UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A model analysis of water resource availability in response to climate change and oil sands operations… Leong, Doris Nian-Shiah 2014

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

Item Metadata


24-ubc_2014_november_leong_doris.pdf [ 7.06MB ]
JSON: 24-1.0135576.json
JSON-LD: 24-1.0135576-ld.json
RDF/XML (Pretty): 24-1.0135576-rdf.xml
RDF/JSON: 24-1.0135576-rdf.json
Turtle: 24-1.0135576-turtle.txt
N-Triples: 24-1.0135576-rdf-ntriples.txt
Original Record: 24-1.0135576-source.json
Full Text

Full Text

A MODEL ANALYSIS OF WATER RESOURCEAVAILABILITY IN RESPONSE TO CLIMATE CHANGEAND OIL SANDS OPERATIONS IN THE ATHABASCARIVER BASINbyDORIS NIAN-SHIAH LEONGB.Sc., The University of British Columbia, 2004M.Sc., Dalhousie University, 2009A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Geography)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)October 2014c© Doris Nian-Shiah Leong, 2014AbstractThe Athabasca River Basin faces challenging tradeoffs between energy produc-tion and water security as climate change alters the seasonal freshwater supplyand water demand from the oil sands mining industry is projected to increase. Ef-fective water management will depend on a physical understanding of the scaleand timing of water supply and demand. This dissertation aims to synthesize theimpacts of water withdrawals and climate change on streamflow in the Athabascaoil sands region, in order to develop a scientific basis for the management of waterresources. The combination of a land surface process model and a hydrologicalrouting model is used to evaluate the influence of water withdrawals and climatechange on streamflow under a variety of different scenarios, and to evaluate theadaptation options.Climate warming is projected to be the primary driver of future streamflowavailability, with little influence from direct water withdrawals. Seasonal patternsthat show a decline in summer flows and an increase in winter flows are con-sistent with the response of a snowmelt-dominated basin to warming. Increasesin the frequency of low flows that are below a threshold of maximum environ-mental protection suggest that daily bitumen production could be interrupted byup to 2-3 months a year by mid-century. It is also projected that water storagewill be required to supplement river withdrawals to maintain continuous bitumenproduction under the impacts of future climate warming. Based on the modelresults, a range of water management options are developed to describe the poten-tial tradeoffs between the scale of bitumen production and industry growth, wateriistorage requirements, and environmental protection for the aquatic ecosystems.This physically-based assessment of future water tradeoffs can inform water pol-icy, water management decisions, and climate change adaptation plans, with ap-plicability to other regions facing trade-offs between industrial development andecosystem water needs.iiiPrefaceThis dissertation is original, unpublished, independent work by the author, DorisNian-Shiah Leong.Chapters 3, 4, and 5 have been written in preparation for submission to peer-reviewed journals. Due to this manuscript format, there is some repetition in theintroductory and methods content between Chapters 2, 3, 4, and 5.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Research problem . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Background on the Athabasca River Basin . . . . . . . . . . . . . 21.2.1 Geography and hydrology . . . . . . . . . . . . . . . . . 21.2.2 Climate change . . . . . . . . . . . . . . . . . . . . . . . 51.2.3 Human activity and the oil sands industry . . . . . . . . . 71.2.4 Water use management . . . . . . . . . . . . . . . . . . . 91.3 Research objectives and contribution . . . . . . . . . . . . . . . . 111.4 Research strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 121.5 Structure of dissertation . . . . . . . . . . . . . . . . . . . . . . . 13v2 Model Validation for the Athabasca River Basin . . . . . . . . . . . 152.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.2 Model description . . . . . . . . . . . . . . . . . . . . . . . . . . 162.2.1 Integrated Biosphere Simulator (IBIS) . . . . . . . . . . . 162.2.2 Terrestrial Hydrology Model with Biogeochemistry (THMB) 192.3 Data sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.3.1 Climate . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.3.2 Soil and vegetation . . . . . . . . . . . . . . . . . . . . . 212.3.3 Geomorphology and hydrology . . . . . . . . . . . . . . . 222.4 Model parameterization . . . . . . . . . . . . . . . . . . . . . . . 272.4.1 Vertical water budget . . . . . . . . . . . . . . . . . . . . 272.4.2 Lateral water budget . . . . . . . . . . . . . . . . . . . . 322.5 Model performance . . . . . . . . . . . . . . . . . . . . . . . . . 352.5.1 Annual and seasonal variability . . . . . . . . . . . . . . . 352.5.2 Interannual variability . . . . . . . . . . . . . . . . . . . . 392.5.3 Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . 462.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473 Sensitivity of Streamflow to Water Withdrawals for the AthabascaOil Sands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.2.1 Model description: IBIS . . . . . . . . . . . . . . . . . . 513.2.2 Model description: THMB . . . . . . . . . . . . . . . . . 533.2.3 Water withdrawals . . . . . . . . . . . . . . . . . . . . . 553.2.4 Streamflow simulations . . . . . . . . . . . . . . . . . . . 603.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603.3.1 Model validation . . . . . . . . . . . . . . . . . . . . . . 603.3.2 Impact of oil sands water withdrawals . . . . . . . . . . . 633.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653.4.1 Streamflow impacts . . . . . . . . . . . . . . . . . . . . . 65vi3.4.2 Data needs . . . . . . . . . . . . . . . . . . . . . . . . . 683.4.3 Water use uncertainties . . . . . . . . . . . . . . . . . . . 683.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704 Streamflow Availability under Climate Warming in the AthabascaOil Sands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754.2.1 Land surface models . . . . . . . . . . . . . . . . . . . . 754.2.2 Future climate projections . . . . . . . . . . . . . . . . . 764.2.3 IBIS-THMB simulations . . . . . . . . . . . . . . . . . . 764.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794.3.1 Climate projections . . . . . . . . . . . . . . . . . . . . . 794.3.2 Streamflow projections . . . . . . . . . . . . . . . . . . . 804.3.3 Frequency of low flows . . . . . . . . . . . . . . . . . . . 874.3.4 Water withdrawals . . . . . . . . . . . . . . . . . . . . . 894.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 924.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 955 Future Water Supply and Demand Management Options in theAthabasca Oil Sands . . . . . . . . . . . . . . . . . . . . . . . . . . . 965.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 965.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 985.2.1 Water management scenarios . . . . . . . . . . . . . . . . 985.2.2 Climate scenarios and models . . . . . . . . . . . . . . . 1015.2.3 Simulating water supply and demand . . . . . . . . . . . . 1025.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1065.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1136 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115vii6.1 Key insights and findings . . . . . . . . . . . . . . . . . . . . . . 1156.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1176.3 Strengths and limitations of the research . . . . . . . . . . . . . . 1186.4 Potential future research directions . . . . . . . . . . . . . . . . . 120Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123viiiList of TablesTable 2.1 Hydraulic and geomorphic observations of river reaches in theARB (from Kellerhals et al. [1972]). . . . . . . . . . . . . . . 23Table 2.2 Hydrometric stations on the Athabasca River measuring monthlydischarge and with recorded data within the 30-year time pe-riod of interest, 1981–2010. . . . . . . . . . . . . . . . . . . . 26Table 2.3 Final saturated hydraulic conductivities (ks) assigned to IBISsoil textures. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Table 2.4 Average annual and seasonal statistics at four gauging stationsbetween 1981–2010, comparing model output to observed data.Values are model discharge as a percentage greater (+) or less (-) than observed discharge, while the shift in peak flow betweenmodel and observations is in days. Seasons are summer (July–October), winter (November–March), and spring (April–June). 38Table 2.5 Evaluation of model performance (streamflow output) usingthree statistics for monthly timesteps. Performance ratings inbrackets are either unsatisfactory (u), satisfactory (s), good (g)or very good (v), as defined by Moriasi et al. [2007]. The modelwas evaluated over the entire time period of interest, as well asover each decade. . . . . . . . . . . . . . . . . . . . . . . . . 44ixTable 2.6 Comparison of month-averaged daily streamflow outputs withmonthly streamflow outputs using three statistics. Performanceratings in brackets are either unsatisfactory (u), satisfactory (s),good (g) or very good (v), as defined by Moriasi et al. [2007].The simulations were evaluated over the entire time period ofinterest, as well as over each decade, at Below McMurray. . . . 45Table 3.1 Water withdrawal source locations for all current and plannedoil sands mining operations. . . . . . . . . . . . . . . . . . . . 56Table 3.2 Average annual and seasonal statistics of streamflow in eachwater use scenario, given as a percentage difference relative tothe control scenario. The shifts in peak and minimum flows aremeasured in days. . . . . . . . . . . . . . . . . . . . . . . . . 64Table 4.1 CMIP5 global climate models used in this study, their equilib-rium climate sensitivities and resolution (from Andrews et al.[2012]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77Table 5.1 Annual water withdrawal rules for the industry-first scenario.For each set of weeks, a water rule (R) defines the maximumpermitted withdrawal rate when the river flow (F) meets a spec-ified threshold (T) condition (adapted from the Option H rulesin Ohlson et al. [2010]). . . . . . . . . . . . . . . . . . . . . . 100Table 5.2 The minimum storage capacity (Mm3) in each managementscenario that is required to maintain the indicated average in-dustry water withdrawal rate across all GCM-driven stream-flow projections for mid-century (2041–2060). The number ofdays that the storage volume can supply demand at the averageindustry withdrawal rate, is also shown. . . . . . . . . . . . . . 104xTable 5.3 Example of the weekly water rules in the environment-first sce-nario for GFDL-ESM2G and RCP4.5. The water rule (R) de-fines the maximum permitted withdrawal rate when the weeklyaverage river flow (F) meets the Q80 threshold (T) condition.Weeks are grouped here for brevity, but R is calculated sepa-rately for each week. . . . . . . . . . . . . . . . . . . . . . . . 106Table 5.4 Matrix showing management options for a range of prioritiesbased on the evaluation of the industry-first and environment-first scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . . 107Table 5.5 Costs and footprint of freshwater pond storage per unit meter[Ohlson et al., 2010] and the calculated costs and footprint ofstorage requirements associated with the different managementoptions considered. . . . . . . . . . . . . . . . . . . . . . . . . 111xiList of FiguresFigure 1.1 The Athabasca River Basin (shaded grey) shown as a sub-basin of the Mackenzie River Basin. Inset shows the basinlocation within Canada. . . . . . . . . . . . . . . . . . . . . . 3Figure 2.1 Schematic representation of IBIS (left) and THMB (right) mod-els showing the interactions between input climate data, sim-ulated land surface physics, and river dynamics (provided bythe Center for Sustainability and the Global Environment (SAGE)at University of Wisconsin). . . . . . . . . . . . . . . . . . . 17Figure 2.2 Map of hydrometric stations in the Athabasca River Basin. . . 25Figure 2.3 Systematic variation of three IBIS parameters (top soil layerthickness (a–d), total soil depth (e–h), saturated hydraulic con-ductivity (j–l)) and the effects on runoff and actual evapo-transpiration (aet) patterns. Each parameter was varied whilekeeping the others at the following default values: top soillayer thickness = 0.2 m, soil depth = 2 m, hydraulic conduc-tivity = 10−7− 10−5 m/s. The IBIS output shown for eachmonth is an average over a five year time interval, 1984–1988. 29Figure 2.4 Model runs over the historical period of 1981–2010 comparedto streamflow observations at the monitoring stations (a) Be-low McMurray, (b) At Athabasca. . . . . . . . . . . . . . . . 37xiiFigure 2.5 Comparison of the surface (surf) and subsurface (sub) runoffgenerated by IBIS for (a) downstream (Below McMurray) and(b) upstream (Near Jasper) locations, along with simulated(sim) and observed (obs) streamflow. Runoff values are summedfrom all grid cells upstream of the station location. . . . . . . 40Figure 2.6 Monthly simulated and observed discharge over the 30-yeartime period 1981-2010, at the Below McMurray location. . . . 41Figure 3.1 Map of the ARB region showing locations of hydrometricstations and model analysis. Inset shows water withdrawalsource locations in the oil sands mining region (see Table 3.1for label references). . . . . . . . . . . . . . . . . . . . . . . 58Figure 3.2 Frequency of occurrence of low flows for each month in eachwithdrawal scenario, at the location Below Ops. In the controlscenario, the lowest 20% of flows (dashed line) defines the lowflow threshold applied to the other water use scenarios. . . . . 66Figure 4.1 Map of the ARB region showing the location of the hydromet-ric station Below McMurray and the location of analysis forstreamflow simulations downstream of oil sands mining oper-ations, Below Ops. Inset shows water withdrawal source lo-cations used in the study (note: there are 20 licenses assignedto 18 physical locations). . . . . . . . . . . . . . . . . . . . . 73Figure 4.2 Change in (a) annual precipitation (∆P), and (b) the ratio ofrain to snow, relative to the change in annual temperature (∆T)projected by the three GCMs and four RCPs in mid-century(2041–2060) and end-of-century (2081–2100). . . . . . . . . 81xiiiFigure 4.3 Streamflow patterns for the Athabasca River at the locationBelow Ops: (a) annual mean streamflow, (b) centroid of flowdistribution, (c) timing of spring runoff, (d) persistence offlow. Dashed lines show the mean values across all GCMs andshaded areas show the range of values across GCMs. Years arethe mid-point of running 20-year time windows over which re-sults are averaged. RCP2.6 and RCP6.0 are omitted for clarity. 82Figure 4.4 Simulated streamflow relative to the (a) change in annual tem-perature (∆T), and (b) change in annual precipitation (∆P)projected by the three GCMs and four RCPs in mid-century(2041–2060) and end-of-century (2081–2100). . . . . . . . . 83Figure 4.5 Changes in streamflow patterns by mid-century (2041–2060)and end-of-century (2081–2100) relative to today (1991–2010)for the Athabasca River at the location Below Ops: (a–b) shiftin the centroid of flow distribution, (c–d) shift in the timing ofspring runoff, (e–f) shift in the persistence of flow. . . . . . . . 85Figure 4.6 Linear relationships between the ratio of rain to snow and the(a) timing of the flow centroid (b) timing of spring runoff (c)persistence of flow. . . . . . . . . . . . . . . . . . . . . . . . 86Figure 4.7 Low flow frequency for each month of the year, for three 20-year time windows: (a) today, 1991–2010 (b) mid-century,2041–2060 (c) end-of-century, 2081–2100. Dashed lines showthe mean values across all GCMs and shaded areas show therange of values across GCMs. RCP2.6 and RCP6.0 are omit-ted for clarity. . . . . . . . . . . . . . . . . . . . . . . . . . . 88xivFigure 4.8 Change in low flow frequency relative to today (1991–2010)for (a) RCP4.5 at mid-century (2041–2060), (b) RCP4.5 atend-of-century (2081–2100), (c) RCP8.5 at mid-century (2041–2060), (d) RCP8.5 at end-of-century (2081–2100). Red showsthe change due to climate change only and blue shows thechange due to water withdrawals. Dashed and dotted linesshow the mean value across all GCMs and shaded areas showthe range of values across GCMs. . . . . . . . . . . . . . . . 90Figure 4.9 Change in the number of months during the mid-century (2041–2060) and end-of-century time (2081–2100) periods that flowfalls below the low flow threshold, relative to today (1991–2010), for each climate scenario. . . . . . . . . . . . . . . . . 