-90 m a1 along their main trunks, risingto <300 m a1 at their termini in 2012 (Van Wychen et al., 2014). Dynamic discharge of glaciersinto the ocean was estimated as 2.6 ± 0.8 Gt a1, indicating that iceberg calving accounts for asmall proportion of total mass loss, with surface melt and runoff responsible for the majority of massloss of glaciers of the QEI. However, only two marine-terminating glaciers (Trinity and Wykehamglaciers on the east coast of Ellesmere Island) accounted for 62% of the total dynamic dischargefor all glaciers of the QEI in winter 2015 (Van Wychen et al., 2016), suggesting that variations inflow of just a few tidewater glaciers could dramatically alter the total dynamic discharge from theQEI. Changes in glacial hydrological forcing, the buttressing effects of sea ice/melange at the glacierterminus, and ocean-driven melt rates, could lead to increased mass loss at tidewater glacier termini,leading to glacier acceleration and increased contribution to sea level rise (Straneo and Cenedese,2015). The nature and causes of tidewater glacier variability have not been well investigated in theCanadian Arctic.Where oceanographic surveys have been conducted on northern Ellesmere Island, including ofthe Nansen Sound fjord system (Ford and Hattersley-Smith, 1965; Hattersley-Smith and Serson,1966), Tanquary Fjord (Keys and Seibert, 1969), Disraeli Fjord (Keys et al., 1968; Keys, 1977),d’Iberville Fjord (Lake and Walker, 1973) and Taconite Inlet (Ludlam, 1996), they show stratifica-4tion of water masses that reflect those of the adjacent Arctic Ocean, subject to modification fromsills, fresh water runoff during the melt period, and interaction with ice shelves and glaciers. Typ-ical stratification consists of three layers distinguished by temperature and salinity characteristics.At the surface is a fresh, well mixed layer, modified by glacial meltwater runoff and local surfaceprocesses. Below is relatively fresh (<34.4 PSU), cold (<0C) Polar Water (PW), extending downto about 200 m depth, derived from river runoff around the margins of the Arctic Ocean and Pa-cific Ocean inflow through Bering Strait (Aagaard et al., 1981; Steele and Boyd, 1998). PW liesabove a layer of warm (0-3C), saline (>34.7 PSU) modified Atlantic Water (AW). AW originatesas an initially warm (>3C) water mass at the surface of the North Atlantic that is cooled by theatmosphere as it is transported poleward through Fram Strait, and submerges to depth, slowly los-ing heat as it is transported around the Arctic, to eventually reach the northern coast of EllesmereIsland. Separating PW and AW is the main Arctic halocline, a layer of low temperature (<-1C)and salinity between ⇠ 30.4 and 34.4 PSU, and arguably the most important feature of the ArcticOcean (Carmack, 1990), due to its effect of limiting convective cooling to the surface ⇠50 m of thewater column and preventing upward flux of heat from the underlying warm AW, processes whichallow sea ice to form.The actual characteristics of ocean waters reaching the fjords of northern Ellesmere Island aredependent on pathways and variability of regional ocean circulation, which are not well constrainedby observations. High-resolution model simulations (Aksenov et al., 2010; Jackson et al., 2014a),geostropic calculations (Steele et al., 2004), and observations of ice island and sea ice drift (Coplandet al., 2007; Pope et al., 2012), indicate surface waters in the Arctic Ocean typically flow westwardalong the Canadian continental shelf. Hydrographic analyses of water masses in the western LincolnSea, to the northeast of Ellesmere Island, indicate that surface waters found there originate from theCanada Basin (Jackson et al., 2014a), are circulated around the Beaufort Gyre and advected by theTranspolar Drift from the Chukchi Sea (Steele et al., 2004). An eastward flowing boundary under-current appears to exist along the continental slope from the base of the mixed layer (30 – 75 m)to the bottom (Newton and Sotirin, 1997) likely an extension of the Arctic Circumpolar Current(Rudels et al., 1999; Aksenov et al., 2010). AW in the Lincoln Sea generally shares characteris-tics with that of the southeastern Canada Basin (i.e. a cold (<0.5C Atlantic layer below 350 mdepth; Steur et al., 2013) further supporting the likelihood of an eastward flowing shelf break cur-rent. Model results indicate North Ellesmere shelf currents are weak (<10 cm s1) and baroclinic(Aksenov et al., 2010).1.2.2 Ice shelvesOne of the most notable features of the northern coast of Ellesmere Island is the presence of iceshelves at the mouths of several fjords. The ice shelves were first observed by a team led by Lieu-tenant Pelham Aldrich during the 1875-1876 British Arctic Expedition. Aldrich noted a unique5undulating surface of snow and ice extending along much of the coast (Nares, 1878). Three decadeslater, polar explorer Robert E. Peary described a broad glacial fringe that extended from Cape Heclawestward some 500 km to at least Nansen Sound (Peary, 1907). This floating glacial fringe, un-officially known as the Ellesmere Ice Shelf, was estimated at 8900 km2 in area (Vincent et al.,2001). The ice shelves are thought to have formed approximately 4000 years ago during a period ofclimatic cooling (Evans and England, 1992; Antoniades et al., 2011) and remained stable for almostthree millenia before undergoing a major fracturing event 1,400 years ago, then reforming 800 yearsago (Antoniades et al., 2011). After the early explorations at the turn of the 20th century the areabecame the focus scientific research in the 1950s when it was discovered that massive ice islandsadrift in the Arctic Ocean originated from the northern coast of Ellesmere (Hattersley-Smith, 1957).The ice islands and subsequent surveys along the coast confirmed that much of the Ellesmere IceShelf had disintegrated since Pearys expedition.In the past century more than 90% of the area of the Ellesmere Ice Shelf has been lost. Atthe turn of the 21st century six major ice shelves remained: the Ayles, Markham, Milne, Petersen,Serson, and Ward Hunt ice shelves, with a total area of⇠1043 km2 (Mueller et al., 2006). From thattime several calving events have resulted in significant loss of the remaining ice shelf area (Fig. 1.1),including a 50% reduction in total ice shelf area, to 563 km2 by 2011 (D. Mueller, pers. comm.).Reductions included the complete loss of the Ayles Ice Shelf in 2005 (Copland et al., 2007), theMarkham Ice Shelf in 2008 (Mueller et al., 2008) and the majority of the Serson Ice Shelf in 2011,combined with further breakup and loss from the Ward Hunt and Petersen ice shelves (White et al.,2015a). In examining the calving of the Ayles Ice Shelf and the breakup of the Petersen Ice Shelf,Copland et al. (2007) and White et al. (2015a) suggested that the ice shelves were weakened bythinning due to long-term negative surface mass balance related to an increase in mean annual airtemperatures over the past 50+ years (+0.5C per decade along the northern coast of EllesmereIsland between 1948 and 2012), the development of fractures, and mechanical erosion due to thepresence of open water along their fronts. The fracturing and calving of ice shelves along this coasthas led to the loss of associated epishelf lakes.1.2.3 Epishelf lakesEpishelf lakes form where ice shelves impose a physical barrier preventing surface meltwater runofffrom flowing freely into the ocean. The ice shelf dam is floating on the ocean, so epishelf lakesare connected to the sea, and thus are tidally influenced. A schematic of Milne Fiord, including arepresentation of an epishelf lake and many other key features and processes discussed in this thesis,is presented in Figure 1.2.Epishelf lakes were first described in Antarctica (Heywood, 1977) and have been divided intotwo types depending on their connection to the ocean: those that form in depressions on land wherethe hydraulic connection to the ocean occurs at a land-ice contact under the ice shelf, or through6cracks in the ice sheet; and those where the freshwater layer floats directly on seawater and thethickness of the freshwater layer is thought to be controlled by the thickness of the ice shelf (Gibsonand Andersen, 2002). The former are numerous in Antarctica and are distributed around the marginsof the continental ice sheet (Heywood, 1977; Gibson and Andersen, 2002; Laybourn-Parry et al.,2006; Smith et al., 2006). The latter, those that float directly on seawater, are not as commonin Antarctica (Wand et al., 2011), but were once relatively numerous along the northern coast ofEllesmere Island (Vincent et al., 2001; Veillette et al., 2008), dammed by ice shelves at the mouthof fjords. Perennial ice cover in this region inhibits wind-mixing, and permits the maintenance of asharp salinity gradient between the freshwater layer and seawater. Epishelf lakes are known to existfor decades, allowing the evolution of unique ecosystem types with freshwater and marine biotacontained within the same water column, vertically separated by a sharp salinity and temperaturegradient (Van-Hove et al., 2001; Vincent and Laybourn-Parry, 2008; Veillette et al., 2011a).The best studied of the northern Ellesmere Island epishelf lakes was the Disraeli Fiord epishelflake, dammed by the Ward Hunt Ice Shelf prior to its complete drainage between 2000 to 2002.When first surveyed in 1954, the water column of Disraeli Fiord consisted of a 63 m deep freshwaterlayer overlying 300 m of seawater, the freshwater-seawater interface only a few meters thick (Crary,1956). A number of authors interpreted the depth of the freshwater layer in Disraeli Fiord as beingequivalent to the draft of the ice shelf (Hattersley-Smith, 1973; Jeffries and Krouse, 1984; Keyset al., 1968; Vincent et al., 2001), as excess freshwater below the minimum draft of the ice wouldflow out of the fjord under the base of the ice shelf. Crary (1956) suggested that outflowing brackishwater could freeze to the ice shelf base, thereby influencing ice shelf mass balance. Keys (1978))observed an outflowing current (9 cm s1) localized at the base of the freshwater-seawater interfaceduring the surface melt season and ice cores from the north-east region of the Ward Hunt Ice Shelfrevealed basal accretion of freshwater ice (Jeffries and Krouse, 1984).Subsequent profiles taken in Disraeli Fiord between 1960 and 1999 revealed the freshwater layerthinned by 33 m during this interval (Mueller et al., 2003) and the steady decrease in thickness ofthe freshwater layer suggested a general thinning of the Ward Hunt Ice Shelf (WHIS). The surfacemass balance of the WHIS was only slightly negative (-1.1 m w.e. from 1967 to 1999; Braun et al.,2004) suggesting most of the thinning was due to basal mass loss. However, Vincent et al. (2001))also suggested that the freshwater could have been preferentially draining via a localized conduitat the base of the ice shelf, and questioned to what extent changes in the thickness of the dammedfreshwater layer are representative of changes in the mean thickness of the entire Ward Hunt IceShelf. The basal mass balance of ice shelves in this region have not been quantified, so the role ofbasal freezing or melting remains unknown.A sinuous fracture running the full north-south extent of the WHIS formed between 2000 and2002, resulting in the complete drainage of the freshwater layer (Mueller et al., 2003). Subse-quent profiles in Disraeli Fiord have shown a complete absence of the freshwater layer as continued7breakup of the ice shelf allowed freshwater to flow unobstructed to the open ocean (Veillette et al.,2008). Similarly, calving and mass loss from the Petersen Ice Shelf resulted in the drainage of itsassociated epishelf lake in August 2005 (White et al., 2015b). From remote sensing data, Veilletteet al. (2008) identified nine ice-dammed or epishelf lakes along the northern coast of Ellesmere Is-land in 2008. Field measurements at the time confirmed the presence of five of those lakes, howeverwith the exception of Milne Fiord, all consisted of a freshwater layer <5 m deep.Apart from the decrease in thickness of the Disraeli Fiord epishelf lake prior to its loss in 2002,observations from epishelf lakes have revealed short-term and interannual variation in depth of thefreshwater-seawater interface (Keys, 1977; Veillette et al., 2008). Keys (1977) suggested a tidally-driven internal wave influenced the depth of the halocline in Disraeli Fiord in 1967, and Veilletteet al. (2008) suggest this may have been the cause of some of the interannual variation observed inother epishelf lakes, although tidal amplitude in this region is generally small (<0.2 m in DisraeliFiord; Fisheries and Oceans Canada; www.tides.gc.ca). Veillette et al. (2008) also suggest that fjordcirculation and shear-induced mixing could further alter the depth of the halocline, although theseprocesses have not been studied.Antarctica retains numerous, and much deeper, epishelf lakes than those in the Arctic. Examplesinclude Beaver Lake, with a freshwater layer estimated to be 170 to 260 m deep overlying seawaterto 435 m (Laybourn-Parry et al., 2001), and the epishelf lakes of the southern Bunger Hills, Tran-skriptsii Gulf, with the deepest having an 85 m deep freshwater layer in 1992 and 2000 (Gibsonand Andersen, 2002). Smith et al. (2006) observed interannual changes in the depth and gradient ofthe freshwater-seawater interface in Moutonne´e Lake, which varied from 66.5 m in 1973, to 64 min 2000, to 68 m in 2001, and suggested this could reflect changes in thickness of the impoundingGeorge VI Ice Shelf, but more likely were the result of either changes in the seasonal supply offreshwater, tidal fluctuation, or changes in the hydraulic connection to the open ocean. The authorsconclude that further long-term monitoring is required to determine the cause and significance ofthese changes, and how epishelf lake depth is related to ice shelf thickness. However, apart from a90-day tidal record from Beaver Lake, Antarctica (Galton-Fenzi et al., 2012), there are no observa-tions of the seasonal variation of an epishelf lake.1.2.4 Fjord oceanographyThe role of the ocean in influencing the cryosphere, including the submarine mass balance of iceshelves and tidewater glaciers, variation in the depth of epishelf lakes, and the export of freshwater tothe open ocean, are determined by the dynamics of fjords. An understanding of factors that influencethe spatial and temporal variation fjord water properties and circulation is therefore essential tounderstanding ice-ocean interactions. The general physics of fjords have been summarized in anumber of reviews (Farmer and Freeland, 1983, Inall and Gillibrand, 2010, Stigebrandt, 2012), withother reviews focusing on the physical oceanography of Arctic fjords (Cottier et al., 2010), and8the dynamics of Greenland fjords with marine terminating outlet glaciers (Straneo and Cenedese,2015).Fjords are typically strongly stratified with distinguishable water masses including a fresh sur-face layer, intermediary water that has a stratification that mirrors offshore waters, and deep basinwater below sill level. Stratification within a fjord is influenced by freshwater input, surface heatfluxes, and density-driven exchange with the open ocean, and tidal action (Inall and Gillibrand,2010). Turbulent mixing and vertical exchange in a stratified fjord is driven by mixing along bound-aries, convective acceleration and hydraulic jumps at constrictions, shear instabilities, internal wavebreaking, and convective overturning (Farmer and Freeland, 1983). In the case of high-latitudefjords with perennial ice-cover wind forcing is negligible within the fjord, but can still be importantfor coastal processes where open water may be present (e.g. coastal upwelling).Variations in density and sea level along the continental shelf will induce horizontal pressuregradients that can drive exchange of waters above sill depth between the fjord and coastal wateroutside the mouth, known as intermediary circulation (Straneo and Cenedese, 2015). Intermediarycirculation can originate from along-shore winds that drive upwelling or downwelling, or densityanomalies advected past the mouth of the fjord (Straneo and Cenedese, 2015). The density contrastbetween fjord waters and coastal waters can result in horizontal pressure gradients that drive baro-clinic flows which can be an order of magnitude larger than estuarine flows (Stigebrandt, 2012).Intermediary circulation has been shown to be important in Greenland’s fjords, explaining the rapidrenewal of the upper water column (Straneo et al., 2010), and the reversing of strongly shearedvelocities every few days (Sutherland and Straneo, 2012; Sutherland et al., 2014). Intermediaryflows have also been shown to dominate fjord circulation and variability in non-summer months,including the advection of shelf anomalies into fjords on submonthly timescales (Jackson et al.,2014b). Variations in intermediary circulation may also explain interannual changes observed insome Greenland fjords (Mortensen et al., 2013; Christoffersen et al., 2011).Replacement of deep water, that is water below sill depth, will occur when outside water at orabove sill depth exceeds the density of water within the basin (Farmer and Freeland 1983). Theresidence time of the deep water depends on the turbulent mixing rate (the density of deep waterwill steadily decrease as near surface water is mixed downward), intermediary circulation, and thevolume of water below sill level (Stigebrandt, 2012). The timescale of renewal can range from tidalperiods to many years and studies from Greenland have shown a wide range of deep water renewalperiods in fjords, ranging from months (Jackson et al., 2014b) to tens of years (Johnson et al., 2011).Explanations for the mechanisms for deep-water renewal vary from tidal forcing, to intermediaryflow, to buoyancy-driven circulation, and may be fjord specific, and are strongly dependent on therate of mixing in the fjord. Once over the sill, replacement waters enter the deep basin as a turbulentgravity current, descending to the bottom of the fjord or spreading out as an interleaving layer onceit reaches a depth at which it is neutrally buoyant (Farmer and Freeland, 1983).91.2.5 Glacial fjordsInterest in the physics of glacial fjords, those with marine terminating glaciers at the head, has grownin recent years due to the widespread retreat and speedup of Greenland’s outlet glaciers, increasingcontributions to sea level rise (Straneo and Cenedese, 2015). As well, most of the freshwater runofffrom surface melting of ice sheets is discharged at the head of glacial fjords, where it is transformedby fjord processes before being exported to the open ocean. There are several unique features spe-cific to glacial fjords that distinguish processes here from non-glacial fjords. Straneo and Cenedese(2015) reviewed the state of knowledge of the dynamics of Greenland’s glacial fjords. This sec-tion provides a brief summary of key processes that are applicable to the glacial fjords of northernEllesmere Island, including Milne Fiord.In classic estuarine circulation, freshwater released at the surface at the head of a fjord mixeswith ambient water, generating an outflowing brackish surface layer and a compensating inflow ofdeep, denser water (Farmer and Freeland, 1983). However, in glacial fjords the freshwater releasecan also occur at depth, via subsurface discharge of runoff at the bed of the glacier, and by sub-marine melting of the submerged portions of the glacier face. The release of freshwater at depthby either mechanism results in an upwelling turbulent plume that entrains ambient water as it rises.Plume theory (Morton et al., 1956; Turner, 1973), dictates that the evolution of the plume is con-trolled by the plume’s buoyancy forcing, which is dependent on the density difference between theplume and ambient water. The buoyancy forcing determines the rate of ascent which regulates theamount of entrainment and mixing of the plume with ambient fjord water. In glacial fjords thefreshwater release that contributes the buoyancy forcing can occur at a fixed depth, as in the caseof subglacial discharge at the glacier bed, or over some depth range, as in the case of submarinemelting along the glacier front. The sediment concentration and particle size distribution within aplume originating from subglacial discharge can alter the density of the plume and fjord stratifica-tion and circulation (Salcedo-Castro et al., 2013). In glacial fjords, the former is likely dominantnear subglacial discharge channels in summer, while the later likely dominates in winter or distantfrom subglacial discharge in summer. The depth at which the plume will detach from the glacier andspread horizontally down the fjord is dependent on the buoyancy flux of the plume and the ambientstratification. If the plume is less dense than ambient water at the same depth, it will continue torise and may reach the free surface. However, where ambient stratification is stronger, or buoyancyflux is low, the plume will not reach the free surface and will form an interflow at some intermediatedepth.The input of freshwater from subglacial discharge at the grounding line provides a source ofbuoyancy and drives convective motion along the ice-ocean interface. The heat brought into theplume through entrainment of ambient water is an additional driver of melting at the base of tide-water glaciers and ice shelves (Jenkins, 2011). The application of the theory of buoyant plumeshas shown that warming ocean waters and increased subglacial discharge both generate increased10melting near the grounding line, with melt rates dependent on the temperature of ambient water, theflux of subglacial discharge, and the slope of the ice-ocean interface (i.e. near-vertical for a calvingglacier terminus, and near-horizontal for an ice shelf or glacier tongue; Jenkins, 2011).Observations in Greenland show glacial fjords in this region are characterized by a stronglystratified water column with intrusions of glacially modified water at mid-depths, often at the inter-face between PW and AW (Mortensen et al., 2011; Straneo et al., 2011; Sutherland et al., 2014),likely giving rise to a multi-cell circulation pattern. In addition, numerical models have shownthe magnitude of subglacial discharge, and geometry of subglacial channels (point source, multiplechannels, or line source) are important in determining the nature of buoyancy-driven circulation inglacial fjords (Sciascia et al., 2013; Xu et al., 2012; Xu et al., 2013; Cenedese and Linden, 2014) andconsequently the submarine melt rate (Xu et al., 2013; Kimura et al., 2013). An improved under-standing the dynamics of subglacial plumes is hampered by the challenge of acquiring observationsnear the face of calving glaciers or at the base glacier tongues.Circulation below ice shelvesKnowledge of circulation in the cavities below ice shelves and glacier tongues, come primarily fromstudies in Antarctica. Thermohaline circulation in the cavities below Antarctic ice shelves can bedriven by the pressure dependency of the freezing point of seawater (Lewis and Perkin, 1986). Asseawater in contact with the ice shelf at depth melts ice, the ascending buoyant plume becomessupercooled relative to the in situ freezing point, and leads to ice formation in the water column,and accretion at the base of the ice shelf at shallower depth. Conversely, descent of waters, afterhaving been in contact with the ice, will have sensible heat available to melt ice at depth. The overalleffect is to remove glacial ice from greater depths and deposit sea ice at a shallower location, withthe effect of evolving a floating ice shelf toward uniform thickness, a mechanisms known as an ’icepump’ (Lewis and Perkin, 1986). This process is driven solely by the pressure dependency of thefreezing temperature, and is independent of melt driven by the transport and entrainment of warmwater to the ice base.To date, there are only a few studies that have addressed the water properties and circulationin glacier tongue cavities in the Arctic, including at Petermann Gletscher (Rignot and Steffen,2008; Johnson et al., 2011) and Nioghalvfjerdsbræ (79NG; Wilson and Straneo, 2015). With theexception of a single water column profile through a borehole in the Petermann Gletscher tongue(Rignot and Steffen, 2008), and through a rift in 79NG glacier tongue, work has primarily beenlimited to observations at the ice front, tens of kilometers from the grounding line.1.2.6 Submarine meltingSubmarine melting is a critical factor in determining the mass balance of ice shelves and glaciertongues and the position of the boundary between grounded and floating ice, known as the grounding11line. The grounding line marks the boundary between ice that has has yet to displace its full weightof water, and thus the potential for glacier contribution to sea level rise. Submarine melting by warmocean waters was demonstrated to be a trigger of the thinning and breakup of the floating tongue ofJakobshavn Isbræin Greenland, which was followed by rapid acceleration of the grounded glacier(Holland et al., 2008a; Motyka et al., 2011).The mass balance of an ice shelf or glacier tongue is determined by the balance of ablation andaccumulation at the surface and base, input from glacier flow, and losses from calving at the icefront. Under steady state conditions, and assuming constant ice density, the area-averaged conser-vation of mass for a glacier tongue reduces to an expression of volume continuity asQgf Qc f as f Agt abAgt = 0 (1.1)where Qgf is the ice volume flow across the grounding line, Qc f is the change in ice volume overtime at the terminus due to calving or terminus advance, as f and ab are the net area-averaged surfaceand basal ablation rates, respectively (expressed as meters of solid ice per unit time and positivefor melting), and Agt is the surface area of the glacier tongue. With estimates of ice velocity, icethickness, changes in terminus position, and surface ablation, an estimate of basal ablation can beobtained.Submarine melt rates in glacial fjords can also, in principle, be estimated from the net heat fluxfrom the ambient water to the ice (Straneo and Cenedese, 2015). However, there are several chal-lenges to getting accurate estimates of heat flux in a glacial fjord, and studies that have tried to do soare based on synoptic surveys that rely on several broad, and perhaps unjustified, assumptions (Rig-not et al., 2010; Sutherland and Straneo, 2012; Xu et al., 2013; Inall et al., 2014). The few existingcontinuous (moored) observations in glacial fjords do not support several of the usual assumptionsmade (Jackson et al., 2014b), and other fluxes not often accounted for, including sea ice formationand melting as well as iceberg melting could strongly influence the overall heat budget (Straneoand Cenedese, 2015). At present, it remains difficult to quantify submarine melt using the heat fluxmethod.Ocean thermodynamic modelsAnother approach to estimate melt rates is to consider the thermodynamics of heat transport at theice-ocean interface. The basic assumption made in all ocean models of submarine ice melt is thatthe temperature and salinity at the ice-ocean interface are related by the equation for the freezingpoint at a given pressure (Holland and Jenkins, 1999). Estimation of the ablation rate then requirescalculating the heat and freshwater fluxes at the ice-ocean interface based on changes in the far-fieldocean properties from freezing point conditions.A conceptual schematic of the relevant temperatures, salinities, and heat and salt fluxes between12layers at the ice-ocean interface are shown in Figure 1.3. The water column is divided into tworegions, the ice-ocean boundary layer, typically a few meters thick, where turbulent mixing is influ-enced by the proximity of the ice boundary, and the ambient ocean mixed layer, a few tens of meterthick, where turbulence is unaffected by the boundary and stratification and rotation control mixing(McPhee et al., 2008). The boundary layer itself consists of an outer turbulent layer (a few metersthick), where rates of exchange of heat and salt are influenced by the velocity shear induced by thestationary ice shelf, and a viscous sublayer (a few millimetres thick) at the ice-ocean interface wheremolecular diffusion regulates exchange. The gradients in temperature and salinity across the bound-ary layer determine the fluxes of heat and salt between the ocean mixed layer and the ice (Hollandand Jenkins, 1999). In addition, the internal temperature gradient within the ice shelf near its basedrives the heat flux between the ice-ocean interface and the ice interior. Assuming the ice-oceaninterface is in thermodynamic equilibrium, a melt (or freezing) rate can be estimated from the heatbalance.The ocean models are based on three physical constraints: that the ice-ocean interface is at thefreezing point, and that heat and salt are conserved at the interface during phase changes (Hollandand Jenkins, 1999). Prescribed quantities are the far-field properties of the ocean and the interiorproperties of the ice shelf, which are then used to determine the characteristics at the interface basedon the above constraints. In the next section we present the fundamental equations governing thefreezing point of seawater and the conservation of heat and salt at the ice-ocean interface, basedlargely on the work of Holland and Jenkins (1999).Fundamental equationsFreezing pointThe freezing point of seawater is a linear function of pressure and a non-linear function of salinity(Millero, 1978). To simplify solving equations for conservation of heat and salt at the ice-oceaninterface, coupled ice-ocean models often use a linearized version of the freezing point dependencyat the ice-ocean interface (Holland and Jenkins, 1999), of the formT fb = l1Sb+l2+l3Pb, (1.2)where Sb and Pb are the salinity and pressure, respectively, at the ice-ocean interface, l1,2,3 areempirical constants. Coefficients l1 and l2 are usually chosen to optimize the the fit at typicalseawater salinities, however the formula is then only valid over a limited salinity range (⇠4-40 psu)and does not apply to freshwater (Holland and Jenkins, 1999).13Heat conservationAt the ice-ocean interface the change in latent heat, QTb , due to melting or freezing must be balancedby the heat flux into the ice, QTi , and the heat flux supplied by the ambient water to the ice QTw:QTb = QTi QTw (1.3)where the latent heat term is given byQTb =rwabLi (1.4)where r is density, L is the latent heat of fusion, and superscript T refers to heat (to distinguishfrom the salt flux below), and subscripts b, i, w refer to the ice-ocean interface, ice, and water,respectively. The ablation rate at the ice-ocean boundary, ab, is a change in thickness of solid iceper unit time (positive for melting).Salt conservationDuring melting at the base of the ice shelf, the ’freshwater’ flux is balanced by the salt flux diver-gence at the ice-ocean interfaceQSb = QSi QSw (1.5)where superscript S is salt. The ’freshwater’ flux isQSb = rwab(SiSb) (1.6)where Si and Sb are the salinities of the ice and the ice-ocean interface, respectively. As salt cannotdiffuse through the solid ice the diffusive flux into the ice shelf, QSi is zero, and the salt balance isdetermined by the flux through the oceanic boundary layer. Si is assumed to be zero.The various models of ablation at the base of an ice shelf differ in the treatment of the turbulenttransfer of heat and salt through the ice-ocean boundary layer, and the flux of heat through the iceshelf (salt cannot diffuse through solid ice so it need not be considered with the ice). Similar modelshave been developed both for sea ice (McPhee et al., 1987; Holland, 1998) and ice shelves (Hollandand Jenkins, 1999; Jenkins et al., 2010) these differ primarily in the estimation of heat flux throughthe ice due to the vast difference in ice thickness between the two, and the resulting importance ofsurface conditions.14Heat flux into iceThe general estimate of the heat flux into the ice shelf can be made using the formQTi = ricik∂Ti∂ zb(1.7)where ri is the density of ice, ci is the specific heat capacity of ice, ki is the thermal diffusivity ofice, and Ti is the internal ice temperature. The density, heat capacity, and thermal diffusivity areassumed constant, and the focus is on estimating the temperature gradient at the base of the iceshelf. The heat transport equation in the ice shelf is∂Ti∂ t+Ui ·—Ti = kTi —2Ti, (1.8)where Ui is the flow field within the ice shelf (Holland and Jenkins, 1999). Without knowledgeof the ice flow field, reduced forms of the equation are considered. The simplest approximation isthat there is no diffusion of heat into the ice shelf. The next level is to assume that there is verticaldiffusion of heat, but no advection. The steady state solution to the heat transport equation thenis a linear temperature profile throughout the thickness of the ice shelf, in which case the gradientcan be defined by the temperature at the surface of the ice shelf, the temperature at the base, andthe thickness of the ice. Although common in sea ice models, this approximation is not entirelyappropriate for ice shelves given the greater thickness and observed non-linear temperature profilesthrough the ice. The heat flux into the ice can be parameterized following Nøst and Foldvik 1994and Holland and Jenkins (1999) using a steady-state, one-dimensional advection-diffusion equation.Heat conduction into the ice is therefore approximated byQTi = riciab(TiT fb ), (1.9)Oceanic heat fluxesThe most sophisticated models make no prior assumptions about conditions at the ice-ocean inter-face and solve the freezing point, heat flux, and salt flux equations (Holland and Jenkins, 1999). Inthis formulation, the fluxes of heat and salt across the boundary layer are expressed asQTw = rwcwu⇤GT (Tfb Tw), (1.10)andQSw =rwu⇤GS(SbSw), (1.11)15where GT and GS are the turbulent transfer coefficients for heat and salt, respectively, that takeinto account the nonlinear profiles of temperature and salinity in the boundary layer resulting fromturbulence, cw is the specific heat capacity of water, and Tw and Sw are the temperature and salinityin the ambient mixed layer, respectively. Note in this formulation the freezing point is determinedusing the salinity at the ice-ocean interface (Sb) as in Eq. 1.19.The friction velocity u⇤ is defined in terms of the shear stress at the ice-ocean interface. It isassumed to be related to the free-stream current beyond the boundary layer U through a quadraticdrag law,u2⇤ =CdU2, (1.12)where Cd is a dimensionless drag coefficient. This formulation assumes a constant shear stress inthe boundary layer and is derived from measurement of flow across a hydraulically smooth surface.The applicability of these assumptions for the base of an ice shelf is largely unknown due to a lack ofobservations, however this parameterization is widely used in numerical models where results havecompared favourably with observations beneath sea ice (McPhee et al., 1999). Determination of theproper drag coefficient and the turbulent transfer coefficients are the main challenges for accuratemodelling of ice shelf ablation rates, and these are discussed further below.Two-equation modelAn alternative approach is to simplify the problem by assuming that the mixed layer salinity and theinterfacial salinity are equal, implying infinite salt diffusion (Holland and Jenkins, 1999). The meltrate is therefore determined by the transfer of heat through the boundary layer which is parametrizedasQTw = rwcwu⇤GTS(T fw Tw) (1.13)where u⇤ is the friction velocity, and GTS is a turbulent transfer coefficient for heat and salt. Inpractice, GTS is derived directly from observations, and therefore accounts for the rate-limitingprocess of salt-diffusion across the boundary layer. Here, the freezing point is evaluated using thefar-field salinity (Sw) (McPhee, 1992; McPhee et al., 1999) calculated asT fw = l1Sw+l2+l3Pb. (1.14)In summary, the two more sophisticated models used to estimate ablation rates at the base of anice shelf are the three-equation model,riabLi = riciab(TiT fb )rwcwu⇤GT (T fb Tw) (1.15)riab(SbSi) =rwu⇤GS(SbSw), (1.16)16T fb = l1Sb+l2+l3Pb, (1.17)or the two equation model,riabLi = riciab(TiT fw )rwcwu⇤G(TS)(T fw Tw) (1.18)T fw = l1Sw+l2+l3Pb, (1.19)with the friction velocity in both models parameterized as per Eq. 1.12. The parameters that areleast well constrained in the above models are the turbulent transfer coefficients GT , GS, and G(TS),and the drag coefficient Cd . In the turbulent boundary layer, heat and salt diffuse at the same rate,whereas in the viscous sublayer the turbulent eddy size is reduced so the turbulent diffusivity isreplaced by molecular diffusivity (Mellor et al., 1986; McPhee et al., 1987; Steele et al., 1989).Therefore, because the transfer coefficients parameterize all transport processes in the boundarylayer, GT must be larger than GS due to the differing rates of molecular diffusivity between heat andsalt.The transport of heat and salt through the boundary layer is also dependent on the shear stressat the ice-ocean interface which is parameterized using the friction velocity u⇤, which is itself de-pendent on the selection of the drag coefficientCd . The above notation follows that of Jenkins et al.(2010) where the turbulent exchange coefficients have a functional dependence on the frictionalvelocity (Jenkins and Doake, 1991). Exchange velocities have also been expressed by others asg(T,S,TS) =C12d G(T,S,TS)U, (1.20)From Eq. 1.10 the turbulent transfer coefficient is defined as,gT =QTbrwcwu⇤[T fb Tw](1.21)which is analogous to a thermal Stanton number, a ratio of heat transferred into a fluid to the thermalcapacity of the fluid, except for the use of the friction velocity rather than the velocity of the bound-ary layer. Stanton numbers derived from laboratory studies of boundary layers on hydraulicallysmooth surfaces have been used in numerical models of the interactions between ice shelves andthe ocean (Jenkins and Bombosch, 1995; Holland and Jenkins, 2001). The application of laboratoryresults to the boundary layer beneath an ice shelf is questionable, however Holland and Jenkins(1999) showed that it yields similar results to coefficients produced by more complex parameteriza-tions that include the effects of the stabilizing buoyancy caused by the release of freshwater at theice shelf base during melting, and the effects of rotation (McPhee et al., 1987). The insensitivity ofthe exchange coefficients to processes that effect mixing in the boundary layer is because exchangerates are largely determined by molecular diffusion across the viscous sublayer. The parameteriza-17tion compared favourably with observations beneath sea ice (McPhee et al., 1999).Jenkins et al. (2010) found that melt rates derived using both the two- and three-equation modelsmatched observations at the base of the Ronne Ice Shelf, Antarctica, within 40%. Adjustmentof the drag coefficient, the parameter that is least well constrained by observation, reduced themismatch below observational error. However, Jenkins et al. (2010) caution that the value of the dragcoefficient used assumes that the turbulent transfer coefficients estimated for sea ice are appropriatefor the base of an ice shelf, which given the different nature of boundary flow beneath ice shelves,forced by buoyancy and tides rather than wind-drift of sea ice, could violate that assumption. Furtherdiscussion of the parameterization of turbulent transfer coefficients can be found in McPhee et al.(1987), Steele et al. (1989), Holland and Jenkins (1999), and Jenkins et al. (2010). Values forturbulent transfer coefficients and drag coefficients have not been validated by measurements belowArctic ice shelves or glacier tongues.1.3 Thesis objectivesThere exist numerous observation gaps that this research aims to address. Recent studies of theoceanography of fjords in the Canadian Arctic are few (Melling et al., 2015). The last comprehen-sive oceanographic study of an ice shelf dominated fjord occurred over 50 years ago (Keys, 1977).For glacial fjords, where studies do exists in other regions they are often limited to brief summer sur-veys, and observations are often restricted to the outer region of the fjord several kilometers from theglacier terminus or grounding line (e.g. Straneo et al., 2010; Straneo et al., 2010; Inall et al., 2014).As well, continuous records of the seasonal cycle of changes in water properties in glacial fjord arerare (Cottier et al., 2005; Jackson et al., 2014b; Mortensen et al., 2014). The factors contributingto variations in epishelf lake depth and how these variation may reflect changes in the thickness ofthe ice shelf are not well understood (Vincent et al., 2001; Smith et al., 2006), due largely to a lackof continual observations in epishelf lakes. Finally, direct estimates of melt rates for ice shelves ormarine terminating glaciers have not been conducted in the Canadian Arctic, so the magnitude ofmelt rates, and factors that effect its spatial or temporal variability are unaddressed.This thesis aims to describe the oceanography of Milne Fiord, the last ice shelf-epishelf lakesystem in the Arctic.The primary objectives of this thesis are:• To provide the first comprehensive physical oceanographic description of Milne Fiord, includ-ing the influence of the Milne Ice Shelf and Milne Glacier on water properties, circulation,and seasonal variation in the fjord.