91Figure 5.1 The percentage of time for each week of the year during mid-century (2041–2060) that river flow withdrawals cannot sup-ply the full average industry withdrawal rate of 16 m3/s for (a)the industry-first scenario and (b) the environment-first scenario.105xvAcknowledgmentsI am grateful for all the guidance, support and encouragement I have received inworking towards the completion of this dissertation. The contributions of the fol-lowing organizations and individuals made this work possible, from conception,through the long hours, to completion.This research was funded by an NSERC Graduate Fellowship, a University ofBritish Columbia graduate fellowship, and an NSERC Discovery Grant. I wish tothank these institutions for their support.My sincere thanks to David Price for dispensing valuable modelling adviceand regional insights. Thank you to Naba Adhikari for thorough and expedientassistance with data collection. Much gratitude goes to Peter Snyder who patientlyentertained my endless modelling questions, and generously helped troubleshootmany model issues as they arose. Thanks also to Dan Moore for early feedbackon my research goals.I would like to thank my committee members, Marwan Hassan, Mark Johnson,and Michael Coe, for their invaluable feedback and encouragement in craftingthis thesis and in parsing out the interdisciplinary elements. Your careful andinsightful comments always kept me on track and challenged me to gain a broaderperspective.Most importantly, thank you to my advisor, Simon Donner, for your unwaver-ing patience and support as I made and re-made a thesis out of a patchwork ofideas and interests. Your insights and inspiration made this work possible, andyour encouragement and eternal optimism made it possible for me.xviFor my familyxviiChapter 1Introduction1.1 Research problemChanges in the magnitude and timing of streamflow can disrupt river ecosystemsand human activities that are accustomed to seasonal water availability. Suchchanges can arise due to shifts in the hydroclimatological regime, and also dueto direct anthropogenic alterations to streamflow. In the Athabasca River Basin,the intersection of climate change and a rapidly expanding oil sands industry willpose future challenges for water management to maintain seasonal water availabil-ity for both ecosystem and industry needs. The effective management of waterresources will require a strong understanding of the climate-driven and human-driven impacts on future water supply, as well as an understanding of projectedwater demand.It is uncertain whether the future Athabasca River streamflow, under changingclimatic conditions, will be sufficient to support forecasted water use by the oilsands industry. As a result, it may be equally important for future water manage-ment to focus on the impacts on industry, in addition to the impacts of industry.Projected climate change impacts on basin streamflow must be integrated with wa-ter use patterns to identify and adapt to future deficits in water availability. Watermanagement can then make informed decisions on tradeoffs between ecosystem1protection and industry growth over long-term century timescales.1.2 Background on the Athabasca River Basin1.2.1 Geography and hydrologyThe Athabasca River Basin (ARB) extends across the Canadian provinces of Al-berta, Saskatchewan and a small area of the Northwest Territories (Figure 1.1). Itis the southernmost subbasin of the Mackenzie River Basin, which drains north-ward into the Arctic Ocean. The ARB’s major artery is the Athabasca River; at1538 km long, it is the third longest undammed river in North America. Alongwith Lake Athabasca, which covers 7,935 km2, the basin river system drains anarea of 269,000 km2 [MRRB, 2004].The Athabasca River originates in the Columbia ice fields and flows northeast,traversing a variety of ecozones including the Cordillera/Rocky Mountains, Bore-al/Interior Plains, and the Canadian Shield. These regions contain diverse ecosys-tems including glaciers, alpine meadows, alpine and boreal forests, and muskegwithin unique landscapes and wildlife habitat including the Peace-Athabasca Delta,the Cardinal River headwaters, McClelland Lake, and the Richardson Sand Dunes[MRRB, 2004, Holloway and Clare, 2012]. In upland areas, coniferous forest,mixed wood, and deciduous forest are the dominant vegetation, with willow brush,shrubs, black spruce, and sphagnum moss dominating lowland areas [Kerkhovenand Gan, 2006]. Between the town of Athabasca and the city of Fort McMurray,extensive muskeg regions occur [Hamilton et al., 1985].2LakeAthabascaGreat BearLakeGreat SlaveLakeLiard RiverPeel RiverLesserSlaveLakePeace -AthabascaDeltaMackenzie RiverPeace RiverSlave R.Fort McMurrayAthabasca RiverBritish Columbiariti  Northwest TerritoriesMACKENZIE RIVER BASINATHABASCARIVER BASINYukonSaskatchewanAlbertaFigure 1.1: The Athabasca River Basin (shaded grey) shown as a sub-basin of the Mackenzie River Basin.Inset shows the basin location within Canada.3The Athabasca River supports over 30 species of fish and serves as an im-portant transportation route (historical trade routes, current recreational activities,etc.). The river environment has also supported large communities of aboriginalpeople for many centuries. The Athabasca River drains into the Peace-AthabascaDelta (PAD), an ecologically sensitive region comprising a 6000 km2 complex ofwetlands and lakes at the western end of Lake Athabasca. The PAD is an impor-tant nesting and staging area for up to one million migratory birds and provideshabitat to roughly 5000 bison, along with many other wildlife.Historically, the ARB has a continental climate with cold, dry winters andshort, cool summers. The average annual temperature of the ARB is 2◦C [Burnet al., 2004]. Winter daily mean temperatures between mid-October to early Aprilare below 0◦C [Kerkhoven and Gan, 2006], with mean January temperatures vary-ing from -15 to -25◦C. In the summer, July temperatures range from 10–15◦C inthe headwaters to 15–17◦C near Fort McMurray [Hamilton et al., 1985]. An-nual precipitation averages 800 mm in the mountains, 500–600 mm in the centralpart of the basin, and 400–500 mm in the northeast [Kerkhoven and Gan, 2006,Hamilton et al., 1985]. The majority of precipitation, up to 75%, falls in the sum-mer months between June and October as major rainstorms [Longley and Janz,1978, Burn et al., 2004].The flow regime of the Athabasca River is typical of northern rivers, charac-terized by low flows in the winter and rising discharge due to snowmelt startingin late April and May [Burn et al., 2004]. The bulk of annual discharge occursin the late spring and early summer, with peak flow in June or July, followed bya gradual recession to low flows in December through February [Choles, 1996].Together with the adjacent Peace River Basin, the Athabasca River Basin sustainsmuch of the low winter flow of the Mackenzie River [Woo and Thorne, 2003].The Athabasca River is fed by four major tributaries which together accountfor under 50% of the total river discharge below Fort McMurray. These are theMcLeod River (10%), Pembina River (6%), Lesser Slave River (8%), and Clear-water River (18%). Lesser Slave Lake, roughly 1,160 km2, drains eastward into4the Athabasca River via the Lesser Slave River. The storage capacity of LesserSlave Lake can dampen the magnitude of peak discharge by delaying the timingof flow from the Lesser Slave River into the Athabasca River, relative to contri-butions from other tributaries, and can contribute to a broader peak in the annualhydrograph [Choles, 1996].The lakes and wetlands in the Peace Athabasca Delta experience annual peri-odic flooding and are highly sensitive to natural variability in river flows and wa-ter levels [Peters et al., 2006]. Flow regulation of the Athabasca River thereforehas important consequences for these ecological systems [Alberta Environment,2007]. Although ice jams on the Athabasca River can lead to flooding in the win-ter that can also contribute to seasonal inundation of the Peace-Athabasca Delta[Kowalcyk and Hicks, 2003, Prowse et al., 2006], most flood activity in the basinoccurs during open water season between May and the end of July, and reflectshigh precipitation or snowmelt runoff [Choles, 1996].1.2.2 Climate changeHigh-latitude regions are especially sensitive to the effects of climate warmingand are expected to warm more quickly than lower latitudes [Hassol, 2004, IPCC,2007]. Temperatures in the ARB have increased on average by 1.5–1.8◦C between1961–2000, three times higher than the global average rise of 0.6◦C [Bruce, 2006].Temperatures are expected to continue increasing considerably in the future, withthe most recent IPCC projections for Northwest Canada showing a mean annualtemperature increase of 2.7◦C by the middle of the century and 3.5◦C by the endof the century. Previous studies have projected temperature increases of up to 4◦Cby mid-century, and up to 6◦C by the end of the century [Gan and Kerkhoven,2004, Prowse et al., 2006, Sauchyn and Kulshreshtha, 2008, Kerkhoven and Gan,2011]. Annual precipitation and potential evapotranspiration are also expectedto rise through the 21st century, while winter precipitation that normally falls assnow is expected to increasingly fall as rain [Schindler and Donahue, 2006]. Themost recent IPCC projections for the Athabasa River Basin region show an annual5precipitation increase of 3% by mid-century and 4% by the end of the century[Christensen et al., 2013].Summer flows in the Athabasca River have been observed to decline by al-most 30% since 1970 due to climate warming [Swainson, 2009]. Warming hasalso led to the rapid shrinking of glaciers in the basin headwaters by 25% in thelast century [Watson and Luckman, 2004] and to the subsequent reduction in flowsfed by glacial sources that will eventually cease to exist [Hopkinson and Young,1998]. Increased temperatures have driven a progressively earlier snowmelt inrecent decades [Serreze et al., 2000, Zhang et al., 2001] and since the mid-20thcentury, observations near Slave Lake show that the number of days winter snowhas remained on the ground has decreased by 25% (39 days) and the maximumsnowpack depth has declined by 54% (27 cm) [Schindler and Donahue, 2006].Observations of reduced snowpack accumulation, periodic and earlier snowmelt,and reduced summer flows [Serreze et al., 2000, Zhang et al., 2001, Sauchyn andKulshreshtha, 2008] are consistent with modelling studies which project that fu-ture warming will result in a lower and earlier spring freshet and reduced summerflows due to low snow accumulation and an earlier snowpack melt [Pietroniroet al., 2006, Schindler and Donahue, 2006, Sauchyn and Kulshreshtha, 2008].Although these general hydroclimatic shifts are projected for all climate sce-narios, the degree and direction of estimated change in monthly streamflow isheavily dependent on the climate change scenario [Toth et al., 2006]. While someprojections show trends toward increased streamflow volumes, others show anoverall decline in streamflow by the end of the century [Toth et al., 2006, Schindleret al., 2007, Swainson, 2009, Kerkhoven and Gan, 2011]. Variability in stream-flow projections is due in part to uncertainty in the balance between potentialevapotranspiration and precipitation as temperatures increase [Sauchyn and Kul-shreshtha, 2008]. Regardless, seasonal declines in streamflow may lead to anincreased potential for drought and water supply problems [Lapp et al., 2005], aswell as a decrease in the frequency of floods that replenish the lakes and wetlandsin the Peace-Athabasca Delta [Prowse et al., 2006, Wolfe et al., 2005, 2008].61.2.3 Human activity and the oil sands industryIn 2001, the human population of the Athabasca River Basin was 155,000, al-though rapid growth and urban and industrial development has likely since in-creased that number significantly. In Fort McMurray, where the majority of thebasin resides, an 80% growth in population occurred between 2000 and 2010; thepopulation of Fort McMurray alone is forecasted to increase to 205,000 by 2028[Regional Municipality of Wood Buffalo, 2010]. There are currently more than200 populated centres with greater than 2000 people in the ARB [Squires et al.,2009].Economic activity in the ARB is led by agriculture, forestry, and coal min-ing in the upstream half of the basin [ERCB, 2010, Holloway and Clare, 2012].Agriculture accounts for roughly 12% of the basin area and consists primarily offorage crops [Wrona et al., 2000, MRRB, 2004]. Forestry is also a major industry,and is active across the basin, including several sawmills and pulp mills. Com-mercial fishing and trapping are also prominent industries which have remainedrelatively static during the past decade. Meanwhile, uranium mining in the basinproduces about a quarter of the world supply of rich, high-grade uranium [Panaˇand Olson, 2009]. In its downstream reaches, alongside a growing conventionaloil and gas industry, the Athabasca River Basin is also home to a burgeoning oilsands industry.Since the late 1990s, the oil sands resource has become an important driverof Alberta’s economy and is expected to continue to play a key role in Canada’sfuture economy as world demand for energy continues to rise. Global demandfor oil is expected to rise from 85.7 million barrels per day in 2008 to 112.2 mil-lion barrels per day in 2035 [Conti and Holtberg, 2011], and Canadian oil sandsproduction is projected to increase from 1.5 to 4.8 million barrels per day overthe same time period [Conti and Holtberg, 2011, CAPP, 2012]. As conventionalcrude oil reserves become depleted, nonconventional sources such as the oil sandshave become more important and now account for 60% of Canadian production[Environment Canada, 2014].7The oil sands deposits in northern Alberta constitute a reliable, long-term sup-ply for the growing global demand for crude oil. The Alberta oil sands are es-timated to contain as much as 1.7 trillion barrels of bitumen, with reserves (theamount recoverable economically with existing technology) estimated at 170 bil-lion barrels of bitumen [Alberta Environment, 2009]. The deposits span approxi-mately 142,200 km2 and are divided into three regions: the Athabasca Wabiskaw-McMurray deposit (∼80% of oil reserves), the Cold Lake Clearwater deposit(∼12% of oil reserves) and the Peace River deposit (∼8% of oil reserves). TheAthabasca Wabiskaw-McMurray deposit lies within the Athabasca River Basin,and surface mineable bitumen covers a 4800 km2 area within this deposit, withroughly 602 km2 currently disturbed [Alberta Environment, 2009].There are several key environmental concerns associated with oil sands bitu-men production. Greenhouse gas (GHG) emissions of carbon dioxide and methaneare produced at every stage of the oil sands production life cycle, and exceedthe emissions of conventional oil production [Alberta Environment, 2009]. Oilsands operations contribute the single largest source of GHG emissions growthin Canada [Woynillowicz et al., 2005], although technological advancements inequipment, along with a decline in upgrading activity due to increased crude bitu-men export, have maintained a generally fixed emissions intensity since 2004 [En-vironment Canada, 2010]. Over 1400 known pollutants are emitted by oil sandsoperations [Weinhold, 2011], with the main contaminants being mercury, arsenic,and polycyclic aromatic hydrocarbons [Alberta Environment, 2009, Schindler et al.,2007]. The disposal of waste containing high concentrations of contaminantsinto tailings ponds may have significant impacts on local soil and groundwaterif seepage occurs, potentially leading to downstream water quality and ecosystemdegradation [Woynillowicz et al., 2005, Droitsch, 2009, Gosselin et al., 2010]. Inaddition to water quality, declining air quality is a concern, with recent monitoringshowing an increasing trend in nitrogen dioxide [Gosselin et al., 2010] and hydro-gen sulphide [Alberta Environment, 2009] due to the burning of fossil fuels duringbitumen production. Land disturbance is another concern, as oil sands deposits in8the basin require the removal of boreal forest and wetland environments to accessdeposits during surface mining [Alberta Environment, 2009]. Land reclamation isan ongoing but challenging process due to the high concentrations of contaminantsinvolved, and the rate of tailings pond creation still exceeds the rate of reclamation[Woynillowicz et al., 2005, Gosselin et al., 2010]. Furthermore, ecosystems arenot expected to be restored to their original state [Woynillowicz et al., 2005].One of the key environmental concerns with oil sands operations is its inten-sive freshwater use. In the ARB, oil sands mining already accounts for the largestconsumption of water from the Athabasca River [Schindler et al., 2007]. Theoil sands mining industry requires a constant supply of freshwater for continuousbitumen production throughout the year, with water necessary at each stage ofoil sands operations, including retrieval, processing, and upgrading. This waterdemand is projected to rapidly increase in the future as operations expand.Surface or open-pit mining is used to recover deposits near the surface, whilein situ methods are used for the recovery of deposits up to 400 m below the sur-face. Both surface and in situ mined deposits require water-intensive processingsteps to extract the bitumen and upgrade it into marketable commodities such asgasoline, diesel, and aviation fuels. However, while the average oil sands surfacemine currently uses roughly two to four barrels of freshwater to produce a bar-rel of oil, the average in situ project uses only about half a barrel of freshwaterto produce a barrel of oil, by making use of recycled and deep-well salt wateras an alternative to freshwater when possible [Alberta Environment, 2007]. As aresult, concerns over freshwater use are primarily associated with surface miningoil sands operations.1.2.4 Water use managementSurface water from lakes and rivers is the main source of water withdrawals fordomestic, agricultural, commercial, and industrial use in the ARB [AMEC Earth& Environmental, 2007, AWRI, 2011]. In 2005, approximately 760 million cu-bic metres of surface water in the ARB was allocated annually for human use9[AMEC Earth & Environmental, 2007]. The petroleum sector was the largest userof surface water (65%) in 2005, with the majority (94%) of that surface waterwithdrawn for oil sands surface mining operations [AMEC Earth & Environmen-tal, 2007]. Projections for 2025 under a high-growth scenario (with many con-siderations including population growth for the municipal and commercial sec-tors, livestock growth for the agricultural sector, forecasted economic activity forthe commercial sector, proposed projects for the petroleum sector, etc.) estimatethat the petroleum sector will continue to dominate surface water withdrawals(78%) with the majority of those withdrawals (86%) for oil sands surface mining[AMEC Earth & Environmental, 2007, Alberta Environment and Sustainable Re-source Development, 2014]. The percentage of surface water withdrawals for allother sectors are projected to decline by 2025 [AMEC Earth & Environmental,2007].Both the provincial and federal governments have jurisdiction over water usein the ARB. Major water users must obtain licenses from the provincial govern-ment which designate the conditions of operation and the amount of permitted wa-ter withdrawals, as well as the quality of returned water. The Athabasca oil sandsare currently mined by five companies who withdraw water from the AthabascaRiver - Canadian Natural Resources Ltd., Imperial Oil Limited, Shell Canada,Suncor Energy Inc., and Syncrude Canada Ltd. These oil sands companies cur-rently comply with Phase One of the Lower Athabasca River Water ManagementFramework, introduced in February 2007, and are currently licensed to withdraw441 million m3 of fresh water from the Athabasca River each year [Alberta En-vironment, 2007]. The framework describes the rules and restrictions on waterwithdrawals by major oil sands operators, in order to sustain in-stream flow needsin the Athabasca River. A weekly cap is placed on the rate at which oil sandscompanies can remove water from the Athabasca River, based on the natural andseasonal variability in river flow. The Phase One framework has been criticizedfor being unenforceable, not establishing incentives for industry to reduce wateruse, and neglecting the impact of climate change on future river flows [Swainson,102009]. Phase Two of the water management framework is currently in develop-ment and aims to gather further scientific and traditional knowledge to assess thepossible limitations of the Phase One framework, as well as to establish a mini-mum base flow below which withdrawals are no longer permitted [Ohlson et al.,2010]. One of the major drawbacks of the current water management systemis that water use reporting from oil sands mining operations remains voluntary[Woynillowicz and Severson-Baker, 2006], and an accurate historical and currentrecord of water demand from these operations is not readily available.Although water withdrawals from the Athabasca River for oil sands operationscurrently represent a small fraction of the total river flow, the continued intensifi-cation of water withdrawals for expanded oil sands resource extraction may pose afuture risk to the sustainable provision of adequate flows [Bruce, 2006, Schindleret al., 2007, Mannix et al., 2010]. In addition, although the magnitude of totalriver withdrawals is relatively small when expressed as a percentage of annualflow (∼ 1−−2%), withdrawals can be large relative to low winter flows [Bruce,2006, AMEC Earth & Environmental, 2007, Alberta Environment, 2007, Wein-hold, 2011]. Flow regulation of the Athabasca River therefore requires the carefulmaintenance of natural flow variation, including seasonal patterns [Alberta Envi-ronment, 2007].1.3 Research objectives and contributionThe overall goal of this research is to synthesize the hydrologic impacts of climatechange and water use in order to advance the understanding of future water man-agement challenges in the Athabasca oil sands. The dissertation addresses thisoverall goal through four overlapping objectives:• Develop a modelling system that can integrate both climate and human-driven impacts on streamflow in the Athabasca River Basin. This con-tributes a new method for a large-scale and physically-based hydrologicalanalysis of the region.11• Examine how spatial and temporal variability in the magnitude and distri-bution of water withdrawals by oil sands operations impacts downstreamstreamflow timing. The body of data on oil sands water use is compiledfrom multiple sources and applied to build a comprehensive range of wateruse scenarios that provide bottom-up estimates of future water demand andthe scale of water use impacts on streamflow patterns.• Explore how climate change will alter future streamflow timing in the basinin combination with, and in contrast to, the impact of water withdrawals.This quantifies the potential range of climate change impacts on futurestreamflow patterns and the projected variability in future water supply.• Investigate how water management can adapt to future water supply anddemand trajectories. This identifies a full range of water management op-tions with different priorities and tradeoffs in environmental protection andindustry growth, which can help to inform future water policy decisions.1.4 Research strategyThis dissertation takes a physically-based modelling approach that is well suitedto simulate the land-surface and hydrological processes and changes in a riverbasin. Two existing models, the Integrated Biosphere Simulator (IBIS) and theTerrestrial Hydrologic Model with Biogeochemistry (THMB) are used togetherand adapted to model the Athabasca River Basin. IBIS is a land surface modelthat simulates the coupled soil-vegetation-atmosphere water and energy budgets[Foley et al., 1996, Kucharik et al., 2000], while THMB is a hydrological rout-ing model that uses prescribed river paths to simulate the storage and transport ofwater [Coe et al., 2002, 2008]. This approach allows for an assessment of futureclimate variability across large spatial and temporal scales and the direct integra-tion of human alterations to streamflow such as water withdrawals.The timing of streamflow is a primary focus throughout the thesis, as it isan important metric of hydrologic alteration that responds to climatic variabil-12ity as well as direct human intervention [Richter et al., 1996, Do¨ll et al., 2009].Changes in the seasonality of streamflow are also important in effective waterresource management where operational decisions often depend on the timing offlow cycles to match supply and demand, as well as a hydrologic baseline that sus-tains the flow regime [e.g., Alberta Environment, 2007]. The modelled projectedchanges in streamflow timing are applied to identify potential threats to futurewater resource availability and to inform the development of water managementadaptation options.1.5 Structure of dissertationThe chapters in this thesis move linearly through an exploration of different stream-flow impacts and their consequences for future water management. Each chapterbuilds upon the next, but is also self-contained and formatted as an individualmanuscript for future submission to target journals. As a result, each chapter in-cludes a description of specific relevant background, concepts, and methods, suchthat some repetition of these details in each chapter occurs.Chapter 1 provides a contextual overview of the research, including the theme,purpose and main research goals of the work.Chapter 2 establishes the modelling methods that are applied to capture thephysical processes that drive basin hydrology. A discussion of model parameteradjustments and an evaluation of the model system performance are presented.The simulated historical streamflow established in this chapter provides abaseline for the following chapters to assess future spatial and temporalalterations to streamflow.Chapter 3 compiles data on current and future oil sands water withdrawals andassesses the streamflow impacts of industry. Different water withdrawalscenarios are examined under historical climate variability.13Chapter 4 simulates the projected climate change impacts on streamflow at themiddle and end of the 21st century using the most recent global climate modelsand climate scenarios available. Climate change impacts are also contrasted withthe scale of water withdrawal impacts from Chapter 3.Chapter 5 integrates the results of Chapter 3 and Chapter 4 to frame an analysisof water management options based on future water supply and demandtrajectories. It explores the tradeoffs that may emerge in adapting to futurestreamflow alterations under a changing climate and given an evolving oil sandsindustry.Chapter 6 concludes the dissertation by discussing its significance andcontributions to the current research field and the stated research problem inChapter 1. The strength and limitations of the research are discussed along withcomments on potential future work that would expand on the dissertation.14Chapter 2Model Validation for the AthabascaRiver Basin2.1 IntroductionThis chapter establishes the modelling methods applied to simulate the hydrologyof the Athabasca River Basin (ARB). A combination of two large-scale land sur-face and ecosystem process models is used. The land surface process model, theIntegrated Biosphere Simulator (IBIS) [Foley et al., 1996, Kucharik et al., 2000],and a river routing algorithm, the Terrestrial Hydrology Model with Biogeochem-istry (THMB) [Coe et al., 2002] are independent models that have been used to-gether in dozens of large-scale studies, including simulations of continental-scalerunoff [Lenters et al., 2000, Coe and Foley, 2001, Li et al., 2005], Amazonianflooding [Coe et al., 2002], and Mississippi nutrient flux [Donner et al., 2002].IBIS produces surface and subsurface runoff outputs which are then used to drivestreamflow routing in THMB (Figure 2.1). Process-based models, in which modelparameters are primarily based on real, physical parameters that can be validatedwith available data, are preferable when extending model application to futurehydroclimatic regimes that may respond differently to calibrated parameters. Adiscussion of IBIS-THMB model parameter adjustments and an evaluation of the15model performance in the ARB are presented.2.2 Model description2.2.1 Integrated Biosphere Simulator (IBIS)IBIS simulates the coupled soil-vegetation-atmosphere water and energy budgetsby modelling a) land surface biophysical processes, b) ecosystem physiology andcarbon balance processes, c) vegetation phenology, d) plant growth, competitionand vegetation dynamics, e) nutrient cycling and soil biogeochemistry, and f) wa-ter cycling among vegetation, atmosphere, and soils [Foley et al., 1996, Kuchariket al., 2000]. IBIS is forced with daily climate inputs such as temperature, pre-cipitation, cloud cover, and humidity, along with land surface characteristics suchas vegetation and soil type and distribution, to yield fluxes of carbon, energy andwater from the land surface to the atmosphere, soil ice and water content, soil tem-perature profiles, and surface and subsurface runoff to streams. These processesare divided into several modules which operate at different timesteps ranging fromminutes to years.16Figure 2.1: Schematic representation of IBIS (left) and THMB (right) models showing the interactions be-tween input climate data, simulated land surface physics, and river dynamics (provided by the Centerfor Sustainability and the Global Environment (SAGE) at University of Wisconsin).17IBIS is scale independent and has been used at sites ranging from one squarekilometre (e.g. farm fields) to hundreds of thousands of square kilometres (e.g.Amazon Basin). The model has been validated against site-specific biophysicalmeasurements (e.g. evapotranspiration, sensible heat flux, vegetation phenology,soil moisture, snow cover and depth, soil temperature, groundwater recharge andriver discharge), as well as spatially extensive ecological data (e.g. total and livingbiomass) [Delire and Foley, 1999, Lenters et al., 2000, Coe and Foley, 2001, Coeet al., 2002, Botta and Foley, 2002, Botta et al., 2002, Vano et al., 2006], includingin cold regions such as Canadian boreal forest ecosystems [El Maayar et al., 2001,Liu et al., 2005] and the Yukon River Basin [Yuan et al., 2010]. The result ofIBIS when forced with climatic data is a comprehensive description of the fluxesof carbon, energy and water from the land surface to atmosphere, the soil iceand water content, soil temperature profile, and surface and subsurface runoff tostreams. In this study, IBIS is run on a 0.375◦ x 0.375◦ geographic grid chosen tomatch the available climate re-analysis resolution for the region.The soil module can be set to any appropriate number and thicknesses of soillayers to describe the diurnal and seasonal cycles of soil ice and water, and thedynamics of soil volumetric ice and water content are simulated for each layer.The soil moisture simulation is based on Richards’ equation, where the change intime of the soil moisture in each layer is a function of diffusion, the soil hydraulicconductivity, and plant water uptake. The plant water uptake is a mechanistic pro-cess governed by stomatal demands and constrained by root water uptake, whichin turn, are complex functions of physical characteristics such as photosyntheticactivity, the canopy structure, atmospheric and surface conditions, root structure,and soil moisture profile [Kucharik et al., 2000, Li et al., 2005]. Soil water infiltra-tion is estimated using Darcy’s law and the number and thickness of soil layers canbe manually adjusted. There are 11 defined soil textures, composed of differentsand, silt and clay fractions [Kucharik et al., 2000]. The various soil parametersfor each texture, such as the saturated hydraulic conductivity, can be adjusted tosatisfy regional characteristics.18For cold-region processes, frozen soils are modelled in IBIS using a soil icefraction parameter and subsurface flow automatically adjusts to the melting offrozen soil. A simple three layer snow model is used to simulate snow tempera-ture, extension and depth. Version IBIS v2.6b4 was employed in this study. Formore details about IBIS, please see Foley et al. [1996] and Kucharik et al. [2000].2.2.2 Terrestrial Hydrology Model with Biogeochemistry(THMB)The Terrestrial Hydrology Model with Biogeochemistry (THMB), formerly HY-DRA, is a river routing algorithm that translates the surface and subsurface runoffoutputs from IBIS into the flow of water through rivers, lakes and floodplains[Coe et al., 2002]. THMB has been extensively applied and validated at globaland continental scales, including Canada’s Arctic-draining rivers [Coe, 2000, Coeand Foley, 2001, Coe et al., 2002, Donner et al., 2002, Donner and Kucharik,2003, Donner et al., 2004, Shankar et al., 2004]. The streamflow routing algo-rithm applies prescribed river paths to simulate the storage and transport of water,where the total water within a grid cell at any point is the sum of the land surfacerunoff, subsurface drainage, precipitation and evaporation over the surface waters,and the flux of water between grid cells. The derived hydrological network andmorphology are linked at 5-minute horizontal resolution to a linear reservoir tosimulate the stage and discharge of rivers at a 1-hour timestep. The streamflowoutput of THMB, when forced with climate data and IBIS runoff are spatiallyexplicit representations of the river discharge.River discharge at a given time step is controlled by the effective velocity ofthe river, u. The effective river velocity is a function of the topographic gradient aswell as the scale of the river [Coe et al., 2008]. This ensures that the river velocityincreases downstream due to the momentum of flow, despite a shallower gradient.The effective river velocity is given by:u = uo[icio·pcpo]0.5(2.1)19where uo (m/s) is the effective reference velocity, ic (m/m) is the downstreamgradient, io (m/m) is a latitude adjusted reference gradient, pc (m) is the wettedperimeter and po (m) is a reference wetted perimeter. The wetted perimeter is afunction of total discharge and constrained according to river bankfull character-istics, wherepmax = 2hi +wi (2.2)and hi and wi are the bankfull height and width respectively.THMB can also estimate the seasonal flood extent over the river floodplain[Coe et al., 2008]. In order to preserve numerical stability in the model, however,floodplain flows are not explicitly subtracted from or added to the river volume,and therefore do not impact river transport. When the river volume rises abovethe flood initiation stage, the excess water amount is allocated from the river to afloodplain reservoir, which can then flow across the land surface to neighbouringgrid cells. Storage and transport of water on the floodplain is given bydWfdt= Fr +∑Fin +(PW −EW )A f −Fout (2.3)where the change in Wf (m3), the floodplain reservoir, over each time step is thesum of the flux between river and floodplain Fr, the contribution from all upstreamfloodplain grid cells ∑Fin, the difference between precipitation and evaporationover the floodplain surface (PW −EW )A f , minus the amount transported to down-stream floodplain grid cells Fout . The inundated area and height are calculatedfrom the floodplain volume based on the sub-grid topography, and the floodplainflow direction is dictated by the water height. The floodplain flow velocity iscalculated in the same manner as the river velocity, with the floodplain wettedperimeter based on the flooded fraction and the grid cell length instead.The version of THMB (v1f) developed by Coe et al. [2008] was employedin this study. For more details on THMB, please see Coe [1998] and Coe et al.