• To understand the dynamics of the Milne Fiord epishelf lake and factors that influence itsseasonal and long-term variability.18• To estimate the magnitude, and spatial and temporal variation, of submarine melt rates inMilne Fiord.To aid in the interpretation of oceanographic data and allow estimates of submarine melt rates asecondary objective is:• To produce a digital elevation model of ice topography and bathymetry of Milne Fiord.1.4 Thesis outlineAn outline of the remainder of this thesis is presented below.Chapter 2 presents observations of the general water properties and circulation of the fjord andhow they are influenced by the presence of an ice shelf and glacier tongue. By conducting extensivehydrographic surveys, including through-ice profiles of water column properties that extend thelength and width of the fjord, including from the grounding line of the Milne Glacier to the outeredge of the Milne Ice Shelf, the spatial variation in the fjord is described. An ice-tethered mooringdeployed through the epishelf lake in the centre of the fjord recorded variation in ocean propertiesover the full water column spanning a period of 3-years. The observations are placed in context withmaps of ice thickness, produced by a ground-based radar survey of ice thickness in combination withremote sensing data, and bathymetric chart, based on spot soundings acquired during water columnprofiling.Chapter 3 presents a description of the seasonal and interannual variations in the epishelf lake,and aims to understand what factors influence its depth and how these are related to the state of theice shelf. All previous water column profiles acquired of the Milne Fiord epishelf lake since the firstobservation in 1983 were obtained and compiled with new profiles collected during this study tounderstand long-term changes in the water column. An ice-tethered mooring provided a continuousrecord of variation of the epishelf lake over a 3-year period. A simple analytical model of hydrauliccontrolled flow is used to simulate and explain the drainage of the lake each winter.Chapter 4 estimates of submarine melt rates of the ice shelf and glacier tongue are presented,and factors determining their spatial distribution and temporal variability discussed. Two indepen-dent methods are used to estimate annual melt rates from 2011 to 2015: a divergence of ice fluxmethod that utilizes remotely sensed ice surface velocities and the ice thickness model produced inChapter 2; and an ocean thermodynamic model that uses field measurements of ocean propertiesand circulation.Chapter 5 provides a summary of the entire body of work in the thesis and discusses potentialareas for future research.191.5 FiguresEllesmere IslandMilneIce ShelfAylesYelvertonInletWard HuntSersonMarkhamM‘ClintockInletTaconiteInletDisraeliFiordPetersenPhillipsInletArctic Ocean50 km AlertNansenSoundMilneGlacierTongue      85 oN  Greenland74 oN100 oW 50o WCape HeclaFigure 1.1Map of the northern coast of Ellesmere Island. Black and grey shading represents the extent of iceshelves in 2003, while the reduced extent as of January 2011 are shown in grey. Over this period the Serson,Ayles, and Markham ice shelves calved completely from the coast, the Petersen and Ward Hunt ice shelveslost large fragments, while the Milne Ice Shelf and Milne Glacier tongue remained largely unchanged duringthis period. Data courtesy of D. Mueller.20PolarWaterAtlanticWaterEpishelf LakeMinimum Draft of Ice ShelfIce Shelf Glacier TongueEpishelf LakeGrounding Line?IntermediaryExchange Flow?ArcticOceanABSubmarine melt?Surface ablation?GlacierFlow?Terminus Advance/Calving?Meltwater Plumes?Deep WaterRenewal?FjordCirculation?Epishelf LakeOutflow?SubglacialDischarge?Surface ablation?Submarine melt?Seabed Bathymetry?Sill?Sea Ice Freshwater IceSurfaceRunoff?FjordGlacierPolarWater?AtlanticWater?Internal Waves?Annual/Seasonal DepthVariation?Mixing?IceThickness?IceThickness?Figure 1.2 Schematic representation of Milne Fiord in A) plan and B) elevation view. Features observedprior to this study are labelled in regular font, while features or processes this study aimed to observe and/orquantify are indicated in bold italic.TbSbTiSiQiTQiSQwSQwTTwSwQbTQbSIceIce-oceanboundarylayerAmbientwaterUViscoussublayerFigure 1.3 Schematic representation of the temperatures, salinities, and the heat and salt balance at the ice-ocean interface at the base of an ice shelf. Modified from Holland and Jenkins (1999) and Straneo andCenedese (2014).21Chapter 2Modification of ocean properties andcirculation by an ice shelf and glaciertongue in Milne Fiord2.1 IntroductionGlacial fjords are the critical gateways between the ocean and ice sheets in the Arctic, and un-derstanding their dynamics is important to understanding and predicting changes to ice sheets andtheir impact on the ocean (Straneo and Cenedese, 2015). Large volumes of freshwater from melt-ing ice sheets are discharged into fjords and modified by fjord processes before being released tothe open ocean, with implications for ocean stratification and circulation. Submarine melting ofoutlet glaciers by warm ocean water has been identified as a plausible mechanism contributing toboth the retreat and increased flow rate of tidewater glaciers (Holland et al., 2008b; Motyka et al.,2011; Joughin et al., 2012; Straneo et al., 2012). These processes are all influenced by the propertiesof the fjord, including its geometry, water mass characteristics, ambient stratification and circula-tion, so an understanding of glacial fjord oceanography is key to predicting ice sheet dynamics,ocean circulation, and their impact on climate.The recent loss of coastal ice and and ice-associated features from fjords along the northern coastof Ellesmere Island, Canada, on the Arctic Ocean, is indicative of a shift in climate in this region.Widespread calving of coastal ice shelves (Vincent et al., 2001; Mueller et al., 2003; Copland et al.,2007; Mueller et al., 2008; White et al., 2015a), the almost complete loss of ice-dammed fjord-lakes(Mueller et al., 2003; Veillette et al., 2008), and the collapse of glacier tongues and the loss ofmulti-year land-fast sea ice (Pope et al., 2012) have been attributed, in part, to regional atmosphericwarming. The warming has resulted in an increased freshwater input to the fjords from surroundingglaciers, with mass loss from glaciers of the Canadian Arctic Archipelago being the single largest22contributor to sea level rise outside of Greenland and Antarctica between 2003-2009 (Gardner et al.,2011; Lenaerts et al., 2013). Despite the fact that fjords along the northern coast of Ellesmere Islandare impacted by climate change, the last physical oceanographic studies of fjords here occurredover 40 years ago (Ford and Hattersley-Smith, 1965; Lake and Walker, 1973; Keys, 1977), with theexception of one study from a small inlet in 1991-92 (Ludlam, 1996).There is growing evidence of the importance of fjord dynamics on determining the mass bal-ance of the Greenland Ice Sheet, and its contributions to sea level rise and influence on the oceanand climate, and these processes are likely relevant to the Canadian Arctic. The recent widespreadacceleration of outlet glaciers in southeast and southwest Greenland is thought to be linked to en-hanced submarine melt due to the migration of warm Atlantic water masses of subtropical origininto deep fjords (Holland et al., 2008b; Straneo et al., 2010; Christoffersen et al., 2011), and therapid collapse of the 15 km long Jakobshavn Isbrae glacier tongue (Motyka et al., 2011) was linkedto a warm water intrusion into the fjord (Holland et al., 2008a). Some estimates suggest that upto 80% of the mass loss of marine-terminating glaciers in Greenland can be attributed to subma-rine melting (Rignot and Steffen, 2008) although melt rates are largely based on parameterizationsthat are dependent on a variety of processes that are poorly constrained in glacial fjords. Theseprocesses include the ambient stratification and vertical distribution of heat in the fjord (Straneoet al., 2011; Mortensen et al., 2013), and the fjord circulation, which in glacial fjords is influencedby a complex interplay of wind stress and tides (Mortensen et al., 2011), baroclinic response toexternal forces (Sutherland and Straneo, 2012; Jackson et al., 2014b), convection from ice-oceaninteractions, and buoyancy-driven flows from seasonal freshwater runoff (Straneo et al., 2011). Intidewater glacial fjords runoff can enter the fjord at the surface or at depth across the groundingline of the glacier, known as subglacial discharge, with implications for vertical heat distribution,ocean stratification and fjord circulation (Straneo et al., 2011; Xu et al., 2012; Sciascia et al., 2013).However, with a few exceptions (Chauch et al., 2014), observations in glacial fjords are often lim-ited to the outer fjord, several kilometers from the glacier terminus, due to the challenges associatedwith sampling in the presence of a calving glacier, icebergs, ice me´lange, or through thick glaciertongues. While some of the Greenland-derived results are likely relevant to other regions, differ-ences in geography, coastal and climatic conditions, suggest glacial fjords in the Canadian ArcticArchipelago warrant special attention, yet they remain largely unexplored.We conducted an oceanographic field study of Milne Fiord, on the northern coast of EllesmereIsland, that is fed by the Milne Glacier, an outlet glacier that terminates in a floating glacier tongue atthe head of the fjord. A unique feature of Milne Fiord is the presence of the Milne Ice Shelf, distinctfrom the glacier tongue, that spans the mouth of the fjord. In this paper we distinguish betweena ’glacier tongue’, defined as the floating extension of an outlet glacier at the head of a fjord, andan ’ice shelf’, defined as a floating platform of thick ice at the mouth of a fjord, formed by insitu sea ice accretion and snow accumulation, sometimes with additional glacial input from low-23lying coastal ice caps and tributary glaciers (Jeffries, 2002). The Milne Ice Shelf forms an ice-damthat traps runoff within the fjord creating a perennially stratified water column with a freshwaterlayer floating on seawater, called an epishelf lake. The thin (<1 m) perennial ice cover of theepishelf lake created a stable platform for field work, permitting access to the glacial fjord foroceanographic investigation. To provide context for our study, we first provide a regional overviewof the northern Ellesmere climate and oceanographic setting with relation to the source waters thatcan access Milne Fiord. Next, because fjord dynamics are highly dependent on the geometry of thefjord we present the first bathymetric survey of the fjord, and present a digital elevation map of icethickness for the entire fjord based on field observations and remote sensing data. From analysis ofextensive through-ice hydrographic profiles, including transects that extend along the entire lengthof the fjord to the glacier grounding line (where the glacier detaches from the bed and is afloat),we describe the vertical and horizontal distribution of water masses, the distribution of submarinemeltwater and subglacial discharge, and the modification water properties in relation to bathymetryand ice topography. Using data collected by a full-depth ocean mooring deployed in the fjord wethen describe the internally and externally forced temporal variability of ocean temperatures in thefjord over the 3-year study. Finally, we discuss the implications of our observations on how loss ofthe ice shelf will impact fjord dynamics and the Milne Glacier.2.2 Methods2.2.1 Field SiteMilne Fiord (8235’N; 81W) lies at the northwest coast of Ellesmere Island on the Arctic Ocean(Fig. 2.1). A combination of low annual average air temperatures (⇠ -18C at Alert between1971-2000; www.ec.gc.ca), low ocean heat flux (Krishfield, 2005), and ocean circulation (Aksenovet al., 2010) create conditions favourable for the formation and persistence of coastal land-fast icealong this coast. The coast maintains the oldest and thickest multiyear land-fast sea ice in theArctic (Maslanik et al., 2011; Laxon et al., 2013), the largest concentration of coastal ice shelvesin the Arctic (total area of 500 km2 at end of 2015; Copland and Mueller, Arctic Ice Shelves andIce Islands, Springer-Verlag., In press); and the lowest glacier equilibrium lines in the NorthernHemisphere (Braun et al., 2004), resulting in extensive floating glacier tongues. The coastal iceconditions prevent ship-based measurements and as a result oceanographic studies of the fjords ofnorthern Ellesmere Island are few (Crary, 1956; Ford and Hattersley-Smith, 1965; Lake andWalker,1973; Keys, 1977; Ludlam, 1996).Existing observations collected offshore of northern Ellesmere Island reveal a stratified wa-ter column consisting of a 30–50 m deep seasonal mixed layer (SML), above comparatively fresh(<34.4 PSU), cold (<0C) Polar Water (PW) to⇠200 m depth (Aksenov et al., 2010 and referencestherein), overlying warmer (up to 0.5C), saline (>34.7 PSU) modified Atlantic Water (AW) below24350 m depth (Jackson et al., 2014a; Steur et al., 2013). A schematic of major currents in the ArcticOcean, including those inferred to exist but not directly observed along northern Ellesmere Island,is shown in Figure 2.1a. Surface waters above 200 m depth originate in the Canada Basin and typi-cally flow westward along the coast Steele et al., 2004; Aksenov et al., 2010). An eastward flowingboundary undercurrent appears to exist along the continental slope from the base of the mixed layer(30 – 75 m) to the bottom (Newton and Sotirin, 1997) likely an extension of the Arctic CircumpolarCurrent (Rudels et al., 1999; Aksenov et al., 2010). Sparse bathymetric soundings along the coastshow the continental shelf shoals to 270 m at the shelf break, separating Milne Fiord, and the 600 mdeep basin at its mouth, from the deep ocean (Fig. 2.1b). Published bathymetry for Milne Fiord inthe International Bathymetric Chart of the Arctic Ocean (IBCAO v3.0; Jakobsson et al., 2012) wasbased solely on an interpolation between offshore soundings and the coastline. Soundings insidethe fjord prior to this study were limited to two small, shallow (77 m and 80 m deep) inlets on thesouthwest and northeast shores of the fjord, the bathymetry was otherwise unknown.The Milne Glacier, a 4–5 km wide, 55 km long outlet glacier that drains ⇠4% by area of thenorthern Ellesmere Island icefields (GLIMS and NSIDC, 2005), flows into the head of Milne Fiord(Fig. 2.2a). Previous work suggested that the terminus of the glacier was probably afloat and formeda glacier tongue (Hattersley-Smith, 1969; Narod et al., 1988; Jeffries, 1984), referred to here as theMilne Glacier tongue (MGT). The MGT is actually a composite of three glacier tongues; the centraltongue comes from the main valley glacier flowing from the highest elevations of the ice cap and hasgreater surface relief and thickness, while the southwest and northeast tongues come from tributaryglaciers that join the main glacier within 10 km of the present-day grounding line. VanWychen et al.(2016) found surface velocities of the Milne Glacier on the order of 100 m a1 near the groundingline, and estimated the mean ice discharge to be 0.05 ± 0.02 Gt a1 between 2011 and 2015. TheMilne Glacier has been identified as a possible surge-type glacier (Jeffries, 1984; Copland et al.,2003).At the mouth of the fjord, spanning 18 km between Cape Egerton and Cape Evans, and distinctfrom the MGT, is the Milne Ice Shelf (MIS; Fig. 2.2a), formed through a combination of sea iceaccretion, snow accumulation and input from surrounding tributary glaciers and low-lying coastalice caps (Jeffries, 1986a). The MIS was mapped using ice-penetrating radar by Mortimer et al.(2012) in 2008/9, and was found to have a mean thickness of 49 m and a maximum thicknessof 94 m. The topology of the ice shelf is quite variable owing to its complex origins, and waspreviously divided into three distinct regions, the Outer, Central and Inner Units, based on surfacemorphology and ice characteristics (Jeffries, 1986a). The Outer Unit at the seaward edge of theice shelf, distinguishable by a series of surface ridges and troughs running parallel to the coast,is generally thick (>50 m), but is bisected by two re-healed fractures that have existed since atleast 1950 (Hattersley-Smith and Fuzesy, 1969). The Central Unit forms the landward edge ofthe ice shelf bordering the epishelf lake, and has a more erratic surface morphology and thickness25(Jeffries, 1986a). The Inner Unit, that was thought to have once been contiguous with the MGT,no longer exists, it has been replaced with epishelf lake ice (Mortimer et al., 2012). As of 2009the epishelf lake consisted of a main basin spanning the width of the fjord between the MIS andMGT, and two arms extending along either side of the MGT, with a total estimated area of 52.5km2 (Mortimer, 2011). The epishelf lake was 14.3 m thick as of 2009, its depth presumed tocorrespond to the minimum draft of the MIS (Veillette et al., 2011b). With the exception of a fewshallow hydrographic profiles collected to study the epishelf lake (Jeffries, 1985; Veillette et al.,2008; Veillette et al., 2011a), there was no information on the general oceanography of the mainfjord prior to this study.2.2.2 Bathymetry and ice thickness mappingBathymetry and ice topography are critical to understanding fjord circulation and exchange withoffshore waters so we produced a digital elevation model (DEM) of Milne Fiord from a combinationof field measurements, remote sensing, and available archived data. Between 2011 and 2015, wecollected 53 bathymetric soundings by lead line and during CTD profiling through augered iceholes in the epishelf lake, steam-drilled holes in the MIS, and through natural fractures in the MISand MGT. Soundings extended from 5 km offshore of the MIS to the hinge crack near the MilneGlacier grounding line. NASA’s Operation Icebridge 2014 Multichannel Coherent Radar DepthSounder (MCoRDS; Leuschen et al., 2010) data was used to constrain bed elevation upstream ofthe grounding line, and the cross-sectional bed profile of the fjord above the grounding line wasmodelled as a U-shaped valley as in Van Wychen et al. (2014) to model glacier thickness at a fluxgate for Milne Glacier. We further constrained fjord bathymetry from analysis of hydrographicdata, and ice radar, discussed in detail below. To reconcile the regional IBCAO 3.0 data with ournew observations we replaced IBCAO grid points in Milne Fiord with our own data and modelbathymetry on a 100 m x 100 m mesh using the Matlab gridfit.m function (J. D’Errico, 2006;www.mathworks.com). Gridfit.m is not an interpolant designed to exactly predict all source data,but rather it is an approximant that produces a smooth surface that represents the behaviour ofthe source data as closely as possible, allowing for gaps and noise in the data, and allows smoothextrapolation beyond the domain of the source data. Bathymetric soundings are sparse in manyareas of the fjord, particularly below regions of the MIS and MGT, so the bathymetric model shouldbe viewed in this context.To constrain ice thickness in the fjord we utilized a combination of ground and aerial radar mea-surements. The thickness of the MIS was mapped by Mortimer et al. (2012) using a snowmobile-towed Ice-Penetrating Radar (IPR) system in 2008 & 2009, and we reproduce this data here, butsupplement it with new sources. We expanded spatial coverage of the MIS in 2011 and 2013 us-ing a snowmobile-towed Sensors and Software Pulse EKKO Pro 250 MHz IPR system, with dataprocessed as in Mortimer et al. (2012). Ground-based measurements of the MGT were acquired26in 2012 and 2013 using a 50 and 70 MHz Blue System Integration IPR system, as described inMingo and Flowers (2010)), towed on foot, with data processing as in Wilson et al. (2013). Icethickness along the centreline of the MGT was further constrained by aerial radar measurementsacquired by the Operation Icebridge 2014 MCoRDS (Leuschen et al., 2010). The location of theMilne Glacier grounding line is assumed to correspond to the point of first hydrostatic equilibrium,where the thickness of the ice derived from Airborne Topographic Mapper (ATM) laser altimeter(Krabill, 2010) surface elevation equals that measured by the Multichannel Coherent Radar DepthSounder (MCoRDS; Leuschen et al., 2010). Error in the ice depth measurements is +- 10 m.For increased spatial coverage, we estimated ice thickness from surface elevation measurementsof the MIS and MGT by assuming that ice downstream of the Milne Glacier grounding line wasafloat. Analysis of IPR transects showed very strong bottom reflections, indicative of an ice-waterinterface, everywhere except a region just south of the confluence of the re-healed fractures in theMIS (2.2c). The ice here was relatively thin (<40 m) and showed intermittent and weak bottom re-flections, possibly due to an ice-bedrock interface. These observation, combined with nearby bathy-metric soundings through the re-healed fracture that revealed the seabed rose to within 28 m of thesurface, suggested the ice shelf may have been partially grounded in this region. Consequently, weexclude an area encompassing the weak bottom reflection IPR transects from the surface elevationto thickness conversion.Where ice was afloat, or where we have no data to the contrary, ice thickness was estimatedfrom surface elevation by assuming the ice was in hydrostatic equilibrium with a mean ice densityof 900 kg m3 and mean seawater density of 1024 kg m3. Existing high resolution (13 m x 23 mgrid size) digital elevation data from Natural Resources Canada (NRCAN; http://www.geobase.ca)was found to be inaccurate over large regions of the MIS and MGT when compared with field data,so we compiled data from various other sources. Ground-based position and elevation data wererecorded using a Topcon HiPerV Dual Frequency Global Navigation Satellite System (GNSS) re-ceiver mounted on the IPR sled sampling at 5-second intervals with precise positioning computedusing the online tool available from Natural Resources Canada (http://www.nrcan.gc.ca). Positionsare accurate to ± 10 mm horizontal and ± 15 mm vertical. We also used the Ice, Cloud, and LandElevation Satellite (ICESat) laser altimetry data (Zwally et al., 2011) from the GLA06 product(Release-33) acquired between 2003 and 2009. The laser altimeter measures surface topographyto a precision of 10 cm with a laser surface-spot diameter of ⇠70 m every ⇠170 m. Additionalelevation data was utilized from a swath along the MGT collected by the Operation Icebridge 2014Airborne Topographic Mapper (ATM) laser altimeter (Krabill, 2010). All elevation data were pro-cessed by manual removal of outliers and smoothed prior to calculating ice thickness by applying a1-dimensional 5-point convolution filter. Tides are not accounted for, however nominal tidal rangeis small, on order of 0.3 m, so we take this as the error in elevation data.A cross-point error analysis was conducted comparing the measured ice thickness from IPR data27to that estimated from surface elevation data for points within a distance of 35 m (the approximateradius of the ICESat laser footprint). Mean absolute discrepancy was 0.66 m, with a standarddeviation of 1.20 m. Given the large ice surface topography variations (± ⇠8 m), tidal heightvariation, and changes in mass balance over time that would affect the elevation data, as well as thetemporal offset of data collection periods, this error was deemed acceptable.Polygons delineating surface features, including the fjord shoreline, margins of the glaciertongue, epishelf lake, and ice shelf and major fractures were identified using a 21 July 2009 Ad-vanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite scene and dig-itized using Matlab (R2012b). Uncertainty in dimensions derived from satellite imagery was esti-mated to be twice the 15 m image resolution. The thickness of the epishelf lake ice was assumed1 m everywhere based on numerous field measurements. IPR transects across the re-healed frac-tures in the MIS revealed a narrow bottom channel at the base of the ice shelf running parallel tothe surface expression of the fracture. The ice thickness at the apex of the channel was measured bysteam drill at 3 locations and varied between 8–11 m. Using the highest quality IPR transects acrossthe fracture as a reference, we modelled the basal channel as an inverted parabola with uniformcross-sectional thickness along its entire length. The depth below sea level of the channel (Zint) wasinterpolated as:Zint =(ZmZc)D21D22+Zc (2.1)where Zc is the depth at the centerline of the channel (10 m), Zm is the depth of the ice at the channelmargin (40 m), D1 is half the width of the base of the channel (70 m), and D2 is the distance fromthe centerline of the channel to the centre of the interpolated ice column. Ice depth was interpolatedat 10 m intervals from the centerline to the margin every 100 m along the length of the channel. Allice thickness data was gridded onto a 100 m X 100 m mesh using the Matlab gridfit.m function (J.D’Errico, 2006) to produce contour maps and raster images of ice thickness for the entire fjord.2.2.3 HydrographyOcean properties were measured over 5 field campaigns in May 2011, May 2012, July 2012,May 2013, July 2013, and July 2014. The 2011 profiles were recorded using a 4 Hz SBE19+Conductivity-Temperature-Depth (CTD) instrument, all others using a 6 Hz RBR XR-620 CTD.The instruments were lowered at 0.5 m s1 using a manual reel through drilled holes, leads in seaice and fractures in the ice shelf and glacier tongue. All profiling was completed on foot or bysnowmobile, except two ⇠50 km along-fjord transects which were completed by helicopter in Julyof 2012 and 2013. During these surveys profiles were collected at up to 12 locations, includingthrough a rift near the grounding line of the Milne Glacier and through leads in the sea ice 5 kmbeyond the seaward edge of the MIS outside the fjord. Data was processed in Matlab following aprocedure that included correction for atmospheric pressure, application of a low-pass filter in time28to the raw pressure, temperature and conductivity, alignment of conductivity and temperature andpressure, a thermal cell mass correction (for the SBE19+ data), and loop editing. Derived vari-ables were calculated using the International Thermodynamic Equation of Seawater 2010 (TEOS-10) Gibbs Seawater Oceanographic Matlab Toolbox (www.TEOS-10.org) and averaged over 0.2 mdepth bins. CTDs were calibrated once every 1–2 years. A single turbidity profile near the cen-ter of the main basin of the epishelf lake was recorded using a Seapoint Turbidity Sensor at 100xgain in July 2013. Additional offshore hydrographic profiles over the northern Ellesmere Islandcontinental shelf break were obtained from the Arctic SwitchYard Project (PI Mike Steel, Uni-versity of Washington). Profiles were collected in May of 2012 and 2013 from Twin Otter andhelicopter by lowering a SBE19plusv2 Seacat CTD through open leads or drilled holes in the seaice. Data was downloaded from the Advanced Cooperative Arctic Data and Information Systemwebsite (www.aoncadis.org) and derived variables calculated as above.To estimate the amount of meltwater present in the water column we use a simple mixing modeldescribed in Jenkins (1999). Conservative properties of a water mixture, c , are determined by therelative mass input of its individual components, c1 and c2:Qc = Q1c1+Q2c2 (2.2)where Q = Q1 +Q2 is the total mass of the mixture. On a bivariate property plot, such as atemperature-salinity diagram, measurements of any two conservative properties of a mixture willfall on a straight line between endpoints defined by the properties of the component water masses.For temperature to be used in the mixing model in the case when ice melts in seawater, heat must beconserved. To do so we must account for 1) sensible heat loss from seawater to ice that warms theice from its internal temperature, Qi, to the freezing point of the ambient water, Q f , 2) absorption oflatent heat during phase change, and 3) sensible heat loss during mixing of the meltwater, at freezingpoint, with ambient seawater. Ice can then be assigned an effective temperature Q⇤i that accounts forheat loss during the melting process by:Q⇤i =Q f Lcp cicp(Q f Qi) (2.3)that allows temperature to be used as a conservative property in (2.2) (Gade, 1979). In Eq. (2.3)L is the latent heat of fusion, and ci and cp are the specific heat capacities of ice and seawater,respectively, at constant pressure.If the ambient water properties are not homogenous, but are a mixture of two water masses,then the meltwater mixture will fall within the triangle bounded by the meltwater lines of eachwater mass endpoint. In Milne Fiord the ambient water is a stratified mixture of Polar Water (PW)and modified Atlantic Water (AW) and melting may occur over various depths in the halocline. A29composite property Y can be defined as in Jenkins (1999):Y2,1 = (c2c2w0) (c1c1w0)✓∂c2∂c1◆(2.4)where superscript i represents temperature or salinity and subscript 0 indicates defined endpointwater mass characteristics (PW with -1.7 C and 31.2 g kg1, and AW with -0.8 C and 34.25 gkg1). The last term in brackets on the right is the gradient of the ambient water line. Values of thecomposite property can be related to meltwater fraction by:Y=Y2,1mixY2,1melt(2.5)where the numerator is the sampled composite property of the mixture, and the denominator is thecomposite property of ice (in fact, pure meltwater with zero salinity and an effective temperature(Q⇤i ) as above) (Eq 6; Jenkins, 1999). The calculation of meltwater concentration assumes directmixing between uniform endpoint water masses, but in practice, the transition between PW and AWis non-linear on a temperature-salinity plot, indicative that other processes influence the interface.The non-linear PW-AW interface is consistent with the cold halocline observed on the northernEllesmere Island continental shelf (Jackson et al., 2014a), that forms in eastern Nansen Basin fromheat loss to the atmosphere, lateral mixing between two types of Atlantic water, and the admixtureof cold, fresh continental shelf water (Steele and Boyd, 1998; Alkire et al., 2010; Rudels, 2013).The curvature of the profiles away from the mixing line produces an offset in calculated meltwaterfractions, so we calculate the meltwater concentration anomaly by subtracting the depth-averagedmeltwater concentration offshore. Other studies have used a similar thermodynamic melt model butwhere ice was assumed to melt into a mixture of homogenous seawater (usually AW) and subglacialfreshwater discharge (Mortensen et al., 2013). However, the grounding line depth in Milne Fiordoccurs within the interface between PW and AW, not deep within the AW layer as in the otherstudies, indicating melting may be driven by any mixture of PW and AW, which suggests the use ofthe Jenkins (1999) heterogeneous water column approach is more appropriate for deriving meltwaterfractions in Milne Fiord.The depth at which the meltwater plume will reach a level of neutral buoyancy can be deter-mined by estimating the salinity (and density) of the rising plume from the concentration of melt-water. Assuming that the plume is a mixture of ambient seawater at the depth of the groundingline and meltwater originating from in situ submarine melting of the glacier with zero salinity andnegligible suspended sediment load, the minimum salinity of the plume is Sa(1Ymax), where Sais the ambient salinity at the grounding line depth and Ymax is the maximum meltwater fraction. Ifwe assume the rising meltwater plume entrains no ambient seawater it will rise to the level of itsisohaline and then separate from the base of the ice shelf or glacier and spread out laterally along30the fjord. The meltwater fraction can be bounded at the lower limit to zero, while the theoreticalmaximum concentration is limited by the heat available in the ambient water to melt ice, so an upperlimit on meltwater fraction is found by:QiQ1+L+ ci(Q f Qi)cp(QwQ f )1(2.6)where Qi is the mass of ice (meltwater), and Q is the total mass of the mixture (Eq 7; Jenkins,1999). In polar waters the meltwater fraction is limited to about 1% per C of temperature abovefreezing (QwQ f ).To measure changes in ocean properties over time a mooring was deployed in the centre of theepishelf lake from May 2011 to July 2014. The mooring initially consisted of 20 RBR TR1050/60thermistors spaced vertically from 0.5 m to 350 m depth, as well as an RBR XR-620 CTD at 355m depth. The mooring was ice-anchored at the surface and suspended down the water column withthe CTD initially just above the bottom. However, the pressure sensor on the CTD indicated itsettled onto the seabed after a week, likely from a combination of the surface anchor melting intothe ice and mooring line stretch. Sediment fouling of the conductivity sensor on the CTD corruptedthe salinity data so only temperature is reported. The mooring was recovered and reconfigured inMay 2012, then serviced in July 2012, July 2013, and July 2014, at which times instruments wererepositioned as availability required.2.2.4 CirculationWater level was recorded using the bottom anchored XR620 CTD at 355 m depth at the mooringsite from May 2011 to May 2012. A 3 hr low pass-filter and a linear detrend was applied to the rawdata. The dominant tidal constituents were determined by harmonic analysis with nodal correctionsusing the T Tide Matlab toolbox (Pawlowicz et al., 2002). Results were compared with predictionsover the same time period at the same location from a two constituent (M2 & K1) barotropic inversemodel of the Arctic Ocean that predicts tidal range and depth-averaged tidal currents on a 5-kmhorizontal grid (Padman and Erofeeva, 2004).Water velocity was recorded over durations of 4, 6, and 10 days in May 2011, July 2012, andJuly 2013, respectively, using an ice-anchored downward looking 300 kHz RDI Workhorse ADCPlocated at the mooring site (Fig. 2.2a). The ADCPwas secured to a rigid mast 1 m below the epishelflake ice. In 2013 an additional downward-looking 300 kHz RDI Sentinel ADCP was moored at 50m depth. Due to the unreliable magnetic field at this latitude the ADCP was deployed on a bifilar(two line) mooring to fix its orientation relative to a known bearing. The ADCPs sampled at 5 minintervals, with 150 pings per ensemble, 2 m depth bins, and had an accuracy of 0.5% water velocity± 0.5 cm s1. The low level of acoustic backscatterers in the water column limited the range ofthe instruments to less than nominal. Noisy data near the practical range limit of the instruments,31defined as falling below a beam correlation threshold of 60, was removed in post-processing inMatlab. Velocities are reported in along- (positive v is flow out of the fjord) and cross-fjord (positiveu is flow toward the northeast shore) components (see Fig. 2.2b).Air temperature, shortwave solar radiation, wind velocity, relative humidity, and barometricpressure were recorded by a HOBO automated weather station located at 10 m elevation on theshore on Milne Fiord (Fig. 2.2a). Air temperature is correlated with surface meltwater production(Hock, 2003), the number of cumulative of positive-degree days (PDDs) were used as a proxy forthe volume of surface meltwater runoff flowing into the fiord each year.2.3 Results2.3.1 MappingThe bathymetry of Milne Fiord (326 km2), defined as extending 43 km from the Milne Glaciergrounding line to the seaward edge of the MIS, is shown in Figure 2.2b. Off the seaward edge ofthe MIS the bed is over 600 m deep and connected to a deep basin that extends out of neighbouringYelverton Bay, west of Cape Evans. Soundings through the re-healed fracture show the bed shoalsupward from the coast under the Outer Unit of the MIS, rising to within 28 m of the surface nearthe confluence of the fractures. Weak bottom reflections from surrounding IPR transects, perhapsindicating an ice-bedrock interface, suggest a seabed ridge rises to within 5 m of the surface andextends about 2 km south of the fracture. Comparison of hydrographic profiles collected off theseaward edge of the ice shelf and through fractures in the Central Unit, indicate the ridge doesnot extend completely across the width of the fjord (i.e. is not a sill) as water properties are thesame down to the seabed at 360 m both to the north and south of the ridge. However, a similarhydrographic analysis (discussed below) reveals a ⇠ 260 m deep bathymetric sill, extending acrossthe width of the fjord, exists below the Central Unit of the MIS, near the northern margin of theepishelf lake. Landward of the sill, a deep basin with a maximum known depth of 436 m existsbelow the main body of the epishelf lake. Below the terminus of the MGT the seabed slopes upwardtoward the head of the fjord, rising to a depth of 150 m near the grounding line. The estimatedlocation of the grounding line of the main branch of the Milne Glacier (82.455N, 80.32W ) issupported by the presence of hinge cracks at this location. The bed topography indicates the glacieris presently grounded on a reverse-sloping bed, and the bed deepens to a maximum of 205 m belowsea level 2 km upstream, and remains below sea level for at least 24 km upstream.The combination of ground, aerial, and satellite radar sources for measurements of ice thickness,and interpolations of ice thickness from surface elevation, provide decent spatial coverage (Fig.2.2c) for production of the ice draft DEM (Fig. 2.2d). Ice thickness at the Milne Glacier groundingline was estimated to be 168 m, slightly thicker than the previous estimate of 152 m by Van Wychenet al. (2014). The 16 km long MGT thins quickly in the first few kilometers downstream of the32grounding line and then more gradually until the terminus where it is <10 m thick. The meanthickness of the MGT is 44 m ± 36 m, with ice of the central branch thicker than the southwest andnortheast branches owing to the different origins of the ice streams that flow into the tongue. The66 km2 MGT covers approximately 20% by area of the fjord and has an estimated total volume of2.6 km3 w.e.The updated thickness data for the MIS shows a maximum ice thickness of 94 m and meanice thickness and standard deviation of 50 m ± 21 m. The MIS covered an area of 206 km2,approximately 65% by area of the fjord, with an estimated total volume of 9.7 km3 w.e., within theerror of that previously estimated (9.8 ± 0.35 km3 w.e.) by Mortimer et al. (2012). Our expandedice thickness coverage generally confirms the model presented by Mortimer et al. (2012), althoughwe find the southern and eastern portions of the Central Unit generally thinner, the difference owingto their lack of data coverage in these regions. Additional IPR transects confirm an area of relativelythin (<40 m), possibly grounded, ice exists south of the re-healed fracture. For relative scale, theice keels of the MGT and MIS combined, occupy an estimated 16% of the total volume of MilneFiord (to mean sea level).The hypsometry of the MIS is such that it must significantly impede water flow into and out ofthe fjord from the surface down to approximately 50 m depth. Export of buoyant freshwater fromthe fjord would be preferentially routed along the thinnest ice which coincides with the re-healedfractures in the Outer Unit. IPR transects across the re-healed fractures revealed that narrow basalchannels exists at the base of the ice shelf that mirror the surface expression of the fractures. Cross-sectional profiles of the channel indicate it has a width of 50-150 m at its base where the ice is 40–60m thick, but narrows to less than 10 m at the crest where the ice is ⇠10 m thick.Thin (<1 m) freshwater ice covers the 52.5 km2 MEL, roughly 15% of the total fjord area. Themain body of the MEL is 5.8 km wide, with long narrow arms running 16 km alongside the MGTto the grounding line, and extending into the three small bays off the main fjord. The volume offreshwater in the lake, using a depth of 10 m (from CTD profiles) and assuming vertical shoresis 0.525 km3. SAR imagery shows isolated pockets of lake ice that occupy small fractures in theMGT and the Central Unit of the MIS, suggesting freshwater ponding occurs extensively throughoutthe fjord. CTD profiling revealed that the depth of the freshwater layer in most, but not all, ofthese fractures is identical to the main body of the epishelf lake, indicating that a network of basalchannels connects these fractures with the main lake. The area and volume of the lake is thereforelikely somewhat larger than reported here.2.3.2 HydrographyHydrographic profiles show the vertical stratification of water masses in Milne Fiord (Fig. 2.3 and2.4). The epishelf lake is evident as a relatively warm (>1C) layer of freshwater at the surface inall profiles collected within the fjord. The epishelf lake is not present seaward of the MIS (dashed33lines in Fig. 2.3), although some surface freshening from sea ice melt is evident in summer profilesoffshore. The thickness of the epishelf lake varied over the study, thinning from⇠13 m in May 2011to ⇠7 m in July 2013 (top salinity inset Fig. 2.3). At the bottom of the epishelf lake salinity risessharply from <3 g kg1 to over 20 g kg1 in 2 m and this strong halocline is associated with thestratification maximum (Fig. 2.3c). The depth of the stratification maximum is associated with theminimum draft of the MIS, water above this depth cannot exit the fjord and is trapped behind the iceshelf. Below the base of the MEL to a depth of 50 m is seawater that is substantially warmer (up to1C ) and fresher (20 g kg1 to 34 g kg1) than that found at equivalent depths offshore, we term thislayer fjord-modified water (FMW). FMW is also associated with a secondary halocline between 20–40 m depth (Fig. 2.3c). Below the FMW is a transition from cold, relatively fresh PW (50 to 200 m)to warm, saline AW (below ⇠200 m), signifying the main Arctic Ocean halocline. Water propertieswithin the fjord over these intermediate depths are generally similar to those offshore, indicatingthere is free exchange below the MIS. There is, however, evidence of cooling and temperatureinterleaving at depths between 50 – 150 m in profiles collected near the head of the fjord, indicativeof modification of fjord waters by submarine melting of ice. Below a depth of 260 m, deep water(DW) properties landward of the MIS are nearly homogenous in temperature, salinity and density,ranging by only 0.044C, 0.028 g kg1, and 0.03 kg m3, respectively, over the bottom 180 m of thefjord. The properties of the weakly stratified deep water are distinct from those at equivalent depthsoffshore (visible in the temperature profile in Fig 2.3a), indicating a topographic sill lies beneath theMIS that limits offshore exchange at these depths.Thermistors moored at 355 m depth in the fjord recorded an abrupt temperature decrease of0.05C between 29–30 January 2012, which indicates a deepwater renewal event occurred (Fig.2.5). A thermistor moored at sill depth (260 m) from May 2011 to May 2012 showed the deepwaterwas initially thermally stratified, but became isothermal to sill depth at the time of the renewal,while thermistors moored above sill depth remained thermally stratified throughout the record (notshown). The change in deepwater properties was confirmed by CTD profiles collected before (May2011) and after (July 2012) the renewal (Fig. 2.5). The deep water was colder, more saline, anddenser after the renewal.The nearly homogenous deep water indicates that the fjord was filled to sill depth by inflowingwaters from a narrow depth range offshore. The level of the source waters can be estimated byfinding the level where seawater density offshore of the sill is equal to the density of fjord bottomwater, known as the effective sill depth. A pair of profiles acquired approximately 3 km apartthrough the inner edge of the MIS in July 2012 show distinct deep water properties, indicating thesill lies between the two profile locations. Water with the same density as bottom water (1027.825kg m3) in the fjord was found at ⇠260 m depth outside the sill, indicating this is the effective silldepth. The actual sill depth could be slightly shallower than the effective sill depth as there areinternal hydraulic mechanisms such as coastal trapped waves or internal tides that could lift dense34water from greater depths up and over the sill during deep water renewal (Farmer and Freeland,1983).The thermistors recorded a steady increase of deep water temperatures in the fjord of approxi-mately 0.037C per year before and after the renewal event. CTD profiles showed that deepwatersalinity increased by 0.015 g kg1 per year after the renewal event, with an associated density in-crease of 0.01 kg m3 per year. Water properties measured at 260 m depth by annual CTD profilesoffshore revealed similar changes of temperature, salinity, and density fromMay 2012 to May 2015.These observations suggest that offshore changes in stratification and circulation altered the sourcewater properties at sill depth, leading to the observed changes in the deep water. The temperatureincrease is steady over time, suggesting that small volumes of dense offshore water