[2002, 2008].202.3 Data sources2.3.1 ClimateClimate input data for the 30-year time period of 1981–2010 was retrieved fromthe NOAA National Operational Model Archive and Distribution System (NO-MADS), which provides access to a 3-hour-averaged North American RegionalReanalysis (NARR) dataset [NOAA, 2013]. The NARR product uses the Na-tional Centers for Environmental Prediction (NCEP) Eta model as its backboneand is output on a 33 km native resolution grid. The 33 km native projection isa Lambert Conformal Conic grid projection which was re-projected in a regularlatitude-longitude geographic grid of 0.375 degree resolution using a simple in-verse distance squared interpolation. The NARR product was chosen over NCEPor Climatic Research Unit (CRU) products due to its improved assimilation of pre-cipitation observations over North America. The seven NARR data fields retrievedwere specific humidity and temperature at 2 m above the surface, meridional andzonal wind speed vectors at 1000 mb, total cloud cover fraction in the atmospherecolumn, and total pressure and precipitation at the surface. This data was thenaveraged into daily data files for input into IBIS.2.3.2 Soil and vegetationSoil surface properties (clay and sand fractions) were obtained from the ISRIC-WISE soil database [Batjes, 2000]. Each layer of the soil column in IBIS is as-signed one of eleven defined soil textures based on these soil surface properties[Kucharik et al., 2000]. The properties of each soil texture, as well as their dis-tribution, can be varied to better represent regional soil characteristics and soilclimate. The dominant surficial soils in the ARB are glacial soils (silt, clay andsands), glaciolacustrine soils (clay loam to heavy clay) and glaciofluvial soils(sandy loam to sands), while peat soils extend over much of the basin rangingfrom 0.3 to 1 m in depth [Kerkhoven and Gan, 2006].21Vegetation type input maps at 0.375 degree resolution were based on theBoston University MODIS (MOD12C1) data set [Friedl et al., 2001] and landcover types were converted to match IBIS vegetation and land cover classifica-tions as follows: evergreen needle leaf forest→ boreal evergreen forest/woodland,mixed forests→ mixed forest/woodland, woody savannas→ savannas, croplands→ grassland/steppe, open shrublands→ tundra, barren/sparsely vegetated→ tun-dra.2.3.3 Geomorphology and hydrologyGlobal geomorphology input files for THMB were retrieved from the Center forSustainability and the Global Environment (SAGE) at the University of Wisconsin-Madison. These files have a 5’x5’ resolution and provide data on basin definition,elevation, river directions, lake area, and lake sill elevation and location [Coe,2000]. Finer resolution 1 km topographic data from the Shuttle Radar Topog-raphy Mission (SRTM) [Farr et al., 2007] was used to define the sub-grid-scaletopography within each 5’ grid cell in order to calculate fractional flooding usinga statistical representation of floodplain morphology, following the method of Coeet al. [2008].The river directions in the SAGE data were derived from the Global DEM5digital elevation model [GETECH, 1995] and modifications were made to im-prove accuracy. The original digital elevation data incorrectly prescribed theflow directions in some parts of the ARB, particularly for the headwaters of theAthabasca River. As a result, river directions in applicable grid cells were manu-ally corrected using physical river maps of the ARB and verified using the knowndrainage area at hydrometric stations. A total of 123 out of 6514 river directionsin the basin were modified. The basin definition was re-calculated based on thenew river directions and lake area and outlet locations were then adjusted to beconsistent with the new basin boundary.22Table 2.1: Hydraulic and geomorphic observations of river reaches in the ARB (from Kellerhals et al. [1972]).Long-term mean 2-yr floodRiver Station Discharge Velocity Discharge Width Depth Velocity Sinuo-(m3/s) (m/s) (m3/s) (m) (m) (m/s) sityAthabasca Near Jasper 90.33 1.13 453.07 114.91 1.74 2.26 1.10Athabasca At Entrance/Hinton 186.89 1.10 906.14 191.11 2.59 1.83 1.00Athabasca At Athabasca 430.42 0.88 1868.91 316.99 3.84 1.55 1.20Athabasca Below McMurray 645.62 1.07 2208.71 539.50 3.14 1.31 1.00Athabasca At Embarras Airport 767.39 0.76 2605.15 441.96 5.33 1.10 1.35Wildhay Near Hinton 8.01 0.67 48.14 39.62 0.85 1.40 1.20McLeod Above Embarras River 20.42 0.34 158.57 67.36 2.07 1.13 2.00McLeod Near Wolf Cr/Edson 38.79 0.55 305.82 110.64 1.68 1.68 1.80Wolf Creek At Highway #16 3.28 0.18 28.32 29.26 1.34 0.73 2.40Freeman Near Fort Assiniboine 8.55 0.55 79.29 60.66 1.01 1.28 1.80Pembina Below Paddy Creek 15.21 0.43 96.28 51.21 2.01 0.94 1.70Pembina Near Entwistle 18.75 0.40 155.74 68.58 1.71 1.34 2.10Pembina At Harvie 41.34 0.58 189.72 79.55 2.44 0.98 2.00Lobstick Near Entwistle/Styal 3.88 0.30 17.56 21.64 1.01 0.79 1.90Paddle Near Rochfort Bridge 2.24 0.43 28.32 17.07 1.07 1.52 1.50Little Paddle Near Mayerthorpe 1.05 0.09 12.74 16.15 1.25 0.58 1.60Lesser Slave Slave Lake/At Highway#243.89 0.55 68.81 50.90 2.16 0.64 2.00West Prairie Near High Prairie 4.62 0.52 62.30 25.91 2.07 1.16 1.80East Prairie Near Enilda 6.51 0.70 79.29 29.87 1.80 1.46 1.40Swan Near Kinuso 12.91 0.46 148.66 42.06 3.78 0.94 1.70Clearwater Above Christina River 80.70 0.76 172.73 115.82 1.52 0.98 1.04Clearwater At Draper 135.07 0.73 424.75 137.77 2.93 1.04 1.5023Optimizing THMB for the ARB requires geomorphological observations fordeveloping relationships between flow and key hydraulic variables, but there arefew published observations in the river basin since the 1970s. Hydraulic and ge-omorphic observations of river depth, width, velocity and sinuosity for 22 riverreaches in the ARB (Table 2.1), were taken from observations conducted for awider channel survey program in Alberta [Kellerhals et al., 1972]. This data wasused to compute rating curves for the bankfull height, width, and initiation vol-umes (see Section 2.4.2).Long-term mean discharge and 2-year flood discharge observations (Table 2.1)were also obtained from the Alberta channel survey program [Kellerhals et al.,1972] to parameterize floodplain flow. The 2-year flood discharge data was usedinstead of bankfull discharge, due to the number of missing values in the recordedbankfull characteristics. The 2-year flood discharge is a reasonable alternativesince bankfull discharge is generally expected at 1.6 to 1.8 year recurrence inter-vals [Leopold, 1994].Observations of monthly-averaged river discharge are available through theWater Survey of Canada Hydrometric Data database [Environment Canada, 2010].Only four hydrometric stations along the Athabasca River contained long-termdischarge observations over the complete 30-year historical time period of inter-est, and these were selected for use (Figure 2.2, Table 2.2).24((((     At Athabasca Below McMurray110°0'0"W110°0'0"W120°0'0"W120°0'0"W60°0'0"N 60°0'0"N50°0'0"NNear JasperAt HintonALBERTABRITISHCOLUMBIASASKATCHEWANNORTHWEST TERRITORIESFigure 2.2: Map of hydrometric stations in the Athabasca River Basin.25Table 2.2: Hydrometric stations on the Athabasca River measuring monthly discharge and with recorded datawithin the 30-year time period of interest, 1981–2010.Station ID Station Name Data Years Latitude (◦N) Longitude (◦W)07AA002 Near Jasper 1913–2010 52.9100 118.058607AD002 At Hinton 1961–2011 53.4242 117.569207BE001 At Athabasca 1913–2011 54.7219 113.287807DA001 Below McMurray 1957–2011 56.7083 111.4019262.4 Model parameterizationThe magnitude and timing of peak annual discharge, as well as the rise and re-cession limbs of the hydrograph are dependent on accurate parameterization ofthe vertical water budget in IBIS, and the lateral water budget in THMB. Pa-rameters that impact the water balance include those that adjust the soil moisturephysics and control flow velocity. Relevant model parameters in both IBIS andTHMB were adjusted based on known physical parameters, if available, or other-wise tuned to improve reproduction of the observed 30-year average hydrograph.Each parameter was systematically tested to determine how it influenced simu-lated results.2.4.1 Vertical water budgetIBIS calculates evapotranspiration, surface, and subsurface runoff through its at-mosphere, soil and vegetation modules. The default soil parameters produced alow subsurface to surface runoff ratio that resulted in a winter streamflow deficitand poor peak flow timing. The shape of the total runoff was sensitive to the par-titioning of surface and subsurface flows, which differed in their annual distribu-tions. For example, subsurface runoff tended to peak roughly one month later thansurface runoff and was responsible for winter runoff. The relative contributions ofsurface and subsurface runoff control the hydrograph response of streamflow inTHMB, changing the timing of peak flows and the magnitude of winter flows.It was expected that some IBIS soil parameters may require adjustment fromtheir default values in order to optimize simulations for the boreal environmentof the ARB (D. Price, pers. comm.). A sensitivity analysis of the various IBISmodule parameters was performed to investigate the dominant controls on thevertical water balance in the ARB in order to simulate more realistic ratios ofsurface and subsurface runoff. Three IBIS soil parameters were identified as thedominant controls which govern:1. the thickness of the top soil layer272. the total soil depth3. the saturated hydraulic conductivity of soil typesThe parameters were systematically tested (Figure 2.3) to determine their ef-fects on average monthly actual evapotranspiration and surface and subsurfacerunoff patterns at the hydrometric station location on the Athabasca River ‘BelowMcMurray’. During testing, the runoff and actual evapotranspiration in all gridcells upstream of this station were summed to represent the total upstream contri-bution to the local water balance over a shorter five year time period 1984–1988.The first parameter, the thickness of the top layer of soil, is expected to controlinfiltration capacity and evapotranspiration. Increasing the layer thickness wasobserved to increase the ratio of subsurface to surface runoff (Figure 2.3a,b) andbroaden the peak runoff (Figure 2.3c). This effect translated into higher stream-flow later in the year. Varying the top soil layer thickness produced little impacton the distribution of actual evapotranspiration (Figure 2.3d).The second parameter, the total soil depth, is expected to control the residencetime of subsurface runoff in the soil column. The total soil depth was adjustedacross a range below the maximum soil depths that have been observed in dif-ferent locations of the basin, as documented in the Alberta Soil Survey Reports27, 29, 31, 42, 43, 44, 58-1, 64-1, and 64-2 ([Dumanski et al., 1972, Hollandand Coen, 1983, Kjearsgaard, 1973, Knapik and Lindsay, 1983, Lindsay et al.,1957, 1963, Turchenek and Lindsay, 1982, Wynnyk et al., 1963, 1969]. Increas-ing the soil depth was observed to decrease the surface runoff (Figure 2.3e) andredistribute the subsurface runoff more uniformly across the seasons (Figure 2.3f).Increasing the soil depth effectively increased the total runoff in the winter monthsby increasing the residence time of water in the soil column. The shape of the totalrunoff distribution was primarily controlled by surface runoff, regardless of soildepth (Figure 2.3g). Again, actual evapotranspiration was not significantly alteredby changing the total soil depth (Figure 2.3h). Although actual soil depths variedacross a basin, IBIS specifies a single soil depth for the entire modelled area toavoid discontinuities between grid cells.28J F MAM J J A S O N D050mmtopsoil thicknessJ F MAM J J A S O N D0102030mmJ F MAM J J A S O N D050mmJ F MAM J J A S O N D050100mmJ F MAM J J A S O N D050total soil depthJ F MAM J J A S O N D0102030J F MAM J J A S O N D050J F MAM J J A S O N D050100J F MAM J J A S O N D050hydraulic conductivityJ F MAM J J A S O N D0102030J F MAM J J A S O N D050J F MAM J J A S O N D050100surfacerunosubsurfacerunototalrunoaet(a)(d)(c)(b)(h)(g)(f)(e)(l)(k)(j)(i)0.05 m0.15 m0.3 m1 m4 m8 m10−910−710−4−10−7−10−5−10−2Figure 2.3: Systematic variation of three IBIS parameters (top soil layer thickness (a–d), total soil depth (e–h), saturated hydraulic conductivity (j–l)) and the effects on runoff and actual evapotranspiration (aet)patterns. Each parameter was varied while keeping the others at the following default values: top soillayer thickness = 0.2 m, soil depth = 2 m, hydraulic conductivity = 10−7−10−5 m/s. The IBIS outputshown for each month is an average over a five year time interval, 1984–1988.29The third parameter, the vertical saturated hydraulic conductivity (ks) is de-fined in IBIS for each soil texture class ranging from sand (highest ks) to clay(lowest ks) and is expected to control infiltration and percolation through eachsoil layer. The magnitudes of ks were adjusted within the observed range of fieldmeasurements reported at a site within the ARB [Coen and Wang, 1989]. Theratio of subsurface to surface runoff increased as ks of all texture classes was in-creased, and the subsurface runoff distribution broadened (Figure 2.3i,j). As kswas increased (by the same order of magnitude for each texture class), a greaterfraction of the total runoff occurred in later months (Figure 2.3k). Meanwhile,actual evapotranspiration decreased (Figure 2.3l), and may indicate that a greaterfraction of runoff percolated into deeper soil layers beyond the rooting depth as kswas increased. The vertical saturated hydraulic conductivities used are on the highend of the observed range of field measurements at the Athabasca site reported inCoen and Wang [1989].Since observations of surface and subsurface runoff were unavailable for theARB, the soil parameter values were systematically adjusted to minimize the dif-ference between the simulated and observed seasonal hydrograph. The final IBISparameter set consisted of a 0.2 m top soil layer thickness, a 1.5 m total soil depth(5 soil layers), and a ks range of 10−5–10−3 m/s (Table 2.3).Other IBIS modules and parameters that could affect the water content of soilswere also considered in tuning the vertical water balance. Snowmelt processeshave been modelled using different approaches of varying complexity, rangingfrom simple methods based only on temperature measurements [Morris, 1985]to complex multilayer models based on an energy balance [Jordan, 1991, Markset al., 1999]. A simple three layer snow model, such as the algorithm implementedin IBIS, is adequate to model snowpack physics on the continental scale, based onground and surface radiation temperatures and accounting for snowpack ripeningthat characterizes snowpack growth and ablation [Lynch-Stieglitz, 1994, Stieglitzet al., 2001]. The density of snow controls the accuracy of snow cover simulationand is important to soil moisture distribution and timing. While snow density is30Table 2.3: Final saturated hydraulic conductivities (ks) assigned to IBIS soiltextures.Texture Class ksSand 5.83x10−3Loamy Sand 1.703x10−3Sandy Loam 7.19x10−4Loam 3.67x10−4Silty Loam 1.89x10−4Sandy Clay Loam 1.19x10−4Clay Loam 6.39x10−5Silty Clay Loam 4.17x10−5Sandy Clay 3.33x10−5Silty Clay 2.50x10−5Clay 1.67x10−5approximated using the density of the snowpack in IBIS, Vano et al. [2006] foundthat it was better to approximate snow density using the density of snowfall despiteresulting overestimates of the snow depth. Decreasing the density of snow in IBISproduced an increase in drainage, however this change was small relative to theeffects of adjusting ks, and the snow density was subsequently left unchanged.Additional IBIS soil parameters were also investigated, but each was observedto have little impact on runoff output. These included the maximum allowed pud-dle depth on the surface of the soil, which is expected to affect the amount ofinfiltration that occurs, and the boundary layer permeability at the bottom of thesoil column, which controls the gravitational drainage of water. The default soilmodule infiltration equations, based on Darcy’s law, were also tested against theGreen-Ampt infiltration equations [Green and Ampt, 1911], which are expectedto increase infiltration into the soil column [Li et al., 2005]. However, the intro-duction of the Green-Ampt equations was not observed to increase the subsurfaceto surface runoff ratio significantly relative to the three main soil parameters, andwas therefore not implemented.31Lastly, various soil data sets were substituted to test if they improved modelperformance. The International Geosphere-Biosphere Programme soils data avail-able with IBIS v2.6b4 (general distribution), is of a coarser resolution and resultsin a lower subsurface to surface runoff ratio than the ISRIC-WISE soils data setused. The soil profiles from the CanSIS database included an organic soil type andno clay soils, which produced a small improvement in the peak runoff timing, butalso led to very high rates of evapotranspiration and were unsuitable. In addition,only one grid cell in the ARB was assigned the organic soil type in this data set.2.4.2 Lateral water budgetAdapting the THMB code to a specific river basin requires identifying suitableparameter values for the river velocity and floodplain algorithms.River velocityThe timing and movement of streamflow between THMB gridcells for the ARBis primarily controlled by the grid cell flow velocity, u, which in turn is controlledby the wetted perimeter. The river bankfull height (hi) and width (wi) which de-fine the wetted perimeter are calculated as power-law functions of the upstreamarea. To adapt THMB to the ARB, simple statistical relationships (p < 0.01) werederived using the empirical measurements of the 2-year flood width and depthcharacteristics for the Athabasca River [Kellerhals et al., 1972] (Table 2.1), andare given bydi = 0.3201 ·A0.2108u (2.4)wi = 0.6765 ·A0.5453u (2.5)(2.6)where Au is the upstream area of each grid cell.River sinuosity (s) also affects the calculation of river velocity by changing theriver length in each grid cell. Based on observations of sinuosity for rivers in the32ARB [Kellerhals et al., 1972], the approximate power-law relationship (p < 0.05)between river sinuosity and upstream area was calculated to bes = 2.7755 ·A−0.0696u (2.7)This was applied to calculate a spatially varying river sinuosity for each gridcellin the basin.Other THMB parameters related to sub-grid drainage and reference velocitieswere set to values used in past studies [Coe, 2000, Donner, 2002]. The surfacerunoff timing constant was set as a function of the average grid cell length andthe effective reference velocity. The subsurface runoff residence time was set to15 days, and the groundwater residence time was set to 180 days; Donner [2002]concluded that the timing of simulated monthly river discharge in continental-scale river basins was not sensitive to these parameters. The effective referencevelocity (uo = 0.35 m/s) in THMB was set based on a global study of continental-scale river basins by Miller et al. [1994]. The reference gradient (io = 1x10−4),reference wetted perimeter (wo), and reference velocities for river and floodplainwere then simultaneously tuned to improve streamflow output timing.Floodplain dynamicsThe calculation of the flood initiation volume followed three steps. First, thebankfull flux in each grid cell was calculated from the simulated average dailyflux over a 10-year period from 1990–1999, using an empirical rating curve re-lationship. This rating curve was updated for the ARB using observations of thelong-term mean discharge and 2-year flood discharge data (Table 2.1), and wasderived (p < 0.01) to be:Fi = 15.0962+4.6560F¯−0.0017F¯2 (2.8)where Fi is the bankfull flux and F¯ is the long-term mean flux. Second, the cross-sectional area of the river was calculated based on the bankfull flux Fi and the33average (hourly) river velocity over the 10-year period. Lastly, the