steadily flowover the sill and continually fill the basin from the bottom. The 355 m temperature record shows adistinct fortnightly spring-neap period (readily visible from October 2011 until the renewal event;Fig. 2.5) suggesting overflow could be tidally driven. This idea is supported by the persistence ofa slight vertical density gradient in deepwaters below sill depth (i.e. the deepwaters are not com-pletely homogenous), suggesting small volumes of dense water are continually flowing over the silland filling the fjord from the bottom up, displacing lighter water upward. In contrast, the January2012 renewal event appeared to be a major event that completely filled the deep basin to sill depth.Dissolved oxygen concentrations from bottle samples collected at 320 m depth in July 2012 providefurther evidence of renewal, as deep water oxygen concentrations (⇠9 mg L1) were only slightlydepleted in comparison to measurements at equivalent depths offshore (⇠9.5 mg L1) collected bythe SwitchYard project in May 2012. The January 2012 renewal event occurred during a neap tide,when tidal energy and flows are low. Deep water renewal has been associated with neap tides inother fjord systems, including the Strait of Georgia (Masson, 2002), Saanich Inlet (Manning et al.,2010), and Puget Sound (Geyer and Cannon, 1982). In these systems reduced turbulent mixing overthe sill generated by weaker tidal flow over the allows stratification to reach a maximum, enhancingdense water overflow into the fjord. A similar process could be occurring in the constriction overthe sill in Milne Fiord and under the Milne Ice Shelf, although as only one complete renewal wasobserved, further investigation of this process is needed.Observations of submarine melt and subglacial runoff near the Milne Glacier grounding line areshown in the temperature-salinity plots of profiles collected in July 2012 and July 2013 (Fig. 2.6).The mixing line connects the ambient endpoint waters masses, PW and AW, any mixture of whichcould be in contact with ice and drive melting. Any ambient water with a temperature above thesurface freezing point (blackline) has the potential to melt ice. The melt line, with a slope of 2.7C(g kg1)1, was calculated using the ambient water properties at the depth of the grounding line(150 m); internal ice temperature , -15C, water temperature, -0.8C, and salinity, 34.25 g kg1. Ifsubmarine melting is occurring, then the mixture of melted glacier and ambient water observed inCTD profiles, called the submarine meltwater mixture, will fall on or close to the melt line (with a35minimum temperature limited to the freezing temperature). Note that the meltwater mixing modelis only valid within the triangle defined by the endpoint waters masses (i.e. between PW at 35m depth and AW at 150 m depth). Profiles collected through hinge cracks at the grounding line(dark blue) exhibit a substantial temperature decrease and fall close to the melt line, indicatingsubmarine melting is occurring. Profiles collected within a few kilometers of the grounding line(light blue), through fractures in the MGT or through the epishelf lake ice along the margins of theMGT, also show a submarine melt signature, but to a lesser degree. At increasing distances fromthe grounding line the meltwater signature diminishes; offshore profiles show little or no evidenceof meltwater. The grounding line profiles also show horizontal excursions away from the melt linetoward lower salinity, most apparent at 120 m depth in July 2013, but evident in both years near thegrounding line. These horizontal excursions are indicative of fresh subglacial discharge enteringthe fjord at depth (Straneo et al., 2011; Chauch et al., 2014). A single turbidity profile collectednear the mooring site in July 2013 shows a turbidity peak at a salinity corresponding to the depth ofthe apparent subglacial discharge inflection (2.6b), further evidence that subglacial runoff exits thegrounding line at this depth. Turbidity reaches a minimum at 35 m depth, but is elevated above thislevel. The source of the turbidity in the MEL (not shown) is likely due to terrestrial surface runoff,while elevated turbidity in the FMW layer could be settling of sediment from the MEL, or spreadingof a turbid subglacial discharge plume.Spatial variation in fjord water properties are shown in hydrographic sections from July 2012(Fig. 2.7) and July 2013 (Fig. 2.8). The overall spatial patterns were similar between years. Waterproperties are contoured through the ice (i.e. shown above the maximum ice depth) due to thepronounced cross-fjord variability in ice thickness that allows some exchange of water above thisdepth to the surface. Along-fjord transects show the water column properties offshore of the iceshelf are typical of the Arctic Ocean, including a SML above PW and AW. The warmest AW at350 m depth extends up the fjord to the location of the inferred topographic sill, below the southernedge of the MIS. The deep waters in the basin inside the inferred sill are nearly homogenous intemperature and salinity, although slightly thermally stratified in 2012. The epishelf lake is apparentas a uniform freshwater layer extending from the Outer Unit of the MIS to the head of the fjord.Water temperatures are generally lower near the head of the fjord, suggestive of cooling by contactwith ice near the grounding line. The highest meltwater concentrations are observed at the head ofthe fjord, with a maximum just above the bed of the glacier. The meltwater fraction is smaller withincreasing distance from the grounding line, suggesting the highest melt rates, and the source ofmeltwater observed in the fjord is from the glacier. The meltwater appears to form several distinctplumes distributed vertically in the water column that spread out horizontally throughout the fjord,and appear to be exported out of the fjord at depth. Some meltwater is observed near the base of theMIS, although the source of the melt, whether produced in situ or advected from up-fjord, cannot bedetermined. This pattern is indicative of an estuarine-type flow, with outflowing meltwater mixing36with seawater inflowing at the same or greater depths.Cross-fjord hydrographic sections extend from the northeast shore to the southwest through themooring site at the centre of the epishelf lake. The epishelf lake extends completely across the fjordat a constant depth and there is minimal cross-fjord tilting of isotherms and isohalines. However,temperature inversions and interleaving layers are visible in the upper water column above 150 mand there is some cross-fjord patchiness in melt water fraction. There is very little change between2012 and 2013, with the exception of a shoaling of isotherms below sill depth in 2013.The layering of water masses in the fjord is consistent over all field campaigns, however thereare changes over time in the properties of each layer. The epishelf lake showed evidence of seasonaland interannual variation, deepening by a few meters fromMay through July 2012, yet showed verylittle change in the depth in 2013. The evidence suggests that the lake depth is linked to the volumeof inflow due to surface melt water production, although a complete discussion of the dynamics ofthe epishelf lake is presented in Chapter 3.Similarly we find evidence that subglacial outflow is linked to seasonal and interannual varia-tions in surface meltwater production (Fig. 2.9). Warm air temperatures facilitate melting of snowand ice in the glacier catchment, which flows down through crevasses and discharges at depth intothe fjord across the grounding line at the bed of the glacier. The subglacial discharge (near 0Cand 0 g kg1) can rise as a turbulent buoyant plume, entraining ambient seawater, until it reaches adepth of neutral buoyancy. Profiles collected over a 10-day period in July 2012 show a decrease ofsalinity by up to 4 g kg1 between 30 and 55 m depth. Above and below these depths there is littlechange in salinity over time, indicating the freshening is due to a lateral, not vertical, transport pro-cess. Temperature interleaving at these depths is also indicative of advection of water, suggesting asubglacial outflow plume. The temperature (<-1C) and salinity (31 g kg1) of the plume indicateit is mainly composed of seawater, and may have been cooled by contact with the base of the glaciertongue. The field campaign in 2012 corresponded to a period of warm atmospheric temperatures(almost 100 PDDs had accumulated by the end of profiling), intense surface melt, and presumably,substantial subglacial discharge. Conversely, the summer of 2013 was cool, with less surface meltwater production (<40 PDDs accumulated by the end of profiling) and no apparent evidence of asubglacial discharge plume in salinity or temperature profiles.2.3.3 CirculationThe tidal range estimated from water level record in Milne Fiord was 0.34 m. Harmonic analysisof 59 tidal constituents showed amplitudes of the M2 (12.42 h), K1 (23.93 h), S2 (12 hr), and O1(25.82 hr) constituents were 0.056 m, 0.042 m, 0.023 m, and 0.023 m, respectively. A ratio ofthese constituents given by (K1+O1)/(M2+ S2), gives 0.82, indicating a mixed, predominatelysemidiurnal tide (Thomson, 1981). Tides accounted for 63.7% of the variance of the original waterlevel data over the 245 day record.37Time-averaged velocities in the upper water column at the mooring location for May 2011, July2012 and July 2013 are shown in Figure 2.10. All years show generally weak (<2 cm s1) baroclinicflow in the upper water column, with cross-fjord currents being of equal or greater magnitude toalong-fjord currents. There is little or no flow at the surface in the epishelf lake. The record fromMay 2011 is a useful baseline, as air temperatures are well below zero and no surface or subglacialrunoff is expected. We find a slight outflow peak below the epishelf lake, and a larger outflow at40 m depth. Velocity profiles in 2012 and 2013 extend further down the water column and showflow reversals. A relatively strong 6 cm1 mean flow to the northeast shore at 60 m depth wasrecorded in July 2012. Summer of 2012 was a strong melt year and this may be related to the exportof subglacial water. Alternatively the current could be associated with exchange below the MIS,which is less obstructed at this depth compared to the restricted flow higher in the water column.The velocity data reveal that circulation in Milne Fiord is fundamentally 3-dimensional and timevarying. The depth-averaged internal Rossby radius within Milne Fiord above the depth of the sillis>8 km, which is larger than the width of the fjord landward of the MIS so the effects of rotationaldynamics on circulation are minimal. Although the stratification of the fjord changes over time,particularly in relation to seasonal freshwater runoff, our data indicate the internal Rossby radiusremains larger than the fjord width. The 3-dimensional nature of the circulation is more likely dueto the complex ice geometry and bathymetry of the fjord.2.3.4 Mooring timeseriesMeteorological conditions recorded by the automated weather station and ocean temperature at themooring site from May 2011 to July 2014 are shown in Figure 2.11. Average air temperature duringthe mooring deployment was -17.6C, with extreme hourly temperatures of +20.2C in July 2012and -51.8C in February 2013. Solar radiation varied from near zero during the polar night (mid-October through February), to a summer daily maximum of 650 W m2 in late June during theperiod of 24 hour sunlight from April to September each year. Cumulative PDDs, a proxy for sum-mer melt conditions, varied substantially among years, with 278, 255, and 92 PDDs in 2011, 2012,and 2013, respectively. Variations in melt water inflow to the fjord were manifested in changes inocean temperatures in the upper water column in the fjord, while temperature fluctuations in deeperwaters are not directly coupled to surface conditions. The highlighted isotherms approximate theinterface between water masses: 0C isotherm around 10 m depth signifies the epishelf lake/FMWinterface; -1.5C isotherm the FMW/PW interface; -0.5C the PW/AW interface; with nearly ho-mogenous DW below 260 m. The warming and thickening of the MEL and FMW are stronglycorrelated with the onset and strength of the surface melt season. During the warm summers of2011 and 2012, when we expect high surface meltwater production and runoff, the MEL and theFMW layer warm and deepen to a much greater degree than during the cool summer of 2013. Thechanges in the FMW layer are consistent with the subsurface freshening and warming attributed to38subglacial discharge plumes (Fig. 2.9), while the deepening of the MEL is consistent with runoffentering the fjord at the surface. The thinning of the MEL and FMW after the surface melt seasonends is much more gradual than the rate of thickening during the melt season, suggesting a slowadvection of heat (and freshwater) out of the fjord under the MIS over the remainder of the year.The PW and AW layers show some seasonal variation as well as pronounced high frequency fluctu-ations above sill depth. The interface between PW and AW in Milne Fiord tends to deepen in winterand shoal in summer, varying by almost 50 m over the year, with excursions of 25 m at periods onthe order of days to weeks. These high frequency fluctuations can be more readily seen in Figure2.12. At 125 m, temperatures seasonally reach a maximum in summer and decrease to a minimumin mid-winter, varying by 0.6C over the duration of the record, but there are large amplitude (up to0.5C), high frequency fluctuations, on timescales of days to weeks at all times during the record.Spectral analysis of temperature fluctuations at 125 m depth (not shown) reveal small peaks at 7–12days, with stronger peaks at 18 days, 35 days, and 59 days, present throughout the timeseries, in-cluding during winter. Temperatures at 25 m and 50 m show smaller amplitude (generally <0.1C)high frequency fluctuations, and appear more influenced by summer runoff. The long-term tem-perature decrease at 25 m from summer is associated with the shoaling of isotherms related to thethinning of the FMW layer. Overall, these observations suggest that waters above 50 m depth areprimarily influenced by the import and export of runoff (both surface and subglacial) from the fjordcatchment, while waters below 50 m depth are strongly influenced by external forcing, particularlyoffshore variations in the depth of the PW-AW interface that translate into the fjord.The heat content of the upper water column of the fjord, from 5 m to 125 m depth, spanning thedepths containing most of the ice in the fjord, increases rapidly during warm summers with highmeltwater inflow (2011 and 2012), with a smaller increase during cool summers (2013) (Fig. 2.13).2.4 Discussion2.4.1 Geophysical settingWe present the first comprehensive bathymetry of Milne Fiord, which reveal a flooded glacial valleywith a maximum known depth of 436 m. Hydrographic evidence indicates the presence of a 260m deep bathymetric sill under the landward edge of the MIS, which could be an ancient terminalmoraine from a previous advance of the Milne Glacier. The bathymetric sill is significantly deeperthan the thickest ice in the fjord (at the Milne Glacier grounding line) so we expect it has little directinfluence on the properties of water that can come in contact with ice and drive melting, though it isfundamental in determining the deep water properties of the fjord. Shallower bathymetric featureshowever, such as the newly discovered ridge that rises to within 28 m of the surface under the MIS,could influence water properties in contact with ice by enhancing mixing, diverting currents, andfurther reducing the cross-sectional area at the mouth of the fjord for exchange flow under the ice39shelf.The full spatial extent of the ridge is unknown, although the weak bottom reflections in IPRdata suggest it could extend up to 2 km south of the re-healed fracture if the ice is grounded inthis region. In an 1981 aerial radar survey, Narod et al. (1988) found a similar band of low bottomreflectivity, extending 4 km south of the re-healed fracture, and stretching across the width of thefjord. However, Narod et al. (1988) interpreted the low-reflectivity as signal attenuation due to brinesoaked or brackish ice freezing onto the bottom of the ice shelf. It is not possible with available datato determine which explanation is correct, so we have chosen not to include data from IPR into ourbathymetric model. Further investigation is required to verify cause of the weak bottom reflections,and if the ice is indeed grounded, to map the spatial extent and topography of the ridge.Although the full extent of the grounded ice is undetermined, our limited bathymetric soundingsand ice thickness data indicate the ice is almost certainly grounded near the confluence of the re-healed fractures as the ice shelf increases to>60 m thick within 200 m of the 28 m deep bathymetricsounding. This finding could explain the origin of the fractures in the MIS (which have existed sincethe first aerial photographs were acquired in 1959 (Hattersley-Smith and Fuzesy, 1969) as stressfractures (or hinge cracks) between grounded and floating ice. The partial grounding of the MISwould also resist forces acting to detach the MIS from the coast, and help explain why the MIS hasbeen relatively stable in contrast to the full or partial collapse of all surrounding ice shelves alongnorthern Ellesmere Island (Mueller et al., 2008; Veillette et al., 2011a)).Our revised DEM of ice thickness in Milne Fiord expands on that constructed by Mortimer etal. (2012) for the MIS by improving spatial coverage of regions of the MIS, and providing partialcoverage of the MGT. Comparison to Mortimer et al.’s 2012 DEM of the MIS reveals that whilethe total ice volume is not significantly different, there are some spatial differences in ice thick-ness. Although most of the difference is due to differences in data coverage, there are likely somedifferences due to methodology in production of the DEM. In particular, the automated griddingfunction that we used may underestimate ice thickness by smoothing over sub-grid scale features,such as the re-healed fractures, producing wider and thinner channels than that manually contouredby Mortimer et al. (2012). Despite these minor differences, our DEM represents the source datawell and provides the critical context for understanding water properties and circulation in the fjord.The thickness of theMGT is highly variable across its width, owing to its composite origins fromthree different ice streams. Based on the grounding line locations estimated by Hattersley-Smith andFuzesy (1969) from surface features in aerial photographs, the grounding lines of the southwest andnortheast ice streams have remained relatively stationary since 1959, while the grounding line ofthe central ice stream appears to have retreated by 2–3 km. These observations suggests the threeice streams have different grounding line depths, which could be associated with subglacial dis-charge channels that enter the fjord at different levels. Although there is some supporting evidencefor freshwater discharge at different depths in the CTD profiles collected through the hinge crack,40further work is needed to confirm this hypothesis.Continued retreat of the Milne Glacier grounding line will push it into deeper water. This couldlead to breakup of the glacier tongue, and acceleration of the Milne Glacier, as observed at theJakobshavn Isbrae glacier tongue in Greenland (Motyka et al., 2011; Joughin et al., 2012). Theappearance of new crevasses and expanded rifts at the hinge line in 2013 suggest the breakup of theMGT may already be underway. As the Milne Glacier is grounded below sea level 24 km upstreamof its present-day grounding line, there is potential for the grounding line to retreat a significantdistance, with important ramifications for Milne Glacier dynamics and its contribution to sea levelrise. We note that the present-day grounding depth of the Milne Glacier is relatively shallow, andindicates understanding factors that affect the properties of the upper water column in Milne Fiordare critical to future understanding of how the glacier will respond to ocean forcing.2.4.2 Freshwater inflow and outflowThe dominant influence on the properties of the upper water column in Milne Fiord is the inflowand outflow of freshwater. Freshwater enters the fjord from three main pathways: 1) surface runoff;2) subglacial discharge; and 3) submarine ice melt. The amount of freshwater runoff entering thefjord, both at the surface and subglacially, is highly dependent on the strength of the melt season. Atthe surface, freshwater enters the fjord via proglacial streams from surrounding tributary glaciers,and flows directly off the Milne Glacier as supraglacial streams. These surface inflows contributeto an increase in volume and depth of the epishelf lake during summer, but the strong stratificationand inhibition of wind-mixing by the perennial ice cover means the influence of surface inflow islargely confined to the epishelf lake.The largest subsurface source of freshwater to the fjord in summer is subglacial discharge. Thesubglacial discharge plume, which entrains ambient water during its ascent, appears to reach a levelof neutral buoyancy below the epishelf lake halocline during years of high meltwater production(e.g. 2012). Observation of current velocities over the same period show flow at this level is directedout of the fjord, suggesting the subglacial plume spreads out horizontally and is exported from thefjord at depth. In July 2013, surface meltwater production was low, and we expect the volume fluxof subglacial discharge was equally small. There was no apparent change in salinity or temperaturebelow the epishelf lake associated with a subglacial discharge plume in 2013, yet the small turbiditypeak at 112 m is suggestive of a deeper subglacial plume. The equilibrium depth of the plume willbe dependent, in part, on the freshwater volume flux across the grounding line. The relatively smallvolume of subglacial discharge in 2013, inferred from cumulative PDDs, means the plume mixture’sdensity may have equilibrated to ambient waters after ascending just meters or a few tens of meters.In contrast, a relatively high volume of subglacial discharge occurred in 2012, imparting to theplume sufficient buoyancy to rise much higher in the water column, its ascent ultimately limited bythe strong stratification near the base of the epishelf lake.41The export of surface and subglacial freshwater fromMilne Fiord is constrained by the thicknessof the MIS. The MIS acts as a dam across the mouth of the fjord, restricting water exchange betweenthe fjord and offshore to a depth of approximately 50 m. Flow above this depth is likely routedthrough a narrow basal channel that runs along a re-healed fracture transversing the ice shelf fromeast to west. All surface runoff that enters the fjord landward of the ice shelf must flow throughthis geometric constriction. Thus, to a depth of 50 m the ice shelf is a slowly leaking dam, allowingthe fjord to store freshwater and heat (supplied by the runoff water and warm water entrained fromdepth in the subglacial plume) over longer periods, gradually releasing it out of the fjord throughthe channel. The hydraulics of the drainage through the channel are discussed in detail in Chapter 3,specifically with regards to the epishelf lake, but similar principles likely apply to drainage of FMW.Substantial volumes of freshwater are exported at depth, and the timing of the export is de-coupledfrom the timing of the surface melt season due to the flow restriction imposed by the MIS. Althoughthis scenario is likely unique to Milne Fiord at present, it may have been more common whenice shelves along northern Ellesmere Island were numerous. The widespread subsurface export offreshwater for weeks or months after the melt season would have implications for the stratificationand circulation of waters along the northern Ellesmere continental shelf.2.4.3 Submarine meltA source for freshwater in the fjord is submarine melting. The highest glacial meltwater concen-trations were observed near the grounding line of the Milne Glacier. A possible source of bias ofobservations collected in the narrow hinge crack is that the water here could be isolated, and notrepresentative of the ambient waters near the grounding line. However, the general water columnproperties in the hinge cracks are consistent with other profiles collected throughout the fiord, show-ing the same depth of the epishelf lake and the structure of the PW-AW interface. At the surface thecracks are up to 50 m wide, extend most of the distance across the fjord and are visibly connectedwith other fractures that could permit water exchange. These observations strongly suggest that thewater within the hinge crack is not isolated, and the high meltwater concentrations observed are rep-resentative of conditions near the grounding line. Melting near the grounding line could be drivendirectly by contact with warm ambient waters at depth, or further enhanced by convection-drivenmelt from subglacial discharge.The submarine meltwater mixture, a mixture of ambient ocean water and freshwater from melt-ing of the glacier, is buoyant and will rise along a vertical ice wall (such as those of the hingefracture) or along the sloped base of an ice shelf (such as the MGT). Assuming the buoyant plumeentrains no additional ambient water during its ascent, the density of the meltwater mixture, and thusits level of neutral buoyancy, is determined by the concentration of fresh meltwater in the plume,which is limited by the amount of heat available to melt ice. The theoretical maximum meltwaterconcentration at the grounding line is ⇠2% (from Eq. (2.6) where QwQ f = 2C), which would42lower the density of the meltwater mixture enough for the meltwater plume to rise ⇠50 m verti-cally given the ambient stratification. The maximum observed meltwater concentration was 0.7%,representing a reduction in buoyancy that would only permit a ⇠25 m vertical displacement of themeltwater plume. In practice, however, a turbulent meltwater plume will entrain ambient seawa-ter, increasing its density, and further reducing the maximum vertical displacement. As a result, ameltwater plume produced solely by submarine melting (no freshwater input from subglacial dis-charge) will rise at most a few 10s of meters in Milne Fiord before separating from the ice faceand spreading out laterally at depth. Indeed, our hydrographic observations show meltwater plumesextending along the fjord at multiple depths, apparently close to their depth of origin, given thatthey extend horizontally from the grounding line (Fig. ??). Interleaving meltwater layers havebeen previously observed adjacent to ice melting in a stratified water column, both in laboratoryexperiments (Huppert and Turner, 1980) and near floating glacier tongues (Jacobs et al., 1981). Theexport of submarine meltwater at multiple depths is likely associated with compensating inflow atother depths, suggesting the submarine melt-driven circulation in Milne Fiord is complex.Subglacial discharge-induced buoyancy-driven circulation has been identified as a key factorin enhancing melt rates of tidewater glaciers in Greenland during summer (Straneo and Cenedese,2015), and is likely partially responsible for observations of high melt rates near the groundingline in Milne Fiord. This process however, may be quite localized across the fjord, dependent onthe location and geometry of the subglacial channels. An even more important factor controllingsubmarine melt rates is the fjord-scale circulation that the localized buoyancy-driven circulationcould drive. Recent observations of horizontal re-circulating cells associated with subglacial plumesin Greenland (Stevens et al., 2015), indicate simple two-layer models of estuarine-like overturningcirculation driven by subglacial discharge, are inappropriate (Straneo and Cenedese, 2015). Themooring timeseries does show a pronounced shoaling of isotherms at the depth of the groundingline coinciding with the surface melt season each year that could be related to enhanced circulationdriven by subglacial discharge, although we lack sufficient data to conclusively link these processes.Regardless of the processes involved, the shoaling of isotherms means warmer water can access thegrounding line and would lead to an increase in submarine melt. While these processes are likelyvery important to summer submarine melt rates, which may be a large portion of the annual massbalance, the effects are confined to the very brief surface melt season, which lasts<2 months at thishigh latitude. For the remaining 10 months each year, submarine melting at the grounding line isdominated by changes in ambient waters, which are largely influenced by processes external to thefjord.2.4.4 External forcesThe rate of melt and the vertical range over which it will occur in Milne Fiord is dependent, in part,on the ambient source water properties in the Arctic Ocean. Johnson et al. (2011) suggested the43depth of the near-freezing SML, which varies between 25 to 60 m (Jackson et al., 2014a), might bean important factor in limiting the extent of basal melt on the ice tongue of the Petermann Glacier.In Milne Fiord, a narrow band at the base of the SML (50–60 m depth) does appear to penetrateunder the MIS to the grounding line. The SML will limit melting over these depths, however,surface properties in the fjord above 50 m are significantly different than those offshore due to theimpoundment of runoff by the MIS. The impact of the SML is therefore largely confined to limitingmelt on the Outer Unit of the MIS.The highest submarine meltwater concentrations were observed at the Milne Glacier groundingline, and the main external influence here is the depth of the interface between PW and AW. Weobserved both seasonal and high frequency temperature fluctuations over the PW-AW interface inMilne Fiord that are likely associated with offshore processes. Thermistors deployed at 160 m depthon the North Ellesmere shelf in the western Lincoln Sea during the Switchyard Project from 2008 to2009, showed similar temperature fluctuations as observed at 125 m depth in Milne Fiord, with vari-ations up to 0.5C per day that occurred several times per month throughout the year (Jackson et al.,2014a). Jackson et al. (2014a) explained temperature fluctuations with periodicities of 12 h and 14days by semidiurnal tides and the spring-neap tide cycle, while they suggested fluctuations with pe-riods of ⇠ 10 days could be explained by synoptic events, including storms. Jackson et al. (2014b)linked fluctuations in water properties in two Greenland fjords on timescales of 3–10 days to densityfluctuations on the continental shelf driven by along-shore down-welling favourable winds inducingvelocity pulses into the fjord. Jackson et al. (2014b) also speculate that other phenomenon suchas coastal trapped waves and eddies could generate the pressure gradients across the continentalshelf and fjord required to generate pulses in the absence of wind forcing. In Milne Fiord, temper-ature fluctuations that we associate with depth variation of the PW-AW interface, occur at severaltime scales, from semidiurnal tidal periods, to seasonal variation. Spectral analysis of tempera-ture fluctuations at 125 m depth (not shown) reveals small peaks at 7–12 days, likely associatedwith the synoptic offshore wind variation, while stronger intraseasonal peaks at 18 days, 35 days,and 59 days, are present throughout the timeseries, including during winter. We speculate theseintraseasonal temperature variations are related to the passage of offshore coastal trapped waves.Although not specifically addressed over the 3-year duration of this study, we suspect long-terminterannual variation due to shifts in regional circulation patterns may be important in determiningwater properties in Milne Fiord. Along the northern coast of Ellesmere Island, waters above 100 mdepth are usually characteristic of the Canada Basin, and show relatively little variability (Jacksonet al., 2014a). However, waters below 100 m can originate in either the Canada Basin or EurasianBasin, which have very different water column properties, resulting in significant interannual varia-tion in temperature and AW depth depending on regional circulation patterns (Newton and Sotirin,1997; Jackson et al., 2014a), which could have a profound influence on the ambient properties inMilne Fiord.44Renewal of the deep water below sill-depth in Milne Fiord can also be driven by changes inoffshore density stratification. The isopycnals at sill depth were 20 m shallower after the renewalevent than prior to it (Fig. 2.5), indicating that shoaling of the halocline offshore allowed denserwater to flow over the sill and replace the deep water in the fjord. As only one deep water renewaloccurred over the 3-year study, it suggests an interval of at least 1–2 years between events. Thecause of the renewal is uncertain; analysis of ice drift patterns and sea ice cover offshore of MilneFiord in the weeks leading up to the event showed no clear evidence of conditions that might haveinitiated upwelling over the continental shelf (i.e. surface ice drift toward the southwest). However,the observed changes in Milne Fiord do indicate that deep water properties are linked to regionalchanges in the depth of the PW-AW interface in the Arctic Ocean.An along-fjord temperature section from 2012 extended over the continental shelf using profilescollected as part of the SwitchYard Project is shown in Figure 2.14. Beyond the continental break,the AW layer has a temperature maximum near 1C at 370 m, almost 3C above the in situ freezingpoint. This warm water is prevented from entering Milne Fiord by the shallow continental shelf andthe sill under the MIS. Although this is important from the perspective of general fjord dynamics,including deep water renewal, the core of the offshore AW layer is substantially deeper than thethickest ice in the fjord, so it will not have a direct contact with the ice. Rather, it is the propertiesof the upper ocean, above 150 m depth, including the SML, PW, and the depth of the PW-AWinterface, that will directly influence ice-ocean interactions in Milne Fiord.The characteristics of AW vary on multidecadal time scales of 50-80 years, likely connected tolarge-scale atmospheric circulation (Polyakov et al., 2004). The upper boundary of AW shoaled byabout 75 m in the late 1980s-90s in the Makarov Basin (Polyakov et al., 2004), and about 40 m in theAmundsen Basin (Steele and Boyd, 1998). If AW along the northern Ellesmere Island continentalshelf break responds similarly, this would suggest the properties of water that could propagate intoMilne Fiord could vary over the long-term as well. The steady warming of deep water in MilneFiord, consistent with warming at sill depth offshore, could in fact indicate a regional shoaling of theupper boundary of AW, although our records are much too short to make any definitive link. Long-term variation of the upper boundary of AW in Milne Fiord could significantly alter the heat contentof water in contact with ice and have repercussions for submarine melt rates and the mass balance ofice shelves and tidewater glaciers. The recent widespread retreat of tidewater glaciers in Greenlandwas linked to propagation of a warm AW anomaly around the coast, indicating link between glacierdynamics and mass balance and regional ocean circulation (Holland et al., 2008a; Straneo et al.,2012). Investigation of a similar link between large-scale atmospheric and ocean conditions in theArctic Ocean and long-term variations in submarine melt rates along northern Ellesmere Islandcould prove an interesting area of future research and provide greater insight into the long-termstability of ice along this coast.452.4.5 OutlookThe ice shelves and glacier tongues of Ellesmere Island are in a state of decline, with little chance ofregeneration in the present and projected climate. The Canadian Arctic Archipelago was the largestsingle contributor to eustatic sea level rise outside of Greenland and Antarctica between 2003–2009(Gardner et al., 2011; Lenaerts et al., 2013), which emphasizes its importance on a global scale.Despite widespread collapse of ice shelves and glacier tongues in surrounding fjords (Mueller et al.,2003; Copland et al., 2007; Mueller et al., 2008; White et al., 2015a; Copland et al., 2015), theMIS and MGT have remained comparatively stable (Mortimer et al., 2012). The presence of theMIS in particular has had a profound impact on the water properties of Milne Fiord, and has likelyinfluenced the stability of the MGT.However, a network of new fractures extending across the inner portion of the MIS was observedin May 2013. The potential collapse of the ice shelf would result in a much more rapid export offreshwater (with its elevated heat content) out of the fjord, shortening the submarine melt season ofthe upper water column to roughly coincide with the length of the surface melt season. The lossof the perennial freshwater surface layer would reduce the stratification of the upper water column,potentially allowing subglacial runoff to reach the surface and alter the depth of freshwater export.As well, the loss of the epishelf lake may increase potential of seasonally ice-free conditions, withenhanced calving activity, as has been seen in Disraeli Fiord since the drainage of the epishelf lake(W. Vincent, pers. comm.). This is particularly important for the Milne Glacier, as the collapse ofglacier tongues have been linked to the acceleration of glacier flows, with repercussion for sea levelrise.2.5 ConclusionRelative to the Arctic Ocean, the water properties and heat content within Milne Fiord are stronglymodified by the presence of a coastal ice shelf, a tidewater outlet glacier, and a bathymetric sill.The modification is primarily due to three factors: 1) the impoundment of freshwater runoff (bothsurface and subglacial) behind the ice shelf at the mouth of the fjord; 2) submarine melting at theglacier grounding line where ice is in contact with warm AW; and 3) the presence of a bathymetricsill under the MIS that prevents the warmest AW from entering the fjord. This work is the firstcomprehensive oceanographic study of Milne Fiord and our findings emphasize the importance ofmapping ice topography and bathymetry to provide context for interpretation of ocean observations.Exchange under the ice shelf, above 50 m depth, is likely restricted to a narrow basal channelextending along a re-healed fracture in the ice shelf. The sub-surface export of freshwater runofffrom the fjord is slowed due to the flow constriction imposed by the basal channel in the MIS,partially de-coupling the export of freshwater from the timing of the summer melt season, a uniqueprocess warrants further investigation. Although the presence and impact of a largely intact coastal46ice shelf is presently unique to Milne Fiord, we expect many of our results to be applicable to othertidewater glacial fjords in the northern Canadian Arctic Archipelago. Tidewater glaciers drain anestimated 47% of the total glaciated area of northern Arctic Canada (i.e. Ellesmere Is., Axel HeibergIs., and Devon Is.; Lenaerts et al., 2013), underscoring the need to understand ocean-ice interactionsin this region, and how it may differ from other regions.The MIS is a critical feature that modifies water properties in Milne Fiord and its inevitablebreakup in the warming climate will fundamentally alter the oceanography of the fjord, with reper-cussions for the stability of the MGT and dynamics of the Milne Glacier. A small number of glaciersdominate discharge to the ocean from the CAA (Van Wychen et al., 2014), and as a consequenceoverall ice discharge may be highly sensitive to changes in the dynamics of only a few marine termi-nating glaciers. Our findings suggest that long-term monitoring of ocean properties near tidewaterglacier termini, along with improved bathymetry and ice topography, will be required to understandglacier-ocean interactions and better predict the influence of these changes in the Canadian Arctic.472.6 Figures  99 o    90o    81o    72o    63o   30’   82 o   30’   83 o   30’   84 o   30’   85 o    Ellesmere IslandArctic Ocean← Milne FiordLongitude (oW)Elevation (m)−1000−800−600−400−2000200400600800>1000Latitude (o N)Canada BasinLomonosov RidgeTranspolar DriftLincolnSeaGreenlandArctic Circumpolar Boundary CurrentMendeleyev-Alpha RidgeABFigure 2.1 Map showing regional setting of Milne Fiord. A) Arctic Ocean bathymetry with major currentsoverlain showing the path of Atlantic Water (red) and Polar Water (white). Dashed line indicates currents notdirectly observed (after Aksenov et al. 2013). B) Bathymetry and topography of northern Ellesmere Island,the region defined by the red box in (A). The white circles mark the University of Washington SwitchYardProject 2012 CTD transect shown in Fig. 2.14. Regional bathymetric data from IBCAO v.3 was combinedwith Milne Fiord soundings collected during this study and re-gridded onto a 2.5 km grid. Contour lines areshown every 200 m to a maximum of depth of 1000 m.48  82o  30’   81o  30’   80o’0382’35’04’54ovu2011 May2012 May2012 July2013 May2013 July2014 July2015 Julyall yearsDepth (m)6003000  8 2o  30’   81o  30’   80o 30’  82 o 35’ 40’ 45’  Ice draft (m)0 50 100 150Ice draft (m)0 50 100 150Outer UnitCentral UnitA BC DsillFigure 2.2 Map of Milne Fiord. A) ASTER image mosaic captured 21 July 2009. The Milne Ice Shelf(MIS), Milne epishelf lake (MEL), and Milne Glacier tongue (MGT) are outlined in black. Also indicatedare the grounding line of the Milne Glacier (red line), the automated weather station (yellow triangle), thelocation of the mooring (red triangle), the re-healed fractures in the Milne Ice Shelf (red arrows), and multi-year land-fast sea ice (MLSI) along the coast. B) Modelled bathymetry of Milne Fiord with locations ofdepth soundings/CTD profiles are indicated. CTD profiles were collected at the mooring location during allfield campaigns. The reference axis for ADCP current velocities (+u flow to the northeast, +v flow out of thefjord) measured at the mooring site are shown. C) Ice draft measurements derived from ground-based IPRand surface elevation, aerial 2014 NASA Icebridge radar and altimetry measurements, and ICESat altimetrydata, overlaid on 21 July 2009 ASTER image mosaic. The dashed rectangle indicates a region of possiblypartially grounded ice. D) Digital ice draft model extending from the Milne Glacier grounding line to theouter edge of the MIS.49−2 −1 0 1 20100200300400Pressure (dbar)Θ (o C)  MELFMWPWAW0 10 20 30SA (g kg−1)0 0.05 0.1N (s−1)May2011May2012July2012May2013July2013offshore20 300204060SA (g kg−1)Pressure (dbar)A B CDWFigure 2.3 Through-ice hydrographic profiles of A) conservative temperature (Q), B) absolute salinity (SA),C) buoyancy frequency (N) in Milne Fiord from all field campaigns. Different water masses are indicated:Milne Fiord epishelf lake (MEL); fiord-modified water (FMW); Polar Water (PW); Atlantic Water (AW); anddeep water (DW). The freezing point of seawater with salinity of 32 g kg1, which is representative of thefjord, is shown (grey line in panel (A)). Dashed lines indicate profiles collected offshore of the Milne IceShelf.500 5 10 15 20 25 30 35−2−1.5−1−0.500.511.52010 20MELPWAWΘ (o  C)SA (g kg−1)Distance from Grounding Line (km)0 10 20 30 40FMWDWFigure 2.4 Bivariate property plots of conservative temperature and absolute salinity in Milne Fiord showingall profiles collected in July 2012 and July 2013. Q SA profiles are coloured by distance along the fjordfrom the Milne Glacier grounding line (KM 0; dark blue) to offshore of the Milne Ice Shelf edge (>KM 45;red). Water masses are labeled (MEL - Milne Fiord epishelf lake; FMW - fjord-modified water; PW - PolarWater; DW - deep water; AW - Atlantic Water). Isopycnals (s0 kg m3; grey lines) and the surface freezingpoint (solid black line) are shown.51Jul/11 Oct/11 Jan/12 Apr/12 Jul/12 Oct/12 Jan/13 Apr/13 Jul/13 Oct/13 Jan/14 Apr/14 Jul/1400.