bankfull floodinitiation volume was calculated as the product of the cross-sectional area and theriver length.The floodplain algorithm in THMB was originally designed to describe thespatial extent of seasonal flooding in the Amazon Basin, but not to simulate theeffect of floodplain inundation on river discharge [Coe et al., 2008]. For thisstudy, the flooding algorithm was updated in order to improve the accountingof exchange between floodplain and river reservoirs, and to test for a possibleinfluence of floodplain dynamics on streamflow. In the initial THMB code, waterin an individual grid cell could be added to a floodplain reservoir, but was notsubtracted from the river reservoir, such that the calculated floodplain flow hadno effect on downstream flow. The algorithm was updated to allow floodplainwater levels to influence the integrated downstream velocity of the river channeland floodplain waters by computing a weighted average velocity based on thefractional grid cell coverage of the river channel and inundated areas:u = uAr +u f A f (2.9)where Ar and A f are the grid cell fractional areas for river and floodplain respec-tively and u f is the floodplain flow velocity. This allowed the floodplain flowto modulate river flow while still maintaining numerical stability in the model byavoiding direct exchanges between the river and floodplain reservoir volumes. Thereference floodplain velocity was arbitrarily set at u f o = 0.27 m/s to roughly rep-resent the slower expected movement of floodplain waters. The floodplain wettedperimeter was then tuned to improve streamflow timing.Ultimately, flood occurrences were infrequent in the ARB and generally ac-counted for a small fraction of the total grid cell volume, such that the streamflowtiming was not sensitive to the temporal and spatial extent of flooding. Althoughobservational data was not available to validate the simulated floodplain area inTHMB, its calculation does serve to parameterize sub-grid surface water flow, asdistinct from the main river movement within a grid cell. The new floodplain al-34gorithm may also be useful in future applications of THMB to other river basinswith available floodplain observations.In an effort to obtain streamflow resolution and estimate inundated area at afiner scale, a simplified version of the THMB model was also tested for the ARB.This simple model required only climate and topography as input and calculatedthe river directions at each time step as a function of water head. Model runswere performed at 1 km resolution using the SRTM topography data, howeverstreamflow flux calculations became unstable for several grid cells, an indicationthat the water volume was too high for this method to be useful in the ARB.THMB was also tested using predicted instead of prescribed lake areas, but thiswas found to reduce the accuracy of streamflow simulations.2.5 Model performanceAn optimal parameterization was derived from the systematic comparison of ad-justed model parameter values in both IBIS and THMB, to the extent that theycan be validated using streamflow observations. The suite of parameter valueswere chosen that generated the observed conditions for average annual stream-flow, seasonality of flow, and timing of peak flow at the hydrometric station lo-cation Below McMurray. Model simulations were evaluated based on calculatederrors between observed and modelled annual and seasonal flow, and by applyingstatistical model performance ratings to the monthly flow time series.2.5.1 Annual and seasonal variabilityThe model output captured the shape of the average hydrograph, including thebroad peak flow, reasonably well at Below McMurray (Figure 2.4a). The flowregime of the Athabasca River is typical of northern rivers with mountainousheadwaters and downstream plains, and is characterized by low flows in the win-ter, followed by rising discharge associated with snowmelt in the lowlands, andleading to a broad peak flow due to convective summer storms and possibly sus-35tained by glacier and high-elevation snowmelt in its headwater areas [Woo andThorne, 2003, Kerkhoven and Gan, 2006]. The simulated hydrograph began torise as expected in (late) April, followed by a sharper rise to a broad peak in Juneand an increase to maximum flow in July.Statistical comparisons between simulated and observed annual, seasonal, peak,and minimum flows showed best model performance at Below McMurray, themost downstream monitoring station (Table 2.4). The simulated mean annualflow and peak monthly flow was within 2.5% of the observations over the 30-yeartime period. Average summer flows (July-October) best agreed with observationsto within 1%, and although average winter flows (November-March) were not aswell simulated, the minimum flow was within 5% of observations.36J F M A M J J A S O N D050010001500discharge(m3 /s)(a)J F M A M J J A S O N D050010001500(b) modelobservedFigure 2.4: Model runs over the historical period of 1981–2010 compared to streamflow observations at themonitoring stations (a) Below McMurray, (b) At Athabasca.37Table 2.4: Average annual and seasonal statistics at four gauging stationsbetween 1981–2010, comparing model output to observed data. Valuesare model discharge as a percentage greater (+) or less (-) than observeddischarge, while the shift in peak flow between model and observationsis in days. Seasons are summer (July–October), winter (November–March), and spring (April–June).Below At At NearMcMurray Athabasca Hinton Jasperannual flow +2.4% -5.8% +17.4% -15.3%peak flow -2.3% -9.7% +22.8% -16.9%peak shift +13 d -18 d -34 d -46 dminimum flow +4.6% +7.8% +43.5% +73.0%minimum shift -4 d +1 d +42 d +80 dspring flow -9.2% +7.7% +101.4% +48.3%summer flow 0.5% -26.4% -48.7% -70.8%winter flow +40.7% +38.3% +91.5% +136.8%Further upstream at the At Athabasca monitoring station, the average annualflow was reasonably well simulated, within 6% of observed magnitudes (Fig-ure 2.3b). The lower model accuracy at the two far upstream stations of NearJasper and At Hinton is expected; these stations drain a small number (3–4) ofmountainous grid cells, an area over which precise streamflow simulation is notrealistic from a large-scale model. Since the mean elevation for a given IBIS gridcell in the mountains is much lower than the actual peak elevations within the cellarea, IBIS cannot describe the heterogeneous processes of ablation and high eleva-tion snowmelt that extend peak flows later into the year. In basins with glaciatedheadwaters, the ablation of glaciers intensifies in the summer and this, togetherwith snowmelt at high elevations, prolongs the high flows into summer [Woo andThorne, 2003]. This results in simulated flow with a narrow, earlier seasonal peakin runoff at the upstream stations (Figure 2.5), which indicate an underestimate ofupstream water storage. Therefore, the spatial variability in model performancelikely arises in the generation of IBIS runoff. This scale problem in the mountain-38ous grid cells has little effect on basin-scale streamflow timing, as evidenced bythe realistic simulation of flow at the At Athabasca and Below McMurray stations.There was limited flow data for locations downstream of Fort McMurray, so themodel performance further downstream could not be verified. For example, theRegional Aquatics Monitoring Program and Water Survey of Canada hydrometricstations for the Athabasca River near the Embarras Airport only have availabledata for less than a third of the historical time period.2.5.2 Interannual variabilityFurther evaluation of model performance focused on the Below McMurray sta-tion, the closest location to the oil sands mining operations. The monthly-averagedsimulated flow at Below McMurray over the 30-year time period was well corre-lated (r = 0.73, p < 0.01) with observations ( Figure 2.6). In the last decade ofthe time period (2001–2010), flows were consistently over-predicted by 25% ormore in the late spring and early summer (June–August), approximately 70–80%more frequently than in the other two decades.No definitive criteria for evaluating model performance have been establishedin the literature yet. However, three quantitative statistical methods for hydro-logical time series analysis performed on a monthly time step, have been recom-mended by Moriasi et al. [2007]: the Nash-Sutcliffe efficiency (NSE), percentbias (PBIAS), and a ratio of the root mean square error to the standard deviationof measured data (RSR). These methods have been used extensively in the statis-tical evaluation of streamflow in both large- and small-scale river basins [e.g., Liet al., 2009, Bekele and Knapp, 2010, Srinivasan et al., 2010], and were used toassess model accuracy of streamflow output from IBIS-THMB (Table 2.5).39J F M A M J J A S O N D0200400600800100012001400discharge/runoff (m3 /s)J F M A M J J A S O N D050100150200250300surf sub sim obs(a) (b)Figure 2.5: Comparison of the surface (surf) and subsurface (sub) runoff generated by IBIS for (a) down-stream (Below McMurray) and (b) upstream (Near Jasper) locations, along with simulated (sim) andobserved (obs) streamflow. Runoff values are summed from all grid cells upstream of the station loca-tion.401980 1985 1990 1995 2000 2005 20100500100015002000250030003500discharge (m3 /s)modelobservedFigure 2.6: Monthly simulated and observed discharge over the 30-year time period 1981-2010, at the BelowMcMurray location.41The NSE statistic determines the relative magnitude of the residual varianceto the measured data variance, and is given byNSE = 1−n∑i=1(Y obsi −Ysimi )2n∑i=1(Y obsi −Ymean)2 (2.10)where Y obsi is the ith observation, Ysimi is the ith simulated value, and Ymean is themean of the observed data for n total observations. NSE indicates how well theobserved and modelled flow fits a 1:1 line, and ranges between −∞ and 1.0 where1.0 is the optimal value.The PBIAS statistic measures the percentage of residual variance, which givesthe average tendency of the simulated data to overestimate or underestimate ob-servations. Low absolute magnitude values indicate higher accuracy, with theoptimal value being 0.0%. PBIAS is expressed as:PBIAS =n∑i=1(Y obsi −Ysimi ) ·100n∑i=1Y obsi (2.11)The RSR statistic is the root mean square error (RMSE) normalized by thestandard deviation of the observations and is given byRSR =√n∑i=1(Y obsi −Ysimi )2√n∑i=1(Y obsi −Ymean)2 (2.12)The RSR provides scale to the error indicated by RMSE. The optimal value ofRSR is 0, with lower values indicating better model performance.When applied over the entire 30-year time period at the Below McMurraylocation, the NSE and RSR statistical tests indicated unsatisfactory model per-42formance (Table 2.5), as defined by Moriasi et al. [2007]. Model performanceimproved to satisfactory or good across all three statistics when tested separatelyover the first and second decades, with best performance in the second decade.The unsatisfactory model performance in the third decade across all three statis-tics, with the PBIAS test showing strong model overestimation, reduced the over-all model performance across the 30-year time period.To examine whether the reduced predictive skill of the model at the BelowMcMurray station during the final decade, 2001–2010, may be driven by clima-tological input data or unrepresented anthropogenic activities (land use and landcover changes or water use and withdrawals related to oil sands or other resourceprojects) that were unaccounted for in the model simulations, model performanceat the At Athabasca station, which lies upstream of all oil sands projects, was alsoevaluated (Table 2.5). Model performance at At Athabasca was satisfactory un-der the NSE and RSR statistical tests over the entire 30-year validation period, incontrast to test results at Below McMurray, suggesting that recent human activityaffecting streamflow that is not accounted for in the model runs may be responsi-ble for lower downstream model performance at Below McMurray.Unsatisfactory model performance at At Athabasca in the third decade, as wellas at Below McMurray, further suggests that some of the discrepancy betweenmodelled and observed monthly streamflow may also be due to the climatologicalinputs. For example, the NARR climate data is known to have regional problemswith the precipitation analysis over Canada, due to a limited set of gauge ob-servations [Mesinger et al., 2006]. A comparison between Environment Canadamonthly precipitation observations at Fort McMurray between 1981-2010 and themonthly NARR precipitation inputs, demonstrated that the NARR precipitationin the last decade of the time period (2001–2010), was higher by 25% or morebetween May and July, approximately 30–70% more frequently than in the othertwo decades. This is consistent with the overpredicted modelled streamflow inthe last decade and suggests that precipitation inputs drive the disagreement withobserved streamflow.43Table 2.5: Evaluation of model performance (streamflow output) using three statistics for monthly timesteps.Performance ratings in brackets are either unsatisfactory (u), satisfactory (s), good (g) or very good (v),as defined by Moriasi et al. [2007]. The model was evaluated over the entire time period of interest, aswell as over each decade.statistic 1981-2010 1981-1990 1991-2000 2001-2010Below At Below At Below At Below AtMcMurray Athabasca McMurray Athabasca McMurray Athabasca McMurray AthabascaRSR 0.81 (u) 0.70 (s) 0.62 (s) 0.67 (s) 0.53 (g) 0.56 (g) 1.30 (u) 0.92 (u)NSE 0.35 (u) 0.50 (s) 0.61 (s) 0.55 (s) 0.72 (g) 0.69 (g) -0.68 (u) 0.16 (u)PBIAS -2.44 (v) 5.79 (v) 18.48 (s) 20.25 (s) 16.83 (s) 18.86 (s) -50.82 (u) -27.06 (u)44Table 2.6: Comparison of month-averaged daily streamflow outputs withmonthly streamflow outputs using three statistics. Performance ratingsin brackets are either unsatisfactory (u), satisfactory (s), good (g) orvery good (v), as defined by Moriasi et al. [2007]. The simulationswere evaluated over the entire time period of interest, as well as overeach decade, at Below McMurray.streamflow driven by monthly IBIS outputstatistic 1981-2010 1981-1990 1991-2000 2001-2010RSR 0.00 (v) 0.00 (v) 0.00 (v) 0.00 (v)NSE 1.00 (v) 1.00 (v) 1.00 (v) 1.00 (v)PBIAS 0.00 (v) 0.00 (v) 0.00 (v) 0.00 (v)streamflow driven by daily IBIS outputRSR 0.13 (v) 0.18 (v) 0.15 (v) 0.12 (v)NSE 0.98 (v) 0.97 (v) 0.98 (v) 0.99 (v)PBIAS 0.58 (v) 0.88 (v) 1.46 (v) -0.20 (v)Daily streamflow was also simulated and compared to the monthly simulatedstreamflow. A comparison was made between1. monthly THMB-simulated streamflow driven by monthly IBIS-simulatedoutput2. month-averaged daily THMB-simulated streamflow driven by daily IBIS-simulated output3. month-averaged daily THMB streamflow driven by monthly IBIS outputsThe RSR, NSE and PBIAS statistics were calculated between 1) and 2), and 1)and 3) (Table 2.6). All comparisons showed only minor differences in simulatedstreamflow between daily and monthly timesteps. As a result, for computationalefficiency, the study results are based on monthly THMB streamflow simulationsdriven by monthly IBIS outputs.452.5.3 UncertaintiesComprehensive observations are needed to fully validate both the vertical andhorizontal water budgets in soils and river transport. In the Athabasca RiverBasin, a more extensive and continuous hydrologic monitoring network is neededto validate model simulations of streamflow, particularly further downstream inthe basin. For IBIS, observations of soil moisture and subsurface runoff in theARB are limited, and parameter adjustments must be validated indirectly and incombination with other parameters, through streamflow observations. For THMB,empirical data on river velocity, floodplain morphology and inundation is neededto better parameterize the flow network, but is also limited for the ARB. Suchobservations are required to validate the new floodplain equations and to confirmthat they improve the accounting of surface waters and streamflow timing. Dataon inundated surface area based on multiple satellite observations [Prigent et al.,2001] so far lack accuracy for western Canada, including the ARB region. Someprogress has been made in the use of satellite data to map soil moisture [e.g.,Temimi et al., 2010] by translating SSM/I passive microwave and MODIS imagesinto a water surface fraction. Future validation of soil moisture may be possiblevia the Soil Moisture Active Passive Mission (SMAP) [Entekhabi et al., 2010],scheduled to launch by 2014.Accurate parameterization and validation of floodplain and soil processes couldalso advance the modeling of other basin processes that affect streamflow, such asthe spatial and temporal variability of wetland areas which play an important rolein flow connectivity across a landscape, and therefore the timing of river flows. Inaddition, although soil physics is expected to play a larger role in driving stream-flow timing, better parameterization of river transport, including accounting fordynamic ice effects, could also refine streamflow simulations. Like other large-scale models of northern rivers, THMB does not represent river-ice freeze, melt,and ice jam cycles, which may influence the flow of the Athabasca River [Andres,1980, Burn et al., 2004, de Rham et al., 2008, Beltaos, 2013]. Ice dynamics andtheir influence on flow are well known to be difficult to simulate due to the lack of46high-resolution data for model development and the difficulty in parameterizingthe exact timing of freeze-up and melt, as well as the mechanics of ice jam re-lease [Prowse and Beltaos, 2002, Prowse et al., 2007, Beltaos, 2007, Peters et al.,2014]. Land use change is also not accounted for in the streamflow simulations,and this may add to model uncertainies. The currently disturbed area of oil sandsmining operations (602 km2) constitute only a fraction of an IBIS grid cell and isexpected to have a negligible impact on runoff and subsequent downstream flow,relative to upstream basin contributions. The recent spread of mountain pine bee-tle in the Upper Athabasca [Forcorp Solutions Inc., 2012] may however result inconsiderable deforestation that can lead to runoff impacts over a wider area.2.6 ConclusionsThis chapter describes the implementation of a two-model