Θ (o C)  260 m thermistor320/355 m thermistor320/355 m CTD profile260 m CTD profile offshore34.6 34.7 34.8 34.9200250300350400SA (g kg−1)−0.1 0 0.1 0.2200250300350400Θ ( o C)27.65 27.75 27.85200250300350400σ0 (kg m−3)   sill depthMay 2011July 2012July 2012 (out)Depth (m)AB C DFigure 2.5 A) Timeseries of bottom-water temperature in Milne Fiord from May 2011 to July 2014 showinga deep water renewal, evident by the rapid 0.05C temperature decrease at 355 m depth on 29–30 January2012 (arrow). Black lines is hourly data from a thermistors moored at 355 m depth until May 2012, thenredeployed at 320 m depth for the remainder of the study (the repositioning was associated with a minor(<0.0025C) temperature discontinuity). Gray line is data from a thermistor moored at 260 m until May2012. Squares indicate temperature measured during CTD profiles at the equivalent depth at the mooringlocation. Triangles indicate temperature measured during CTD profiles at 260 m depth offshore. Profilesof B) conservative temperature, C) absolute salinity, and D) potential density before (May 2011; grey line)and after (July 2012; black line) the deep water renewal. Also shown is a profile collected in July 2012 justoutside the sill (dashed line). Effective sill depth is estimated as the level at which density outside the sillcorresponds to bottom water density in the fjord in July 2012.5230 30.5 31 31.5 32 32.5 33 33.5 34 34.5−2−1.8−1.6−1.4−1.2−1−0.8120 m150 m35 mSA (g kg−1)July 20122425 26 27 Melt lineMixing line   30 30.5 31 31.5 32 32.5 33 33.5 34 34.5−2−1.8−1.6−1.4−1.2−1−0.8120 m150 m35 mSA (g kg−1)July 201325 26 27  00.10.2112 mTurbidity (FTU) Melt lineMixing line Distance from Grounding Line (km)0 10 20 30 40Θ (o  C)Θ (o  C)A BFigure 2.6 Bivariate property plots of conservative temperature and absolute salinity in Milne Fiord fromA) July 2012, and B) July 2013. Q SA profiles are coloured by distance along the fjord from the MilneGlacier grounding line (KM 0; dark blue) to offshore of the Milne Ice Shelf edge (>KM 45; red). In (B) theturbidity vs salinity profile shown was collected near the mooring location. The peak in turbidity correspondsto a depth of 112 m. In (A) and (B) the approximate depths of the Milne Glacier grounding line (150 m), asubglacial runoff outlet (120 m), and the base of the seasonal mixed layer (35 m), and mixing and melt lines(dashed black line) are shown. Isopycnals (s0 kg m3; grey lines) and the surface freezing point (solid blackline) are shown. See Section 2.3.2 for the methodology related to the melt and mixing lines.53122626.426.827.227.627.827.9Depth (m)0100200300400500Depth (m)Distance (km)45 40 35 30 25 20 15 10 5 001002003004005002426 426.827.227.627.8Depth (m) 0100200300400Θ (o C)−1.5−1−0.50Distance (km)Depth (m)  64200100200300400Melt fraction (x10−3 )01234567A CB DFigure 2.7 Along- (left panels) and cross-fjord (right panels) hydrographic sections from July 2012 of A)and C) conservative temperature with isopycnals overlain, and B) and D) melt water fraction. Along-fjordsections extend from the Milne Glacier grounding line (KM 0) to offshore of the Milne Ice Shelf (KM45). The cross-fjord maximum ice draft of the MGT (KM 0 to KM 16) and the MIS (KM 20 to KM 43)(black lines) and bathymetry along the centre of the fjord (gray) are shown. CTD casts (black squares) werecollected through natural fractures in the ice or through drilled holes in the epishelf lake ice by helicopter over24 hrs. Hydrographic data is shown above the maximum draft of the ice because cross-fjord variability inice thickness allows water exchange at all depths within the fjord (inward of the epishelf lake ice shelf dam).The ice dam for the epishelf lake is assumed to be within the Outer Unit of the Milne Ice Shelf, so propertiesbeyond KM 37 are fixed constant to the offshore profile. Cross-fjord hydrographic sections extend from thenortheast shore (KM 0) to the southwest shore (KM 5.5) of the epishelf lake, looking toward the glaciergrounding line. Bathymetry (gray) is shown. CTD casts (black squares) were collected through drilled holesin the epishelf lake ice on foot over 16 hrs. The epishelf lake ice (not shown) is approximate 0.6 m thick.White areas indicate data gaps. Note that the meltwater fraction is only valid between the chosen PW andAW endpoint water masses at approximately 35 m depth and 150 m depth. The arrow in each panel indicatesthe mooring site and the intersections of the along- and cross-fjord transects.542626.4 26.827.227.627.827.827.9Depth (m)0100200300400500Depth (m)Distance  (km)45 40 35 30 25 20 15 10 5 00100200300400500242626.426.8 827.227.6 627.8Depth (m) 0100200300400Θ (o C)−1.5−1−0.50Distance (km)Depth (m)  64200100200300400Melt fraction (x10−3 )01234567A CB DFigure 2.8 Along- (left panels) and cross-fjord (right panels) hydrographic sections, as in Fig. 2.7, but fromJuly 2013. Along-fjord CTD profiles were collected by helicopter and foot over 72 hrs, the cross-fjord profilesby foot over 16 hrs.5525 30 20 30 40 50 60 Depth (m ) S A  (g kg −1 ) −1.5 −1 20 30 40 50 60 Θ ( o C) 180 190 200 0 50 100 150 Cum. PDD  ( o C day) DoY 2012 25 30 20 30 40 50 60 Depth (m ) S A  (g kg −1 ) −1.5 −1 20 30 40 50 60 Θ ( o C) 180 190 200 0 50 100 150 Cum. PDD  ( o C day) DoY 2013 Figure 2.9 Comparison of subglacial runoff in 2012 and 2013. Plotted are the cumulative positive degree-days (PDDs), a proxy for surface melt production, and changes in water column salinity and temperatureproperties over 2-weeks in July 2012 (top panels) and July 2013 (bottom panels). Profiles are coloured byelapsed time from initial profile each year and the time of profiling is marked on the PDD panel.56−0.0500.05020406080100OUT    IN2011Depth (m)−0.05 0 0.05020406080100SW    NE−0.0500.050204060801002012Depth (m)−0.05 0 0.05020406080100−0.0500.050204060801002013Depth (m)v (m s−1)−0.05 0 0.05020406080100u (m s−1)Figure 2.10Mean along- (v) and cross-fjord (u) current velocities at the mooring location in May 2011, July2012, and July 2013. Duration over which ADCP measurements were averaged varied among years from 4days in May 2011, to 6 days in July 2012, to 10 days in July 2013. Grey area indicates one standard deviationfrom the mean.57−40 0 20 Air Temp ( o C) 0 500 Solar Rad. (W m2 ) 300 Cum. PDD ( o C day) A B C 0 −1.5 −1. 5 − 0 .5 0 −0.5 0 −1. 5 −0.5 0 −1.5 −0.5 Depth (m ) 2012 2013 2014 50 100 150 200 250 300 D MEL FMW PW AW DW Temperature  ( o C) −1. 6 −1. 4 −1. 2 −1 −0. 8 −0. 6 −0. 4 −0. 2 0 Figure 2.11 Timeseries of A) air temperature, B) shortwave solar radiation, C) cumulative positive degreedays (PDDs), and D) ocean temperature at the mooring site from May 2011 to July 2014. In (D) the high-lighted isotherms approximate the interface between the epishelf lake (MEL), fjord-modified water (FMW),Polar Water (PW), Atlantic Water (AW), and deep water (DW). Filled black circles indicate thermistor depthsat each mooring redeployment. Unfilled circles indicate the sill depth where temperature was interpolated byassuming it was equivalent to temperatures at 320 m depth. White regions indicate data gaps during mooringservicing.582012 2013 2014−1.7−1.6−1.5−1.4−1.3−1.2−1.1−1−0.9−0.8DateTemperature (o C)  25m50m125mFigure 2.12 Variation in ocean temperature in Milne Fiord at 25 m, 50 m, and 125 m depths from May 2011to July 2014.592012 2013 20141.3351.33551.3361.33651.3371.33751.338x 1011Heat content (J m−2 )Date  Figure 2.13 Variation in ocean heat content of Milne Fiord between 5 m and 125 m depth from May 2011 toJuly 2014.60Distance from grounding line (km)Depth (m)  −250 −200 −150 −100 −50 0 50−300−200−1000100200300400500600700800Θ (o C)−1.5−1−0.500.5Figure 2.14 Hydrographic section of conservative temperature extending from the Milne Glacier groundingline across the continental shelf. Profiles offshore of the MIS were collected during the University of Wash-ingtons SwitchYard Project (PI Mike Steele) in May 2012 (see Fig. 2.1b for locations), while profiles withinMilne Fiord were collected in July 2012 during this study. An ice thickness profile along the centreline ofthe Milne Glacier from the grounding line to the summit is shown. Offshore bathymetry is from IBCAO v3,while bed topography and ice thickness upstream of the grounding line are from IceBridge 2014 data.61Chapter 3Dynamic response of the last remainingArctic epishelf lake to seasonal andlong-term forcing3.1 IntroductionIn bays at high latitudes, thick floating ice can create a physical barrier, or dam, behind whichmeltwater runoff from the surrounding glacial catchment accumulates. Where the dam is formedby an ice shelf (a floating sheet of ice attached to the coast) the impounded freshwater is knownas an epishelf lake. Epishelf lakes are freshwater bodies but are tidally influenced and can bedivided into two types depending on their connection to the ocean: those that form on land but havehydraulic connection to the ocean below the ice shelf at the land-ice interface; and those wherethe freshwater layer floats directly on seawater. The former are numerous in Antarctica and aredistributed around the margins of the continental ice sheet (Heywood, 1977; Gibson and Andersen,2002; Laybourn-Parry et al., 2006; Smith et al., 2006), while the latter, those that float directly onseawater, are not as common in Antarctica (Wand et al., 2011), but were once relatively numerousin the Arctic, specifically along the northern coast of Ellesmere Island, Canada. Here, remnantsof a once expansive coastal ice shelf (Vincent et al., 2001) blocked the mouths of fjords creatingseveral epishelf lakes (Veillette et al., 2008), the deepest of which was the Disraeli Fiord epishelflake dammed by the Ward Hunt Ice Shelf which had a maximum depth of 63 m (Crary, 1956).Over the last two decades, several of the Ellesmere ice shelves have experienced thinning, frac-turing, and collapse (Mueller et al., 2003; Copland et al., 2007; England et al., 2008; Vincent et al.,2009; White et al., 2015a) and this has resulted in the loss of all but one epishelf lake along thiscoast (Mueller et al., 2003; Veillette et al., 2008). The collapse of the ice shelves has been largelyattributed to climate forcing, as this region has warmed at twice the global average (IPCC, 2013).62The existence and evolution of epishelf lakes is dependent on the interaction between atmospheric,oceanographic, and cyrospheric conditions, and thus epishelf lakes have been identified as sensi-tive indicators of climate change (Veillette et al., 2008). They are also a unique type of aquaticecosystem, containing fresh, brackish and sea water biota within a single water column, with strongvertical gradients in biological as well as chemical properties (Laybourn-Parry et al., 2006; Veilletteet al., 2011a).Studies of Ellesmere epishelf lakes have shown that the transition from freshwater to seawateroccurs over only a few meters depth, and the strong halocline can persist for decades (Keys et al.,1968; Vincent et al., 2001), with perennial ice cover and low tidal action limiting vertical mixing.The depth of the halocline has been interpreted as equivalent to the minimum draft of the impound-ing ice shelf (Hattersley-Smith, 1973; Jeffries and Krouse, 1984; Gibson and Andersen, 2002), andinterannual depth changes have been used to infer long-term changes in thickness of the ice shelf(Vincent et al., 2001; Mueller et al., 2003; Veillette et al., 2008). Questions remain however, as towhat extent changes in the thickness of the dammed freshwater layer are representative of changesin the mean thickness of an ice shelf (Vincent et al., 2001; Smith et al., 2006).Prior to its complete drainage between 1999 and 2002 due to fracturing of the Ward Hunt IceShelf, the Disraeli Fiord epishelf lake had been thinning since 1954, at a rate roughly correspond-ing to overall mass loss from the Ward Hunt Ice Shelf (Veillette et al., 2008). Other epishelf andice-dammed lakes in the region have shown similar long-term thinning, however, with interannualvariability on the order of a meter (Veillette et al., 2008). This variability has been suggested to berelated to tidal cycles, internal waves, and fjord circulation (Veillette et al., 2008). However, the onestudy that addressed some of these factors found that internal waves excited by tidal flow were verysmall in amplitude (<20 cm; Keys et al., 1968). Despite lingering questions over the dynamics ofepishelf lakes, observations have largely been limited to single water column profiles collected onan opportunistic basis every few years. A targeted study of epishelf lake physical limnology has notbeen conducted since a campaign to study the Disraeli Fiord epishelf lake in the 1960s (Keys et al.,1968; Keys, 1978), prior to the recent phase of widespread decline and loss of epishelf lakes alongthis coast.The objectives of this study were to evaluate the physical structure and temporal dynamics ofthe Milne Fiord epishelf lake, the last known epishelf lake in the Arctic. We investigated interan-nual changes in epishelf lake depth, area, and volume over 10 years from analysis of water columnprofiles and remote sensing imagery. To better understand how the lake changes on sub-seasonaltimescales we conducted a targeted 3-year field study that involved periods of intensive water col-umn profiling, collection of tide heights and current velocities, and the collection of a continuousmulti-year mooring time series of epishelf lake temperature and salinity, as well as meteorologicaldata. We measured the evolution of the thermal and salt stratification of the lake and its spatialand temporal variation, with the aim to understand the controlling factors for lake depth and the63implication of recent changes for the future of this ecosystem.3.2 Methods3.2.1 Study siteThe 436 m deep Milne Fiord (8235N, 8035W) lies on the northern coast of Ellesmere Islandadjacent to the Arctic Ocean (Fig. 3.1). The Milne Glacier flows into the head of the fjord andforms the tapered 16 km long Milne Glacier tongue (MGT), varying from 150 m thickness at thegrounding line to 10 m at the terminus (Chapter 2). Spanning the mouth of the fjord, but separatefrom the MGT, is the Milne Ice Shelf (MIS), which varies in thickness from <10 m to 100 m. TheMIS originated from a combination of tributary glacier input, in situ snow accumulation and marineice accretion, and is readily distinguished from other ice types by the undulating surface marked bya series of parallel troughs and ridges.The MIS forms a floating dam across the mouth of the fjord, trapping surface runoff within thefjord. The first water column samples in the fjord were collected in 1983 through 3.19 m of ice,and revealed an ⇠17.5 m deep freshwater layer separated from seawater by a sharp halocline onlya few meters thick (Jeffries, 1985). Although water profiles are lacking prior to this time, based onaerial photographs and aerial radar measurements of thin (<9.7 m thick) ice in the inner fjord in1981 (Narod et al., 1988), suggests the epishelf lake was present prior to 1983. The next samplingconducted in 2004 showed the freshwater layer had deepened to 18.3 m, but then thinned to 14.3 mby 2009 (Veillette et al., 2011b).In examining changes in the area of the MIS, Mortimer et al. (2012) suggested that prior to1959 the MIS and MGT were connected, and over the next few decades the epishelf lake developedfrom small ice-marginal lakes along the side of the fjord to eventually replace the Inner Unit ofthe MIS by 2009, severing the connection between the MIS and MGT. Mortimer (2011) estimatedthe area of the epishelf lake based on identification of lake ice in satellite imagery as 52.5 km2 in2009, although the actual extent of the lake had not been confirmed with field observations. Thelake appeared to consist of a 6 km wide main basin between the inner edge of the MIS and theterminus of the MGT, and two narrow arms extending 16 km along the sides of the MGT to thegrounding line (Fig. 3.1d). The otherwise steep walls of the fjord are punctuated by three shallowbays with inflowing streams at their heads. The lake is fed by snow and glacial runoff from the⇠1500 km2 catchment of the Milne Glacier and its tributaries (GLIMS and NSIDC, 2005). Icethickness mapping of the MIS (Mortimer et al., 2012; Chapter 2) indicates the only ice thin enoughto provide an outlet to the lake is along the re-healed fractures in the MIS, however, the drainagepathway of the lake has not been confirmed by observations.643.2.2 Area and volumeChanges in the area of the MEL from 1959 to 1988 were estimated from optical imagery acquiredby aerial and satellite platforms, and from 1992 onward from Synthetic Aperature Radar (SAR)imagery. The 1 m thick epishelf lake ice was discriminated from other surrounding ice types, in-cluding ice shelf, glacier ice and marine ice, in optical imagery by its lack of surface topography, andin SAR imagery by its high backscatter signal (>-6 dB), produced by its lack of surface topographyand the freshwater underneath perennial lake ice (regions underlain by salt water have a darker re-turn; Veillette et al., 2008; White et al., 2015a). The epishelf lake included non-contiguous regionsof lake ice in fractures in the MIS and between calved pieces of the MIS and MGT. The epishelflake was digitized in ArcGIS 10.2.2 at an image scale of ⇠1:20,000, with a pixel size of 6.5 m.Lake volume was estimated from area and depth, assuming a spatially uniform depth (see Sec-tion 3.3.8) and vertical shores. The volume estimated included the volume of surface ice. The depthwas taken from the hydrographic profiles collected the May following image acquisition (usuallyin Feb or March) when available, otherwise from the profile collected closest to the date of imageacquisition.3.2.3 HydrographyNear-annual sampling of water properties in Milne Fiord commenced in 2004, with a directed andintensive sampling program from 2011 to 2014. Water properties were measured through drilledholes or natural leads in the ice, including fractures through the ice shelf and glacier tongue, ac-cessed by foot, snowmobile, or helicopter. Where possible, lake ice thickness was measured andvaried from a maximum of 3.19 m in 1983 (Jeffries, 1985) to a minimum of 0.65 m in July 2010.Profiles of temperature and salinity were collected using a conductivity-temperature-depth (CTD)profiler lowered at ⇠0.5 m s1 using a manual reel, with the exception of the May 1983 profile col-lected by Jeffries (1985) that was measured using reversing thermometers, a 1 L Knudsen bottle anda Radiometer CDM80 Conductivity Meter. Opportunistic profiles were collected in August 2004,June 2006, and July 2007 using a 1 Hz RBR XR-420 CTD. Subsequent profiles were collected inMay & July of 2009, July 2010, May & July 2011, 2012, and 2013, and July 2014 with a 6 Hz RBRXR-620 CTD. Profiles from May 2011 were collected using a 4 Hz Seabird SBE19+ CTD, and inJuly 2011 using a 1 Hz Hydrolab HLX. CTDs were calibrated once every 2 years after 2011, priorto this the CTDs were not regularly calibrated so we interpret data prior to 2011 with this caveaton its absolute accuracy. Profiles collected between 2004 and 2009 were previously published inVeillette et al. (2011b), although we have reprocessed all these data from raw conductivity and tem-perature (where available) for consistency. Prior to 2011, profiles were collected in either PurpleValley Bay or Neige Bay (unofficial names), two small inlets on the west and east sides of the fjord,respectively, with water depths of⇠80 m. Hydrographic data suggests Neige Bay has a topographicsill at⇠30 m depth so water properties below this level in the bay are not representative of the fjord.65An exception is the 2004 profile, which was located approximately 2 km southwest of Neige Baynear the MGT terminus.From 2011 onward multiple profiles were collected during each field campaign throughout themain fjord. Full depth profiles were usually collected to the bottom of the fjord, however here wefocus on the upper 25 m of the water column. CTD data were processed in Matlab following a pro-cedure that included: correction for atmospheric pressure, application of a 3 point low-pass filter intime to the raw pressure, temperature, and conductivity; alignment of conductivity and temperaturewith respect to pressure; a thermal cell mass correction (for the SBE19+ CTD data only); and loopediting (removal of pressure reversals), and bin averaged to 0.2 m intervals. Derived variables werecalculated using the International Thermodynamic Equation of Seawater 2010 (TEOS-10) GibbsSeawater Oceanographic Matlab Toolbox (www.TEOS-10.org). Temperature and salinity are re-ported here as Conservative Temperature (Q) and Absolute Salinity (SA), with the exception of the1983 data, which are presented as originally published as practical salinity (ppt) and in situ temper-ature.For freshwater lakes, the calculation of salinity from temperature, conductivity, and pressureis dependent on the chemical composition of the water, data which we lack. If the chemical com-position of the lake varies substantially from seawater, the error in calculated salinity could be asmuch as 30% for salinities <3 g kg1 (Pawlowicz, 2008). The source of the epishelf lake water isprimarily snow and meteoric glacial ice melt, with low ionic concentrations (and conductivity). Themeasured conductivity of two proglacial meltwater streams entering the fjord in 2012 and 2013 were<0.06 mS cm1, while the conductivity of the lake was generally >0.15 mS cm1. The obvioussource of salt required to increase the conductivity of the lake water is entrainment of underlyingseawater, meaning the bulk chemistry of the lake should reflect that of diluted seawater. The errorin calculated salinity is therefore likely <<30%.In order to compare changes in the thickness of the freshwater layer over time it was necessaryto define the bottom of the epishelf lake, which was actually a continuum from freshwater to seawa-ter. Previous studies delineated the bottom of an epishelf lake using the depth of the 3 ppt isohaline(Mueller et al., 2003; Veillette et al., 2008), or the depth of the halocline (which the authors qualita-tively define as the zone of abrupt salinity change between freshwater and sea water; Veillette et al.,2011b). We formalized the definition of Veillette et al. (2011b) by defining the bottom of the lake(DEL) as the depth of the stratification maximum as defined by the Brunt Vaisala frequency:DEL = z(N2max), (3.1)where z is depth (positive downward) and N2 is the Brunt Vaisala frequency:N2 =gr∂r∂ z, (3.2)66where g is gravitational acceleration, and r is density of the water. Profiles of N2 were averagedusing a 12-point depth window before calculating the maximum. Water density is dominated bysalinity at these low temperatures, so DEL represents the maximum salinity gradient. The advantageof this method is that the epishelf lake depth calculation is clearly defined, quantitative, and appli-cable to other epishelf systems (regardless of the absolute salt content of the epishelf lake, whichwould affect the 3 ppt method). Due to the bottle sampling method used by Jeffries (1985) we couldnot calculate N2 for that profile, so we defined the bottom of the lake as the depth of the samplecollected nearest the apparent stratification maximum. The depth of the primary halocline varied by± 0.15 m in a series of 18 profiles collected at a single location over a 24 hr period in May 2009,so this was considered to be the error in the depth of the epishelf lake as determined by the CTDprofiling method.3.2.4 Current velocitiesWater velocities of the upper water column were measured using an ice-anchored, downward-looking 300 kHz RDI acoustic Doppler current profiler (ADCP) at the mooring site over 4 daysin May 2011, 7 days in July 2012 and 10 days in July 2013. The ADCP sampled at 2 min intervalsand data were processed in Matlab.3.2.5 Tidal heightA bottom anchored RBR XR-620 CTD was deployed at 355 m depth from May 2011 to July 2012sampling at 2 minute intervals to measure changes in water level (accuracy is± 0.37 dbar, drift 0.74dbar a1). A 3 hr low pass-filter and a linear detrend was applied to the raw pressure data. Thedominant tidal constituents were determined by harmonic analysis with nodal corrections using theT Tide Matlab toolbox (Pawlowicz et al., 2002).3.2.6 Meteorological time seriesMeteorological conditions were measured by a HOBO automated weather station (AWS) locatedat 10 m elevation on the shore of Milne Fiord (Fig. 3.1 a). Only air temperature (at 1 m and2 m above ground) and shortwave solar radiation are reported here. Cumulative positive-degreedays (PDDs), the daily integrated air temperatures above 0C, were calculated to provide a directproxy for summer melting, which has been directly linked to air temperature (Hock, 2003). Prior tothe AWS installation in 2009, we estimated summer air temperatures in Milne Fiord from recordsfrom Eureka, Nunavut (www.ec.gc.ca). Linear regression showed that air temperature in MilneFiord TMilne, was related to air temperature in Eureka by TMilne = 0.47⇤TEureka+0.48. TMilne. Thepredicted temperatures were then used to calculate PDDs prior to 2009. The interpolation was onlyvalid for TEureka 0C, and had a root-mean-squared error of 1.98C.673.2.7 Mooring time seriesMilne Fiord water properties were measured fromMay 2011 to July 2014 by a mooring deployed inthe centre of the epishelf lake. The mooring was anchored to the epishelf lake ice and suspended inthe water column. The mooring consisted of 20 RBR TR1050/60 , 2 RBR XR420-freshwater CTs,2 Seabird SBE37 CTs, and 1 RBR XR620 CTD from May 2011 to July 2012, then was reduced to7 TR1060s, 1 XR420 CT, and 1 XR620 CTD for the remainder of the study. Calibrated instrumentaccuracy is ±0.002C, ± 0.003 mS cm1, and ± 0.37 dbar, and nominal drift is ±0.002C a1,± 0.012 mS cm1a1, and ± 0.7 dbar a1. Some time series records were truncated for variousreasons, including salinity going beyond the maximum calibrated range (3 PSU) of the freshwaterinstruments (XR420s), or instrument malfunction. The mooring was serviced once or twice peryear and instruments were repositioned to track the halocline. Initially, the instruments were spacedevery meter from the surface to 20 m depth, with increasing depth intervals below 25 m. Althoughthe mooring instruments extended to the full depth of the fjord, in this paper we focus on the top25 m of the water column. Instruments sampled at 30 to 120 second intervals and were calibratedbefore and after deployment. CTD profiles collected during deployment and recovery were used tocorrect for instrument drift, which was within manufacturer specifications.Seasonal changes in the depth of the epishelf lake were estimated from an inferred verticaldisplacement of the halocline from the mooring time series, where salinity changes at a fixed depthover time are projected onto the initial vertical salinity profile. The accuracy of this method isgreatest where dS/dz is highest (i.e. in the halocline), while the error increases substantially wherethe salinity gradient is weaker (i.e. above or below the halocline). This constraint on the method isacceptable as the lake bottom is defined above as the level of the stratification maximum, so onlythe instrument initially positioned within the halocline (instrument at 13 m depth from May 2011to May 2012) is included in the analysis. The method assumes a constant salinity gradient that isdisplaced downward by inflow (surface runoff) and displaced upward by outflow (drainage underthe MIS); it neglects the effects of horizontal advection of freshwater (from subglacial runoff) andvertical mixing processes that alter the salinity gradient.We also estimated epishelf lake depth using temperature data from the temperature recorders,which were much more closely spaced in depth. Lake depth was estimated from the depth ofthe isotherm corresponding to the average temperature at the depth of the N2max measured by theCTD profile at the beginning and end of the mooring deployment. The process was repeated eachtime the mooring was serviced and CTD profiles were collected. This method assumed isothermdisplacement was due to displacement of the halocline, however we acknowledge other processescould have altered the depth of the isotherms that were not related to a depth change of the halocline,such as in situ heating due to solar radiation, horizontal advection of heat, and vertical heat fluxacross the halocline. However, the results of the isotherm proxy showed good agreement with thesalinity proxy and the CTD estimates of epishelf lake depth, providing confidence in the method.683.3 Results3.3.1 AreaThe area of the lake increased substantially since the first aerial imagery in 1959, reaching 71.2 km2as of 27 March 2015 (Fig. 3.1; Table 3.1). The largest change in area occurred sometime between1959 and 1988, as the epishelf lake replaced the Inner Unit of the MIS. Since 1988 the lake existedin close to its present form, with a main fjord-wide basin and two arms extending along the sides ofthe MGT. Increases in lake area after 1992 were due the retreat of the southern margin of the MIS,including calving of the MIS into the fjord, the creation of small satellite lakes in fractures of theMIS, and wastage along the margins of the MGT. Where available, the satellite lakes show watercolumn structure that is nearly identical to that of the main basin of the MEL (see Section 3.3.8),suggesting the satellite lakes are connected to the main basin by a network of small fractures orbasal crevasses that allow water exchange, so we included them in the area estimates. Some gainsin area were partially offset by losses due to the advance of the terminus of the MGT, which rangedfrom 56.4 m a1 to 173.2 m a1 between 1950 and 2009 (Mortimer 2011).3.3.2 StratificationThe most striking feature of all water column profiles collected in Milne Fiord was the presence ofa several meter thick freshwater (defined here as salinities <0.5 g kg1) layer at the surface, theepishelf lake, which was never present offshore of the MIS. The epishelf lake was a conspicuousfeature of the first profile obtained in the fjord in 1983 (Jeffries, 1985), but it has clearly thinnedthrough time. Despite changes in the depth of the lake, several distinct layers in the epishelf lakeand upper water column could be identified based on salinity and temperature characteristics. Insummer, a 1-2 m thick stratified layer with salinity approaching zero and temperature approachingthe freshwater freezing point (0C) was present just below the ice-water interface. We termed thisthe surface melt layer. Below the thin surface melt layer was a layer of nearly constant salinity(approximately 0.2 g kg1), the mixed layer, extending from the base of the surface melt layer tothe top of the halocline. The mixed layer was up to 8 m thick, however it was not present in allyears, and usually only evident in summer, at other times the lake was weakly salinity stratified.The salinity gradient below the mixed layer could be divided into an upper halocline and a lowerhalocline. The upper halocline is the transition from the base of the mixed layer to the bottom ofthe lake (i.e. the N2max), while the lower halocline is defined as extending below the N2max to the levelat which properties within the fjord are equivalent to those at the same depth offshore (between 25- 50 m). A subsurface temperature maximum (up to 3C) was usually associated with the upperhalocline. In some profiles the mixed layer was not present, and the upper halocline extended to thebase of the surface ice melt layer (if present) or to the ice-water interface. The gradient and thickness69of the upper halocline varied among years, with a thicker and more gradual salinity gradient apparentprior to 2009 (e.g. in 2004 the upper halocline was 15 m thick and extended almost to the surface),while after 2009 the salinity gradient was thin and sharp (e.g. in June 2012 the upper halocline was<3 m thick). Temperatures in the lower halocline decreased rapidly with depth toward the freezingpoint of seawater (-1.8C). The gradient and properties of the lower halocline were dependent onlocal fjord processes, including interactions with ice and advection of subsurface glacial meltwaterrunoff, so the lower halocline was also referred to as fjord-modified water (Chapter 2).The stratification at the base of the epishelf lake is very strong, in all years the N2max is >0.1s2 (or 103 cycles per hour). For comparison, typical buoyancy frequencies in the open ocean aregenerally<20 cycles per hour, so the stratification in the epishelf lake halocline is one to two ordersof magnitude stronger by this measure. Density in the halocline is determined primarily by salinity,changes in temperature are only important in determining stratification in the isohaline mixed layer.For example, for water at 0.2 g kg1 and 2C, to change the density by 0.1 g kg1 requires a 4Cchange in temperature, but only a 0.15 g kg1 change in salinity. As a result, density profiles (notshown) are nearly identical to the salinity profile. It is important to note that temperatures in theepishelf lake are everywhere below the temperature of maximum density, so a temperature increaseresults in a density increase.3.3.3 Current velocitiesADCP deployments revealed a quiescent system with currents <2 cm s1 in the upper 25 m of thewater column (not shown). The currents were weakly baroclinic, with velocities near zero in theepishelf lake above the level of the halocline, increasing to 1-2 cm s1 just below the halocline. Theresults support the view that the MIS forms a barrier to flow above the halocline, and acts as a dampreventing offshore exchange at the surface. The potential for velocity shear stress to generate ver-tical mixing in the water column can be determined by calculating the gradient Richardson Number,a ratio of stratification to velocity shear:Ri=N2⇣∂u∂ z⌘2 (3.3)where u is horizontal velocity (m s1) and z is depth (m; positive z down). During all three periods ofobservation Ri >>1 across the halocline, indicating that stabilizing buoyancy forces dominate andturbulent mixing is not expected. This finding helps explain the persistence of strong stratificationin the fjord, as the transfer of heat and mass across the pycnocline is likely limited to moleculardiffusion. In the mixed layer of the epishelf lake, however, stratification is much weaker and Ri <1,indicating shear induced turbulent mixing is possible in this layer.703.3.4 Lake depth: interannual variationThe salinity profiles show a clear long-term thinning of the epishelf lake, from a maximum depthof 18.3 m in 2004 to a minimum of 8.0 m depth in 2013 (Fig. 3.2); Table 3.1). There was littlechange in the depth of the epishelf lake between 1983 and 2004, however it is unknown how muchvariation occurred in the intervening 20 years. Between 2004 and 2014 the epishelf lake thinned atan average rate of 1.08 m a1, however, the rate and direction of change was not constant. An abruptthinning occurred between 2011 and 2012, when the lake depth decreased by almost 4 m, quadruplethe decadal average thinning rate. Conversely, the lake appeared to increase in depth between someyears (e.g. between 2006-2007, 2009-2010, and 2013-2014), with a maximum increase of 1.3 mbetween 2013 and 2014. However, we argue below that the apparent increase in lake depth on aninterannual basis is an artifact due to differences in the timing of profiling relative to the melt season.3.3.5 Lake depth: seasonal variationCTD profiles collected in Milne Fiord within the same year reveal a pronounced seasonal increasein the depth of the lake (Fig. 3.3). In 2012, the halocline deepened over the summer by 1.9 m,with a 0.7 m increase between May 5th and June 28th, then a further 1.2 m progressive increaseover the following 11 days until July 9th. In 2013, however, the lake depth changed very littlebetween May 10th and July 5th, with a small (<0.5 m) depth increase apparent over the subsequent2-weeks of profiling in July. Changes in depth of the epishelf lake are significantly correlated withthe cumulative PDDs (n = 226, R2 = 92%, p = 0.005), with a ratio of 1.7 cm C1 day1 (Fig. 3.4),which are a proxy for the volume of surface meltwater production from the glacier catchment. Thesummer of 2012 was very warm, 50 PDDs were accumulated between May and June, and increasedto a total of 93 PDDs by the time the final profile was collected that year. The summer of 2013,however, was quite cool, only 10 PDDs had accumulated between May and July, and a total of only38 PDDs accumulated by the time the final profile was collected that year (despite the final 2013profile being collected almost 2 weeks later in the year than the final profile of 2012). Neither fieldcampaign spanned the duration of the entire summer melt season. For example, in 2012, profilingended July 9th when 92 PDDs had accumulated, yet a seasonal total of 253 PDDs accumulated byAugust 15, so the lake likely deepened substantially more by the end of the melt season which wasnot captured in the profiles (see Section 3.3.12).3.3.6 TidesHarmonic analysis of the water level record reveals the tide in Milne Fiord is mixed, predominatelysemidiurnal, with a range of 0.31 m. Amplitudes of the dominant M2 (12.42 h) and K1 (23.93 h)tidal constituents are 0.056 m and 0.040 m, respectively, which together account for 86.7% of thevariance of the original water level data. The low tidal energy available for mixing in Milne Fiord71was likely to be a factor in the long-term persistence of the epishelf lake halocline.3.3.7 Lake depth: internal wavesIn June 2011, during the 2-weeks prior to the onset of the melt season, the salinity at 13 m depthshowed ⇠2 g kg1 fluctuations in salinity (Fig. 3.6), equivalent to a vertical displacement of thehalocline of ⇠15 cm, that we associate with the passage of internal waves. This suggests that mostof the error in synoptic CTD depth estimates is due to vertical displacement of the halocline by thepassage of internal waves.Spectral analysis of the salinity time series (not shown) reveals energy peaks at diurnal andsemi-diurnal tidal periods in the halocline, with a strong non-tidal peak at 48 min, and secondarypeaks at 70 min and 5.7 hrs. Based on the ratios of the amplitude of the basin-scale wave (⇠15 cm)to the thickness of the epishelf lake (⇠10 m) and the depth ratio of the epishelf lake to the full depthof the fjord (⇠400 m), the epishelf lake falls within a regime where only linear waves are expected(Horn et al., 2001). Application of linear wave theory allows calculation of the nth mode internalseiche period for a two-layer fluid as:T =2Lnqg0h1h2(h1+h2), (3.4)where n is the wave mode, L is the average width of the basin (m) at the depth of the halocline h1(13 m), overlying a layer of depth h2 (440 m), and g0 is reduced gravity:g0 = gr2r1r(3.5)where r1 is the density of the upper layer (1000 kg m3), r2 the density of the lower layer (1025 kgm3), and r the average density of the two layers. A range of values could be chosen for L given thecomplex geometry of the lake, although the width of the main basin, 5.8 km, is an obvious startingpoint. Taking this distance as the length scale L gives a first mode internal seiche period of 129min and a second mode period of ⇠54 min, which is very similar to the 48 min periodicity of thestrongest non-tidal salinity signal. Alternatively, using the long axis of the epishelf lake, ⇠20 kmfrom the grounding line to the inner edge of the MIS, results in a first mode internal seiche periodup to ⇠8 hrs, suggesting a range of internal seiche periods are plausible.These observations are similar to measurements in Disraeli Fiord, where small 17 cm ampli-tude internal waves propagating along the sharp epishelf lake halocline, with periods ranging from⇠6 min to ⇠6 hrs (Keys, 1977). The driving force for internal waves in epishelf lakes is almostcertainly the tidal current, although other possible sources are kinetic energy from inflowing melt-water streams (both surface and subglacial), atmospheric pressure changes, and offshore baroclinicfluctuations propagating into the fjord. The relatively low energy of the background internal wave72field in Milne Fiord, suggests their role is limited to inducing a small amount of mixing across thehalocline at the lake boundaries, and for our purposes, they are largely of concern for the error theyimpart to lake depth estimates derived from CTD profiles.Temperature fluctuations also occurred above the halocline, confined to the freshwater layer ofepishelf lake, and were indicative of long-period internal waves, perhaps excited by the spring-neaptidal processes. An in-depth analysis of these waves, however, is reserved for a separate publicationbecause the focus here is on changes to the depth of the halocline, and how this may influence theinterpretation of lake depth based on synoptic CTD measurements.3.3.8 Spatial extentSynoptic CTD profiles collected at distant locations over a short time period (i.e. the same colourin Fig. 3.3) show the MEL is spatially uniform in depth (the level of the halocline varies <20 cmbetween profiles >20 km apart) and extends throughout the fjord. For example, the depth of theN2max along a 5.8 km transect across the width of the fjord, from 6 profiles collected in a single 24 hrperiod in July 2012 (green lines, Fig. 3.3a), varied by only ± 5 cm, well within instrument error.While the depth of the N2max varied by only ± 10 cm along a 23 km transect extending down thelength of the fjord, from near the grounding line of the Milne Glacier to a fracture in the Central Unitof the MIS, collected by helicopter on 29 June 2012 (dark blue lines, Fig. 3.3a). That the epishelflake was present even in fractures in the MGT and MIS suggests a network of basal channels permitexchange of surface waters throughout much of the fjord. Each of the profiling locations where theepishelf lake was present mapped onto regions identified as epishelf lake in SAR imagery, providingverification of the remote sensing method used to map the area of the lake. However, remote sensingcannot determine how far under the MIS the lake extends, nor be used to determine the location ofthe ice dam. To address these questions, field observations are required.Our CTD profiling indicates that the ice dam must be located somewhere in the Outer Unit ofthe MIS, as the epishelf lake was observed in all profiles collected through fractures in the CentralUnit of the MIS, but was not present in any profile collected offshore of the seaward edge of the MIS(as expected). Ice thickness maps of the MIS (Mortimer et al., 2012; Chapter 2) indicate the onlyice of the Outer Unit