system, IBIS and THMB,to simulate the movement of water in the Athabasca River Basin between the at-mosphere and soil, through streams and lakes, and over topography. Model pa-rameters were calibrated to reflect the physical characteristics of the basin, wherethe water budget is largely controlled by the vertical movement of water throughsoils in IBIS and the velocity of water through the river network in THMB. TheTHMB routing algorithm was modified to improve the accounting of floodplainwaters, and to enable floodplain flow to impact river flow velocity and thereforethe timing of discharge.The IBIS-THMB modelling system reproduced the annual hydrograph wellat the basin location that will be important to assessing water use in the follow-ing chapters. The simulated time series of monthly flow, which represents thesensitivity of streamflow to interannual variations in climate, was over-predicted70–80% more frequently in the most recent decade, possibly due to a combinationof consumptive water use in the lower basin and input data.While model performance is limited by a lack of available observations on pa-rameters that control the vertical and horizontal water balance, the IBIS-THMBsimulations of streamflow can be applied to assess relative streamflow impacts.47The realistic sensitivity of model-simulated streamflow to climate variability sug-gests the models may be well-suited to simulate streamflow in different climateregimes. The historical streamflow simulation established in this chapter providesa baseline for the following chapters to assess future spatial and temporal changesdue to human water withdrawals and climate change in the Athabasca River Basin.48Chapter 3Sensitivity of Streamflow to WaterWithdrawals for the Athabasca OilSands3.1 IntroductionOne of the major environmental concerns associated with oil sands operations isthe intensive freshwater demand. Extracting, processing and upgrading crude bi-tumen from oil sands (a mixture of sand, water, bitumen, heavy metals and othercontaminants), all require a constant water supply. In the Athabasca River Basin(ARB), oil sands mining operations already account for the largest sectoral waterallocations (62%) and actual water use (57%), with roughly 93% of actual wateruse as surface water volume [AMEC Earth & Environmental, 2007]. Oil sandsoperations draw water from a variety of sources including surface water (rivers,runoff), fresh groundwater wells, saline aquifers, recycled water, and storage wa-ter. The majority of the surface water use is withdrawn by mining operations,which primarily divert water from the Athabasca River. The Athabasca River, atributary of the Mackenzie River, drains into the Peace-Athabasca Delta (PAD),an ecologically sensitive region of wetlands and lakes that is highly responsive49to natural variability in river flows and water levels [Peters et al., 2006]. Alongwith the Peace River, the Athabasca also supplies much of the winter low flowof the Mackenzie River, which drains northward into the Arctic Ocean [Woo andThorne, 2003]. Increased water withdrawals may affect the volume and timingof downstream flows required to sustain freshwater and estuarine ecosystems andthe goods and services that those ecosystems provide.The impact of oil sands water withdrawals on the Athabasca River streamflowwill depend upon the scale of future operational expansion. The oil sands depositsin northern Alberta constitute a reliable, long-term supply for the growing globaldemand for crude oil. In 2012, annual crude bitumen production from surfacemining reached 338 million barrels per year [ERCB, 2013] and at these rates,mining production could last for over a century. New operations and plannedexpansions for existing mines are expected to increase production capacity bynearly 500% through 2035 [The Oil Sands Developers Group, 2013]. As miningactivity expands, surface water use demand is projected to rapidly increase, addingpressure to water availability in the ARB [Natural Resources Canada, 2009].Under compliance with Phase One of the Lower Athabasca River Water Man-agement Framework, oil sands operators are currently licensed to withdraw 1-2%of annual flow from the Athabasca River each year in order to maintain historicalmedian flows [Alberta Environment, 2007], although water use reporting remainsvoluntary. While the overall annual allocations are relatively small, sustained bitu-men production rates throughout the year, together with seasonality in flow, meansthat withdrawals can be large relative to low winter flows [Swainson, 2009]. PhaseTwo of the Lower Water Management Framework is currently under developmentand aims to establish best management practices under a high-growth scenario[Ohlson et al., 2010], but stakeholders have been unable to achieve consensus ona final set of water management rules. While studies have shown that a long-termapproach to applying water restrictions is needed [e.g., Mannix et al., 2010], thereis a lack of physically-based analyses that identify the response of streamflow towithdrawals within the context of the basin’s hydrological regime and the location50and timeline of current and planned water withdrawals.This chapter describes a first attempt to explore the sensitivity of streamflow inthe Athabasca River to water withdrawals for oil sands operations by linking twoindependent, process-based models: the Integrated Biosphere Simulator (IBIS)[Foley et al., 1996, Kucharik et al., 2000], a land surface process model, and theTerrestrial Hydrology Model with Biogeochemistry (THMB) [Coe et al., 2002], ahydrological routing model. IBIS and THMB have been used together in dozensof global, large-scale studies, including simulations of continental-scale runoff inNorth America [Lenters et al., 2000, Coe and Foley, 2001] and Africa [Li et al.,2005], Amazonian flooding [Coe et al., 2002], and Mississippi nutrient flux [Don-ner et al., 2002]. Here, the models are first validated for the ARB and then appliedto simulate streamflow under new oil sands water use scenarios developed fromdata on current and planned oil sands projects. The impacts of water withdrawalsare assessed based on relative changes to streamflow magnitude and the frequencyof occurrence of low flows.3.2 Methods3.2.1 Model description: IBISThe Integrated Biosphere Simulator (IBIS) is a process-based land surface modelthat simulates the coupled soil-vegetation-atmosphere water and energy budgets[Foley et al., 1996, Kucharik et al., 2000]. IBIS is scale independent and hasbeen used at sites ranging from one square kilometre (e.g. farm fields) to mil-lions of square kilometres (e.g. Amazon Basin). The model has been validatedagainst site-specific biophysical measurements (e.g. evapotranspiration, sensibleheat flux, vegetation phenology, soil moisture and ice, snow cover and depth, soiltemperature, groundwater recharge and river discharge), as well as spatially ex-tensive ecological data (e.g. total and living biomass) [Delire and Foley, 1999,Lenters et al., 2000, Coe and Foley, 2001, Coe et al., 2002, Botta and Foley, 2002,Botta et al., 2002, Vano et al., 2006] including in cold regions such as Canadian51boreal forest ecosystems [El Maayar et al., 2001, Liu et al., 2005] and the YukonRiver Basin [Yuan et al., 2010].IBIS is forced with daily climate inputs and land surface characteristics includ-ing vegetation and soil type. Climate data was retrieved for the 1981–2010 periodfrom the 3-hour-averaged North American Regional Reanalysis (NARR) datasetof the NOAA National Operational Model Archive and Distribution System (NO-MADS) [NOAA, 2013] and averaged into daily files. The NARR product usesthe National Centers for Environmental Prediction (NCEP) Eta model as its back-bone with output on a native 33 km Lambert Conformal Conic projection grid,which was re-projected in a regular latitude-longitude geographic grid of 0.375degree resolution using a simple inverse distance squared interpolation. The re-trieved NARR data fields were specific humidity and temperature at 2 m above thesurface, meridional and zonal wind speed vectors at 1000 mb, total cloud coverfraction in the atmosphere column, and total pressure and precipitation at the sur-face.Vegetation type maps at 0.375 degree resolution were developed from theBoston University MODIS (MOD12C1) data set [Friedl et al., 2001] and con-verted to match IBIS vegetation and land cover classifications. Each layer of thesoil column was assigned one of 11 defined soil textures [Kucharik et al., 2000]based on soil surface properties (clay and sand fractions) obtained from the ISRIC-WISE soil database [Batjes, 2000]. The properties of each soil texture, as well astheir distribution, can be varied to better represent regional soil characteristics andsoil climate.Runoff in the ARB was primarily controlled by three IBIS soil parameters thatgovern the thickness of soil layers, the total soil depth, and the saturated hydraulicconductivity of soil types. These parameters were systematically tested to de-termine their relative controls on average monthly actual evapotranspiration, andsurface and subsurface runoff patterns (see Figure 2.3 in Chapter 2). Based onthis analysis, the soil parameter values were adjusted to minimize the differencebetween streamflow simulations and observations. The final IBIS simulations em-52ployed a top soil layer thickness of 0.2 m, a soil depth of 1.5 m and a final hy-draulic conductivity range of ks = 10−5−10−3 m/s for the 11 soil texture classes,with all parameter values within the range of regional field observations.Version IBIS v2.6b4 was employed in this study. For more details about IBIS,please see Foley et al. [1996], Kucharik et al. [2000].3.2.2 Model description: THMBThe Terrestrial Hydrology Model with Biogeochemistry (THMB) [Coe et al.,2002] is a river routing algorithm that translates the surface and subsurface runoffoutputs from IBIS into the flow of water through rivers, lakes and floodplains.The output of THMB is a spatially explicit representation of river discharge andflooding extent. THMB has been extensively applied and validated at global andcontinental scales, including Canada’s Arctic-draining rivers [Coe, 2000, Coe andFoley, 2001, Coe et al., 2002, Donner et al., 2002, Donner and Kucharik, 2003,Donner et al., 2004, Shankar et al., 2004]. The streamflow routing algorithm ap-plies prescribed river paths to simulate the storage and transport of water, wherethe total water within a grid cell at any point is the sum of the land surface runoff,subsurface drainage, precipitation and evaporation over the surface waters, andthe flux of water between grid cells. The derived hydrological network and mor-phology are linked at 5-minute horizontal resolution to a linear reservoir model tosimulate the stage and discharge of rivers at a 1-hour time step.Global geomorphology at 5’x5’ resolution for THMB was retrieved from theUniversity of Wisconsins Center for Sustainability and the Global Environment(SAGE) to define the basin drainage area, elevation, river directions, and lakearea, location, and sill elevation in THMB [Coe, 2000]. The river directions data,derived from the Global DEM5 digital elevation model GETECH [1995], wasmanually corrected using physical river maps at grid cells where flow directionswere inaccurate. Finer resolution 1 km topographic data from the Shuttle RadarTopography Mission (SRTM) [Farr et al., 2007] was used to define the sub-grid-scale topography within each 5’ grid cell in order to calculate fractional flooding53using a statistical representation of floodplain morphology, following the methodof Coe et al. [2008].The timing and movement of streamflow between gridcells in THMB is pri-marily controlled by the effective velocity of the grid cell or river, u, which is afunction of the topographic gradient as well as the size of the river [Coe, 2000,Coe et al., 2008]. The river velocity is estimated in part by calculating the wettedperimeter, which is a function of the total discharge volume in the river and isconstrained by the bankfull depth (di) and width (wi). Simple statistical relation-ships for bankfull depth and width as a function of upstream area (p < 0.01) werederived from empirical 2-year flood width and depth characteristics collected for22 river reaches in a channel survey program of Alberta [Kellerhals et al., 1972](Table 2.1), and given bydi = 0.3201 ·A0.2108u (3.1)wi = 0.6765 ·A0.5453u (3.2)(3.3)where Au is the upstream area (km2) of each grid cell. The 2-year flood datawas used instead of bankfull data, due to the number of missing values in therecorded bankfull characteristics, and was a reasonable alternative since bankfulldischarge is generally expected at 1.6 to 1.8 year recurrence intervals [Leopold,1994]. The river sinuosity (s) also affects the calculation of river velocity bychanging the river length in each grid cell. Based on observations of sinuosity forAlberta rivers [Kellerhals et al., 1972], the approximate power-law relationship(p < 0.05) between river sinuosity and upstream area wass = 2.7755 ·A−0.0696u (3.4)and was applied to calculate a spatially varying river sinuosity for each gridcell inthe basin.The THMB version THMB.v1f Coe et al. [2008] used in this study was pro-54vided by M. Coe (pers. comm.). For more details on THMB, please see Coe[1998], Coe et al. [2002, 2008].3.2.3 Water withdrawalsThe Athabasca Wabiskaw-McMurray deposit within the Athabasca oil sands areais the only deposit where crude bitumen occurs near the surface and can be re-covered economically by open-pit mining. This deposit is contained within anarea north of Fort McMurray that constitutes 3% of the total oil sands area, withroughly 1% of the available area currently disturbed and 99% of the area underlease [Alberta Energy, 2013]. Between 1981 and 2010, five companies operatedsix oil sands surface mining sites, and four additional mine sites have since beengranted regulatory approval (Table 3.1). In total, twenty licensed source locationsof water withdrawals for these mine sites were included in the study (Figure 3.1).Although licensed water allocations exist for each mine, water use reportingremains voluntary and the actual withdrawal amounts used in this study were com-piled from a combination of reported and licensed water use and oil production.Four data sources were used to assess current and future water use for oil sandsoperations. First, historical annual water use data by operator for 2005–2012 wasobtained from the Data Library hosted by Alberta Environment and SustainableResource Development [Energy Resources Conservation Board, 2010] and sup-plemented with information on the breakdown of water use by source (N. Ad-hikari, pers. comm.).55Table 3.1: Water withdrawal source locations for all current and planned oil sands mining operations.Operator Mine Start WithdrawalDate Source LocationSuncor Energy Inc. Millennium 1967 Athabasca River (SL1) 57.0068◦N, -111.4612◦Esurface runoff (SL2) 56.9910◦N, -111.3607◦Esurface runoff (SL3) 56.8614◦N, -111.2329◦Esurface runoff (SL4) 56.8941◦N, -111.3872◦ENorth Steepbank 2012 surface runoff (SB1) 57.0104◦N, -111.4141◦EFort Hills 2016 Athabasca River (SF1) 57.3121◦N, -111.6655◦Esurface runoff (SF2) 57.4234◦N, -111.3917◦ESyncrude Canada Ltd. Mildred Lake 1978 Athabasca River (YD1) 57.0269◦N, -111.5006◦Esurface runoff (YD2) 57.0431◦N, -111.5549◦EAurora North 2001 surface runoff (YA1) 57.2831◦N, -111.4217◦EShell Albian Sands Jackpine 2010 Athabasca River (HJ1) 57.2540◦N, -111.6385◦Etributaries and surface runoff (HJ2) 57.2540◦N, -111.3676◦EMuskeg River 2002 Athabasca River (HG1) 57.2540◦N, -111.6385◦Esurface runoff (HG2) 57.2526◦N, -111.5333◦ECanadian Natural Horizon 2008 Athabasca River (CH1) 57.3268◦N, -111.6789◦EResources Limited Tar River, tributaries, surface runoff(CH2)57.3849◦N, -111.9495◦EImperial Oil Limited Kearl 2012 Athabasca River (IK1) 57.5630◦N, -111.4967◦EMuskeg & Firebag Rivers, tribu-taries, surface runoff (IK2)57.5630◦N, -111.4967◦ETotal E&P Canada Ltd. Joslyn North 2018 Athabasca River (TJ1) 57.2758◦N, -111.6656◦Esurface runoff (TJ2) 57.3013◦N, -111.6993◦E56The second source of data was water allocations by source, available throughproject water licenses [Alberta Energy Regulator, 2010]. Allocations generallyexceeded actual use, in part because they were granted based on volumes neededfor operation start-up, which were substantially greater than the volume requiredfor continuous operation [Griffiths et al., 2006]. It was assumed that all licenseswould remain valid through 2035, regardless of current expiry dates. Licensedamounts were often delineated based on an expected project phase schedule, andthe start year for each phase and associated licensed withdrawal were estimated inthese cases.The third source of data was actual bitumen production by project between1996–2012 [Alberta Environment and Sustainable Resource Development, 2013],and the fourth source of data was the expected bitumen production capacity byproject [The Oil Sands Developers Group, 2013]. Both were used to estimatewater withdrawals using the ratio of water use per barrel of oil production, alsoknown as the water use efficiency (WUE), calculated from the most recent actualreported withdrawals and production. The estimated future production capacityin 2035 was based on proposed operations and expansions that had been eitherrecently announced, were under application, or had been granted approval.The data sources were used to build four scenarios of oil sands expansion inorder to assess the sensitivity of streamflow to water withdrawals, relative to acontrol scenario in which no water is withdrawn. The first oil sands water usescenario assumes that annual withdrawals remain constant at 2010 levels, with‘no-growth’ over the model run time period. This was calculated asW En = WAn (2010) (3.5)where subscript n denotes the nth water source, W E , is the estimated water useand W A is the actual reported water use for the year in brackets.The second and third scenarios estimate water use based on future oil sandsexpansion scenarios. The ‘capacity-based’ expansion scenario estimates annualwater withdrawals based on each project’s expected oil production capacity in573333 