of the MIS thin enough to allow drainage of the epishelf lake lies along thetwo re-healed fractures (Fig. 3.5). Several attempts in 2012, 2013, and 2014 to profile through there-healed fractures, to constrain the location of the ice dam and confirm the drainage pathway of theMEL, were unsuccessful. Investigation into the location of the ice dam and the drainage pathwayof the MEL is ongoing.3.3.9 Lake volumeThe observed spatial uniformity of the depth of the lake (at a moment in time) provides an easymeans to estimate the volume of the lake when both area and depth are known (Table 3.1). The73earliest reliable estimate of the lake volume is 1.04 km3 in 2006, after which the volume decreasedsubstantially, reaching a minimum of 0.54 km3 in 2013. The decrease in volume is largely due tothe decrease in thickness of the lake; the area of the lake varied by only 10% between 2006 and2014, while the depth varied by over 50% during this period. This observation is evidence thatinterannual depth changes of the lake are related to changes in the thickness of the ice dam, ratherthan area changes. We note that the surface ice thickness decreased from a maximum of 3.19 m in1983 (Jeffries, 1985) to a minimum of 0.65 m in July 2010, varying annually by ⇠1 m thereafter,however, changes in surface ice thickness do not affect estimated volumes as the surface ice is inhydrostatic equilibrium.3.3.10 Meteorological time seriesThe meteorological data recorded by the automated weather station in Purple Valley for the durationof the mooring deployment are shown in Figure 3.6. Average air temperature during the mooringdeployment (May 2011 to July 2014) was -17.6C, with extreme hourly temperatures of +20.2Cin July 2012 and -51.8C in February 2013. Solar radiation varies from zero during the polar night(mid-October through February) to a summer daily maximum of⇠650Wm2 in late June. The sitereceives 24 hrs of sunlight from April to September. Summer melt conditions varied substantiallyamong years, with cumulative positive degree days of 278, 253, 92 and 110 in 2011, 2012, 2013,2014, respectively (note that for display purposes the meteorological record is truncated to matchthe mooring record in Fig. 3.6, but we collected meteorological data through to the end of 2014).The melt season occurs between early-June and mid-August, although the onset and duration varyby up to 2-weeks among years.3.3.11 Salinity time seriesThe most striking feature of the salinity time series (Fig. 3.6d) is the seasonal change in salinity ofinstruments deployed in the halocline (instruments at 13 m and 15 m in 2011-2012, and 10 m in2012-2013; no instruments were positioned in the halocline in 2013-2014). The substantial fresh-ening at these depths is consistent with a deepening of the halocline from mid-June to mid-Augustdue to surface meltwater inflow to the epishelf lake. The rebound of salinities commencing in mid-August of each year suggests the halocline shoals again after meltwater inflows cease. Focusing onthe instrument with the longest continual record (13 m depth), shows the instrument was initiallydeployed in the strong halocline (at 8 g kg1 in May 2011). Salinity was relatively stable beforea suddenly decrease that coincided with the commencement of the melt season. By the end of themelt season, salinity at 13 m depth had decreased to 0.3 g kg1, suggesting the instrument wasthen in the mixed layer of the epishelf lake (salinity at 8 m, 11 m, and 13 m overlap, indicatingan unstratified water column). After the melt season ceased, salinity increased to >22 g kg1 byJanuary 2012, overlapping with the salinity of the lower halocline, consistent with thinning of the74epishelf lake. The sudden drop in salinity in January 2012 is a unique feature, and is examined inmore detail in Section 3.3.13. Instruments spaced 2 m apart vertically showed a temporal offset inthe seasonal evolution of salinity with depth (on the order of 5-weeks), which is further evidencethat the halocline progressively deepened over summer, then gradually shoaled over winter. In themixed layer, salinity increased over time, the instrument moored at 5 m depth showed an increaseof 0.2 g kg1 from August 2012 to July 2014, and the 7 m instrument showed an increase of 0.1 gkg1 from 2013 to 2014.3.3.12 Temperature time seriesThe temperature time series (Fig. 3.6e) provides a comprehensive view of the thermal evolution ofthe epishelf lake over the 3-year deployment period. The most readily apparent features of the timeseries are that the temperature of the epishelf lake and the depth of the thermocline vary substantiallyon seasonal and interannual timescales. The lake warms from mid-June to mid-August each year,although the magnitude varies between years, reaching a subsurface maximum of 2.5C, 4.0C, and2.5C, in 2011, 2012 and 2013, respectively. The strongest warming was observed in summer of2012 when water temperatures between 5 m and 10 m depth were almost isothermal above 3C.After peaking in mid-August, the temperatures of the epishelf lake gradually decreased until thefollowing summer, although temperatures remained significantly above freezing all year despite theextreme low air temperatures in mid-winter. The strong thermocline at the base of the epishelf lake,which corresponds to the epishelf lake halocline, varies by several meters each year. The deepeningof the epishelf lake commenced between 2-14 June of each year, corresponding to the when airtemperatures increased above freezing, and thus the commencement of surface meltwater inflow.The maximum depth of the thermocline occurred at the end of the melt season each year. Thethermocline deepened by 3.0 m, 3.3 m, and 1.0 m in summer of 2011, 2012, 2013, respectively.After the melt season ended the lake gradually thinned each winter, reaching a minimum inearly June the following year, just prior to the commencement of the subsequent melt season. On aninterannual basis, the depth of the thermocline, measured on June 1st each year, shoaled by 4.1 mbetween 2011-2012, 1.5 m between 2012-2013, and 1 m between 2013-2014. This interannualthinning of the lake, despite increases in lake thickness of the same magnitude in summer, suggeststhe MIS is in a state of negative mass balance, and the ice dam is thinning. However, the rate of lakethickness change over time is not constant, as apparent by an abrupt shoaling of the thermocline thatoccurred in January of 2012, which is examined in more detail in Section January 2012 eventA sudden change in temperature and salinity at the bottom of the epishelf lake occurred on 11January 2012 06:00 UTC (Fig. 3.7). Over a duration of 18 hrs the salinity at 13 m depth droppedfrom 22 to 12 g kg1, and remained below 15 g kg1 for the remainder of the winter. At the same75time, the heat content of the upper 25 m of the water column was relatively steady (apart from somefluctuations during the actual event), the slow rate of heat loss was not substantially different fromthe long-term average over winter. During the event the upper portion of the thermocline (above11 m depth) was displaced upwards 1.5 m, while isotherms in the lower portion of the thermocline(below 11 m depth) spread apart vertically. The event was associated with rapid vertical fluctuationsof isotherms of much larger amplitude then those observed in the days prior to the event. Isothermfluctuation associated with the event were recorded from the uppermost thermistor (at 2 m depth)down to at least 50 m depth (not shown), and possibly deeper. Profiles collected during the fieldcampaigns months before and after the event show a marked change in the depth of the epishelflake and the gradient of the lower halocline/thermocline (Fig. 3.7 c & d). These observationsindicate that a sudden mixing event occurred at the bottom of the epishelf lake that entrained warm,relatively fresh water from the lake downward, and cool, salty water upward. The result was anabrupt, irreversible, thinning of the lake by 1.5 m and a change in the gradient of the lower halocline.The negligible change in the heat content of the upper water column implies a conservation of heat,indicating the observations show either a) an in situ vertical mixing event, presumably uniformthroughout the lake, or b) the advection of a water masses with the same overall heat content, or c)some combination of the two. The latter could probably only be achieved if the advected water massoriginated within the fjord where the water column properties were similar (e.g. if vertical mixingoccurred in the, presumably spatially uniform, lake at some distance from the mooring site andthe resulting mixed water mass, with the same overall heat content, was advected past the mooringlocation).3.3.14 Long-term lake depthThe long-term record of lake depth over the past decade is shown in Fig. 3.8. Plotted is the depth ofthe lake measured by synoptic CTD profiling, as well as that inferred from the continuous mooringrecords of temperature and salinitiy between 2011 and 2014. Also plotted are the PDD-correctedlake depths, which reveal that during the period 2004 to 2011 the lake thinned at an average rate of0.51 m a1. The PDD-corrected depths also show that the lake steadily thinned almost every yearduring this period, contrary to the apparent deepening in some years as indicated by the uncorrecteddepths. The steady rate of thinning was interrupted in 2011-2012, when the lake thinned by 4.1m, in part due to the mixing event in January 2012. From 2012 to 2014 the lake thinned at a rate0.34 m a1, although the latter rate is based on only 3 years of CTD profiles, too short a record toconclude the rate has changed from the period prior to 2011.Continuous observations from 2011 to 2014 reveal that the depth of the lake fluctuated by severalmeters over a year. Such seasonal changes in the depth of an epishelf lake have not previouslybeen reported, and highlights the importance of continuous records to understand the timescale ofvariation in a system, so as not to alias long-term trends from annual sampling programs. The76sudden decrease in the thickness of the epishelf lake in January 2012 highlights the importance ofepisodic events to lake dynamics.3.4 DiscussionOur compilation of historical data along with new observations shows that the Milne Fiord epishelflake is a longstanding feature of the northern coast of Ellesmere Island, but with large changes inits area and depth over the last three decades. The lake has thinned over the long-term, at a rate thatis likely indicative of thinning of the Milne Ice Shelf. At shorter time intervals, our observationsshow the lake is seasonally dynamic, increasing and decreasing in depth by several meters a year.We have also shown lake is subject to abrupt mixing events that can dramatically shift the haloclineover very short time periods.3.4.1 Area expansionThe Milne Fiord epishelf lake was first observed in 1983, although it is likely that the lake existedbelow the Inner Unit (which was 9.7 m thick in 1981) prior to this time (Mortimer et al., 2012).Although the timing of the origin of the lake is unknown, it appears to have expanded in the decadesafter 1959 as the Inner Unit of the MIS transitioned to lake ice. Since 1988 changes in the area ofthe lake have been due to a balance between expansion processes, including mechanical breakupand melting of the ice margins, and contraction processes, primarily the advance of the terminusof the MGT. The terminus of the MGT advanced ⇠2 km between 1988 and 2014, associated witha reduction in lake area, however, this was offset by gains due to erosion of the ice margins ofthe MIS and MGT. An example of expansion processes was observed in the summer of 2012. InAugust 2012, satellite imagery revealed that the lake ice in the northwest region of the epishelf lakepartially broke up, and fragments of the MIS calved into the fjord. Summer of 2012 was very warm,air temperatures reached an hourly maximum of +20.2C in July, and lake temperatures peaked atover 3C, the highest recorded during the 3-year mooring deployment. This strong thermal forcingwould have induced substantial surface and submarine melting, which was apparent in the 0.7 mthickness of the surface lake ice, the second thinnest recorded during this study. This event is likelyindicative of the future of Milne Fiord under a warming climate, with enhanced thermal erosionweakening the ice cover, leading to further break up of the MIS and MGT, and expansion of theMEL.3.4.2 Depth changesAnnual CTD profiling in Milne Fiord has shown that the epishelf lake thinned approximately 9 mfrom 2004 and 2014. The overall decrease in thickness of the lake strongly indicates the ice damis thinning over time, roughly on the order of 0.5 m a1. Increases in depth of the lake were77significantly correlated with the number of cumulative PDDs, strongly indicating that meltwaterinflow from the surrounding glacial catchment was responsible for the observed depth increase.The gradual thinning of the lake after the cessation of meltwater inflow, suggests the outflow offreshwater from the fjord was restricted by the geometry of the MIS.The depth of the lake at any given time is is determined by the balance of the volumetric inflowand outflow rates. If drainage of the lake is continuous and hydraulically controlled under the iceshelf, rapid inflow during an intense surface melt season will generate a deeper lake than slowinflow, for the same volume of water added.3.4.3 Observational errorSeasonal fluctuations in lake depth indicate that timing of observations is critical to prevent aliasingthe long-term, interannual record of epishelf lake depth. This is especially important if epishelflake depth is to be used as a climate indicator, as suggested by Veillette et al. (2011b). Our resultshave shown that in the absence of continuous records, the best long-term estimate for the lakedepth is obtained from CTD profiling each year just prior to the initiation of the melt season (i.e.approximately June 1st). Observations at this time capture the annual minimum depth of the epishelflake prior to inflow. Although the lake depth may not have reached equilibrium by June 1st, thedepth measured at this time is arguably the most reliable indicator of the long-term state of the lake,and the closest indicator of the actual depth of the ice dam. Profiles collected at other times of theyear must account for the variations in summer meltwater inflow and drainage hydraulics under theice shelf.For Milne Fiord, we have found that depths measured at other times of year can be correctedusing the number of PDDs accumulated that season up to the date of profiling. We demonstrate theimportance of this timing by examining the apparent increase in lake depth during the periods 2006-2007, 2009-2010, and 2013-2014. Without knowledge of the seasonal deepening of the lake, theseobservations could be interpreted as periodic thickening of the ice dam, suggesting a fluctuatingice shelf mass balance. However, once the number of cumulative PDDs at the timing of profilingare accounted for, the corrected lake depths reveal thinning of the lake, and by proxy the ice dam,every year between 2004 and 2011, indicating mass loss from the ice shelf occurred at a nearlyconstant rate during this period. This example clearly demonstrates that to establish a long-termtrend, synoptic observations must be placed in the context of the temporal variation of the underlyingphenomenon. Ultimately, however, the interpretation of synoptic observations is limited withoutcontinuous observations due to the possibility of episodic events, such as the mixing event recordedin January 2012.783.4.4 January 2012 mixing eventThe January 2012 event was associated with the vertical mixing of a stable water column, consist-ing of warm freshwater above cold seawater. The mixing event appears to have been widespreadthroughout the lake, as all profiles collected after the event show a consistent change in the gradientof the lower halocline and the depth of the epishelf lake. To induce such a widespread mixing wouldhave required a substantial input of energy. We can estimate the amount of energy required to mix-ing the stable water column from the change in gravitational potential energy of the water columnbefore and after the event. The closest profiles in time were collected 6 months prior to (July 2011),and 4 months after (May 2012) the event. These profiles are not completely representative of thewater column at the time of the event because of changes in the density profile due to seasonalmeltwater input and epishelf lake drainage under the ice shelf during the intervening months.Instead, we estimate the change in potential energy by assuming the water column was an ide-alized 2-layer system, with freshwater above and seawater below, and the two layers completelymixed, resulting in a homogenous column. The change in potential energy (DPE) of a water col-umn per unit area due to mixing isDPE = 12h1h2(r2r1)g (3.6)where h1 is the thickness of the upper layer involved in mixing, h2 is the thickness of the lower layerinvolved in mixing, r1 is the density of the upper layer (1000 kg m3), r2 is the density of the lowerlayer (1025 kg m3), and g is gravitational acceleration (9.81 m s2). From the mooring temperaturerecord we choose a range of values for h1 (0.5 m and 1.5 m) and h2 (2 m and 7 m), which results in achange in potential energy per unit area of order 101 to 103 J m2. This estimate assumes completemixing into a homogenous water column over the height h1+ h2. The water column was not fullyisothermal after the event, so the calculated change in potential energy is considered the upper limitof that required to induce the observed mixing. Assuming the mixing was uniform across the fullarea of the lake (64.4 km2), then the total energy required for mixing was of the order 108 to 1010 J.What are the possible sources of energy for mixing?Tidal motion is an obvious source of kinetic energy in Milne Fiord, and the interaction of tidallydriven oscillations with bathymetry or ice keels could induce mixing. However, tidal oscillationis continuous and provides no explanation for the episodic nature of this event. In addition, thebottom pressure sensor recorded no water level anomalies prior to or during the onset of the mix-ing event, meaning tides are unlikely the source. By the same reasoning, any flow associated withbarotropic changes in water level, such as tsunamis, are unlikely to have occurred. A sufficient mag-nitude earthquake could release enough energy to the water column, however there were no sub-stantial earthquakes recorded at this time (http://earthquaketrack.com/r/ellesmere-island-nunavut-canada/recent), and an earthquake would be expected to induce some degree of seiching or mixing79over the full depth of the water column, which is not apparent in the temperature records (the tem-perature signal associated with this event appears to be limited to depths above ⇠50 m).Another possible source of energy to mix the upper water column in a glacial fjord, is the capsizeor calving of an iceberg. In a laboratory study, Burton et al. (2012) found that most (approximately84%) of the total energy released during an iceberg capsizing event was transferred to the watercolumn via hydrodynamic coupling, viscous drag, and turbulence (with an additional ⇠15% to ki-netic energy, and⇠1% to radiated surface wave energy). The authors suggest that iceberg capsizingis a potentially important source of mixing in the stratified ocean proximal to marine ice margins.The energy release from a capsizing event (Ecap) of a free-floating iceberg with an idealized cuboidgeometry can be estimated asEcap =12rigLiH3i e(1 e)✓1 rirw◆, (3.7)where Li is the length of the iceberg parallel to the axis of rotation, Wi is the width, Hi is the pre-capsize height of the iceberg, and e =Wi/Hi (Burton et al., 2012). For a large iceberg in MilneFiord, we estimate Li is 100 m, Hi is 150 m, Wi is 50 m, and set ri to 900 kg m3, rw to 1025 kgm3, resulting in an energy release of 4 x 1010 J. If 80% of this total energy liberated from capsizewas dissipated through turbulent mixing of the water column, that would have been of sufficientmagnitude to induced the mixing observed in Milne Fiord. However, the mixing event appearedto be limited to the upper 50 m of the water column, suggesting perhaps the capsize of a smallericeberg. Even for a much smaller iceberg, where Li is 25 m, Hi is 50 m, Wi is 25 m, the energyreleased is >108 J, which is still of the correct order of magnitude to induce the observed mixingthroughout the lake.Breakup and calving of the MGT and inner margin of the MIS have increased over the pastdecade, resulting in dozens of icebergs in Milne Fiord with large enough dimensions that theircapsize could have released sufficient energy for mixing. Icebergs are typically frozen into fast-iceyear-round, which may reduce the frequency of capsizing and help explain the episodic nature ofthis event (only one major mixing event recorded in 3 years). Iceberg capsizing and capsize mightbe expected to be more frequent during the summer melt season, however, preferential melting ofthe iceberg keels by warm water, with negligible mass from the surface in winter, suggests wintercapsize is possible. Although we lack sufficient evidence to exclude other possible mechanismsfor the mixing event, such as a landslide, or the propagation of an offshore anomaly into the fjordbelow the ice shelf, the iceberg capsizing mechanisms appears to be a plausible explanation that canaccount for the observations.The abrupt depth reduction of the epishelf lake in Milne Fiord might be considered a drainageevent, similar to that recorded in Disraeli Fiord in 2001, caused by the fracturing of the Ward HuntIce Shelf (Mueller et al., 2003). However, the observations suggest this is not the case, and a few80simple calculations provide evidence to support this argument. If, for example, we assume the 1.5m depth change was due to a rapid drainage of epishelf lake water under the MIS, precipitated by afracturing of the ice shelf, then the total volume change across the ⇠64.4 km2 lake was 9.8x107 m3.For all of this water to drain out of the fjord over the observed 18 hr duration of the event, wouldrequire a volume flux of 5.4 x 106 m3 hr1, requiring outflow velocities of>100 m s1, through the⇠10 m wide basal channel in the MIS. This result is clearly not realistic, and given the outflow rateis likely hydraulically controlled (see Section 3.4.6), we consider it unlikely that the January 2012event was related to a rapid drainage event.3.4.5 Freshwater budgetThe volume and residence time of the epishelf lake are determined by the balance between inflowsand outflows, and changes in area of the epishelf lake. Mass inputs to the epishelf lake comefrom surface runoff and basal melting of the MIS and MGT. Although the melting of floating icearound the lake margins contributes a volume of freshwater equivalent to the volume of ice melted(adjusting for density differences), and will affect the residence time of the lake, it will not alter thedepth of the lake. Changes to the depth of the lake are primarily determined by the balance betweeninflow from the fjord catchment and outflow under the MIS.If we assume the magnitude of deepening of the halocline is equal to the volume of water perunit area that enters the lake from surface runoff each summer, then the volume of the lake thatwould have been replaced each summer ranges between 13% and 35%. Thus the residence timeof the lake is between 3-8 years. However, this estimate is conservative in that it assumes there isno outflow during the deepening of the halocline. If outflow occurs during summer inflow, as wewould expect, then the volume entering the lake is potentially much larger, and the residence timewill be reduced. As the lake thins over time it appears more likely the residence time will be furtherreduced.We compare our findings with published glacier mass balance rates for the northern CanadianArctic as an independent measure of the volume of freshwater entering Milne Fiord. Gardner et al.(2013) estimated an average glacier mass budget of -310 ± 40 kg m2 a1 from 2003-2009 forthe northern Canadian Arctic, including Ellesmere, Axel Heiberg, and Devon Islands. The MilneGlacier catchment has a glaciated area of 1108 km2, so the estimated annual mass budget for theMilne Glacier is -0.343 Gt a1. We can convert this into a volume of freshwater entering MilneFiord by assuming that all mass is lost as meltwater runoff that enters Milne Fiord at the surface.It must be noted however, the Gardner et al. (2013)estimate does not overlap in time with our data,and is a net glacier mass loss, meaning it does not account for additional runoff from snow melt.Acknowledging this, the volume of freshwater input from glacier melt is equivalent to ⇠5 m a1increase in the depth of the lake (assuming an average lake area of ⇠65 km2). This is roughly50% larger than the maximum depth increase recorded by the mooring between 2011 and 2014,81suggesting there is ample meltwater entering the fjord to replenish the epishelf lake at estimatedrates.We have determined that there is sufficient surface runoff entering the fjord in summer to ac-count for the increase in lake depth during the melt season. Once inflow ceases, however, the lakebegins to thin, and continues to thin until the following summer. The rate of thinning is linked tothe hydraulics of drainage through the basal channel under the MIS.3.4.6 Outflow hydraulicsThe mooring data showed that the lake thinned each year from the time the summer melt seasonended, through winter, until meltwater inflow commenced again the following spring. The outflowrate was non-linear and appeared to be dependent on the relative depth difference between the lakeand the ice dam, suggesting hydraulically controlled flow.An idealized schematic of epishelf lake outflow through a basal channel in the ice shelf is shownin Figure 3.9, represented as a simple two-layer system, with freshwater overlying seawater (Dr =25 g kg1). Under steady state conditions the ice dam acts as a hydraulic control, limiting two-waytransport below the ice shelf. If the depth of the seawater layer is much greater than the depth ofthe freshwater layer, then the situation is analogous to single layer flow through an inverted weir,but here the horizontal pressure gradient is supplied by the density difference between freshwaterand seawater. If we assume a rectangular channel geometry, the volumetric outflow discharge (Qd ;m3 s1) can be estimated using a modified form of the Kindsvater-Carter rectangular weir equation(Kindsvater and Carter, 1959):Qd =23p2g0Cebh32 (3.8)where g0 is reduced gravity (g0 = g(Dr/r)), Ce is an empirically derived discharge coefficient, bis the effective width of the outlet channel (m), and h is the effective depth of the lake below theice dam (m). b and h account for the effects of viscosity and wall friction and will be thereforesomewhat larger than the actual physical dimensions.Assuming vertical sidewalls, the change in volume of the epishelf lake over time (dV/dt) is:dVdt= ALdhdt(3.9)where dh/dt is the change in thickness of the lake over time, and AL is the area of the epishelf lake.During winter, inflow is negligible, so the change in volume is equal to the volumetric outflow (i.e.dV/dt = Qd). Equating Eq 3.8and Eq. 3.9, solving for dh/dt, and integrating gives:h(t) =✓12at+1ph0◆2(3.10)82wherea=23p2g0CebAL, (3.11)and h0 is the initial depth of the lake below the ice dam at t = 0.From Eq. 3.10 we modelled the change in depth of the epishelf lake over each winter of thethree year mooring record, from approximately September to May of 2011-2012, 2012-2013, and2013-2014. We assume changes to the draft of the ice dam occurred during the surface summer meltseason (which we did not model), so the draft of the ice dam remained constant each winter. Thestart time (t = 0) for each run was chosen as the date when air temperatures fell below zero for thewinter, assumed to be when meltwater inflow ceased, and the initial depth of the lake at that time(z(t = 0)) was estimated from the mooring temperature record. The model was run for 275 days overeach of the three winters. In the model freshwater drains under the ice dam at a rate proportionalto h, until the lake depth z(t) shoals to the level of the ice dam zi (i.e. when z(t) zi = h = 0).The mooring records show the lake depth was still shoaling when meltwater input commenced thefollowing spring, so we could not directly estimate zi, nor could we directly calculate the initialdepth of the lake below the ice dam (h0). Instead, we estimated h0 as the sum of the differencebetween the initial lake depth and the depth when t = 240 elapsed days (the maximum duration ofthe shortest mooring record), plus some unknown offset ch (i.e. h0 = z(t = 0)z(t = 240days)+ch).We run the model to find values for parameters ch andCeb that gave the best fit to the observed depthchanges each year, keeping the parameters constant for all years. We found the best fit was achievedfor all years when ch = 1.6 m andCeb = 4.5 m.Modelled and observed changes to the depth of the lake relative to the ice dam are shown inFigure 3.10. Despite the initial depth of the lake below the ice dam varying by over a factor of twoamong the different years (from 3.3 m in 2013-2014 to 7.5 m in 2011-2012), the simple drainagemodel simulated the observed pattern of changes in the depth of the lake each winter well. Themodel could not account for the abrupt thinning of the lake in January 2012 (elapsed day 140 for2011-2012) due to the mixing event, so observed and modelled values differ accordingly after thisdate. Overall, however, the results indicated outflow drainage from the epishelf lake through thebasal channel could generally be well simulated by weir outflow hydraulics.To assess the model we need to determine if the selected values for the parameters are physicallyrealistic. If ch is 1.6 m then the actual draft of the ice dam zi varied between 9.4 m and 7.5 m from2011 to 2014. This is consistent with field measurements of 8 to 11 m thick ice along the re-healedfracture in 2015. Next, for typical weirs Ce varies between 0.55 and 0.8 (ISO, 1980). If we assumethis range for Ce is broadly appropriate for the MEL system, then the width of the channel b isbetween 5.6 and 8.2 m. This value is comparable to the minimum width of the surface expression ofthe re-healed fracture at its narrowest point (2 – 8 m) from field observations. We conclude that thevalues chosen for the parameters are physically realistic and appropriate for this system, although83further work is required to validate these parameters.The model does have some limitations. It does not account for the possibility, and likelihood, ofchanges in the depth of the ice dam during winter, which could be driven by submarine melting dueto outflow of warm epishelf lake water, or alternatively, basal accretion due to freezing of brackishwater. These processes are dependent on the actual properties of water in the channel, and theexchange of heat between the outflowing water and the ice, which are dependent on mixing and thecharacter of the boundary layer within the channel, factors which are unknown. Investigation ofthese processes is ongoing, but beyond the scope of this study. It is also questionable whether theapplication of a standard weir equation derived for a thin plate weir is appropriate for the epishelflake system, particularly given the length of the outflow channel (on the order of 20 km). However,the constriction along the channel that is the hydraulic control point may actually be several ordersof magnitude shorter (on the order of a few meters). As well, the values of b and h will accountfor some of the differences between this system and a standard weir, particularly the increasedfriction likely along the long channel. Further consideration of frictional exchange flow through along channel is ongoing, and we consider the model presented here as a first step in understandingthe hydraulics of the outflow. Acknowledging the model’s limitations we attempted to minimizethe possibility of over-tuning the model by using the same values for the unknown parameters in allthree years. The only variable that changed between years was the initial depth of the lake below theice dam, and this was determined from mooring observations using a standardized method that wasconsistent for all years. In summary, despite its simplicity, the model, using apparently physicallyrealistic parameters, simulates the observed depth changes well, indicating drainage of the epishelflake is hydraulically controlled by the geometry of the outflow channel under the ice shelf.3.4.7 Implications for MISThe epishelf lake depth has been considered a proxy for the minimum draft of the MIS. We notethat in Milne Fiord interannual depth changes are likely only related to the depth of the ice damalong the outflow channel, not necessarily the overall thinning of the MIS. It is apparent that theice dam is thinning over time and this is likely due to a combination of surface ablation and basalmelting. Mortimer et al. (2012) found a change in thickness along a re-profiled transect that crossedthe region of the suspected ice dam of the MIS of 2.63 ± 2.47 m, or 0.10 ± 0.09 m a1 between1981 and 2008/2009. Over roughly the same period (May 1983 to May 2009) the epishelf lakeshoaled by 2.8 m, at an average rate of 0.11 m a1, indicating that changes in epishelf lake depthare valid proxies for changes of ice dam thickness. Extending the time series using our observationssuggests the ice dam has thinned a further 5.4 m between 2009 and 2014. This is an average rateof 1.08 m a1, an order of magnitude faster than the period from 1983 to 2009. However, we knowthat 4.1 m, or 75%, of the thinning over this period occurred between 2011 and 2012, with at least1.5 m of thinning, or 27%, related to the mixing event in January 2012. This observation suggests84that 2011-2012 was an anomalous year. The summer of 2011 did record the highest cumulativePDDs on record at Milne Fiord, indicating that the strong melt season may have played a role in thechanging dynamics of the ice shelf-epishelf lake system that year.Observations from July 2014 indicate the ice dam is approximately 8-9 m thick. If the averagerate of thinning observed between 2004 and 2014 persists, the MEL will cease to exist as a perenniallake by 2034. However, recent observation of increase fracturing of the Central Unit of the MISindicate that a sudden fracturing of the MIS could result in the abrupt drainage of the MEL atany time. In addition, enhanced iceberg calving and capsize could result in further mixing of thehalocline, quickly eroding the base of the epishelf lake.3.5 SummaryThe MEL has existed since at least 1983, and possibly as early as the 1950’s. The lake has grownin extent primarily through melting and breakup of the MIS, likely due in part to submarine meltingby the heat retained in the epishelf lake over winter by its salinity stratification. The recent arealexpansion has been balanced by losses due to the advance of the terminus of the MGT. Based onestimates of freshwater residence time in the lake, it appears there is sufficient volume of runoff fromthe fjord catchment to renew the entire volume of the lake at least every 3 years, yet the volume ofthe lake has decreased since 2004, owing largely to the decrease in depth of the lake.The summer inflow of meltwater leads to a substantial seasonal increase in depth of the epishelflake, which has not previously been reported. We found that the magnitude of deepening is directlycorrelated with the cumulative number of PDDs, a proxy for the volume of surface meltwater inflow.Outflow of the epishelf lake likely follows a basal channel along a re-healed fracture in the MIS.Observations along the hypothesized drainage pathway of the epishelf lake are needed to improveunderstanding of the hydraulics of the system, and constrain the location of the ice dam.The seasonality of the epishelf lake suggests that synoptic annual profiles could alias the long-term depth record of the epishelf lake, and thus bias inferred changes in the thickness of the MISdam. Continuous time series observations are required to place the synoptic profiles in context,however if only annual profiling is feasible in the future we suggest observations collected just priorto the beginning of the melt season (i.e. around June 1st for Milne Fiord) provide the most reliableindicator of steady state epishelf lake depth and ice shelf thickness. The depth of the epishelflake does appear to have been a reliable indicator of the long-term mass balance of the MIS in thepast, however the rapid changes observed in 2011/2012 indicate continual monitoring is required tounderstand the mechanisms influencing lake depth.Drainage of an epishelf lake below an ice shelf is hydraulically controlled, dependent on thedensity between lake water and seawater offshore, the depth of the lake below the draft of the ice,which is determined by the rate and volume of inflow and the mass balance of the ice shelf, and thegeometry of the outflow channel. The magnitude of epishelf lake depth variability, and therefore its85utility as an indicator of ice shelf mass balance, will be dependent on the particular epishelf lake-iceshelf system, and will likely change over time, necessitating continual and repeat observations tounderstand the evolving hydraulics of the system.The existence of epishelf lakes is highly sensitive to the interactions of the atmosphere, cryosphereand hydrosphere. The warming climate of the Arctic has resulted in an increase freshwater flux tocoastal fjords, with the potential to increase the volume of epishelf lakes. However, an overall thin-ning of the lake has been driven by mass loss of the MIS. At current rates of thinning the Milne Fiordepishelf lake, the last known epishelf lake in the Arctic, will be lost by 2034, however continuedbreakup of the Milne Ice Shelf suggests a catastrophic drainage could occur at any time.863.6 TablesTable 3.1Milne Fiord epishelf lake depth, area, volume and related observations.Date(s) CTD No. of Depth Area Volume PDDs Image AcquisitionProfile CTD mean (range) (km2) (km3) to Date Source DateLocation Profiles (m) (yr total) for (yyymmdd)(C days) Area Est.1959/08/17 - - - 13.5 - - Aerial photo 195908171963/08/29 - - - 13.5 - - Corona 196308291983/05/25 PV 1 17.5a - - 0 (158)* - -1988/08/08 - - - 67.3 - - SPOT-1 19880808SPOT-1 198808081992/01/29 - - - 60.6 - - ERS-1 19920129ERS-1 199203161998/01/13 - - - 61.0 - - Radarsat-1 199801132003/01/11 - - - 59.3 - - Radarsat-1 200301112004/08/06 NB 1 18.3b - - 98 (132)* - -2006/06/03 NB 1 16.0b 65.0 1.04 4 (153)* Radarsat-1 200601142007/07/13 NB 1 16.5b - - 78 (215)* - -2009/05/29 - 05/30 PV 18 14.7 (14.5-14.9) 65.2 0.96 0 (274) Radarsat-2 200901042009/07/04 NB 1 14.6 - - 34 (274) - -2010/07/09 NB 1 15.3 - - 100 (185) - -2011/05/10 MM 1 13.6 67.6 0.92 0 (278) Radarsat-2 20110103Radarsat-2 201102282011/07/05 NB 1 14.4 - - 94 (278) - -2012/05/05 - 05/14 MM 3 9.5 (9.4-9.6) 64.4 0.61 0 (253) Radarsat-2 20120203Radarsat-2 201204172012/06/28 - 07/09 MM, ML 23 10.6 (10.2-11.5) - - 50 (253) - -2013/05/11 - 05/18 MM, ML 11 8.0 (7.8-8.1) 67.0 0.54 0 (92) Radarsat-2 20130427Radarsat-2 201304272013/07/04 - 07/22 MM, ML 46 8.1 (7.8-8.5) - - 10 (92) - -2014/07/12 - 07/24 MM, ML 13 9.3 (9.1-9.4) 71.2 0.66 39 (110) Radarsat-2 20150327Radarsat-2 20150327PV - Purple Valley BayNB - Neige BayMM - Milne Fiord mooringML - multiple locations inside and outside Milne FiordaJeffries, 1985bVeillette et al., 2008*PDDs calculated from air temperatures interpolated from the Eureka weather station where TMilne = 0.47*TEureka + 0.47 (RMSE = 1.98 C)873.7 FiguresMISMGTMELMLSIArcticOceanMGCape EvansCape EgertonN2015 1959 19882003  99 o    90o    81o    72o    63o    80 o    81 o    82 o    83 o    84 o  Ellesmere Is.Milne FiordDisraeli FiordEurekaAlertArctic Ocean20115 kmABC DE FNBPVCentral UnitOuter UnitFigure 3.1 Map of Milne Fiord study area. A) RADARSat-2 image of Milne Fiord showing the extent ofthe Milne Fiord epishelf lake (MEL) in 2015, corresponding to the region of high backscatter (light gray)outlined in blue. The Central and Outer Units of the Milne Ice Shelf (MIS), as well as two re-healed fractures(red arrows), are indicated, as well as the Milne Glacier (MG), grounding line (red line), Milne Glacier tongue(MGT), multiyear landfast sea ice (MLSI), the met station (red square), mooring (red triangle), and two smallinlets unoffically named Purple Valley Bay (PV) and Neige Bay (NB). B) Regional map of Ellesmere Island,Canada. The sequence of four panels on right show the increase in area of the MEL (grey) estimated fromaerial and satellite imagery from C) 1959, D) 1988, E) 2003, and F) 2011, based on data from Mueller et al.(2016). The coastline of Milne Fiord is outlined in black. White areas inside the coastline are glacier or iceshelf.880 10 20 300510152025Depth (m)SA (g kg−1)0 0.2 0.40510−2 −1 0 1 2 3Θ (oC)0 1 5 10 20 30IceSurface meltMixed layerUpper halocline N2 maxLower haloclineSA (g kg−1)−1.5 −1 −0.5 0 0.5 10−2 10400Θ (oC)35 Epishelf lakeFjord Modified Water1983−052004−082006−062007−072009−052009−072010−072011−052011−072012−052012−062013−052013−072014−07offshoreA B CFigure 3.2 Changes in A) salinity and B) temperature properties of the upper 25 m of Milne Fiord from allfield campaigns from 1983 to 2014. A single representative profile collected at the mooring site from eachfield campaign is shown when multiple profiles were collected. Inset in A) shown a zoom in of epishelflake salinities. Dashed line in A) and B) indicates a representative profile collected offshore of the MIS.C) Idealized salinity (black line) and temperature (grey line) profiles showing the layers of the upper watercolumn. Note the non-linear salinity scale. The epishelf lake is defined as extending from the surface to thebuoyancy frequency maximum (N2 max). The inset shows the full water column properties of the fjord to440 m depth.890 5 10 156789101112Depth (m)S  (g kg−1)0 10 20507090oC days) Cum. PDDs (Elapsed time (days)0 10 20103050 Cum. PDDs (o C days)Elapsed time (days)20122013A0 5 10 153456789Depth (m)SA (g kg−1)BFigure 3.3 Salinity profiles showing seasonal changes in depth of the MEL halocline in a) 2012 and b) 2013.Profiles are coloured by time over the duration of each summer field campaign (10-days in June/July 2012and 18-days in July 2013). A single profile collected in May of each year, prior to the onset of the meltseason, is shown (black line). In each panel the left inset shows the PDD accumulated during each summerfield campaign (a proxy for the volume of surface meltwater inflow), with the timing of profiles indicated(coloured circles). Note the different y-axes range between a) and b), although the incremental scales areconsistent. The right inset shows the profiling locations on a map of Milne Fiord (the MEL, MIS, and MGTare outlined). Multiple profiles were collected at the mooring site (black circle) during each campaign.900 50 100 150 200 250 300−1012345Slope = 0.017 m oC−1 day−1Cum. PDDs (oC days)Depth change (m)  201120122013Figure 3.4 Correlation between the cumulative number of PDDs and the change in depth of the epishelf lake.Depth is estimated from the isotherm proxy during the melt season in 2011, 2012, and 2013.  Ice draft (m)050100150350 400060Distance (m)Depth (m)MISFigure 3.5 Map of ice thickness in Milne Fiord indicating the likely drainage pathway of the MEL along are-healed fracture in the MIS. The 35 m ice thickness contour is shown to highlight the outflow restrictionthe MIS imposes across the mouth of the fjord. The inset shows measured ice thickness along a 400 mice-penetrating radar transect across the re-healed fracture from July 2013.91−40 −20 0 20 Air Temp ( o C) 0 500 Solar Rad. (W m −2 ) 0 100 200 300 Cum. PDD ( o C day) 0. 5 5 10 20 8m 13m 11m 5m S A (g k g −1 ) Dept h (m ) Date 2011−2014     07/11 10/11 02/12 05/12 08/12 11/12 03/13 06/13 09/13 01/14 04/14 2 4 6 8 10 12 14 16 18 20 Temperature  ( o C) −1 0 1 2 3  11m A B C D E Figure 3.6 Meteorological conditions and epishelf lake properties from May 2011 to July 2014 in MilneFiord. A) Air temperature. B) Shortwave solar radiation. C) Cumulative positive degree days. D) Absolutesalinity from instruments moored at depths between 5 m and 15 m. Note the logarithmic scale. E) Temper-ature time series from thermistors moored between 1 m and 20 m depth. Black circles indicate thermistordepths at each mooring deployment and white areas in indicate data gaps. Grey regions outline the start andend of the surface melt season each year. Arrows in D) and E) indicate the timing of the mixing event shownin Fig. 3.7.9210152025S A(g kg−1)  13 m3276327732783279Heat content(J m−2 )  0 m to 25 mDepth (m)DoY 2012  9 10 11 12 13 14 15 16681012141618Temperature (o C)−1−0.500.510 5 10 15 20 2502468101214161820Depth (m)SA (g kg−1)−1 −0.5 0 0.5 102468101214161820CT (oC)  July 2011May 2012D EABCFigure 3.7 January 2012 MEL halocline mixing event. time series of A) salinity at 13 m depth, and B) heatcontent between 0 m and 25 m, C) temperature between 6 m and 18 m depth, and profiles of D) salinityand E) temperature from field campaigns 6 months before (July 2011) and 4 months after (May 2012) thedrainage event. In C) the isotherm increment is 0.2C and the 0C isotherm (grey) is highlighted as a proxyfor the bottom of the epishelf lake. Black squares on y-axis indicate depth of thermistors.932004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 201581012141618YearDepth (m)  Slope = −0.51 m a−1Slope = −0.34 m a−1↑CTD profile N2max (uncorrected)CTD profile N2max (corrected)Salinity proxyIsotherm proxyFigure 3.8 Interannual and seasonal changes in the depth of the MEL from 2004 to 2014. Plotted are theuncorrected lakes depths from CTD profiles, and depths corrected for cumulated PDDs (the corrected depthis shown only for the first profile from each field campaign after the continuous mooring records commencedin 2011), as well as depth determined from moored salinity and temperature records from May 2011 to July2015. The arrow indicates the mixing event in January 2012. Average thinning rates are shown for the periodsbefore and after the mixing event.94freshwater1000 kg m-3hicedambhicedamABALicedamzizbepishelf lakeArctic OceanArctic Oceanepishelf lakeseawater1025 kg m-3CzziQdFigure 3.9 Schematic representation of epishelf lake outflow through a basal channel in the ice shelf dam inA) plan, B) elevation, and C) cross-sectional views. The volumetric discharge is modelled using a modifiedform of the rectangular weir equation.950 50 100 150 200 250 300012345678Elapsed time (days)Effective depth below ice dam (m)  2011−20122012−20132013−2014Figure 3.10 Change in the effective depth (h) of the epishelf lake below the ice dam over time during thewinter of 2011-2012, 2012-2013, and 2013-2014. Elapsed time is measured from the end of the surface meltseason each year. Observed depths (solid lines) are based on the isotherm proxy from the mooring record,while modelled depths (dashed lines) are based on a weir equation. Arrow indicates the mixing event thatoccurred in January 2012. See Section 3.4.6 for details on methodology.96Chapter 4Depth-dependent submarine melt ratesof a glacier tongue and ice shelf in aHigh Arctic fjord4.1 IntroductionThe presence of floating glacier tongues and ice shelves at the coast slows the rate of dischargeof grounded ice into the ocean. Changes in the thickness or extent of the floating ice buttressesmay, if sustained, alter the rate of ice discharge to the ocean and thereby alter the rate of change ofsea level. The mass balance of ice shelves is determined by supply from land, calving of icebergsfrom the ice front, and melting and accumulation at their top and bottom surfaces. Ice shelves areparticularly sensitive to change because they are in contact with both the atmosphere and the ocean,and therefore vulnerable to changes in the temperature or circulation pattern of either. Melting at thebase of the ice shelf driven by seawater above its in situ freezing temperature is responsible for 80%of total mass loss from the Petermann Gletscher tongue in Greenland (Rignot and Steffen, 2008),and accounts for three quarters of the mass loss via ablation from Antarctic ice shelves (Rignotet al., 2013). These high fractions highlight the importance of ice-ocean interactions to ice-shelfmass balance. The transformation of seawater via ice interaction within the under-ice shelf cavitiesalso affects the ocean, via its impact on ocean stratification and circulation. Understanding basalmelt is of interest to oceanographers as well as glaciologists.Ice shelves and glacier tongues were, until recently, extensive along the northern coast ofEllesmere Island in the Canadian High Arctic. In this region a glacier tongue, which is the floatingextension of an outlet glacier at the head of a fjord, needs to be distinguished from an ice shelf,which is a thick platform of ice formed in situ at the mouth of a fjord. Ice shelves here develop viaaccretion of sea ice on the underside and accumulation on the top side, with additional mass acquired97from tributary glaciers and low-elevation ice caps (Jeffries, 2002). Ice shelves along this coast donot play a large role in directly buttressing grounded ice, however their presence dramatically altersthe oceanographic properties of the fjords where they exist (Chapter 2 and 3). The presence ofice shelves at the mouth of fjords can result in perennial freshwater surface layers within the fjord,known as epishelf lakes. The year-round ice cover and stratification associated with the epishelflake alters surface heat fluxes and eliminates wind mixing, with implications for the properties, andthus melt potential, of the upper water column. The presence of ice shelves at the fjord mouthscould also trap icebergs with implications for surface heat fluxes and freshwater content within thefjord. The thinning or loss of ice shelves can impact the mass balance of marine terminating glaciersthrough changes in ocean forcing.The recent widespread collapse of ice shelves and glacier tongues on Ellesmere Island has beenattributed, in part, to atmospheric warming (Vincent et al., 2001; Mueller et al., 2003; Coplandet al., 2007; Mueller et al., 2008; White et al., 2015a). In the Canadian High Arctic atmosphericwarming has been twice the global average (IPCC, 2013). Atmospheric warming in this region hasresulted in a sharp increase in the rate of surface mass loss from glaciers and ice caps of the QueenElizabeth Islands (QEI; including Ellesmere, Devon, and Axel Heiberg Islands); values averaged-7 ± 18 Gt a1 between 2004 and 2006, and -61 ± 18 Gt a -1 between 2007 and 2009 (Gardneret al., 2011; Gardner et al., 2013). In addition, a widespread acceleration of tidewater glaciersin this region, and increased discharge from a few individual glaciers — Trinity and WykehamGlaciers accounted for 60% of all dynamic discharge in the QEI during 2011-2014 — and evidenceof dynamically induced thinning (Van Wychen et al., 2016) has increased the need to understandthe causes. Missing in studies to date in this region have been observations of ocean properties nearthe termini of tidewater glaciers, estimates of basal melting, and how changes in ocean forcing mayinfluence glacier mass balance.Milne Fiord on northern Ellesmere Island is a notable exception; ocean properties here havebeen monitored since 2011. Milne Fiord is also of interest due to its unique geophysical features.The Milne Glacier (MG) terminates at the head of the fjord in the ⇠16 km long Milne Glaciertongue (MGT), while the separate Milne Ice Shelf (MIS) spans the width of the fjord at its mouth.The presence of the MIS strongly alters the water structure in the fjord because it forms a dam thattraps seasonal runoff from land to form a highly stratified upper water column, wherein seawater iscapped by a perennial freshwater layer up to 18 m thick; this feature is known as the Milne Fiordepishelf lake (MEL). The vertical distribution of heat in the fjord is likewise affected by the ice shelf;in addition to a temperature maximum below 200 m depth associated with the Atlantic layer of theArctic Ocean, a near-surface temperature maximum exists at⇠10 m depth that is associated with theepishelf lake (Chapter 2). Analytical and numerical modelling studies suggests melt rate has a lineardependence on ocean temperature (Jenkins, 2011; Sciascia et al., 2013) and therefore the thermalstratification of the water column very likely influences the vertical distribution of submarine melt98in Milne Fiord. This study is, to our knowledge, the first to directly estimate basal melt rate and itsspatial distribution for an ice shelf or a tidewater glacier in the Canadian Arctic.Our goal is to estimate submarine melt rates of the MGT and MIS, and understand how meltrates vary with location and time. We use independent oceanographic and glaciological methodsto estimate melt rates for the MGT. We then apply the oceanographic method to map the spatialdistribution of melt rate across the MGT and MIS.We first locate the grounding zone of the Milne Glacier from measurements of ice thicknessand elevation using the assumption of hydrostatic equilibrium. We then calculate the annual area-averaged submarine melt rate for the floating portion of the MGT from 2011 to 2015 using the lawof mass conservation — ice discharge across the grounding line minus terminus advance equalsthe total of ice loss at the top and bottom surfaces. Next we calculate submarine melt rates eachyear along the length of the MGT from the divergence of ice flux — from surface ice velocitydetermined via speckle tracking of satellite imagery — and from a digital elevation model of icethickness produced from field and remote sensing data. Independently we estimate submarine meltrates along the length of the MGT, over the same period, using measured vertical profiles of oceantemperature, salinity, and current speed. The method is based on the application of a two-equationice-ocean thermodynamic melt model. We then apply the ice-ocean thermodynamic model to theentire fjord, including the MIS, to investigate spatial and interannual variability of submarine melt.Finally, we discuss the role of submarine melting in recent changes observed in Milne Fiord, andthe implications for the future stability of the MIS, MGT, and the dynamics of the Milne Glacier.4.2 Methods4.2.1 Site descriptionMilne Fiord (8235’N; 8035’W) lies at the northwest coast of Ellesmere Island on the Arctic Ocean(Fig. 4.1). Milne Fiord is perennially ice-covered, the majority of the fjord covered by the relativelythick ice of the MGT and the MIS, with the thin (<1 m thick) freshwater ice of the MEL filling theregion between. The Milne Glacier (MG), a 4-5 km wide, 55 km long outlet glacier that drains⇠4%by area of the northern Ellesmere Island icefields, flows into the head of Milne Fiord. Van Wychenet al. (2016) estimated the mean ice discharge of the Milne Glacier to be 0.06± 0.02 Gt a1 between2011 and 2015. The MGT varies from approximately 150 m thick near the grounding line to<10 mthick at the terminus (Chapter 2). The terminus of the Milne Glacier advanced >5 km between1950 and 2009 (Mortimer, 2011), and it has been identified as a possible surge-type glacier (Jeffries,1984; Copland et al., 2003).At the mouth of the fjord, spanning 18 km between Cape Egerton and Cape Evans, and distinctfrom the MGT, is the MIS, formed through a combination of sea ice accretion, snow accumulationand input from surrounding tributary glaciers and low-lying coastal ice caps (Jeffries, 1986a). The99MIS has an average thickness of 50 m and a maximum thickness of 94 m (Chapter 2). Based onsurface morphology and ice characteristics the MIS has been divided into two regions (Jeffries,1986a), an Outer Unit consisting of uniformly thick ice bisected by two re-healed fractures, and aCentral Unit consisting of variable, but generally thinner, ice. By comparing radar measurementsof ice thickness from 1981 and and 2009, Mortimer et al. (2012) estimated the ice shelf thinnedon average by 8.1±2.8 m over this period, an average rate of 0.26 ± 0.09 m w.e. a1, althoughthey noted substantial spatial variability. By assuming surface mass balance was equivalent to thatmeasured on the Ward Hunt Ice Shelf, 100 km east of Milne Fiord, over approximately the sameperiod, the authors indirectly inferred that submarine melting may have accounted for ⇠73% of theoverall thickness change of the MIS, suggesting ocean melting was an important factor in ice shelfmass balance.Water properties and bathymetry in the fjord have been discussed in detail in Chapter 2 and 3,so are only described here briefly. The near-freshwater MEL lies above a layer of seawater modifiedby the presence of the MIS to a depth between 35 - 50 m. Below this level, to the approximate depthof the Milne Glacier grounding line, water in the fjord shares similar characteristics to that offshorein the Arctic Ocean, consisting of cool, relatively fresh Polar Water near the freezing point, abovewarm, saline, modified Atlantic water to the bottom of the fjord at 436 m depth. A ⇠260 m deepbathymetric sill lies under the MIS (Chapter 2), however it is substantially deeper than the thickestice in the fjord, and is not expected to influence submarine melt rates. Limited bathymetric data inthe fjord indicate the MIS maybe partially grounded on an seabed ridge near the confluence of there-healed fractures (Chapter 2) and that the sea bed slopes upward below the MGT toward the headof the fjord.4.2.2 Bulk mass balance of MGTWe calculate a bulk annual area-averaged annual basal mass balance, ab (m a1; positive for melt-ing) under steady state conditions for the MGT from conservation of mass as:ab =Qgf Qc fAMGTas f , (4.1)where Qgf is the ice volume flow across the grounding line, Qc f is the change in ice volume overtime (m3 a1) at the terminus (positive for calving, negative for terminus advance), AMGT is thesurface area of the MGT (66 km2), and as f (m a1) is the area-averaged annual surface mass balance(positive for melting).4.2.3 Ice thickness, motion and terminus positionThe digital elevation model (DEM) of ice surface and bed elevation for Milne Fiord produced inChapter 2 is used to constrain ice thicknesses for melt calculation. Qc f is estimated from changes in100terminus position of the MGT, assuming no change in ice thickness over time. Terminus positionswere digitized from Radarsat-2 imagery acquired each winter from 2011 to 2015. Qgf is estimatedas the ice flow through a cross section at the grounding line (i.e. a flux gate) from ice surfacevelocities and ice thickness. Ice surface velocities were calculated by Van Wychen et al. (2016)using a custom speckle-tracking algorithm applied to Radarsat-2 fine beam (8 x 8 m resolution) andultrafine beam (3 x 3 m resolution) image pairs collected in early spring (November-April) . Weassume displacement over the period of image pair acquisition was representative of the entire year,and that there is no vertical velocity shear. Error in velocity measurements are roughly 10 m a1.Surface velocities of the central ice stream were also measured by the displacement of three ablationstakes, MG01, located 1 km upstream of the grounding line, MG02, located 1 km downstream ofthe grounding line, and MG03, located 8 km downstream of the grounding line, between July 2012and July 2013.Van Wychen et al. (2016) calculated ice discharge through a flux gate near the Milne Glaciergrounding line, however we further constrain the position of the grounding line using data fromNASA’s Operation Icebridge 2014. The grounding line is assumed to correspond to the point of firsthydrostatic equilibrium, where the thickness of the ice derived from Airborne Topographic Mapper(ATM) laser altimeter (Krabill, 2010) surface elevation equals that measured by the MultichannelCoherent Radar Depth Sounder (MCoRDS; Leuschen et al., 2010). Upstream of the grounding linethe thickness of the glacier, estimated from surface altimetry assuming (incorrectly) that the iceis in hydrostatic equilibrium, is much greater than the actual glacier thickness measured by radar.Downstream of the grounding line, where the ice is floating, the thickness estimated from surfaceelevation is equal to the actual thickness measured by radar. Therefore, moving from upstream todownstream the first location where the two thickness estimates are equal indicates where the icegoes afloat (i.e. the grounding line; Rignot et al., 2001). Error in the ice depth measurements is± 10 m. The cross-sectional bed profile of the fjord at the grounding line was modelled as a U-shaped valley as in Van Wychen et al. (2014), with the maximum cross-fjord ice thickness from theMCoRDS data at the grounding line position.4.2.4 Surface mass balanceAverage annual surface mass balance was estimated from 6 ablation stakes on the MIS betweenMay 2009 and May 2011 and from 3 ablation stakes on the MGT and MG between July 2012 andJuly 2013. The resulting annual surface melt of 0.78 ±0.64 m a1 is assumed constant in time andspatially uniform over the entire fjord.1014.2.5 Divergence of ice flux along MGTThe depth integrated conservation of mass∂H∂ t+— · (uH) =as f ab (4.2)relates the change in ice thickness over time, ∂h/∂ t, and the flux divergence (— · (uH)), where uis the depth-averaged along-stream velocity, to the surface mass balance as f and the basal massbalance ab. We refer to this model as the divergence of ice flux model throughout this chapter.Assuming steady state, with no change in ice thickness over time (∂h/∂ t = 0), and constantsurface melt, then basal mass balance can be calculated from the divergence in ice flux along theflow line. Seroussi et al. (2011) showed that gridded ice thickness maps based on interpolation ofsparse ice thickness source data, produce strong anomalies in ice flux divergence if the interpolationschemes do not conserve mass. Regions of the Milne Fiord DEM produced in Chapter 2 are basedon sparse source data (up to 2 km gaps between source data measurements) and did not include anymass conservation scheme, so we refrain from calculating the two-dimensionally resolved ice fluxdivergence for the entire MGT. Instead, we calculate a one-dimensional width-averaged basal meltrate along the length of the MGT. For free-floating ice we can assume surface velocity is equal todepth-average velocity, i.e. us = u, and calculate the along-stream velocity component. Velocityand ice thickness data were then averaged over the width of the MGT (⇠4.2 km) and a distance of2.5 km along-stream, and ice flux and melt rates were calculated every 500 m from the groundingline to the terminus of the glacier tongue.4.2.6 Ocean thermodynamic melt modelThermodynamic models of ice-ocean interaction aim to obtain a realistic prediction of melt rate atthe ice shelf base from ocean properties. The models use prescribed interior properties of the iceshelf and the ambient water column to estimate the characteristics exactly at the ice-ocean interfacewhere there are three physical constraints: the interface is at the freezing point, and both heat andsalt are conserved during phase changes (Holland and Jenkins, 1999). The transfer of heat andsalt through the oceanic boundary layer are parameterized as functions of the bulk differences invelocity, heat, and salt between the ice-ocean interface and the far-field ocean mixed layer.A hierarchy of models have been formulated to calculate the heat and freshwater fluxes thatresult from deviations in the far-field ocean properties from freezing point conditions, and theseare reviewed in Holland and Jenkins (1999). The three-equation formulation makes no assumptionabout the conditions at the ice-ocean interface and solves equations for each of the freezing pointdependency and the conservation of heat and salt (Chapter 1). This formulation is the most sophis-ticated, and has been widely used to predict melt rates of ice shelves in ice-ocean thermodynamicmodels (e.g. Hellmer and Olbers, 1989; Holland and Jenkins, 2001; Losch, 2008; Kimura et al.,1022013). The three-equation model is generally preferred owing to its wider applicability over a rangeof thermal forcing (the elevation of the mixed layer temperature above freezing). Alternatively, atwo-equation model assumes the interface salinity and mixed layer salinity are identical (implyinginfinite salt diffusivity across the boundary layer) and the model is evaluated using the far-field salin-ity. In this formulation the rate at which the mixed layer temperature relaxes toward the freezingpoint is governed by the transport of heat through the oceanic boundary layer. The simpler two-equation model is advantageous in reduced or analytic models, and Jenkins et al. (2010) showedthat observations of melting under moderate thermal forcing at the base of the Ronne Ice Shelf couldbe fitted equally well by either the three-equation or two-equation model. Similarly, McPhee 1992and McPhee et al. (1999) have shown that beneath sea ice the two-equation formulation producesheat fluxes that agree well with measurements over wide range of basal roughness characteristics.In any model of ice-ocean interaction, a calculation of the freezing temperature at the ice-oceaninterface is required. The freezing point of seawater is a non-linear function of salinity and a lin-ear function of pressure. Solving the fundamental equations requires simultaneously solving threeequations with three unknowns, and while the complex polynomial form of the freezing point couldbe used, a linearized version of the freezing point relation is often invoked to simplify solving theequation analytically and is the standard form used in much of the ice-ocean thermodynamic lit-erature (Hellmer and Olbers, 1989; Holland and Jenkins, 1999; Jenkins et al., 2001; Losch, 2008;Kimura et al., 2013). In this formulation the freezing point Tf equation has the formTf = l1Sw+l2+l3pb (4.3)where Sw is the salinity in the mixed layer, pb is the sea pressure at the interface, and l1,2,3 areempirical constants (Millero, 1978). The linearized formulation is only valid over a limited salin-ity range (e.g. 4-40 psu in Holland and Jenkins (1999)), acceptable for typical seawater salinitiesunder ice shelves. However, the low salinity of the epishelf lake in Milne Fiord lies outside ofthis range (<0.2 g kg1), and freezing temperatures calculated using the linearized formulation atfreshwater salinities could be incorrect by up to 0.1C, introducing error into the calculated subma-rine melt rates. To avoid this error we use the International Thermodynamic Equation of Seawater2010 (TEOS-10) Gibbs Seawater Oceanographic Matlab Toolbox (www.TEOS-10.org) polynomialfunction to calculate the temperature (ITS-90) at which seawater freezes as a function of AbsoluteSalinity, SA (in g kg1) and sea pressure, p (dbar) (we neglect the influence of dissolved gases onthe freezing temperature).The need for the polynomial form of the Tf equation means solving the three-equation model(??) would be non-trivial, so we use the simpler two-equation model of the formriabLi = riciab(TiTf )rwcwu⇤G(TS)(Tf Tw) (4.4)103where Tf is calculated using the TEOS-10 function instead of the typical linear form (Eq. 4.3). In theabove equations r is density, L is latent heat of fusion, c is specific heat capacity, T is temperature,S is salinity; the subscripts i, b, w, and f refer to ice, ice-ocean boundary, water, and the freezingpoint, respectively. The ablation rate at the ice-ocean boundary, ab, is expressed as a change inthickness of solid ice per unit time, is positive for ablation, and is determined by the divergenceof the sensible heat flux at the phase change interface. The turbulent transfer coefficient for heatand salinity G(TS) is derived from a single observation beneath the Ronne Ice Shelf in Antarctica(Jenkins et al., 2010). We refer to this model as the ice-ocean model throughout this chapter.Friction velocity (u⇤) is calculated as:u2⇤ =CdU2, (4.5)whereU is the free-stream velocity beyond the ice-ocean boundary layer andCd the drag coefficient(Table 4.2). The drag coefficient is based on measurements below sea ice and is not constrainedby observations below ice shelves. Its use implies that the turbulent transfer coefficients estimatedfor sea ice are appropriate for the base of an ice shelf. Temporal variability of the boundary layerand the differing nature of the boundary flow beneath ice shelves, where the forcing comes frombuoyancy and tides, and beneath sea ice, where the primary forcing is the wind-driven drift of the icecover, could violate these assumptions. However, use of the values for Cd and G(TS) recommendedby Jenkins et al. (2010) is required until further observations below ice shelves are available. Byapplying the two-equation model Jenkins et al. (2010) found these values were able to reproduceobserved melt rates at the base of the Ronne Ice Shelf, Antarctica, within observational error.The model does not explicitly account for circulation induced by buoyant convection of themeltwater plume along the base of a sloping ice shelf. One consequence of this is that the modeldoes not directly account for the so-called ‘ice pump’, the possibility of ice accretion from the as-cending meltwater plume on the base of the ice shelf due to the pressure dependency of the freezingtemperature (-7.53 x 104 C dbar1). Therefore the possibility of ice accretion on the shallowerportions of the ice shelf or glacier tongue is not accounted for in the model. However, the relativelyshallow maximum draft (150 m) of ice in Milne Fiord means this effect will be relatively small (thefreezing point changes by ⇠0.1C over the water column). This is in contrast to Antarctica whereice shelves are an order of magnitude thicker, and a change in freezing point of>1C over the watercolumn in contact with ice means the ice pump can be the main driver of cavity circulation and basalmass balance.We also note that the ice-ocean model used here was developed for the near-horizontal slope atthe base of an ice shelf and may underestimate the melt rate near vertical ice walls, such as those ofthe MGT. Using a model based upon one-dimensional buoyant plume theory, Jenkins et al. (2010)showed that changing the slope of the ice-ocean interface from that of a floating ice tongue (basal104slope of⇠4%) to a vertical wall, resulted in a near doubling of the melt rate. This suggests the actualmelt rates along the near vertical margins of the MGT could be twice as high as those calculatedhere.We use vertical profiles of ocean properties and currents collected each year from 2011 to 2015in Milne Fiord to derive a depth-dependent melt rate. Then, assuming water properties and currentsmeasured at the mooring site were uniform throughout the fjord, we project the melt rate profileonto the ice draft DEM.To assess the effect of changing the input values to the calculated melt rates, a sensitivity analysiswas performed whereby the model was run adjusting one variable over its likely range while allother variables were held constant. A sensitivity analysis was run for all variables, however themodel is only very weakly sensitive to the values of internal ice temperature (a value of - 15Cwas used based on 10 m borehole temperature of the Ward Hunt Ice Shelf (Jeffries, 1991)) and thedensities of water and ice so we do not include the results of those runs. The range of values for thetemperature of the ambient water above the freezing temperature, or thermal driving, (TwT fw ) andU were determined from the variation of observed properties from field measurements. Estimatingthe appropriate range of possible values for Cd and G(TS) was more uncertain, given the paucity ofobservations under ice shelves. We chose to test model sensitivity to these parameters using a rangevalues derived from observations under ice shelves and sea ice, varying Cd from 1 x 104 to 1 x102 and GTS from 1 x 104 to 8 x 103 (McPhee, 1990; Shirasawa and Ingram, 1991; Holland andJenkins, 1999). Further investigation will be required to determine the most appropriate values forMilne Fiord.4.2.7 Ocean propertiesTo run the ice-ocean model we used conductivity-temperature-depth (CTD) profiles collected Mayof 2011 and 2012, and July of 2013, 2014, and 2015 at the mooring site in Milne Fiord using a 6HzRBR XR-620 CTD (except in 2011 when a 4Hz SBE19+ was used). To investigate the influencethat the MIS had on melt rates in the fjord, we also run the model using profiles collected a fewkilometers off the seaward edge of the MIS (see Fig. 4.1) in the same year. In 2011 and 2012the offshore profiles were collected as part of the SwitchYard Project (P.I. M. Steele, University ofWashington), while in all other years offshore profiles were collected as part of this study. Derivedvariables were calculated using the TEOS-10 Gibbs Seawater Matlab Toolbox.4.2.8 Ocean circulationWater speeds were measured using an ice-anchored, downward-looking 300 kHz RDI acousticDoppler current profiler (ADCP) at the mooring site in May 2011, July 2012 and July 2013. Wecalculated the time-averaged water speed over the duration of each deployment: 4, 7, and 10 days,in 2011, 2012, and 2013, respectively. The current profile was consistent between each deployment105so we calculated a time-average current profile over all years. The depth range of the ADCP waslimited to <100 m, so we estimate speeds below 100 m by assuming a constant speed with depthequal to the measured average between 75 m and 100 m. Speeds are taken to represent the free-stream current beyond the ice-ocean boundary layer. Finally, the vertical speed profile was assumedto be uniform throughout the fjord and invariant in time, so the same averaged speed profile wasused every year.4.2.9 Temperature timeseriesTheoretical considerations of thermodynamics at the ice-ocean interface suggest submarine meltrate varies in proportion to the thermal driving of the ambient water (Holland and Jenkins, 1999; Jenk-ins, 2011). To understand the variability of ocean temperature in Milne Fiord over time we deployedan ice-tethered mooring deployed through the epishelf lake from May 2011 toJuly 2014. Thermis-tors were suspended at 5 m, 25 m, 50 m, and 125 m depths. To calculate T fw salinity is also required,however we did not have salinity sensors moored at each depth for the full timeseries. Instead weuse the average salinity measured at each depth from all CTD profiles collected at the mooring sitebetween May 2011 and July 2015. The range of salinity then gives an estimate of uncertainty in thecalculation of T fw . Salinity at 5 m, 25 m, 50 m, and 125 m depths varied by 0.1, 10, 6, and 2 g kg1,respectively. The saline coefficient of the freezing point in the TEOS-10 equation is approximately-59 mK (g kg1)1, resulting in an error in the calculation of T fw due to salinity variation of approx-imately 0.001, 0.059, 0.035, and 0.012C at 5, 25, 50, and 125 m, respectively. The largest errorin T fw , at 25 m depth, was equal to approximately 10% the total variation in thermal driving at thisdepth over the full timeseries.4.3 Results4.3.1 Milne Glacier grounding lineThe Milne Glacier is presently grounded on a reverse-sloping bed at about 150 m depth, the beddeepens upstream of the grounding line (Fig. 4.2). The glacier thins rapidly within the first 5 kmdownstream of the grounding line and then tapers more slowly toward the terminus. We note theaerial survey line flown by the 2014 Icebridge mission crossed over from the central ice stream ofthe Milne Glacier to the highly fractured north-east and south-west ice streams so surface elevation-derived thicknesses for some sections are highly variable (e.g. from 200 m to 5 km, and 10 km to12 km). The DEM shown in Fig. 4.1 is better constrained with additional data and shows the cross-fjord variability in the thickness of the MGT. Upstream of the grounding line the bed over-deepensto 200 m below sea level within 2–3 km, and remains below sea level for 26 km inland (not shown).1064.3.2 Glacier surface velocitiesGlacier surface velocities showed substantial spatial variability and interannual variation from 2011to 2015 (Fig. 4.3). From 2011 to 2015, glacier velocities averaged 100 m a1 at the groundingline, with a maximum of 150 m a1 in 2011, and a minimum of 70 m a1 in 2012. These valuesare broadly consistent with the measured displacement of the ablation stake just upstream of thegrounding line (MG01), which showed an average displacement of 95 m just upstream of thegrounding line from 2012 to 2015 (Table ??). Downstream of the grounding line velocities of theMGT generally decreased rapidly in the first 5 km to about 60 m a1 and gradually slowed towardthe terminus. However, velocities and flow patterns in 2012 and 2013 were somewhat different,with nearly uniform velocities of 60 m a1 along the full length of the MGT in 2012, while in 2013velocities were <25 m a1 over most of the MGT. Ablation stake displacements on the MGT from2012-2013 were anomalously high, with both MG02 and MG03 showing 290 m displacement overthat interval. Measurements from 2013 to 2015 showed a displacement of 60 m a1, consistentwith the speckle tracking data in those years.A review of available satellite imagery shows the main transverse fracture at the groundingline, which is first apparent in 2006, grew and extended completely across the width of the centralice stream by 2012, perhaps even 2011. This observation suggests that the MGT became partiallydetached from the grounded glacier sometime during or prior to 2012. If so, this would help explainthe anomalous inter annual velocities for the MGT in 2012 and 2013, and suggest that the melt ratescalculated in these years must be viewed in with caution as the MGT was not in steady state.4.3.3 Ocean propertiesAnnual CTD profiles revealed the vertical distribution of heat available to drive submarine meltingchanges annually (Fig. 4.4). The strong salinity stratification in the fjord above 50 m depth, thatwas not present in offshore profiles, was due to the MIS damming runoff within the fjord. Thenearly fresh surface waters (<15 m depth) indicated the presence of the epishelf lake. The epishelflake was associated with a near-surface temperature maximum between 5 m and 15 m depth, thatvaried between 0.25C and 2C above freezing in 2011 and 2015, respectively. Offshore profilescollected in July of 2013-2015 showed a slight surface freshening and warming due to sea ice melt,but temperatures remained at the in situ freezing point from the surface to 30 m depth. Below 50m depth properties in Milne Fiord were similar to properties offshore that year. The temperatureabove freezing, or thermal driving, increased steadily with depth below 50 m, the heat supplied bywarm waters of Atlantic origin. Deep fjord water near the depth of the Milne Glacier grounding lineshowed temporal variation in thermal driving of 0.5C among years, 2013 had the lowest thermaldriving (0.9C) at depth, while 2015 had the highest (1.4C). Temperature changes above 50 mand below 50 m were not correlated (e.g. in 2013 near-surface waters had the highest thermaldriving, while deep waters had the least of all years). The lack of correlation over depth suggests107that surface waters are influenced by local surface conditions (i.e. air temperature, snow cover,solar radiation, surface runoff), while deep waters are influenced by other processes (e.g. offshorevariation transmitted into the fjord by intermediary exchange flows). Some of the differences intemperature among years, particularly near the surface, were due seasonal differences in the timingof profiling so interannual trends should not be inferred from the results.Measured ocean current speeds revealed flow in the fjord was weak and baroclinic, speedsincreased from around 1 cm s1 near the surface to a maximum of 3 to 6 cm s1 between 50-60 m,then decreased below this depth to about 2 cm s1 (Fig. 4.5). Tidal oscillations on the order of 2cm s1 accounted for a substantial portion of the flow speeds in Milne Fiord. However, the MISformed an obstacle to flow, limiting speeds to ⇠ 1 cm s1 in the upper 25 m of the water column.We have assumed constant flow below 100 m, however we acknowledge that other processes, suchas intermediary exchange below the MIS could lead to variable and potentially higher flow speedsat these depths (Jackson et al., 2014b).4.3.4 Ocean model sensitivityThe ice-ocean model shows a linear dependence on TwT fw , U , and GTS, and a non-linear depen-dence on Cd (Fig. 4.6). Estimated melt rates are most sensitive to values of TwT fw and U , whichin our model are based on field observations. TwT fw is known from CTD profiles whileU is basedon ADCP measurements. We assumed the properties measured at the mooring site were represen-tative of properties throughout the fjord. CTD transects collected at many sites throughout the fjord(Chapter 2) have shown water properties to be relatively homogenous, so our values of TwT fw arefairly robust. However, we only have current speed measurements from a single location, and onlyover the top 100 m, so horizontal or vertical differences in current speeds from those used in themodel could change the estimated melt rates accordingly. The model is slightly less sensitive to thechosen values of Cd and GTS, however the values used here are based on other studies, so furtherobservations are required to validate their use in this system. This study is most concerned with theeffect of depth variation of ocean properties on the vertical distribution of melt rates, so although theactual magnitude of the melt rates calculated have some degree of uncertainty, the relative changeswith depth are based on field observations and should be fairly robust. We reiterate that the resultsof the model must be interpreted with proper consideration of above uncertainties.4.3.5 Submarine melt ratesWidth-averaged MGT melt ratesWidth-averaged submarine melt rates calculated by the divergence of ice flux and the ocean thermo-dynamic model along the length of the MGT for 2011 to 2015 are shown in Fig 4.7. The MGT thinsrapidly from 150 m depth to<75 m depth within the first 2.5 km downstream of the grounding line,108then thins more gradually toward the terminus where the ice is <10 m thick. The bottom profile ofthe MGT is similar to that of other glacier tongues in Greenland (Rignot et al., 2001). Melt ratescalculated by both methods showed a similar pattern of high melt rates averaging ⇠4 m a1 withinthe first 2.5 km downstream of the grounding line, then decreasing to near zero within 5 km. Theice-ocean model indicated a slight increase in melt rate near the terminus, while the ice flux modeldid not. There was an ⇠1 m a1 difference in melt rate between the two methods over the mid-portion if the glacier tongue. Given the assumption inherent in each method it is difficult to quantifythe uncertainty in the absolute accuracy of the calculated values, however the offset is likely withinerror limits. In 2012 and 2013 high variability in ice surface velocities along the MGT suggested theglacier may not have been in steady state, and the values estimated by the ice flux method in thoseyears may not be reliable. Overall, the two method show general agreement in both the magnitudeand distribution of melt along the glacier tongue, and relatively little interannual variation (with thenoted exception of 2012 and 2013).4.3.6 Area-averaged melt ratesAnnual area-averaged submarine melt rates for the MGT and MIS from 2011 to 2015 calculatedusing the grounding line flux, divergence of ice flux, and the ice-ocean model are shown in Table4.3. The average melt rates for the MGT calculated by the grounding line flux method and thedivergence of ice flux method each year are comparable, showing <1 m a1 melt, and possiblenet ice accretion in some years (negative melt rates). In contrast, the ice-ocean model results areconsistently higher, showing a total average melt rate of 1.6 m a1 for the MGT. Average melt ratescalculated by divergence of ice flux and the ice-ocean model over the first 2.5 km downstream ofthe grounding line are consistently higher, averaging 3.5 m a1. Annual melt rates for the first 2.5km calculated by the divergence of ice flux method are quite variable, ranging from a maximum of6.0±1.9 m a1 in 2011, to a minimum of 0.9±0.3 in 2012, owing to the difference in ice surfacevelocities in those years. Neglecting 2012 and 2013, when velocities of the MGT may have beenanomalous due to a partial detachment from the glacier, gives an average melt rate for the first 2.5km of the MGT of 5 m a1. Submarine melt rates calculated by the ocean thermodynamic modelfor the MIS averaged 1.4 m a1 over all years with only 0.2 m a1 variation among years.4.3.7 Spatial distribution of melt ratesThe spatial distribution of submarine melt rates in Milne Fiord from 2011 to 2015, calculated fromthe ice-ocean model using the CTD profiles acquired at the mooring site are shown in Fig. 4.8).Melt rates are spatially heterogenous in both the magnitude and interannual variability. Thermo-dynamic melt models predict a linear or quadratic dependence of melt rates with thermal driving(Holland and Jenkins, 1999; Jenkins, 2011), and consequently the vertical distribution of heat inMilne Fiord results in melt rates that vary vertically over the water column. The melt rate is thus109largely dependent on ice draft (see the Discussion for a consideration of other processes that influ-ence melt rate). High melt rates occur where the ice is thick and penetrates deep into warm AW(e.g. the Milne Glacier grounding line and Outer Unit of the MIS), and where the ice is quite thin(<15 m deep) and in contact with warm water of the epishelf lake (i.e. along the margins of theepishelf lake). At mid-depths, for ice drafts between approximately 25 m and 60 m, the melt rateis generally lower (<1 m a1), but shows more interannual variability. The Central Unit of theMIS shows a high degree of both spatial and interannual variability. The spatial variability is dueto the heterogeneity of ice thickness across the unit, owing to its composite origin of marine iceaccretion and input from tributary glaciers. Much of the ice of the Central Unit is between 30 mand 50 m thick and shows much more interannual variability than the >80 m thick ice of the OuterUnit. The variability is caused by the high current speeds at 50 m depth (Fig. 4.5) amplifying theeffects of small temperature variations at these depths (Fig. 4.4), leading to pronounced differencesin calculated melt rate for ice with a draft around 50 m.The influence of the MIS and the epishelf lake on submarine melting in Milne Fiord is apparentin the results of the ice-ocean