3333333333333((SB1((((((      110°0'0"W110°0'0"W120°0'0"W120°0'0"W60°0'0"N 60°0'0"N50°0'0"NALBERTANORTHWEST TERRITORIESNear JasperAt Hinton At AthabascaBelow McMurrayBelow OpsAt OutletBelow OpsBelow McMurraySL4SL3SL1YD2HG1HJ1,TJ1SF2 CH1CH2TJ2IK1,IK2YA1HJ2HG2SF1SL2YD1SASKATCHEWANBRITISHCOLUMBIAFigure 3.1: Map of the ARB region showing locations of hydrometric sta-tions and model analysis. Inset shows water withdrawal source loca-tions in the oil sands mining region (see Table 3.1 for label references).582035, and is given byW En = PC(2035) ·W An (2012)PA(2012)(3.6)where PC is the expected production capacity and PA is the actual reportedproduction for the closest available year (2012, from the available data). The‘license-based’ expansion scenario estimates annual water withdrawals based oneach project’s licensed water allocations in 2035, and is given byW En = WLn (2035) (3.7)where W L is the licensed water use.The fourth scenario allows for an improvement in the efficiency of water usewithin the same parameters of the capacity-based scenario. Reductions in wa-ter withdrawals are a current focus for oil sands operators and companies havecommitted to improving their water use efficiency in order to expand operationswithout increased water allocations. The water use estimate for this ‘improved-efficiency’ scenario is given byW En = PC(2035) ·WUE(2035) ·W An (2012)∑nW An (2012)(3.8)Since 2011, two-thirds of active mining operations have already exceededplanned efficiency targets of a 2:1 WUE [Natural Resources Canada, 2009]. In2011, estimates from water use and production data show that Suncor and Syn-crude mining operations had achieved a WUE of 1.8:1. An extremely optimisticWUE of 1:1 was chosen in the improved-efficiency scenario in order to representthe maximum possible improvement in WUE. This 1:1 ratio is lower than industryprojections.593.2.4 Streamflow simulationsThe control scenario (natural flow with no water withdrawals) was used to vali-date the model against observations of monthly-average river discharge from theWater Survey of Canada Hydrometric Data (HYDAT) database at four locationsalong the Athabasca River with available long-term observations over the 1981–2010 period (Figure 3.1, Table 2.2). The model was then run under the withdrawalconditions of each oil sands scenario using the same climate input data, in order tocapture the range of streamflow responses that could occur under a fixed envelopeof typical recent climate variability. For each scenario, the hourly water use rate(m3/s) for each withdrawal source location was calculated from annual water useassuming sustained water use for 24 hour operations over 365 days of the year,and spatially mapped into model input. Withdrawals were subtracted from the ap-propriate grid cells at each hourly time step in THMB. The simulated streamflowin each withdrawal scenario was evaluated in terms of changes relative to the con-trol scenario at two selected locations, one immediately downstream of all miningoperations (Below Ops), and the second located near the outlet of the ARB (AtOutlet), which flows into the PAD.3.3 Results3.3.1 Model validationStatistical comparisons of simulated annual, seasonal, peak, and minimum flowsshowed best model performance at McMurray, the most downstream monitoringstation (Table 2.4). The simulated mean annual flow and peak monthly flow waswithin 2.5% of the observations over the 30-year time period. Average summerflows (July-October) best agreed with observations to within 1%, and althoughaverage winter flows (November-March) were not as well simulated, the minimumflow was within 5% of observations.The model output captured the shape of the average hydrograph, including the60broad peak flow, reasonably well at McMurray (Figure 2.4a). The flow regimeof the Athabasca River is typical of northern rivers and is characterized by lowflows in the winter, followed by rising discharge associated with snowmelt in thelowlands, and leading to a broad peak flow due to snowmelt in the mountainousheadwaters of the basin and convective summer storms [Woo and Thorne, 2003,Kerkhoven and Gan, 2006]. The simulated hydrograph began to rise as expectedin (late) April, followed by a sharper rise to a broad peak in June and an increase tomaximum flow in July. The interannual variability in simulated flow at McMurrayis also well correlated (r = 0.73, p < 0.01) with observations (Figure 2.6).Further upstream at the Athabasca monitoring station, the average annual flowwas reasonably well simulated, within 6% of observed magnitudes (Figure 2.4b).The lower model accuracy at the two far upstream stations of Jasper and Hintonis expected; these stations drain a small number (3–4) of mountainous grid cells,an area over which precise streamflow simulation is not realistic from a large-scale model. Since the mean elevation for a given IBIS grid cell in the mountainsis much lower than the actual peak elevations within the cell area, IBIS cannotdescribe the heterogeneous processes of ablation and high elevation snowmeltthat extend peak flows later into the year. This results in simulated flow witha narrow, earlier seasonal peak in runoff at the upstream stations (Figure 2.5),which indicate an underestimate of upstream water storage. This scale problemin the mountainous grid cells has little effect on basin-scale streamflow timing,as evidenced by the realistic simulation of flow at the Athabasca and McMurraystations.Further evaluation of model performance focused on the McMurray station,the closest location to the oil sands mining operations. Model accuracy was as-sessed using three quantitative statistical methods for hydrological time seriesanalysis recommended by Moriasi et al. [2007]: the Nash-Sutcliffe efficiency(NSE), percent bias (PBIAS), and a ratio of the root mean square error to the stan-dard deviation of measured data (RSR) (Table 2.5). When applied over the entire30-year time period at the McMurray location, the NSE and RSR statistical tests61indicated unsatisfactory model performance, as defined by Moriasi et al. [2007].Model performance improved to satisfactory or good across all three statisticswhen tested separately over the first and second decades, with best performance inthe second decade. Unsatisfactory model performance in the third decade acrossall three statistics, with the PBIAS test showing strong model overestimation, re-duced the overall model performance across the 30-year time period.To examine whether the reduced predictive skill of the model at the McMurraystation during the final decade, 2001–2010, may be driven by poor climatologicalinputs or anthropogenic impacts (land use and land cover changes or water use andwithdrawals related to oil sands or other resource projects) that were unaccountedfor in the model simulations, model performance at the Athabasca station, whichlies upstream of all oil sands projects, was evaluated (Table 2.5). Model per-formance at Athabasca was satisfactory under the NSE and RSR statistical testsover the entire 30-year validation period, in contrast to test results at McMurray,suggesting that recent human activity affecting streamflow may be responsiblefor lower downstream model performance at McMurray. Unsatisfactory modelperformance at Athabasca in the third decade, as well as at McMurray, furthersuggests that some of the discrepancy between modelled and observed monthlystreamflow may also be due to climatological inputs.Daily streamflow was also simulated and compared to the monthly simu-lated streamflow. A comparison was made between 1) monthly THMB-simulatedstreamflow driven by monthly IBIS-simulated output 2) month-averaged dailyTHMB-simulated streamflow driven by daily IBIS-simulated output and 3) month-averaged daily THMB streamflow driven by monthly IBIS outputs. The RSR,NSE and PBIAS statistics (Moriasi et al., 2007) were calculated between 1) and2), and 1) and 3) (Table 2.6. All comparisons showed only minor differences insimulated streamflow between daily and monthly timesteps. As a result, for com-putational efficiency, the study results are based on monthly THMB streamflowsimulations driven by monthly IBIS outputs.623.3.2 Impact of oil sands water withdrawalsThe impact of oil sands water withdrawals was evaluated by comparing the controlscenario, simulated during model validation, to each withdrawal scenario, in orderto control for bias between model simulations and observations. The annual andseasonal flow characteristics of simulated streamflow under each water use sce-nario were evaluated as a percentage change relative to the control scenario (Ta-ble 3.2). Withdrawals based on expected licensed allocations in 2035, the license-based scenario, produced the greatest decrease in streamflow magnitude acrossall statistics (by 1.3–12.3%), followed by the capacity-based (by 1.0–5.7%), no-growth (by 0.5–2.7%), and improved-efficiency (by 0.5–2.5%) scenarios in orderof decreasing impacts.Across all scenarios, winter and minimum flows experienced the greatest rela-tive decrease in magnitude (by 1.4–6.7% and 2.5–12.3% respectively) due to waterwithdrawals. In the license-based scenario, winter flows and minimum flows re-spectively decreased by about two and four times that of mean annual flow. In allscenarios, decreases in the spring, summer, and peak flows were less than thoseof the mean annual flows by up to 1%. The decreases in streamflow due to waterwithdrawals were consistently larger at the Below Ops location than the At Outletlocation across all scenarios (by 0.1–5.4%), suggesting that additional flows fromtributaries and lakes downstream of oil sands disturbance may mitigate the impactof withdrawals.63Table 3.2: Average annual and seasonal statistics of streamflow in each water use scenario, given as a per-centage difference relative to the control scenario. The shifts in peak and minimum flows are measuredin days.‘no growth’ ‘capacity-based’ ‘license-based’ ‘improved efficiency’BelowOpsAt Out-letBelowOpsAt Out-letBelowOpsAt Out-letBelowOpsAt Out-letannual flow -0.71% -0.41% -1.48% -0.85% -3.15% -1.81% -0.65% -0.37%peak flow -0.30% -0.17% -0.63% -0.36% -1.34% -0.78% -0.27% -0.16%peak shift 0 d 0 d 0 d 0 d 0 d 0 d 0 d 0 dminimum flow -2.67% -1.50% -5.74% -3.22% -12.26% -6.87% -2.50% -1.40%minimum shift 0 d 0 d 0 d 0 d 0 d 0 d 0 d 0 dspring flow -0.56% -0.39% -1.18% -0.82% -2.51% -1.75% -0.52% -0.36%summer flow -0.49% -0.26% -1.02% -0.55% -2.18% -1.18% -0.45% -0.24%winter flow -1.49% -0.77% -3.15% -1.63% -6.69% -3.46% -1.39% -0.72%64In each scenario, the models estimate an increase in the frequency of low flowsat the Below Ops location. The low flow threshold for each month of the year wasdefined as the monthly flow magnitude (Q80) that is exceeded 80% of the timeover the 30-year time series of simulated flow in the control scenario. To calculatethis, the monthly mean flows in each of the 30 different years were ranked andthe 80th percentile in flow magnitude was defined as the Q80 threshold for thatmonth. This threshold represents the minimum flow level prescribed by the Al-berta Desktop Method to preserve in-stream flow needs [Locke and Paul, 2011],defined as the quantity, timing and quality of water that is required to sustain ahealthy aquatic ecosystem [Alberta Environment, 2007]. The low flow frequencywas then defined as the percentage of months that flows below the Q80 thresholdoccur over the 30-year time period.The models project that the low flow frequency increases the most in thelicense-based scenario (Figure 3.2), with increases in 8 months of the year, and upto 37% of streamflow occurring as low flows in March, the minimum flow month.The frequency of low flows in the capacity-based scenario is projected to increasefor seven months of the year, with low flows occurring up to 33% of the timefor some months. The no-growth and improved-efficiency scenarios project thesame relative increases in annual low flow occurrence, for five months of the year,up to a 27% low flow frequency. For the four months of June, July, August andNovember, the models project no increase in the frequency of low flows in anyscenario.3.4 Discussion3.4.1 Streamflow impactsDecreases in the magnitude of streamflow due to oil sands water withdrawals weresmall in all scenarios, demonstrating that water withdrawals will have relativelylittle impact on overall streamflow magnitudes in the future, even under conditionsof maximum permitted withdrawals. Constant water withdrawals throughout the65J F M A M J J A S O N D20253035frequency of low ows (%)  no growthcapacity−basedlicense−basedimproved efficiencyFigure 3.2: Frequency of occurrence of low flows for each month in eachwithdrawal scenario, at the location Below Ops. In the control sce-nario, the lowest 20% of flows (dashed line) defines the low flowthreshold applied to the other water use scenarios.year reduce a greater fraction of the streamflow during low winter flow periods asexpected.The frequency of low flows was more sensitive to the intensity of water with-drawals than the flow magnitude was. The low flow frequency is also a betterindicator of ecosystem stress, as this determines the length of time that in-streamflow needs are not met. The timing of low flows is an important considerationin water resource management, where operational decisions often depend on theavailability of a pre-determined baseline flow at a given time [e.g., Alberta En-vironment, 2007]. An increase in the frequency of low flows below a thresholdsuch as the Q80 may translate into more frequent restrictions of operational with-drawals, which may disrupt production depending on the regulatory environment.The higher frequency of low flows during eight months of the year in all scenarios66suggest that a seasonal operation schedule, in contrast to the current, constant op-erations year-round, may help to minimize withdrawals during these months, aswell as sudden interruptions to production schedules.A comparison of the no-growth and improved-efficiency scenarios shows thatan improvement in water use efficiency under projected 2035 oil sands productionlevels would reduce the frequency of low flows to those associated with current(2010) practices. The improved WUE of 1:1 used in this study was intentionallyoptimistic in order to bookend possible future scenarios, as it exceeds plannedefficiency targets of 2:1 for most mining operations. It is more likely that futurewithdrawals in the ARB will involve water use efficiencies between those of theimproved-efficiency and capacity-based scenarios, resulting in an up to 10% in-crease in low flows for some months. As a result, in order to continue meetingin-stream flow needs, limits to the growth of operational withdrawals may need tooccur alongside improvements to water use efficiency.Adaptation of future water use patterns to the projected expansion of wateruse demand should also consider in-stream flows far downstream of local distur-bances. While contributions from tributary flows and lakes did appear to mitigatestreamflow reductions in the ARB, even the impacts of low intensity withdrawalsunder the no-growth scenario did propagate downstream and could become impor-tant for larger streamflow reductions. Seasonal recharge in the PAD, for example,depends on a complex network of floodplain lakes, wetlands and channels that aresupplied by the Athabasca River [Pavelsky and Smith, 2008].One of the challenges in interpreting streamflow impacts and adapting man-agement strategies to projected impacts lies in defining the low flow threshold andthe level of acceptable risk to in-stream needs. The low flow threshold recentlyproposed for Phase Two of the Lower Athabasca water management framework[Ohlson et al., 2010] is less restrictive than the Q80 threshold used in this study,yet Phase Two has been unable to achieve consensus on implementing water userestrictions. Although the selection of a low flow threshold should be a functionof acceptable mean flow conditions, in reality it will require a balance between67maintaining in-stream flow needs with achievable restrictions on oil sands waterdemand.3.4.2 Data needsComprehensive hydrologic observations including soil moisture, runoff, river andfloodplain morphology, and river discharge data are needed to fully validate boththe vertical and horizontal water budgets in soils and river transport. However,such information is currently limited for the basin and especially lacking down-stream of water withdrawal activities. Observations of soil moisture, soil depthand characteristics, and surface and subsurface runoff would be needed to cali-brate soil parameter adjustments in IBIS or any land surface model, while obser-vations of inundated areas would help to better parameterize the flow network inTHMB.A more extensive and continuous hydrologic monitoring network is neededto improve model simulations of future streamflow impacts and the sensitivity towithdrawals. The provincial and national governments have recently invested in amajor oil sands monitoring program to be implemented by 2015, which includesa planned expansion of water quantity monitoring sites. Recent industry-fundedinitiatives like the Regional Aquatics Monitoring Program are also now adding tothe observational network.3.4.3 Water use uncertaintiesThe calculated water use estimates in each scenario involved a number of un-certainties. First, project timelines of production and associated water use fromstart-up to final production were estimated based on operator projections. Suchprojections, however, are dependent on the economic viability of retrieval. Thefuture growth of oil sands operations will be highly dependent on oil prices andchanging international markets, which are motivated by concerns about global oilsupply [National Energy Board, 2006]. Production capacity could be overesti-mated if project timeframes are delayed, as has previously occurred with Phase681B of the Jackpine Mine [Alberta Utilities Commission, 2010], or it could beunderestimated if projects proceed ahead of schedule.Second, future water use will vary as technological innovations improve wa-ter use efficiency by reducing the amount of freshwater needed or by increasingwater returns to the