model run using profiles acquired outside the fjord (Fig. 4.9). Similarto the model run using the mooring profile, melt rates are consistently high where the ice is thick andin contact with AW. However, where ice is very thin melt rates are low compared to the model runwith the mooring CTD profile. A comparison between melt rate calculated from the mooring CTDprofile versus the offshore CTD profile are shown in Figure 4.10, revealing the highest melt rateanomalies occur where the ice is thin. This is evidence that the presence of the MIS, and its effecton the water column structure within the fjord results in higher near-surface melt rates than wouldotherwise be expected if the ice shelf was not present. The Central Unit of the MIS also shows ahigh melt rate anomaly in 2012. This due to the substantial difference in salinity, temperature, andthermal driving down to 40 m depth between the mooring CTD profile and the offshore CTD profilethat year (Fig. 4.4).We note that the high near-surface ablation rates calculated using the mooring profile are onlyvalid where the water column structure and flow speeds are well represented by measurement takenat the mooring site. CTD profiles acquired throughout the fjord indicate water properties are similareverywhere landward of the Outer Unit, including under the Central Unit (Chapter 2 and 3). Noprofiles have been acquired through the Outer Unit, the water properties here are unknown. Some-where below the Outer Unit water properties of the upper water column must transition from havingcharacteristics similar to the mooring site, with a strong stratification and an elevated heat content,to having characteristics similar to offshore, where the near-freezing surface mixed layer extends to30–50 m depth. It is therefore uncertain whether melt rates estimated for the relatively thin portionsof the Outer Unit (i.e. along the re-healed fracture and just to the south of the western extremeof the fracture) are best represented by those calculated using the mooring profile or the offshoreprofile. In general, melt rates for the Outer Unit are best represented by those calculated using the110offshore profile (Fig. 4.9), while melt rates for the Central Unit and MGT are best represented bythose calculated using the mooring profile (Fig. 4.8).4.3.8 Thermal driving depth-dependenceIn Figure 4.11 the variability of thermal driving over 3-years at different depths is shown. The wa-ters of the epishelf lake, at 5 m depth, showed the highest average thermal driving and the greatestmagnitude of variation, with seasonal increases >2C. This seasonal variation was due to inflowof relatively warm meltwater and solar heating. Waters at 25 m depth showed an abrupt increaseof 0.5C in June 2011, then a long-term gradual decrease over the remaining 3 years of the record,with very little seasonal variation. The flow of water at this depth was largely restricted by the MIS,and was influenced by the accumulation of subglacial runoff within the fjord and its slow drainageunder the ice shelf. Water at 50 m depth had the lowest thermal driving and the least interannualvariability. At this depth the flow of water under the ice shelf was less restricted and had character-istics of the surface mixed layer of the Arctic Ocean. At 125 m depth there was substantial seasonaland short-term (days to weeks) variation in thermal driving, with annual variation of 0.5C. Wateras this depth was at the interface between Polar Water and Atlantic Water, and the variation waslikely representative of offshore depth variation of the main Arctic halocline being transmitted intoMilne Fiord under the MIS. Overall, the records indicated a strong depth-dependence on both themagnitude and variability of thermal driving in Milne Fiord, that could substantially alter subma-rine melt rates over time. The variability of thermal driving at 125 m is of particular interest, as thisrecord is representative of changes in thermal driving of waters in contact with the grounding lineof the Milne Glacier.4.4 DiscussionSubmarine melt rates are determined by the heat content and circulation of ocean waters in contactwith ice. We have shown that the heat content in Milne Fiord varies with depth, and the spatialdistribution of melting in the fjord is therefore largely dependent on ice thickness. The heat contentin Milne Fiord has a bimodal distribution, with a near-surface maximum in the epishelf lake, and agradient toward a deeper maximum in the Atlantic layer, and this has repercussions for the vertical,and thus spatial, distribution of melting in the fjord.4.4.1 Grounding line meltUnsurprisingly, melt rates are highest near the Milne Glacier grounding line, a finding consistentwith other studies of submarine melting of glacier tongues and ice shelves (Rignot et al., 2001; Rig-not and Steffen, 2008). The average melt rates calculated within 2.5 km of theMilne Glacier ground-ing line are substantially less than those estimated near the grounding lines of other glacier tongues111in Greenland. Rignot et al. (2001) estimated melt rates between 22 m a1 and 26 m a1 within10 km of the grounding line for six glaciers with ice tongues in northern Greenland. However, thedepths of the glacier grounding lines surveyed in Rignot et al. (2001) ranged from 460 to 650 m,so the combined effect of the pressure-induced depression of the freezing point and the increase oftemperature of the Atlantic water with depth (to a maximum between 300 m and 800 m; Carmack,1990) means that thermal driving generally increases with depth in the Arctic, and the higher meltrates for the thicker glaciers in Greenland are expected.Although Milne Glacier appears to have a substantially lower melt rate than Greenland coun-terparts, the melt could still cause retreat of the grounding line if basal melt is greater than surfaceaccumulation and dynamic thickening. Retreat of the grounding line into deeper water could in-crease ice flux, and increase the glacial contribution to sea level. Neglecting other effects, if theaverage basal melt rate of 5± 2 m a1 induces an equivalent thinning of the glacier this could causethe grounding line to retreat approximately 400 ± 200 m a1 given the glacier surface slope (1.5%from 2014 Icebridge ATM data) and bed slope (-3% from MCoRDs data). The appearance in 2013of new transverse fractures a few hundred meters upstream of the previous hinge fracture suggestedthe grounding line has retreated. The average rate of basal melt at the grounding line is fives timesgreater than the estimated surface melt rate for the fjord, suggesting basal melt is a major factor inretreat of the grounding line. The contribution from dynamic thinning, however, remains unclear.The Milne Glacier is a suspected surge-type glacier (Jeffries, 1984; Copland et al., 2003), andthe differences in surface velocities recorded at the grounding line between 2011 and 2012 mayindicate the end of a surge cycle, which in the QEI are characterized by quasi-periodic fast flows(7–15 years), followed by periods of slow flow (30–40 years; Van Wychen et al., 2016). Recentevidence also suggests the existence of ‘pulse’ type tidewater glaciers, which are characterized bybrief periods of accelerated motion (2–5 years) (Van Wychen et al., 2016). In addition, the possibledetachment of the MGT from the glacier suggest the melt rates calculated by the divergence of iceflux method at the grounding line in 2012 and 2013 are artifacts of unsteady dynamics. The lack ofoceanographic data for the fjord before 2011 means it is unclear what role changes in ocean forcingmay have had in influencing glacier dynamics prior to this time.4.4.2 Near-surface meltThe enhanced near-surface melt rates in Milne Fiord due to the strong heat and salt stratificationcreated by the MIS dam are a unique feature of this fjord. Seasonal warming of surface waters fromsolar radiation and runoff are expected in other glacial fjords, and may lead to seasonally enhancednear-surface melting, however the effect is transient as surface waters return to near-freezing duringwinter. In contrast, the strong stratification in Milne Fiord is present year-round, and temperaturesremain well above above freezing even in mid-winter (e.g. almost 3C above freezing at 5 m inJanuary 2013; Fig. 4.11), so there is sufficient heat available to drive near-surface melting in Milne112Fiord all year. Melting during winter could account for a significant portion of the total annualnear-surface melt in Milne Fiord. It is important to note that the calculated melt rates assume steadystate ice thickness. Melting and thinning of ice over time will substantially alter the melt rate overtime for ice drafts less than 20 m due to the strong gradients in thermal driving near the surface.The elevated heat content of the upper water column in Milne Fiord is due to the existence of theMIS. The effect of the MIS was shown by the lower near-surface melt rates calculated when the ice-ocean model was run with the profiles collected offshore. Using the offshore profile was analogousto removing the MIS and its influence on the water properties of the fjord, as free-exchange at thefjord mouth would mean surface properties within the fjord would be very similar to those offshore.Surface temperatures in the Arctic Ocean remain close to the freezing point down to the base of themixed layer (30-50 m depth) year-round. The year-round enhanced near-surface melt rates in MilneFiord are then entirely dependent on the presence, and structural integrity, of the MIS dam.MIS meltingEnhanced near-surface melting in Milne Fiord means the MIS may be contributing to its owndemise. Elevated melt rates of the inner margin of the Central Unit of the MIS are owing to itscontact with the epishelf lake, whose existence is dependent upon the integrity of the MIS. Analysisof satellite imagery reveals the southern edge of the MIS retreated northward in recent years, andthe Central Unit has undergone widespread fracturing and calving (Chapter 3). Our results suggestthat weakening and retreat of the southern edge of the Central Unit, is due, in part, to submarinemelting (likely in combination with mass loss from surface ablation). This observation also suggestsa mechanism for the formation of the epishelf lake, something not explicitly addressed in previousstudies (Jeffries, 1984; Mortimer et al., 2012; Chapter 3). The epishelf lake is thought to have ex-panded as a region of the MIS, known as the Inner Unit, was replaced with freshwater lake ice.We have provided evidence that indicates enhanced submarine melting of the ice shelf, driven bytrapped surface runoff behind the ice shelf, was a plausible mechanism for the retreat of the ice shelfmargin, and the expansion of the epishelf lake. The melt rate for regions of the MIS with a draftless than 35 m showed a 50% increase when the ice-ocean model was run with the mooring profilecompared to the offshore profile, highlighting the impact of the ice dam on melt rates in the fjord.Continued melting induced by the epishelf lake could eventually lead to a compromise of the MISdam, and a loss of the epishelf lake.We estimated a surface ablation rate of 0.78 m a1 in Milne Fiord and an areal average basal meltrate for the MIS of 1.4 m a1, indicating basal melt accounted⇠64% of total melt between 2011 and2015. This is similar to the estimate of Mortimer et al. (2012) who suggested submarine meltingmay have accounted for ⇠73% of the overall thickness change of the MIS over the period 1981 -2009. Longer-term surface mass balance records from the Ward Hunt Ice Shelf show that surfacemelt increased substantially in recent years, from 0.08 m a1 between 1989-2003, to an average of1130.5 m a1 between 2002-2005 (Braun et al., 2004; Mueller et al., 2006). These observations areconsistent with the dramatic regional increases of surface mass loss across the CAA over the period2003 to 2009 (Gardner et al., 2011; Gardner et al., 2013). The general trend suggests that althoughsubmarine melting appeared to have been the dominant factor in thinning of the ice shelves in pastdecades, surface melt has become an increasingly important factor in the past decade.The rapid increase in surface melt over the past decade suggests changes in basal melt shouldalso be considered. Our 4-year study is too brief to identify a long-term trend in basal melt rate forthe MIS, but we did find interannual variability in the spatial distribution of melting. This was due toboth local changes in the water column in Milne Fiord, driven largely by annual variation in runoff,but also changes in the properties of ambient source waters offshore. Jackson et al. (2014a) showedthat the water column structure in the Lincoln Sea, north-east of Ellesmere Island, from 1991 to2012 varied yearly owing to the complex circulation in the region. The region is a bifurcationpoint, where waters from the Canadian and Eurasian Basins flow toward Nares or Fram Strait, orare advected westward along the continental shelf. Minor shifts in circulation patterns thereforesubstantially altered the water properties along northern Ellesmere Island. How regional circulationpatterns influenced basal melt rate of ice shelves along northern Ellesmere Island in the past, andhow future circulation and water column changes may in the future is worthy of further investigation.Although this is beyond the scope of the present study, what is apparent is that ocean-driven meltrates in Milne Fiord are influenced by a variety of factors over a range of spatial and temporal scales.4.4.3 Circulation in Milne FiordOne of the key factors in determining melt rates is ocean circulation. The water column in MilneFiord has sufficient thermal driving to induce melt rates substantially higher than those estimatedhere, however the transfer of heat to the ice boundary appears to be limited by relatively weakcurrents. The main factors driving circulation in a typical ice covered fjord are tides, intermediaryexchange, and buoyancy-driven flow from freshwater discharge at the head of the fjord. Belowan ice shelf or glacier tongue, additional thermohaline processes can also drive circulation. MilneFiord is somewhat atypical, in that regardless of the forcing, the MIS impedes flow in the upperwater column. However, below the draft of the MIS, we expect the main drivers of circulation arethe same as a typical fjord. ADCP records showed that tidal flows below 80 m, deeper than the draftof the MIS, were on the order of 2 cm s1. Lacking observations of current speed below 100 mdepth we therefore estimated a constant speed of 2 cm s1. But what are the potential contributionsof intermediary exchange and buoyancy-driven circulation to flow speeds at depth?We can obtain a rough idea of the significance of buoyancy-driven flow by using a simple modelof estuarine circulation. To do so, we assume a simple 2-layer return flow driven by the release offreshwater at the head of the fjord. Although recent studies, including this one, have shown that cir-culation in glacial fjords is more complex than a simple 2-layer model (Straneo and Cenedese,1142015), the exercise is still useful in providing a conceptual understanding of the magnitude ofbuoyancy-driven flow as a factor in fjord circulation. We assume freshwater input at the head ofthe fjord drives an outflowing surface layer with an inflowing bottom layer. From a considera-tion of the steady state salt and volume continuity of a fjord, and neglecting diffusion, Knudsen’shydrographical theorem provides an estimate of the volume flow in the upper and lower layersQ1 =QfwS2S11 (4.6)andQ2 =Qfw1 S1S2(4.7)where Q1, S1 and Q2, S2 are the volume flows and salinities of the upper and lowers layers, respec-tively, Qfw is the freshwater (assumed zero salinity) discharged at the head of the fjord. We cantherefore estimate volume flow in the two layers from CTD profiles of salinity and an estimate ofQfw. In a glacial fjord Qfw =Qs f +Qb, where Qs f is freshwater runoff from surface melt and Qb isfreshwater input from basal submarine melt. In reality, a portion of surface melt enters the fjord atthe surface, and feeds the epishelf lake, while the remainder enters the fjord at depth at the bed ofthe glacier. However, for the purpose of this estimate we will use the combined total to understandthe influence of total runoff on driving estuarine circulation.We estimate Qs f from regional surface mass loss for the CAA from 2006-2009 (Gardner et al.,2011). Applying an average loss of 310 kg m2 a1 over the 1108 km2 glaciated area of the MGcatchment gives an annual average runoff of 3 x 108 m3 a1, equivalent to 12 m3 s1, or 73 m3 s1if only averaged over the 2 month melt season.The other source of freshwater at the head of the fjord is submarine melting of glacial ice nearthe grounding line. The average melt rate within 2.5 km of the grounding line was 3.8 m a1 overall years, equating to roughly 4 x 107 m3 a1 of freshwater input from submarine melt. Submarinemelt therefore accounts for only 10% of the total freshwater input at the head of the fjord, 90%originates as surface runoff.Based on salinity and velocity profiles we estimate the upper layer is between 30 and 60 mdepth, and the lower layer is between 60 and 260 m depth (the sill depth). The average salinitiesat these depths give S1 = 30 g kg1 and S2 = 33 g kg1. Inputting the above values into Eq. 4.7results in upper and lower layer volume flows on the order of 100 m3 s1. Assuming a rectangularcross-section equal to the height of each layer, and a fjord width of 5 km, gives estimated currentspeeds of <<1 cm s1. Even if Qfw is an order of magnitude greater, the estimated current speedsare still <1 cm s1. It therefore appears unlikely that the small volume of freshwater dischargedinto Milne Fiord will drive strong fjord-scale currents, and may therefore play only a minor rolein determining average melt rates across the fjord. While subglacial discharge could induce high115melt rates at the grounding line, we expect the effect will be localized along subglacial dischargechannels and limited to the brief summer melt season. Similarly, outflow along the basal channelin the MIS may enhance local melt rates in the channel, but the effect on fjord-scale circulation issmall given the slow export over several months.Another process that could enhance flows in Milne Fiord is intermediary exchange below theMIS. Jackson et al. (2014b) observed frequent velocity pulses that exceeded 50 cm s1 in the upper200 m of Sermilik Fjord, Greenland, lasting several days between September and May 2011-2012.The pulses were largely attributed to wind-driven upwelling along the continental shelf setting upbaroclinic exchange between the fjord and the continental shelf. The authors suggested that glaciermelt rates could vary substantially throughout the year as a result. Similarly, Stigebrandt (1990)found externally forced baroclinic exchange driven by density fluctuations at fjord mouth was anorder of magnitude higher than tidal or estuarine flow in a Scandinavian fjord.Although we lack long-term current measurements at depth that can directly address variabilityin flow due to offshore exchange, we can obtain a first order estimate of exchange flow in MilneFiord from changes in isotherm depths from the multiyear temperature timeseries presented in Chap-ter 2. Those observations showed that isotherms around 125 m depth varied by up to 50 m depthover the course of several days. If the change in height of the isotherm was uniform throughoutthe inner ⇠66 km2 of the fjord, this is equivalent to an inflow volume of 3 x 109 m3. Immediately,we see this is an order of magnitude larger than the total annual freshwater inflow from surface andsubmarine melting. If we assume the inflow of water occurred over a depth of 200 m across the 5km width of the fjord, and occurred over 5 days, this would have resulted in an average inflow speedon the order of 1 cm s1, although peak flows could be substantially higher. This is an admittedlycrude calculation, but it does suggest that intermediary exchange flow is of the same order of mag-nitude as tidal flow, and further investigation is this process is required to accurately estimate basalmelt rates in Milne Fiord.In summary, it appears that tidal and intermediary exchange are important drivers of circulationin Milne Fiord over most of the year. Thermohaline circulation in the ice cavity is a factor that wehave not addressed here, but will discuss further in the next section. It is likely that buoyancy-drivenflow from subglacial discharge is an important factor in determining localized melt rates in summer,but does not appear to play a large role in fjord-scale circulation and can assumed to be negligibleover most of the year. Overall, melt rates are then largely determined by tidal and intermediaryexchange flows and the thermal driving of the ambient water column.4.4.4 Future workThe thermodynamic melt model used here was based on the assumption that water properties andcurrents speeds were uniform throughout the fjord, and that physical parameterizations of heat trans-fer through the ice-ocean boundary layer based on observations below a single Antarctic ice shelf116were appropriate for Milne Fiord. Clearly, this is a substantially oversimplified view of the fjordand the results of the melt model presented here must be viewed in this context, as a useful firstattempt that can inform future studies.There are several improvements that could be made to the ocean melt model. First, circulationunder the glacier tongue (and ice shelf) can be driven by the thermohaline differences which resultfrom mass and energy exchange at the ice-ocean interface. This process has been modelled using aone-dimensional flow-line model based on buoyant plume theory, where the rising plume along theice base is treated as a turbulent gravity current initiated at the grounding line, by a flow of subglacialwater or in situ melting (Jenkins, 1991; Jenkins, 2011). The subsequent evolution of the plume asit ascends along the ice base has been used to explain the observed distribution of melting andfreezing along several ice shelves in Antarctica and glacier tongues and tidewater glacier faces inGreenland and Alaska (Jenkins, 1991; Jenkins, 2011). However, given the complex ice topographyand bathymetry of Milne Fiord, circulation may be best derived from 3-dimensional numericalmodelling, as has been done for other tidewater glacier fjords in Greenland (Xu et al., 2012; Kimuraet al., 2013; Sciascia et al., 2013; ; Xu et al., 2013; Cenedese and Linden, 2014). A numericalmodelling approach is likely required for Milne Fiord because melt rates for the MIS are also ofinterest, and the complex ice topography may defy reduction to a one- or two-dimensional model.Three-dimensional numerical models, however, still rely on turbulent transfer and drag coefficientsthat have not been validated for these systems, so it is apparent that further observations are needed.In this study we chose to use the two-equation formulation of ice-ocean thermodynamics ratherthan the three-equation formulation, due primarily to the error that a linearization of the freezingpoint equation would introduce across the large salinity range in Milne Fiord. However, in doingso the effects of salinity stratification below the ice, both ambient stratification and that induced bymelting at the ice base, have largely been neglected. The value of the combined turbulent transfercoefficient for heat and salt (GTS) does in theory account for this, however it was validated belowan Antarctic ice shelf, where ambient stratification is much less than that observed in Milne Fiord(Jenkins et al., 2010). Although Jenkins et al. (2010) emphasized that there was no difference in thefit between calculated and observed ablation rates using the two-equation formulation compared tothe three-equation formulation for their limited set of measurements, they do state that the three-equation formulation is likely the best parameterization to use over a broader range of oceanographicconditions given its explicit treatment of salt transfer through the boundary layer. A comparison ofmelt rates calculated between the two- and three-equation formulations for the conditions observedin Milne Fiord would be a useful avenue for future research. Fundamentally, however, improve-ments to the accuracy of derived melt rates from measured ocean properties can only come withvalidation against direct observations.Although the actual magnitude of melt rates calculated using the ice-ocean model must beviewed with caution given the limitations to the model discussed above, the depth-dependence is117likely a robust finding. Thermodynamic melt models predict a linear or quadratic dependence ofmelt rates with thermal driving (Holland and Jenkins, 1999; Jenkins, 2011), and the vertical dis-tribution of heat in Milne Fiord indicates melting will vary substantially over the water column.Variations in the vertical and horizontal distribution of melt, assuming values of Cd and G(T,S,TS)are constant throughout the fjord, is therefore dependent on the spatial variation in thermal drivingand current speed. Current speeds are the least well constrained of these two measurable variables,and suggests these would be a valuable focus for future field studies. In particular, current speeds atthe grounding line of the Milne Glacier could be substantially higher in the vicinity of a subglacialdischarge channel, potentially leading to much higher melt rates in localized regions. In addition,current speeds at the base of the MGT, and along the basal outflow channel in the MIS, could behigher and result in higher melt rates at these locations.4.5 Summary and conclusionWe have estimated submarine melt rates in Milne Fiord using independent methods, including adivergence of ice flux model and an ocean thermodynamic model. Melt rates calculated by bothmethods were broadly consistent, and revealed a depth-dependence. This was related to the verticaldistribution of heat in the fjord and vertical velocity variation. The depth-dependence created spatialheterogeneity of melt rates owing to the variability of ice thickness in Milne Fiord. Melt rates werehighest where thick ice penetrated into warm AW, including at the grounding line of the MilneGlacier and below the Outer Unit of the MIS. The ice-ocean model also predicted enhanced near-surface melt rates within the fjord caused by the elevated heat content of the highly stratified watersdammed by the MIS. We found substantial interannual and spatial variability of melt rates aroundthe Central Unit of the MIS, caused by its mean draft being at the depth of maximum current speeds,so minor variations in thermal driving caused large changes in melt rate.This study provides the first direct estimates of submarine melt rates for an ice shelf or tidewaterglacier in the Canadian Arctic Archipelago. The oceanography of glacial fjords in the CAA is ingeneral poorly known. If an effort is made to understand or predict how ocean forcing influencestidewater glaciers in the region, measurements of thermal driving, which are relatively easy to ob-tain by CTD profiling, must be accompanied by observations of circulation patterns, which requiresustained, long-termmonitoring. In Milne Fiord, tidal flows and intermediary exchange appear to bethe main drivers of circulation, however as surface meltwater production increases, buoyancy-drivencirculation may be an increasingly important factor. Although these findings are generally transfer-able to other fjords in the region, we expect different fjords will respond in different ways owing tovariations in bathymetry, ambient stratification, the magnitude and seasonality of meltwater input,ice-cover, and glacier dynamics.Milne Fiord contains the last intact ice shelf along the northern coast of Ellesmere Island, retainsone of only a few floating glacier tongues in the Arctic, and the last known epishelf lake in the118Northern Hemisphere. Our results have revealed that the evolution, and eventual fate, of thesefeatures is interlinked. The MIS impounds freshwater within the fjord, which enhances melt rateswithin the fjord, which in turn is leading to a thinning and erosion of the MIS. The thinning of theMIS could eventually contribute its own demise, and the subsequent loss of the epishelf lake. Theloss of the epishelf lake will reduce submarine melting along the margins of the MGT, however,the associated loss of perennial lake ice cover that has been observed with the loss of other epishelflakes (e.g. in Disraeli Fiord; W. Vincent, pers. comm.), would likely result in increased calving ofthe Milne Glacier. As the Milne Glacier is currently grounded on a reverse-sloping bed, increasedcalving could result in further grounding line retreat into deeper water, further enhancing submarinemelting, and accelerated glacier flux to the ocean.1194.6 TablesTable 4.1 Annual horizontal displacement of ablation stakes on the Milne Glacier.Stake Latitude* Longitude* Time Interval DisplacementName (N) (W) (m)MG01 82.438880 -80.234604 07/2012 - 07/2013 9907/2013 - 07/2014 9407/2014 - 07/2015 93MG02 82.463561 -80.429472 07/2012 - 07/2013 28807/2013 - 07/2014 6407/2014 - 07/2015 61MG03 82.528931 -80.673765 07/2012 - 07/2013 29507/2013 - 07/2014 6307/2014 - 07/2015 na* Positions are given at the time of ablation stake installation in July 2012.120Table 4.2 Values of physical constants and parameterizations used in ocean thermodynamic ice shelf ablationmodel.Symbol Value Units Descriptionri 900 kg m3 Density of iceLi 334 000 J kg1 Latent heat of fusion of icerw 1024 kg m3 Density of seawatercw 3974 J C1 kg1 Specific heat capacity of seawaterci 2009 J C1 kg1 Specific heat capacity of iceTi -15 C Internal ice shelf temperatureCd 0.01 Drag coefficientG(TS) 0.006 Turbulent transfer coefficientTable 4.3 Submarine melt rates (m a1) in Milne Fiord from 2011 to 2015.Grdln flux Divergence of ice flux Ocean therm. modelYear MGT MGT MGT MGT MGT MIS(2.5 km) (2.5 km)2011 0.2±0.4 0.6±2.5 6.0±1.9 1.2±1.1 3.0±1.3 1.2±0.82012 0.0±0.4 -0.2±0.5 0.9±0.3 1.5±1.1 3.3±1.4 1.6±0.82013 -0.2±0.3 -0.1±1.4 3.0±0.9 1.7±1.0 2.6±1.0 1.6±0.72014 0.0±0.3 0.3±2.0 4.6±1.4 1.6±1.3 3.3±1.6 1.1±0.82015 0.2±0.4 0.2±1.9 4.4±1.7 2.2±1.6 3.7±1.8 1.3±1.0The three methods used to calculate submarine melt rates are the grounding line flux,the width-averaged divergence of ice flux, and the ocean thermodynamic melt model.Values reported are area-averaged for the entire Milne Glacier tongue (66 km2),the first 2.5 km downstream of the grounding line where melt rates are highest, andthe Milne Ice Shelf (215 km2). One standard deviation is shown.1214.7 Figures    Ice draft (m ) 0 50 100 150 2011 2012 2013 2014 2015 Mooring Grd. line A B Outer Unit Central Unit 5 km5 kmFigure 4.1Map of Milne Fiord. In A) locations of ocean profiles (coloured circules) collected between 2011and 2015 are overlain on a RADARSat-2 image acquired in January 2015. Note that profiles were collectedat the mooring site (black triangle) in Milne Fiord in all years. Labeled features include the Milne Glacier(MG), the Milne Glacier tongue (MGT), the Milne Fiord epishelf lake (MEL), the Milne Ice Shelf (MIS),and multiyear land-fast sea ice (MLSI). The grounding line of the Milne Glacier is shown in red. B) Digitalelevation model of ice draft in Milne Fiord. Grey area indicates thin (⇠1 m) epishelf lake ice.122−10 −5 0 5 10 15−250−200−150−100−50050100↓Distance (km)Elevation (m)Figure 4.2 Surface and bed elevations along a profile of the MGT. Data are from the NASA IceBridge 2014aerial survey, with surface elevations (thin solid line) from the airborne topographic mapping (ATM) andbottom elevations (thick solid line) from the Multichannel Coherent Radar Depth Sounder (MCoRDS). Theposition of the grounding line is the point of first hydrostatic equilibrium of the ice where the thicknesscalculated from the surface elevation (dashed line) crosses that measured by the MCoRDs. Note the aerialsurvey line crossed over from the central stream of the Milne Glacier to the highly fractured north-easttributary stream⇠100 m downstream of the grounding line for 5 km, so surface elevation-derived thicknessesalong this section are variable and are not representative of the central ice stream.12320115 km20125 km20135 km20145 km20155 km  Surface ice velocity (m a−1 )020406080100Figure 4.3 Surface ice speed for the Milne Glacier from 2011 to 2015. The model domain used for area-averaged basal melt calculations for the MGT (shown in Fig. 4.7) is marked by the black line. Velocity datais reproduced with permission from Van Wychen et al. (2016).1240 10 20 30020406080100120140160180200Depth (m)SA (g kg−1)Max depth MGTMax depth MIS−1 0 1 2Tw (oC)0 0.5 1 1.5 2Tw − Tf (oC)  20112012201320142015A B CFigure 4.4 Profiles of A) Absolute Salinity (SA), B) in situ temperature (Tw), and C) temperature abovefreezing (TwTf ) in Milne Fiord from 2011 to 2015. Solid lines indicate profiles collected in Milne Fiord atthe mooring site, dotted lines are profiles collected offshore (see Fig. 4.1 for locations).0 0.01 0.02 0.03 0.04 0.05 0.06 0.07050100150200Current speed (m s−1)Depth (m)  2011 ADCP2012 ADCP2013 ADCPmeanextrapolatedFigure 4.5 Time-averaged current speeds in Milne Fiord. Profiles were collected at the mooring site usingan 300 kHz ADCP in May 2011, and July of 2012 and 2013. Input to the ice-ocean model is the averageobserved velocity from all years from 1 to 100 m depth (solid black line), and an extrapolated velocity below100 m (dashed black line).1250 1 2 30510Tw − Twf (oC)a b (m a−1 )0 0.02 0.040510U (m s−1)a b (m a−1 )0 0.005 0.010510Cda b (m a−1 )0 2 4 6 8x 10−30510ΓTSa b (m a−1 )A BC DFigure 4.6 Melt rate dependence on the A) thermal driving (TwT fw ), B) mixed layer speed (U), C) dragcoefficient (Cd), and D) turbulent transfer coefficient (GTS). The x-axis limits of each variable are equivalentto the likely range of values in Milne Fiord or in the literature. For each run other variables were fixed at theactual value used in the model (circle), except for TwT fw which was fixed at 1C, andU which was fixed atthe depth-integrated average of the velocity profile shown in Fig. 4.5.126−150−100−500Ice draft (m)0100200Ice speed (m a−1 )00.050.1Ice flux (Gt a−1 )0 5 10 1502468Basal ice melt (m a−1 )Distance from grounding line (km)  20112012201320142015ABCDFigure 4.7 Area-averaged submarine melt along the length of the MGT for years 2011 to 2015. Cross-fjordaverage A) ice draft, B) ice speed, C) ice flux, and D) submarine melt rate vs distance from the Milne Glaciergrounding line. The domain used for calculation is shown in Fig. 4.3. Lines in B), C), & D) are coloured byyear. Shown in D) are the submarine melt rates calculated by the divergence of ice flux method (solid lines)and the ocean thermodynamic method (dashed lines) for each year.127Figure 4.8 Submarine ice melt rates in Milne Fiord from 2011 to 2015 calculated using the ocean thermody-namic model based on ocean profiles collected at the mooring site in the fjord. The grey area indicates thearea of the epishelf lake where melt rates were not calculated.128Figure 4.9 Submarine ice melt rates in Milne Fiord from 2011 to 2015 calculated using the ocean thermody-namic model based on profiles collected offshore. The grey area indicates the area of the epishelf lake wheremelt rates were not calculated.129Figure 4.10 Difference in submarine ice melt rates in Milne Fiord from 2011 to 2015 calculated using theocean thermodynamic model between profiles collected at the mooring site in the fjord versus profiles col-lected offshore. The grey area indicates the area of the epishelf lake where melt rates were not calculated.130123T w − T wf  (o C)  5 m00.20.4T w − T wf  (o C)  25 m00.20.4T w − T wf  (o C)  50 mJul/11 Jan/12 Jul/12 Jan/13 Jul/13 Jan/14 Jul/140.60.81DateT w − T wf  (o C)  125 mABCDFigure 4.11 Variation in thermal driving (TwT fw ) over from May 2011 to July 2014 at A) 5 m, B) 25 m,C) 50 m, and D) 125 m depths. The y-axes scales are equal in B), C) and D), allowing for direct comparisonof the magnitude of variability at these depths, although the origins are offset. The y-scale in A) is differentgiven the substantially larger variation.131Chapter 5Conclusions5.1 SummaryThis thesis has demonstrated the importance of ice-ocean interactions in determining the waterproperties and circulation in a glacial fjord. Field and analytical methods were used to describethe spatial and temporal variation of water column structure, circulation, and submarine meltingin Milne Fiord between 2011 and 2015. The thesis began by setting the geophysical context ofthe fjord, then provided a general description of the horizontal and temporal variation of waterproperties and how they were influenced by interaction with ice. Next, factors influencing theseasonal and long-term dynamics of the Milne Fiord epishelf lake, the last remaining epishelf lakein the Arctic, were investigated. Finally, the magnitude and spatial variations of submarine melt rateof an ice shelf and glacier tongue were estimated.Chapter 2 provided an overall description of the geophysical and oceanographic properties ofMilne Fiord. By undertaking an extensive field study I was able to produce the first comprehensivemap of both the ice topography and bathymetry over the entire fjord. This provided the criticalcontext in which the oceanographic observations could be interpreted. The vertical distribution ofwater masses in the fjord reflected the general structure of the adjacent Arctic Ocean, consisting ofPolar Water above Atlantic Water. However, I found that the Milne Ice Shelf had a profound impacton both the vertical structure and circulation in the fjord. Inflow of freshwater at the surface wastrapped by the ice shelf, creating the epishelf lake, while upwelling of subglacial runoff created anintermediate layer termed fjord-modified water. Export of fjord-modified water was restricted to anarrow basal channel in the ice shelf, resulting in a prolonged (by several months) drainage fromthe fjord over winter. This effectively de-coupled the timing of freshwater export from the seasonalcycle of meltwater production at the surface. Intermediate water properties varied substantially overbroad timescales, from days to months, likely reflecting depth variation of the main Arctic haloclineoffshore being transmitted into the fjord under the ice shelf. I also inferred the presence of a sill132under the ice shelf from hydrographic data, and presented evidence of a renewal of deep water belowsill depth in Milne Fiord.The work presented in Chapter 2 demonstrated the strong impact that an ice shelf could have onfjord oceanography. The loss of theMilne Ice Shelf will fundamentally alter the water properties andcirculation of the fjord, with consequences for the stability and dynamics of the Milne Glacier. TheMilne Ice Shelf is the last remaining intact ice shelf along the northern coast of Ellesmere Island,and the observations presented here provide a valuable reference to understand future oceanographicchanges in the fjord after the inevitable collapse of the ice shelf.Chapter 3 related an investigation of the seasonal and long-term dynamics of the last remainingArctic epishelf lake. Continuous mooring records from the lake revealed that the halocline variedby several meters each year. It was determined this was due to the balance between the volume ofmeltwater inflow, a function of the intensity of summer surface melt, and hydraulically controlledoutflow under the ice shelf, a function of the geometry of the outflow channel. On a long-term basisthe depth of the lake was dependent on the thickness of the ice shelf, and the interannual shoaling ofthe halocline suggested a long-term thinning of the Milne Ice Shelf. However, it was also shown thatepisodic events that provide sufficient kinetic energy to the system, such as the calving or capsizeof an iceberg, could induce mixing of the halocline, thereby altering the depth of the lake.The findings in Chapter 3 strongly suggest that a full consideration of the hydrology and hy-draulics of an epishelf lake must be undertaken where hydrographic profiles are used to infer thethickness of the impounding ice shelf. Epishelf lakes are much more dynamic than previouslythought, which also has important implications for the ecology of these rare aquatic ecosystems.Chapter 4 discussed efforts to estimate submarine melting in Milne Fiord. An average melt rateof ⇠4 m a1 near the grounding line of the Milne Glacier was found, as was an enhanced near-surface melt rate of ⇠2 m a1 for ice in contact with the epishelf lake. This finding indicated adepth-dependence of melt rates due, in part, to the vertical distribution of the water temperatureabove freezing, or thermal driving, in the fjord. The melt rate depth-dependence suggested a com-plex spatial patterns of basal melting on the Milne Ice Shelf, owing to its variable ice thickness.However, the relationship between melt rate and thermal driving in Milne Fiord was not linear (e.g.the epishelf lake had higher average thermal driving than water near the grounding line, yet thegrounding line showed higher melt rates), indicating other factors, notably circulation, were alsoimportant in determining melt rates.The work in Chapter 4 has confirmed that knowledge of current speeds and their variation arefundamental for predicting melt rates and ice-ocean interaction processes in glacial fjords. Theresults suggest efforts to measure circulation, and its spatial and temporal variation, will be invalu-able for more accurate prediction of melt rates. This is particularly important as surface runoff ispredicted to continue to increase in the Canadian Arctic Archipelago (CAA; Lenaerts et al., 2013)which is likely to impact buoyancy-driven circulation in tidewater glacial fjords. The submarine133melt rates presented in Chapter 4 are, to the best of my knowledge, the first direct estimates of sub-marine melt rates for an ice shelf or tidewater glacier in the CAA. The work indicates that processesthat alter the vertical distribution of heat in the fjord or circulation patterns, including the collapseof the Milne Ice Shelf, will change the rate and distribution of melting of the Milne Glacier tongue.The present state of knowledge of Milne Fiord, based on this thesis, is summarized schemati-cally in Figure 5.1. The Milne Ice Shelf is partially grounded on a bathymetric ridge near the mouthof the fjord. There is no freshwater outflow or exchange above the minimum draft of the ice shelf(⇠10 m). The epishelf lake, which seasonally deepens by several meters during summers of intensesurface melt, drains over several months after the summer melt season along a basal channel in theice shelf, its rate hydraulically controlled by the dimensions of the channel and the excess depth ofthe lake below the ice. Warm waters of the epishelf lake are enhancing submarine melting alongthe landward edge of the ice shelf and around the margins of the glacier tongue. Water exchangeacross the ice shelf is restricted down to approximately 50 m depth, creating a relatively fresh, warmlayer in the fjord that is termed fjord-modified water. Fjord-modified water is also slowly exportedalong the basal channel in the ice shelf. The source of the fjord-modified water is largely subglacialdischarge plume, which enter the fjord at the grounding line of the Milne Glacier, mix with ambientwater on their ascent, but do not reach the surface due to the strong stratification imposed by thepresence of the epishelf lake, and are consequently exported at depth through the basal channel inthe Milne Ice Shelf. Relatively free exchange of water occurs below the ice shelf to the depth of thebathymetric sill, resulting in vertical excursions of water masses, particularly the interface betweenPolar Water and Atlantic Water, in the fjord that are driven by offshore processes. The warm waterof the Atlantic layer drives the highest submarine melt rates at the glacier grounding line, leadingto a retreat of the grounding line. Below the sill, deep water of the fjord is relatively homogenous,with evidence of complete deep water renewal events that occur with an as yet unknown frequency.Milne Fiord presently retains the last intact ice shelf, the last deep epishelf lake, and the last ex-tensive glacier tongue on the northern coast of Ellesmere Island. However, until very recently thesefeatures were widespread on this coast: in 2000 there were 6 major ice shelves (Ayles, Markham,Milne, Petersen, Serson, and Ward Hunt), at least two (Milne and Ward Hunt) were known todam deep (>10 m) epishelf lakes, while the others were associated with shallower epishelf or ice-dammed lakes (Veillette et al., 2008; White et al., 2015b), and there were at least two extensive(>10 km) glacier tongues (Milne and Yelverton Bay; Copland et al., 2015). Even these featureshowever, are just remnants of what was once a continuous ice shelf that extended at least 500 kmalong this coast in the late 19th century (Vincent et al., 2001), that Veillette et al. (2008) suggestmay have retained up to 17 epishelf lakes, and a larger number of the 11 major tidewater glaciersalong this coast (Copland et al., 2015) may have terminated in glacier tongues. The present-daygeophysical setting of Milne Fiord, and the oceanographic processes described in this thesis, maybe representative what was once a common system along this coast.1345.2 Future directionsThe findings presented in this thesis have answered many of the questions that originally motivatedthe research. There remain, however, a number of unanswered questions that ongoing and futureefforts aim to resolve. The findings have also raised a number of important new questions that couldmotivate future research directions.One of the most unexpected findings, presented in Chapter 2, was the discovery that the MISmay be partially grounded on a sea bed ridge. If so, this could explain both the origin of the mainre-healed fractures in the ice shelf and provide an explanation for the ice shelf’s relative stabilitycompared to others along this coast. Confirmation that the ice shelf is actually grounded, and overwhat extent, will help constrain the fate of the ice shelf and inform the possible mechanisms offuture breakup.The results of the hydraulic drainage model presented in Chapter 3 strongly suggest that epishelflake outflow occurs along the basal channel in the MIS, which preliminary field observations ap-pear to support. If confirmed, the geometry of the channel must be further constrained along itslength, and measurements of water properties and flow rates will be needed to validate, or refute,the drainage theory. Another important aspect necessitating further study is to investigate how themechanics and thermodynamics of stratified flow influence melting or freezing along the channelwalls and ceiling. A better understanding of this process will provide insight into the variation of thedepth of the epishelf lake and preferential melting in basal channels as a mechanism contributingto the breakup of Ellesmere Island ice shelves (Mueller et al., 2003) and leading to highly complexpatterns of melt on Antarctic ice shelves (Dutrieux et al., 2013).Submarine melt rates are strongly influenced by the transport of heat along a fjord and throughthe boundary layer at the ice-ocean interface. The melt rates estimated using the ocean model inChapter 4 were based on limited field measurements, requiring broad assumptions about the waterproperties and circulation under the glacier tongue and ice shelf. The model also required the useof parameterizations that have not been validated for Milne Fiord. Despite these acknowledgedlimitations the magnitude and distribution of melt predicted for the Milne Glacier tongue generallycompared well with those based on the independent divergence of ice flux method, suggesting theassumptions applied were not unrealistic. However, it is clear that improved measurements of cir-culation, both at the scale of the entire fjord, including intermediary exchange flows and buoyancy-driven circulation, but also proximal to the ice face, including near subglacial outlet channels, arerequired to more accurately model melt rates. Direct measurements of basal ablation, ocean prop-erties, and currents through the oceanic boundary layer below ice shelves are also required to refinethe turbulent transfer and drag coefficients used in the ocean model. These measurements must beacquired in systems appropriate for use in Milne Fiord, that is highly stratified, low energy environ-ments, with similar ice roughness characteristics. If parameterizations are validated, then numericalmodelling could provide a valuable tool for understanding the contribution of basal melt to total ice135mass balance in Milne Fiord.In addition to improved estimates of basal ablation, proper mass balance estimates for the MilneIce Shelf and Milne Glacier tongue require better resolved, in space and time, measurements of sur-face ablation. The ablation stake network used in this study was extremely sparse, but the availabledata indicated there is an along-fjord gradient in surface mass balance with greater accumulationtoward the coast on the Outer Unit of the Milne Ice Shelf. Production of repeat, annual high res-olution surface elevation models from remote sensing data, calibrated against the existing ablationstake network, would dramatically enhance understanding of surface mass balance, and thus furtherconstrain the contribution of basal melt to total mass balance of the ice shelf and glacier tongue.Expansion of surface mass balance estimates throughout the Milne Glacier catchment wouldallow for more precise estimates of surface meltwater into the fjord. This would improve the under-standing of both inflow to the epishelf lake, as well as the volume of subglacial discharge. As surfacemelt is predicted to increase with climate warming, buoyancy-driven circulation and convection-driven melting at the grounding line of the Milne Glacier may be increasingly important, indicatinga well-constrained estimate of freshwater input to the fjord is critical to understanding how thedynamics of the fjord will evolve in the future, and how this may effect the stability of ice in thefjord.Water properties and circulation in Milne Fiord are also strongly dependent on variations of wa-ter mass structure and circulation on the continental shelf. How large-scale and long-term changesto ocean properties in the Arctic Ocean are transmitted into Milne Fiord, and what effect this mayhave on the dynamics of the fjord is unknown. The dynamics of fjords on Ellesmere Island arelinked to regional changes in the ocean, atmosphere, and cryosphere, so an interdisciplinary broad-scale view of these processes could help inform the factors driving the observed local changes alongthis coast.The eventual collapse of the Milne Ice Shelf, and the loss of the epishelf lake, will have aprofound impact on fjord water properties and circulation. These changes are shown schematicallyin Figure 5.2. Some of the changes predicted to occur include weaker stratification of the upperwater column, seasonal export of runoff at the surface, and a transition from perennial to seasonalice cover. These changes will impact, among other processes, the vertical distribution of heat,buoyancy-driven circulation, wind-mixing of the water column, and wave erosion of the glaciertongue. Ultimately, the loss the Milne Ice Shelf is expected to lead to a collapse of the Milne Glaciertongue with implications for the dynamics of the Milne Glacier. Long-term monitoring of MilneFiord during its transition from an ice shelf-epishelf lake-glacier tongue system to a seasonally icecovered tidewater glacier system could provide rare insight into the dynamics of glacial fjords.1365.3 FiguresPWAWEpishelf LakeIce Shelf Glacier TongueEpishelf LakeGrounding LineArcticOceanABDWSea Ice Freshwater IceSurfaceRunoffFjordGlacierPWAWFMWSea IceFigure 5.1 Schematic representation of Milne Fiord at present in A) plan and B) elevation view showingknown ice features and oceanographic properties based on the work detailed in this thesis. See text fordiscussion. Water masses are labelled: PW - Polar Water, AW - Atlantic Water, FMW - fjord modified water,DW - deep water.137PWAWArcticOceanABDWSea IceFjordPWAWSea IcePlumeIcebergsIce IslandIce IslandIcebergsSea IceSea IcePlumeOpen WaterFigure 5.2 Schematic representation of the future of Milne Fiord in A) plan and B) elevation view showingcollapse of ice features and associated oceanographic changes. See text for discussion. Water masses arelabelled: PW - Polar Water, AW - Atlantic Water, DW - deep water.138BibliographyAagaard, K., Coachman, L. K., and Carmack, E. (1981). On the halocline of the Arctic Ocean.Deep Sea Research Part A. Oceanographic Research Papers, 28(6):529–545. ! pages 5Aksenov, Y., Bacon, S., Coward, A. C., and Holliday, N. P. (2010). Polar outflow from the ArcticOcean: A high resolution model study. Journal of Marine Systems, 83(1-2):14–37. ! pages 5,24, 25Alkire, M., Falkner, K., Boyd, T., and Macdonald, R. (2010). Sea ice melt and meteoric waterdistributions in Nares Strait, Baffin Bay, and the Canadian Arctic Archipelago. Journal ofMarine Research, 68(6):767–798. ! pages 30Antoniades, D., Francus, P., Pienitz, R., St-Onge, G., and Vincent, W. F. (2011). Holocenedynamics of the Arctic’s largest ice shelf. Proceedings of the National Academy of Sciences ofthe United States of America, 108(47):18899–18904. WOS:000297249800016. ! pages 6Braithwaite, R. J. (2005). Mass-balance characteristics of arctic glaciers. Annals of Glaciology,42(1):225–229. ! pages 4Braun, C., Hardy, D. R., Bradley, R. S., and Sahanatien, V. (2004). Surface mass balance of theWard Hunt Ice Rise and Ward Hunt Ice Shelf, Ellesmere Island, Nunavut, Canada. Journal ofGeophysical Research: Atmospheres, 109(D22). ! pages 7, 24, 114Burton, J. C., Amundson, J. M., Abbot, D. S., Boghosian, A., Cathles, L. M., Correa-Legisos, S.,Darnell, K. N., Guttenberg, N., Holland, D. M., and MacAyeal, D. R. (2012). Laboratoryinvestigations of iceberg capsize dynamics, energy dissipation and tsunamigenesis. Journal ofGeophysical Research: Earth Surface, 117(F1):n/a–n/a. ! pages 80Carmack, E. C. (1990). Large-Scale Physical Oceanography of Polar Oceans. In Smith, W. O.,editor, Polar Oceanography, pages 171–222. Academic Press, San Diego. ! pages 5, 112Cenedese, C. and Linden, P. F. (2014). Entrainment in two coalescing axisymmetric turbulentplumes. Journal of Fluid Mechanics, 752:R2 (12 pages). ! pages 11, 117Chauch, N., Hubbard, A., Gascard, J.-C., Box, J. E., Bates, R., Koppes, M., Sole, A.,Christoffersen, P., and Patton, H. (2014). Iceocean interaction and calving front morphology attwo west Greenland tidewater outlet glaciers. The Cryosphere, 8(4):1457–1468. ! pages 23, 36Christoffersen, P., Mugford, R. I., Heywood, K. J., Joughin, I., Dowdeswell, J. A., Syvitski, J.P. M., Luckman, A., and Benham, T. J. (2011). Warming of waters in an East Greenland fjord139prior to glacier retreat: mechanisms and connection to large-scale atmospheric conditions. TheCryosphere, 5(3):701–714. ! pages 9, 23Copland, L., Dalton, A., White, A., Van Wychen, W., and Dawson, J. (2015). Patterns of glacierdisintegration and iceberg production in Yelverton Bay and Inlet, northern Ellesmere Island. InArcticNet Annual Scientific Meeting, Vancouver, Canada. ! pages 1, 2, 46, 134Copland, L., Mueller, D. R., and Weir, L. (2007). Rapid loss of the Ayles Ice Shelf, EllesmereIsland, Canada. Geophysical Research Letters, 34(L21501):6. ! pages 2, 5, 6, 22, 46, 62, 98Copland, L., Sharp, M. J., and Dowdeswell, J. A. (2003). The distribution and flow characteristicsof surge-type glaciers in the Canadian High Arctic. Annals of Glaciology, 36(1):73–81. !pages 25, 99, 112Cottier, F., Tverberg, V., Inall, M., Svendsen, H., Nilsen, F., and Griffiths, C. (2005). Water massmodification in an Arctic fjord through cross-shelf exchange: The seasonal hydrography ofKongsfjorden, Svalbard. Journal of Geophysical Research-Oceans, 110(C12).WOS:000234248800001. ! pages 18Cottier, F. R., Nilsen, F., Skogseth, R., Tverberg, V., Skarhamar, J., and Svendsen, H. (2010).Arctic fjords: a review of the oceanographic environment and dominant physical processes.Geological Society, London, Special Publications, 344(1):35–50. ! pages 8Crary, A. (1956). Geophysical studies along northern Ellesmere Island. Arctic, 9(3):154–165. !pages 7, 24, 62Dutrieux, P., Vaughan, D. G., Corr, H. F. J., Jenkins, A., Holland, P. R., Joughin, I., and Fleming,A. H. (2013). Pine Island glacier ice shelf melt distributed at kilometre scales. The Cryosphere,7(5):1543–1555. ! pages 135England, J. H., Lakeman, T. R., Lemmen, D. S., Bednarski, J. M., Stewart, T. G., and Evans, D.J. A. (2008). A millennial-scale record of Arctic Ocean sea ice variability and the demise of theEllesmere Island ice shelves. Geophysical Research Letters, 35(19). WOS:000259803300002.! pages 62Evans, D. and England, J. (1992). Geomorphological evidence of Holocene climatic change fromnorthwest Ellesmere Island, Canadian high arctic. The Holocene, 2(2):148–158. ! pages 6Farmer, D. M. and Freeland, H. J. (1983). The physical oceanography of Fjords. Progress InOceanography, 12(2):147–219. ! pages 8, 9, 10, 35Ford, W. L. and Hattersley-Smith, G. (1965). On the oceanography of the Nansen Sound fiordsystem. ARCTIC, 18(3):158–171. ! pages 4, 23, 24Gade, H. G. (1979). Melting of Ice in Sea Water: A Primitive Model with Application to theAntarctic Ice Shelf and Icebergs. Journal of Physical Oceanography, 9:189–198. ! pages 29Galton-Fenzi, B. K., Hunter, J. R., Coleman, R., and Young, N. (2012). A decade of change in thehydraulic connection between an Antarctic epishelf lake and the ocean. Journal of Glaciology,58(208):223–228. ! pages 8140Gardner, A., Moholdt, G., Wouters, B., Wolken, G., Burgess, D., Sharp, M., Cogley, J., Braun, C.,and Labine, C. (2011). Sharply increased mass loss from glaciers and ice caps in the CanadianArctic Archipelago. Nature, 473(7347):357–360. ! pages 2, 4, 23, 46, 98, 114, 115Gardner, A. S., Moholdt, G., Cogley, J. G., Wouters, B., Arendt, A. A., Wahr, J., Berthier, E.,Hock, R., Pfeffer, W. T., Kaser, G., Ligtenberg, S. R. M., Bolch, T., Sharp, M. J., Hagen, J. O.,van den Broeke, M. R., and Paul, F. (2013). A Reconciled Estimate of Glacier Contributions toSea Level Rise: 2003 to 2009. Science, 340(6134):852–857. ! pages 81, 98, 114Geyer, W. and Cannon, G. (1982). Sill processes related to deep water renewal in a fjord. Journalof Geophysical Research, 87(C10):7985–7996. ! pages 35Gibson, J. and Andersen, D. (2002). Physical structure of epishelf lakes of the southern BungerHills, East Antarctica. Antarctic Science, 14(3):253–261. ! pages 7, 8, 62, 63GLIMS and NSIDC (2005). GLIMS Glacier Database, Version 1. Technical report, NSIDC:National Snow and Ice Data Center, Boulder, Colorado USA. ! pages 25, 64Hattersley-Smith, G. (1957). The Ellesmere Ice Shelves and the Ice Islands. CanadianGeographer, 3(9):65–70. ! pages 6Hattersley-Smith, G. (1969). Results of radio echo sounding in Northern Ellesmere Island, 1966.Geographical Journal, 135(4):553–557. ! pages 25Hattersley-Smith, G. (1973). Ice Shelf and Fiord Ice Problems in Disraeli Fiord, NorthernEllesmere Island, NWT. Technical report, Defence Research Board, Department of NationalDefence Canada, Ottawa, Canada. ! pages 7, 63Hattersley-Smith, G. and Fuzesy, A. (1969). Glacier Depths in Northern Ellesmere Island:Airborne Radio Echo Sounding in 1966. Technical Note, Defence Research Board, Departmentof National Defence Canada, Ottawa, Canada. ! pages 25, 40Hattersley-Smith, G. and Serson, H. (1966). Reconnaissance oceanography over the ice of theNansen Sound Fiord System. Technical Report 28, Defense Research Board of Canada,Directorate of Physical Research, Geophysics, Hazen, Ottawa. ! pages 4Hellmer, H. and Olbers, D. (1989). A two-dimensional model for the thermohaline circulationunder an ice shelf. Antarctic Science, 1(04):325–336. ! pages 102, 103Heywood, R. B. (1977). A Limnological Survey of the Ablation Point Area, Alexander Island,Antarctica. Philosophical Transactions of the Royal Society of London. B, Biological Sciences,279(963):39–54. ! pages 6, 7, 62Hock, R. (2003). Temperature index melt modelling in mountain areas. Journal of Hydrology,282(1-4):104–115. ! pages 32, 67Holland, D. M. (1998). On the parameterization of basal heat flux for sea-ice modelling.Geophysica, 34(1-2):1–21. ! pages 14141Holland, D. M. and Jenkins, A. (1999). Modeling thermodynamic ice-ocean interactions at thebase of an ice shelf. Journal of Physical Oceanography, 29(8):1787–1800. ! pages 12, 13, 14,15, 16, 17, 18, 102, 103, 105, 106, 109, 118Holland, D. M. and Jenkins, A. (2001). Adaptation of an isopycnic coordinate ocean model for thestudy of circulation beneath ice shelves. Monthly Weather Review, 129(8):1905–1927. ! pages17, 102Holland, D. M., Thomas, R. H., de Young, B., Ribergaard, M. H., and Lyberth, B. (2008a).Acceleration of Jakobshavn Isbrtriggered by warm subsurface ocean waters. Nature Geoscience,1(10):659–664. ! pages 12, 23, 45Holland, P. R., Jenkins, A., and Holland, D. M. (2008b). The Response of Ice Shelf Basal Meltingto Variations in Ocean Temperature. Journal of Climate, 21(11):2558–2572. ! pages 2, 22, 23Horn, D. A., Imberger, J., and Ivey, G. N. (2001). The degeneration of large-scale interfacialgravity waves in lakes. Journal of Fluid Mechanics, 434:181–207. ! pages 72Huppert, H. and Turner, J. (1980). Ice blocks melting into a salinity gradient. Journal of FluidMechanics, 100(2):367–384. ! pages 43Inall and Gillibrand, P. (2010). The Physics of mid-Latitude Fjords: A Review. Geological SocietySpecial Publication, 344(1):17–33. ! pages 8, 9Inall, M. E., Murray, T., Cottier, F. R., Scharrer, K., Boyd, T. J., Heywood, K. J., and Bevan, S. L.(2014). Oceanic heat delivery via Kangerdlugssuaq Fjord to the south-east Greenland ice sheet.Journal of Geophysical Research: Oceans, 119(2):631–645. ! pages 12, 18IPCC (2013). Climate Change 2013: The Physical Science Basis. Contribution of Working GroupI to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. CambridgeUniversity Press, Cambridge, United Kingdom and New York, NY, USA. ! pages 62, 98ISO (1980). International Organization of Standards. ISO 1438/1-1980(E). Water flowmeasurement in open channels using weirs and venturi flumes - Part 1: Thin plate weirs. !pages 83Jackson, J. M., Lique, C., Alkire, M., Steele, M., Lee, C. M., Smethie, W. M., and Schlosser, P.(2014a). On the waters upstream of Nares Strait, Arctic Ocean, from 1991 to 2012. ContinentalShelf Research, 73:83–96. ! pages 5, 25, 30, 44, 114Jackson, R. H., Straneo, F., and Sutherland, D. A. (2014b). Externally forced fluctuations in oceantemperature at Greenland glaciers in non-summer months. Nature Geoscience, 7(7):503–508.! pages 9, 12, 18, 23, 44, 108, 116Jacobs, S. S., Huppert, H., Holdsworth, G., and Drewry, D. J. (1981). Thermohaline Steps InducedBy Melting of the Erebus Glacier Tongue. Journal of Geophysical Research,86(C7):6547–6555. ! pages 43142Jakobsson, M., Mayer, L., Coakley, B., Dowdeswell, J. A., Forbes, S., Fridman, B., Hodnesdal, H.,Noormets, R., Pedersen, R., Rebesco, M., Schenke, H. W., Zarayskaya, Y., Accettella, D.,Armstrong, A., Anderson, R. M., Bienhoff, P., Camerlenghi, A., Church, I., Edwards, M.,Gardner, J. V., Hall, J. K., Hell, B., Hestvik, O., Kristoffersen, Y., Marcussen, C., Mohammad,R., Mosher, D., Nghiem, S. V., Pedrosa, M. T., Travaglini, P. G., and Weatherall, P. (2012). TheInternational Bathymetric Chart of the Arctic Ocean (IBCAO) Version 3.0. GeophysicalResearch Letters, 39(L12609):1–6. ! pages 4, 25Jeffries, M. (1991). Massive, ancient sea-ice strata and preserved physical-structural characteristicsin the Ward Hunt ice shelf. Annals of Glaciology, 15:125–131. ! pages 105Jeffries, M. and Krouse, H. R. (1984). Arctic ice shelf growth, fiord oceanography and climate.Zeitschrift fur Gletscherkunde und Glazialgeologie, 20:147–153. ! pages 7, 63Jeffries, M. O. (1984). Milne Glacier, Northern Ellesmere Island, N.W.T., Canada: A SurgingGlacier? Journal of Glaciology, 30(105):251–253. ! pages 25, 99, 112, 113Jeffries, M. O. (1985). Physical, Chemical and Isotopic Investigations of Ward Hunt Ice Shelf andMilne Ice Shelf, Ellesmere Island, Nwt. Ph.D. Thesis, University of Calgary, Calgary. ! pages3, 26, 64, 65, 67, 69, 74, 87Jeffries, M. O. (1986a). Glaciers and the Morphology and Structure of Milne Ice Shelf, EllesmereIsland, N.W.T., Canada. Arctic and Alpine Research, 18(4):397–405. ! pages 25, 26, 99, 100Jeffries, M. O. (1986b). Ice Island Calvings and Ice Shelf Changes, Milne Ice Shelf and Ayles IceShelf, Ellesmere Island, N.W.T. Arctic, 39(1):15–19. ! pages 3Jeffries, M. O. (2002). Ellesmere Island Ice Shelves and Ice Islands. In Satellite Image Atlas ofGlaciers of the World, page 147. United States Geological Survey, usgs professional paper 1386edition. ! pages 1, 24, 98Jenkins, A. (1991). A one-dimensional model of ice shelf-ocean interaction. Journal ofGeophysical Research: Oceans, 96(C11):20671–20677. ! pages 117Jenkins, A. (1999). The impact of melting ice on ocean waters. Journal of physical oceanography,29(9):2370–2381. ! pages 29, 30, 31Jenkins, A. (2011). Convection-Driven Melting near the Grounding Lines of Ice Shelves andTidewater Glaciers. Journal of Physical Oceanography, 41(12):2279–2294. ! pages 10, 11,98, 106, 109, 117, 118Jenkins, A. and Bombosch, A. (1995). Modeling the effects of frazil ice crystals on the dynamicsand thermodynamics of Ice Shelf Water plumes. Journal of Geophysical Research: Oceans,100(C4):6967–6981. ! pages 17Jenkins, A. and Doake, C. S. M. (1991). Ice-ocean interaction on Ronne Ice Shelf, Antarctica.Journal of Geophysical Research: Oceans, 96(C1):791–813. ! pages 17Jenkins, A., Dutrieux, P., Jacobs, S. S., McPhail, S. D., Perrett, J. R., Webb, A. T., and White, D.(2010). Observations beneath Pine Island Glacier in WestAntarctica and implications for itsretreat. Nature Geoscience, 3(7):468–472. ! pages 14, 17, 18, 103, 104, 117143Jenkins, A., Hellmer, H. H., and Holland, D. M. (2001). The Role of Meltwater Advection in theFormulation of Conservative Boundary Conditions at an IceOcean Interface. Journal of PhysicalOceanography, 31(1):285–296. ! pages 103Johnson, H., Mnchow, A., Falkner, K., and Melling, H. (2011). Ocean circulation and properties inPetermann Fjord, Greenland. Journal of Geophysical Research, 116(C01003):1–18. ! pages 9,11, 43Joughin, I., Alley, R. B., and Holland, D. M. (2012). Ice-sheet response to oceanic forcing.Science, 338(6111):1172–1176. ! pages 2, 22, 41Keys, J. E. (1977). Water regime of ice-covered fjords and lakes. Ph.D. Thesis, McGill University,Montreal. ! pages 4, 8, 18, 23, 24, 72Keys, J. E. (1978). Water regime of Disraeli Fjord, Ellesmere Island. Technical Report 792,Defense Research Establishment Ottawa, Ottawa. ! pages 7, 63Keys, J. E., Johannssen, O., and Long, A. (1968). On the oceanography of Disraeli Fiord onNorthern Ellesmere Island. Technical Report 6, McGill University, Montreal. ! pages 4, 7, 63Keys, J. E. and Seibert, G. H. (1969). Operation Tanquary, Ellesmere Island, N.W.T. 1963-1966.Canadian Oceanographic Data Centre, pages 99–100. ! pages 4Kimura, S., Candy, A. S., Holland, P. R., Piggott, M. D., and Jenkins, A. (2013). Adaptation of anunstructured-mesh, finite-element ocean model to the simulation of ocean circulation beneathice shelves. Ocean Modelling, 67:39 – 51. ! pages 11, 102, 103, 117Kindsvater, C. and Carter, R. (1959). Discharge characteristics of rectangular thin-plate weirs.Transactions of the American Society of Civil Engineers, 124(1). ! pages 82Krabill, W. B. (2010). IceBridge ATM L1b Elevation and Return Strength, Version 2. DigitalMedia, NASA DAAC at the National Snow and Ice Data Center, Boulder, Colorado USA. !pages 27, 101Krishfield, R. A. (2005). Spatial and temporal variability of oceanic heat flux to the Arctic icepack. Journal of Geophysical Research, 110(C7). ! pages 24Lake, R. and Walker, E. (1973). Notes on the Oceanography of d’Iberville Fiord. Arctic,26(3):222–229. ! pages 4, 23, 24Laxon, S. W., Giles, K. A., Ridout, A. L., Wingham, D. J., Willatt, R., Cullen, R., Kwok, R.,Schweiger, A., Zhang, J., Haas, C., Hendricks, S., Krishfield, R., Kurtz, N., Farrell, S., andDavidson, M. (2013). CryoSat-2 estimates of Arctic sea ice thickness and volume. GeophysicalResearch Letters, 40(4):732–737. ! pages 24Laybourn-Parry, J., Madan, N. J., Marshall, W. A., Marchant, H. J., and Wright, S. W. (2006).Carbon dynamics in an ultra-oligotrophic epishelf lake (Beaver Lake, Antarctica) in summer.Freshwater Biology, 51(6):1116–1130. ! pages 7, 62, 63144Laybourn-Parry, J., Quayle, W., Henshaw, T., Ruddell, A., and Marchant, H. (2001). Life on theedge: the plankton and chemistry of Beaver Lake, an ultra-oligotrophic epishelf lake, AntarcticaRID D-4634-2011. Freshwater Biology, 46(9):1205–1217. WOS:000171284900005. ! pages 8Lenaerts, J., Angelen, J. H., Broeke, M. R., Gardner, A. S., Wouters, B., and Meijgaard, E. (2013).Irreversible mass loss of Canadian Arctic Archipelago glaciers. Geophysical Research Letters,40(5):870–874. ! pages 2, 4, 23, 46, 47, 133Leuschen, C., Gogineni, P., Rodriguez-Morales, F., Paden, J., and Allen., C. (2010). IceBridgeMCoRDS L2 Ice Thickness, Version 1. Digital Media, NASA National Snow and Ice DataCenter Distributed Active Archive Center, Boulder, Colorado USA. ! pages 26, 27, 101Lewis, E. L. and Perkin, R. G. (1986). Ice pumps and their rates. Journal of GeophysicalResearch: Oceans, 91(C10):11756–11762. ! pages 11Losch, M. (2008). Modeling ice shelf cavities in a z coordinate ocean general circulation model.Journal of Geophysical Research: Oceans, 113(C8). C08043. ! pages 102, 103Ludlam, S. (1996). Stratification patterns in Taconite Inlet, Ellesmere Island, NWT. Journal ofPaleolimnology, 16(2):205–215. ! pages 4, 23, 24Manning, C. C., Hamme, R. C., and Bourbonnais, A. (2010). Impact of deep-water renewal eventson fixed nitrogen loss from seasonally-anoxic Saanich Inlet. Marine Chemistry, 122(1-4):1–10.! pages 35Maslanik, J., Stroeve, J., Fowler, C., and Emery, W. (2011). Distribution and trends in Arctic seaice age through spring 2011. Geophysical Research Letters, 38(L13502):1–6. ! pages 24Masson, D. (2002). Deep Water Renewal in the Strait of Georgia. Estuarine, Coastal and ShelfScience, 54(1):115–126. ! pages 35McPhee, M. G. (1990). 6 - Small-Scale Processes A2 - Smith, Walker O. In Polar Oceanography,pages 287–334. Academic Press, San Diego. ! pages 105McPhee, M. G. (1992). Turbulent heat flux in the upper ocean under sea ice. Journal ofGeophysical Research: Oceans, 97(C4):5365–5379. ! pages 16, 103McPhee, M. G., Kottmeier, C., and Morison, J. H. (1999). Ocean Heat Flux in the Central WeddellSea during Winter. Journal of Physical Oceanography, 29(6):1166–1179. ! pages 16, 18, 103McPhee, M. G., Maykut, G. A., and Morison, J. H. (1987). Dynamics and thermodynamics of theice/upper ocean system in the marginal ice zone of the Greenland Sea. Journal of GeophysicalResearch: Oceans, 92(C7):7017–7031. ! pages 14, 17, 18McPhee, M. G., Morison, J. H., and Nilsen, F. (2008). Revisiting heat and salt exchange at theice-ocean interface: Ocean flux and modeling considerations. Journal of Geophysical Research,113(C6). ! pages 13Melling, H., Haas, C., and Brossier, E. (2015). Invisible polynyas: Modulation of fast ice thicknessby ocean heat flux on the Canadian polar shelf. Journal of Geophysical Research: Oceans,120(2):777–795. ! pages 18145Mellor, G. L., McPhee, M. G., and Steele, M. (1986). Ice-Seawater Turbulent Boundary LayerInteraction with Melting or Freezing. Journal of Physical Oceanography, 16(11):1829–1846.! pages 17Millero, F. J. (1978). Annex 6: Freezing point of seawater. Eighth report of the Joint Panel ofOceanographic Tables and Standards. Technical Report 28, UNESCO. ! pages 13, 103Mingo, L. and Flowers, G. E. (2010). Instruments and Methods: An integrated lightweightice-penetrating radar system. Journal of Glaciology, 56(198):709–714. ! pages 27Mortensen, J., Bendtsen, J., Lennert, K., and Rysgaard, S. (2014). Seasonal variability of thecirculation system in a west Greenland tidewater outlet glacier fjord, Godthbsfjord (64N):Godthbsfjord. Journal of Geophysical Research: Earth Surface, 119(12):2591–2603. ! pages18Mortensen, J., Bendtsen, J., Motyka, R. J., Lennert, K., Truffer, M., Fahnestock, M., and Rysgaard,S. (2013). On the seasonal freshwater stratification in the proximity of fast-flowing tidewateroutlet glaciers in a sub-Arctic sill fjord. Journal of Geophysical Research: Oceans,118(3):1382–1395. ! pages 9, 23, 30Mortensen, J., Lennert, K., Bendtsen, J., and Rysgaard, S. (2011). Heat sources for glacial melt ina sub-Arctic fjord (Godthbsfjord) in contact with the Greenland Ice Sheet. Journal ofGeophysical Research, 116(C01013):1–13. ! pages 11, 23Mortimer, C. A. (2011). Quantification of Changes for the Milne Ice Shelf, Nunavut, Canada,1950 - 2009. M.Sc. thesis, University of Ottawa, Ottawa. ! pages 3, 26, 64, 99Mortimer, C. A., Copland, L., and Mueller, D. R. (2012). Volume and area changes of the MilneIce Shelf, Ellesmere Island, Nunavut, Canada, since 1950. Journal of Geophysical Research,117(F04011):1–12. ! pages 3, 25, 26, 33, 40, 46, 64, 73, 77, 84, 100, 113Morton, B. R., Taylor, G., and Turner, J. S. (1956). Turbulent Gravitational Convection fromMaintained and Instantaneous Sources. Proceedings of the Royal Society of London A:Mathematical, Physical and Engineering Sciences, 234(1196):1–23. ! pages 10Motyka, R. J., Truffer, M., Fahnestock, M., Mortensen, J., Rysgaard, S., and Howat, I. (2011).Submarine melting of the 1985 Jakobshavn Isbr floating tongue and the triggering of the currentretreat. Journal of Geophysical Research, 116(F01007):1–17. ! pages 2, 12, 22, 23, 41Mueller, D., Vincent, W., and Jeffries, M. (2006). Environmental gradients, fragmented habitats,and microbiota of a northern ice shelf cryoecosystem, Ellesmere Island, Canada. Arctic,Antarctic, and Alpine Research, 38(4):593–607. ! pages 6, 114Mueller, D. R., Copland, L., Hamilton, A., and Stern, D. (2008). Examining Arctic Ice ShelvesPrior to the 2008 Breakup. Eos, Transactions American Geophysical Union, 89(49):502–503.! pages 1, 6, 22, 40, 46, 98Mueller, D. R., Copland, L., and Jeffries, M. (2016). Changes in Canadian Arctic ice shelf extentsince 1906. In Arctic Ice Shelves and Ice Islands. ed. L. Copland and D.R. Mueller. Dordrecht:Springer SBM. ! pages 88146Mueller, D. R., Vincent, W. F., and Jeffries, M. O. (2003). Break-up of the largest Arctic ice shelfand associated loss of an epishelf lake. Geophysical Research Letters, 30(20):1–4. ! pages 1,2, 7, 22, 46, 62, 63, 66, 80, 98, 135Nares, G. (1878). Narrative of a voyage to the polar sea during 1875-6 in HM ships ‘Alert’ and‘Discovery’. Sampson Low, Marston, Searle & Rivington, London. ! pages 6Narod, B., Clarke, G., and Prager, B. (1988). Airborne UHF radar sounding of glaciers and iceshelves, northern Ellesmere Island, Arctic Canada. Canadian Journal of Earth Sciences,25(1):95–105. ! pages 3, 25, 40, 64Newton, J. L. and Sotirin, B. J. (1997). Boundary undercurrent and water mass changes in theLincoln Sea. Journal of Geophysical Research, 102(C2):3393–3403. ! pages 5, 25, 44Nst, O. A. and Foldvik, A. (1994). A model of ice shelf-ocean interaction with application to theFilcher-Ronne and Ross Ice Shelves. Journal of Geophysical Research: Oceans,99(C7):14243–14254. ! pages 15Padman, L. and Erofeeva, S. (2004). A barotropic inverse tidal model for the Arctic Ocean.Geophysical Research Letters, 31(L02303):1–4. ! pages 31Pawlowicz, R. (2008). Calculating the conductivity of natural waters. Limnology andOceanography: Methods, 6(9):489–501. ! pages 66Pawlowicz, R., Beardsley, B., and Lentz, S. (2002). Classical tidal harmonic analysis includingerror estimates in MATLAB using T tide. Computers & Geosciences, 28(8):929–937. ! pages31, 67Peary, R. E. (1907). Nearest the Pole. Hutchinson, London. ! pages 6Polyakov, I. V., Alekseev, G. V., Timokhov, L. A., Bhatt, U. S., Colony, R. L., Simmons, H. L.,Walsh, D., Walsh, J. E., and Zakharov, V. F. (2004). Variability of the Intermediate AtlanticWater of the Arctic Ocean over the Last 100 Years. Journal of Climate, 17(23):4485–4497. !pages 45Pope, S., Copland, L., and Mueller, D. (2012). Loss of multiyear landfast sea ice from YelvertonBay, Ellesmere Island, Nunavut, Canada. Arctic, Antarctic, and Alpine Research,44(2):210–221. ! pages 1, 5, 22Rignot, E. (1998). Hinge-line migration of Petermann Gletscher, north Greenland, detected usingsatellite radar interferometry. Journal of Glaciology, 44(148):469–476. ! pages 2Rignot, E., Gogineni, S., Joughin, I., and Krabill, W. (2001). Contribution to the glaciology ofnorthern Greenland from satellite radar interferometry. Journal of Geophysical Research:Atmospheres, 106(D24):34007–34019. ! pages 101, 109, 111, 112Rignot, E., Jacobs, S., Mouginot, J., and Scheuchl, B. (2013). Ice-Shelf Melting AroundAntarctica. Science, 341(6143):266–270. ! pages 2, 97Rignot, E., Koppes, M., and Velicogna, I. (2010). Rapid submarine melting of the calving faces ofWest Greenland glaciers. Nature Geoscience, 3(3):187–191. ! pages 12147Rignot, E. and Steffen, K. (2008). Channelized bottom melting and stability of floating ice shelves.Geophysical Research Letters, 35(L02503):1–5. ! pages 2, 11, 23, 97, 111Rudels, B. (2013). Arctic Ocean circulation, processes and water masses: A description ofobservations and ideas with focus on the period prior to the International Polar Year 20072009.Progress in Oceanography. ! pages 30Rudels, B., Friedrich, H. J., and Quadfasel, D. (1999). The Arctic circumpolar boundary current.Deep Sea Research Part II: Topical Studies in Oceanography, 46(6):1023–1062. ! pages 5, 25Salcedo-Castro, J., Bourgault, D., Bentley, S. J., and deYoung, B. (2013). Non-hydrostaticmodeling of cohesive sediment transport associated with a subglacial buoyant jet in glacialfjords: A process-oriented approach. Ocean Modelling, 63:30–39. ! pages 10Sciascia, R., Straneo, F., Cenedese, C., and Heimbach, P. (2013). Seasonal variability of submarinemelt rate and circulation in an East Greenland fjord. Journal of Geophysical Research: Oceans,118(5):2492–2506. ! pages 11, 23, 98, 117Seroussi, H., Morlighem, M., Rignot, E., Larour, E., Aubry, D., Ben Dhia, H., and Kristensen, S. S.(2011). Ice flux divergence anomalies on 79north Glacier, Greenland: THE 79north GLACIER.Geophysical Research Letters, 38(9). ! pages 102Sharp, M., Burgess, D. O., Cogley, J. G., Ecclestone, M., Labine, C., and Wolken, G. J. (2011).Extreme melt on Canada’s Arctic ice caps in the 21st century. Geophysical Research Letters,38(11):n/a–n/a. L11501. ! pages 2, 4Shirasawa, K. and Ingram, R. G. (1991). Characteristics of the turbulent oceanic boundary layerunder sea ice. Part 1: A review of the ice-ocean boundary layer. Journal of Marine Systems,2(1):153–160. ! pages 105Smith, J. A., Hodgson, D. A., Bentley, M. J., Verleyen, E., Leng, M. J., and Roberts, S. J. (2006).Limnology of Two Antarctic Epishelf Lakes and their Potential to Record Periods of Ice ShelfLoss. Journal of Paleolimnology, 35(2):373–394. ! pages 7, 8, 18, 62, 63Steele, M. and Boyd, T. (1998). Retreat of the cold halocline layer in the Arctic Ocean. Journal ofGeophysical Research: Oceans, 103(C5):10419–10435. ! pages 5, 30, 45Steele, M., Mellor, G. L., and Mcphee, M. G. (1989). Role of the Molecular Sublayer in theMelting or Freezing of Sea Ice. Journal of Physical Oceanography, 19(1):139–147. ! pages17, 18Steele, M., Morison, J., Ermold, W., Rigor, I., Ortmeyer, M., and Shimada, K. (2004). Circulationof summer Pacific halocline water in the Arctic Ocean. Journal of Geophysical Research,109(C02027):1–18. ! pages 5, 25Steur, L., Steele, M., Hansen, E., Morison, J., Polyakov, I., Olsen, S. M., Melling, H., McLaughlin,F. A., Kwok, R., and Smethie, W. M. (2013). Hydrographic changes in the Lincoln Sea in theArctic Ocean with focus on an upper ocean freshwater anomaly between 2007 and 2010.Journal of Geophysical Research: Oceans, 118(9):4699–4715. ! pages 5, 25148Stevens, L. A., Straneo, F., Das, S. B., Plueddemann, A. J., Kukulya, A. L., and Morlighem, M.(2015). Linking catchment-scale subglacial discharge to subsurface glacially modified watersnear the front of a marine terminating outlet glacier using an autonomous underwater vehicle.The Cryosphere Discussions, 9(5):4583–4624. ! pages 43Stigebrandt, A. (1990). On the response of the horizontal mean vertical density distribution in afjord to low-frequency density fluctuations in the coastal water. Tellus A, 42(5):605–614. !pages 116Stigebrandt, A. (2012). Hydrodynamics and Circulation of Fjords. In Bengtsson, L., Herschy,R. W., and Fairbridge, R. W., editors, Encyclopedia of Lakes and Reservoirs, pages 327–344.Springer Netherlands, Dordrecht. ! pages 8, 9Straneo, F. and Cenedese, C. (2015). The Dynamics of Greenland’s Glacial Fjords and Their Rolein Climate. Annual Review of Marine Science, 7(1):89–112. ! pages 2, 4, 9, 10, 12, 22, 43, 114Straneo, F., Curry, R. G., Sutherland, D. A., Hamilton, G. S., Cenedese, C., Vge, K., and Stearns,L. A. (2011). Impact of fjord dynamics and glacial runoff on the circulation near HelheimGlacier. Nature Geoscience, 4(5):322–327. ! pages 11, 23, 36Straneo, F., Hamilton, G. S., Sutherland, D. A., Stearns, L. A., Davidson, F., Hammill, M. O.,Stenson, G. B., and Rosing-Asvid, A. (2010). Rapid circulation of warm subtropical waters in amajor glacial fjord in East Greenland. Nature Geoscience, 3(3):182–186. ! pages 9, 18, 23Straneo, F., Sutherland, D. A., Holland, D., Gladish, C., Hamilton, G., Johnson, H., Rignot, E., Xu,Y., and Koppes, M. (2012). Characteristics of ocean waters reaching Greenlands glaciers.Annals of Glaciology, 53(60):202–210. ! pages 2, 22, 45Sutherland, D. A. and Straneo, F. (2012). Estimating ocean heat transports and submarine meltrates in Sermilik Fjord, Greenland, using lowered acoustic Doppler current profiler (LADCP)velocity profiles. Annals of Glaciology, 53(60):50–58. ! pages 9, 12, 23Sutherland, D. A., Straneo, F., and Pickart, R. S. (2014). Characteristics and dynamics of twomajor Greenland glacial fjords. Journal of Geophysical Research: Oceans, 119(6):3767–3791.! pages 9, 11Thomson, R. E. (1981). Oceanography of the British Columbia Coast. Number 56 in CanadianSpecial Publication of Fisheries and Aquatic Sciences. Ottawa, Canada. ! pages 37Turner, J. S. (1973). Buoyancy Effects in Fluids. Cambridge University Press. Cambridge BooksOnline. ! pages 10Van-Hove, P., Swadling, K., Gibson, J., Belzile, C., and Vincent, W. (2001). Farthest north lakeand fjord populations of calanoid copepods Limnocalanus macrurus and Drepanopus bungei inthe Canadian high Arctic. Polar Biology, 24(5):303–307. ! pages 7Van Wychen, W., Burgess, D. O., Gray, L., Copland, L., Sharp, M., Dowdeswell, J. A., andBenham, T. J. (2014). Glacier velocities and dynamic ice discharge from the Queen ElizabethIslands, Nunavut, Canada. Geophysical Research Letters, 41(2):484–490. ! pages 4, 26, 32,47, 101149Van Wychen, W., Davis, J., Burgess, D. O., Copland, L., Gray, L., Sharp, M., and Mortimer, C.(2016). Characterizing interannual variability of glacier dynamics and dynamic discharge(1999-2015) for the ice masses of Ellesmere and Axel Heiberg Islands, Nunavut, Canada.Journal of Geophysical Research: Earth Surface, 121(1):39–63. ! pages 4, 25, 98, 99, 101,112, 124Veillette, J., Lovejoy, C., Potvin, M., Harding, T., Jungblut, A. D., Antoniades, D., Chenard, C.,Suttle, C. A., and Vincent, W. F. (2011a). Milne Fiord epishelf lake: A coastal Arctic ecosystemvulnerable to climate change. Ecoscience, 18(3):304–316. ! pages 3, 7, 26, 40, 63Veillette, J., Martineau, M., Antoniades, D., Sarrazin, D., and Vincent, W. (2011b). Effects of lossof perennial lake ice on mixing and phytoplankton dynamics: insights from High Arctic Canada.Annals of Glaciology, 51(56):56–70. ! pages 26, 64, 65, 66, 78Veillette, J., Mueller, D. R., Antoniades, D., and Vincent, W. F. (2008). Arctic epishelf lakes assentinel ecosystems: Past, present and future. Journal of Geophysical Research,113(G04014):1–11. ! pages 2, 3, 7, 8, 22, 26, 62, 63, 65, 66, 87, 134Vincent, W., Gibson, J., and Jeffries, M. (2001). Ice-shelf collapse, climate change, and habitatloss in the Canadian high Arctic. Polar Record, 37(201):133–142. ! pages 2, 6, 7, 18, 22, 62,63, 98, 134Vincent, W. F. and Laybourn-Parry, J. (2008). Polar Lakes and Rivers: Limnology of Arctic andAntarctic Aquatic Ecosystems. Oxford University Press, Oxford. ! pages 7Vincent, W. F., Whyte, L. G., Lovejoy, C., Greer, C. W., Laurion, I., Suttle, C. A., Corbeil, J., andMueller, D. R. (2009). Arctic microbial ecosystems and impacts of extreme warming during theInternational Polar Year. Polar Science, 3(3):171–180. ! pages 62Wand, U., Hermichen, W.-D., Brggemann, E., Zierath, R., and Klokov, V. D. (2011). Stableisotope and hydrogeochemical studies of Beaver Lake and Radok Lake, MacRobertson Land,East Antarctica. Isotopes in Environmental and Health Studies, 47(4):407–414. ! pages 7, 62White, A., Copland, L., Mueller, D., and Van Wychen, W. (2015a). Assessment of historicalchanges (19592012) and the causes of recent break-ups of the Petersen ice shelf, Nunavut,Canada. Annals of Glaciology, 56(69):65–76. ! pages 1, 6, 22, 46, 62, 65, 98White, A., Mueller, D., and Copland, L. (2015b). Reconstructing hydrographic change in PetersenBay, Ellesmere Island, Canada, inferred from SAR imagery. Remote Sensing of Environment,165:1–13. ! pages 8, 134Wilson, N. J., Flowers, G. E., and Mingo, L. (2013). Comparison of thermal structure andevolution between neighboring subarctic glaciers. Journal of Geophysical Research: EarthSurface, 118(3):1443–1459. ! pages 27Wilson, N. J. and Straneo, F. (2015). Water exchange between the continental shelf and the cavitybeneath Nioghalvfjerdsbrae (79 North Glacier). Geophysical Research Letters,42(18):7648–7654. ! pages 11150Xu, Y., Rignot, E., Fenty, I., Menemenlis, D., and Flexas, M. M. (2013). Subaqueous melting ofStore Glacier, west Greenland from three-dimensional, high-resolution numerical modeling andocean observations. Geophysical Research Letters, 40(17):4648–4653. ! pages 11, 12, 117Xu, Y., Rignot, E., Menemenlis, D., and Koppes, M. (2012). Numerical experiments onsubaqueous melting of Greenland tidewater glaciers in response to ocean warming and enhancedsubglacial discharge. Annals of Glaciology, 53(60):229–234. ! pages 11, 23, 117Zwally, H., Schutz, R., Bentley, C., Bufton, J., Herring, T., Minster, J., Spinhirne, J., and Thomas,R. (2011). GLAS/ICESat L2 Antarctic and Greenland ice sheet altimetry data V031. DigitalMedia, NASA Distributed Active Archive Center at the National Snow and Ice Data Center,Boulder, Colorado USA. ! pages 27151


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