river. A fraction of the total water use (∼ 7%) used in coolingand drainage diversion processes is currently returned by mining projects [AMECEarth & Environmental, 2007], but there was insufficient information on the quan-tity or location of return flows to explicitly include return flows in water use esti-mates. Water returns can, however, be indirectly accounted for when applying animproved water use efficiency to water withdrawal scenarios. Return flows maybecome important in water accounting if technologies are developed to allow thereturn of water used in the processing stages.Third, the water use efficiency for surface water may also decrease if ground-water use increases. Groundwater is primarily withdrawn from wells or deepsaline aquifers for in situ projects, and accounts for less than 10% of water usein mining projects [AMEC Earth & Environmental, 2007]. Such withdrawalswere not simulated in THMB, which does not explicitly model deep groundwaterflow. Accounting for groundwater withdrawals, along with in situ projects, maybe important in assessing future streamflow impacts, as the majority of plannedoil sands operations are in situ projects which use primarily groundwater (∼ 78%[Ko and Donahue, 2011]).Lastly, the future distribution and timeline of some licensed allocations wereuncertain or unknown. For proposed mine sites currently in the application stage,approval for surface water licenses are pending, and there is insufficient informa-tion to determine potential water use or the location of withdrawals. Three minesites, the Teck Resources Limited Frontier Mine (4 phases scheduled for 2021,2024, 2027, and 2030), the Shell Albian Sands Pierre River Mine (2 phases withthe first one beginning 2018), and the Suncor Energy Inc. Voyageur South Mine(no start date scheduled), were therefore excluded from this study, which mayunderestimate future water use. For operations with active water licenses, the ma-69jority expire well before 2035, and future allocations for these operations wereassumed to continue. This may either underestimate or overestimate future waterwithdrawals, depending on how future licensing rules evolve.In the range of scenarios created, the projected impacts on streamflow magni-tude are small, as are the seasonal impacts on the low flow frequency, even at thehighest withdrawal intensities. As a result, these overall findings are unlikely tobe highly sensitive to the assumptions made in building each scenario.3.5 ConclusionsPlanned growth in oil sands production will continue to increase water use overthe next few decades, but the scale of streamflow impacts is uncertain. This studywas a first attempt at examining the response of streamflow in the Athabasca Riverto water use by oil sands surface mining operations. A physically-based mod-elling approach consisting of a land-surface process model linked to a hydrologi-cal transport model was used to simulate the natural flow regime of the AthabascaRiver Basin together with spatial and temporal patterns of water withdrawals.Overall, the impact of surface mining water withdrawals on streamflow mag-nitude was small, even under maximum projected growth and water use intensity.An increase in the intensity of water withdrawals tends to exacerbate already lowin-stream flows and these impacts can propagate further downstream. The fre-quency of low flows, which increases for most months of the years in all scenar-ios, is more sensitive to the intensity of withdrawals and can be used to indicatean increased threat to in-stream flow needs. The modelled impacts suggest that acombination of increased water use efficiency and restricted growth in oil produc-tion will be needed to prevent future increases in the frequency of low flows. Inparticular, winter flows should be a management priority and may require adaptingthe timing of water use, and therefore production schedules, to minimize periodsof low flows that fall below in-stream thresholds.Accurate predictions of future streamflow impacts will require a more compre-hensive network of observations in the Athabasca River Basin to better validate70the models, particularly downstream of oil sands operations. In the meantime, themodelling approach employed in this study provides a useful tool to assess therange of streamflow impacts, based on relative differences between streamflowscenarios, that may occur under different water withdrawal trajectories related tofuture water allocations or intended production growth.71Chapter 4Streamflow Availability underClimate Warming in the AthabascaOil Sands4.1 IntroductionThe Athabasca River Basin (ARB) in northern Canada spans a 269,000 km2 area(Figure 4.1) drained by Lake Athabasca and the Athabasca River, its main arterywhich originates in the Columbia ice fields and eventually flows into the Peace-Athabasca Delta. Along the way, the Athabasca River crosses diverse ecosystemsincluding glaciers, alpine meadows, alpine and boreal forests, and muskeg thatcontain unique landscapes and vital wildlife habitat [MRRB, 2004, Holloway andClare, 2012]. The ARB is subject to a warming climate as well as increasing wateruse for an expanding oil sands industry. While the impact of climate warming onthe ARB’s hydrological regime has been well studied [e.g., Zhang et al., 2001,Prowse et al., 2006, Kerkhoven and Gan, 2011], it is uncertain whether futureconsumptive water demand will exacerbate and/or be threatened by the impactsof climate change.Temperatures in the ARB have increased on average by 1.5–1.8◦C between72      110°0'0"W110°0'0"W120°0'0"W120°0'0"W60°0'0"N 60°0'0"N50°0'0"NALBERTANORTHWEST TERRITORIESSuncor Energy IncSyncrude Canada LtdShell Albian SandsCanadian Natural Resources LtdImperial Oil LtdTotal E&P Canada Ltd Below OpsBelow McMurraySASKATCHEWANBRITISHCOLUMBIABelow OpsBelow McMurrayFigure 4.1: Map of the ARB region showing the location of the hydrometricstation Below McMurray and the location of analysis for streamflowsimulations downstream of oil sands mining operations, Below Ops.Inset shows water withdrawal source locations used in the study (note:there are 20 licenses assigned to 18 physical locations).731961–2000, three times higher than the global average rise of 0.6◦C [Bruce, 2006],and past studies have predicted a continuing rise by up to 3.5–4◦C by 2050 [Ganand Kerkhoven, 2004, Sauchyn and Kulshreshtha, 2008]. The most recent IPCCprojections for Northwest Canada show a mean annual temperature increase of2.7◦C by mid-century and 3.5◦C by the end of the century, along with an annualprecipitation increase of 10% by mid-century and 14% by the end of the cen-tury [Christensen et al., 2013]. Observations of snowpack decline and periodicwinter melting in recent decades in the ARB [Zhang et al., 2001, Sauchyn andKulshreshtha, 2008] are consistent with model projections of an earlier springfreshet and reduced summer flows under future warming [Pietroniro et al., 2006,Schindler and Donahue, 2006]. These climate-driven changes to streamflow pat-terns have been linked to a decrease in the frequency of floods that replenish thelakes and wetlands in the Peace-Athabasca Delta [Prowse et al., 2006, Wolfe et al.,2005, 2008], an ecologically sensitive region that provides important nesting andstaging areas and habitat for a diverse wildlife population.In addition to climate change, land use changes and increased industrial wateruse can also alter streamflow patterns and may explain declining summer flowsdespite increased flow from melting glaciers [Burn et al., 2004, Schindler andDonahue, 2006, Squires et al., 2009]. The growing Athabasca oil sands miningindustry depends on water withdrawals from the Athabasca River in order to ex-tract, process, and upgrade crude bitumen from surface-mined oil sands deposits.In situ mining also occurs under less intensive freshwater usage. Current wateruse by oil sands operations is licensed to ensure that in-stream flow needs, definedas the quantity, timing, and quality of water that is required to sustain a healthyaquatic ecosystem, are met [Alberta Environment, 2007]. The combined impactsof both climate and industrial drivers on the flow of the Athabasca River is cur-rently unknown.This study examines the impacts of both climate change and oil sands waterwithdrawals on streamflow availability for industrial and in-stream flow needs.Two independent, physically-based models are linked to simulate streamflow re-74sponse in the ARB under multiple future climate and water use scenarios. Alarge-scale, process-based modelling approach is used here in order to representthe upstream landscape and fluvial processes that are necessary to capturing thesensitivity of downstream flow to natural climate variability and climate change.Impacts on streamflow patterns are assessed as a change in the frequency of oc-currence of low flows, which is then applied to estimate future water availabilityfor oil sands mining production.4.2 Methods4.2.1 Land surface modelsTwo models are used together to simulate the land surface processes and stream-flow in the Athabasca River Basin. The Integrated Biosphere Simulator (IBIS) isa land surface model that simulates the coupled soil-vegetation-atmosphere waterand energy budgets [Foley et al., 1996, Kucharik et al., 2000]. The TerrestrialHydrology Model with Biogeochemistry (THMB) is a hydrological routing al-gorithm that uses prescribed river paths to simulate the storage and transport ofwater [Coe et al., 2002]. IBIS and THMB have been used together in dozensof global, large-scale studies, including simulations of continental-scale runoff inNorth America [Lenters et al., 2000, Coe and Foley, 2001] and Africa [Li et al.,2005], Amazonian flooding [Coe et al., 2002], and Mississippi nutrient flux [Don-ner et al., 2002]. IBIS has also been applied to cold, northern regions includingCanadian boreal forests [El Maayar et al., 2001, Liu et al., 2005].IBIS is driven with daily climate inputs at a 0.375◦ x 0.375◦ lat-long reso-lution that matches the available climate re-analysis used to validate the model.Its modules operate at different timesteps ranging from minutes to years and themonthly-averaged surface and subsurface runoff outputs are used here. IBIS andTHMB are linked by driving THMB with the runoff outputs from IBIS to sim-ulate the hourly flow of water through rivers, lakes and floodplains at a 5’ x 5’lat-long resolution, and subsequently output a spatially explicit representation of75monthly river discharge. Validation of both models for the Athabasca River Basinis described in detail in Chapter 2, and demonstrates that IBIS-THMB simulationscapture the average hydrograph shape well at the ‘Below McMurray’ streamgaugelocation (Figure 2.4), including low flows in the winter followed by rising dis-charge leading to a broad, late-spring peak in flow. The interannual variability insimulated flow was also well correlated (r = 0.73) with observations (Figure 2.6).4.2.2 Future climate projectionsProjected climate output for a 120 year period from 1981–2100 was obtained fromthe Coupled Model Intercomparison Project (CMIP5) [Taylor et al., 2012], whichprovides global climate model (GCM) output using the four IPCC RepresentativeConcentration Pathway (RCP) climate scenarios [Moss et al., 2010]. The climatescenarios range from RCP2.6, an extreme mitigation scenario (with a mid-centurypeak in radiative forcing), to RCP8.5, the highest radiative forcing scenario, whichmatches the trajectory of greenhouse gas emissions for the past decade. IBIS re-quires seven daily climate variables as input: near surface specific humidity, nearsurface air temperature, eastward and northward near surface wind speed, totalcloud fraction in the atmosphere column, precipitation, and surface air pressure.Only three of the CMIP5 GCMs (GFDL-ESM2G, MIROC5, and IPSL-CM5A-LR) provided all seven of the required climate output variables in all four RCPsand at the required temporal resolution, and were therefore selected for use in thisstudy. The three GCMs cover a range of equilibrium climate sensitivities for theregion (Table 4.1). Output variables from each GCM were re-projected from anative grid onto the IBIS grid using bilinear interpolation. All variables were ob-tained at a daily time step with the exception of surface pressure, which was onlyavailable in 6-hour intervals and was averaged into daily intervals.4.2.3 IBIS-THMB simulationsFor all simulations in this study, IBIS was driven by daily GCM climate outputover a 120-year time period, 1981–2100, to yield monthly average surface and76Table 4.1: CMIP5 global climate models used in this study, their equilibriumclimate sensitivities and resolution (from Andrews et al. [2012]).EquilibriumModel Climate Sensitivity (◦C) Resolution (◦)GFDL-ESM2G 2.4 2.5 x 2.0MIROC 5 2.7 1.4 x 1.4IPSL-CM5A-LR 4.1 3.8 x 1.9subsurface runoff. Outputs at daily timesteps were considered, however the sim-ulated hydrograph produced with daily IBIS output did not differ significantlyfrom that produced from monthly simulations, so the latter was chosen for com-putational speed (see Chapter 2).All GCM-driven monthly outputs from IBIS were adjusted to a historical base-line, before driving THMB. First, IBIS was driven by observation-based NorthAmerican Regional Reanalysis (NARR) data over a 30-year historical time pe-riod between 1981–2010 (see details in Chapter 2). A NARR-driven historicalclimatology for the IBIS output variables, NARRclim(m¯), was calculated from thisIBIS output as an average for each month (m) over all 30 years (y). Second, IBISwas driven by GCM outputs over a 120-year time period between 1981–2100 toyield a monthly simulated time series, GCMsim(m,y). A GCM-driven historicalclimatology, GCMclim(m¯), was calculated from the 1981–2010 period of this IBISoutput. A default anomaly correction (Equation 4.1) was then applied by multi-plying the IBIS simulated outputs by the ratio of NARR-driven and GCM-drivenhistorical climatologies [Arnell and Reynard, 1996], to yield the future projectedIBIS outputs for each month and year (GCMpro j(m,y)).GCMpro j(m,y) = GCMsim(m,y) ·NARRclim(m¯)GCMclim(m¯)(4.1)where m¯ =1302010∑y=1981my77In cases where the GCM-driven historical climatology was zero, the deltachange method [Hay et al., 2000] was applied instead by subtracting the GCM-driven historical climatology from the GCM-driven simulation time series, andthen adding the NARR-driven historical climatology (Equation 4.2). This wasthe secondary method for anomaly correction, since the delta-change method canproduce negative values for positive-only variables like precipitation and runoff.GCMpro j(m,y) = GCMsim(m,y)−GCMclim(m¯)+NARRclim(m¯) (4.2)Since Equation 4.1 can lead to very large adjusted values when GCMclim(m¯)NARRclim(m¯), a maximum value for each grid cell was defined to be ten times themaximum of the NARR-driven historical time series. If the anomaly-correctedvalues using Equation 4.1 exceeded this maximum, the correction was appliedusing Equation 4.2. This factor of ten threshold was tested on the time series ofmultiple climate variables and found to be appropriate in removing anomalousspikes that resulted from the default anomaly correction method. The projected(i.e. anomaly-corrected) IBIS outputs were then used to drive THMB to simulatethe time-varying volume and flow of surface water through lakes and rivers in theARB over the 120 year period.Streamflow impacts were simulated using a combination of different futureclimate scenarios and different water withdrawal scenarios. Each IBIS-THMBsimulation employed one of the four RCPs and either no withdrawals or licensedwithdrawals, in order to assess the range of possible streamflow impacts. A totallicensed withdrawal rate of approximately 21 m3/s represented estimates of themaximum future withdrawals and were based on licensing agreements with indi-vidual oil sands mining operations (see Chapter 3) at the 20 known withdrawallocations (Figure 4.1). This withdrawal rate was applied over the entire time pe-riod in the licensed water withdrawal scenario. Together with the scenario thatinvolves no withdrawals, this effectively bookends the minimum and maximumlikely impacts of water withdrawals.78A total of 24 IBIS-THMB simulations were run using a combination of thethree GCMs, four RCPs and two water withdrawal scenarios. Water withdrawalswere simulated in THMB by extracting the water requirements at each timestep,for each grid cell that corresponds to a licensed withdrawal location. The stream-flow output was evaluated at the location ‘Below Ops’ (57.7083◦N,−111.4583◦E),which lies downstream of all surface mining oil sands operations. IBIS-THMBoutputs were analyzed as running averages over 20-year time windows, with afocus on changes in mid-century (2041–2060) and end-of-century (2081–2100),relative to today (1991–2010).4.3 Results4.3.1 Climate projectionsRelative to today, the mid-century annual mean air temperature is projected by thethree GCMs to rise by 0.9◦C to 3.1◦C (from 0.9–1.9◦C in RCP2.6 to 2.0–3.1◦Cin RCP8.5), while the end-of-century annual mean air temperature is projected torise by 0.5◦C to 7.0◦C (from 0.5–1.9◦C in RCP2.6 to 4.4–7.0◦C in RCP8.5) (Fig-ure 4.2). The projected change in mean annual precipitation generally increaseslinearly with temperature (r2 = 0.3, p < 0.01), however it is variable across thethree GCMs (Figure 4.2a). IPSL-CM5A-LR projects the largest increase of 56mm (12%) in RCP8.5 by end-of-century and GFDL-ESM2G and MIROC5 projectthe largest decrease of 28 mm (6%) in RCP6.0 by mid-century. Overall, IPSL-CM5A-LR projects the greatest increase in warmth and moisture, while GFDL-ESM2G projects the lowest increase. By mid-century and end-of-century, all threeGCMs project an increase in precipitation in RCP8.5. The ratio of rain to snowincreases linearly in response to warming (r2 = 0.6, p < 0.01), with less variabil-ity than the precipitation response. All three GCMs project the largest rain tosnow ratios in RCP8.5 by end-of-century (Figure 4.2b). Previous climate changeanalysis for the ARB conducted by Kerkhoven and Gan [2011] used the Modi-fied Interactions between the Soil-Biosphere-Atmosphere (MISBA) model forced79by seven major GCMs using the four IPCC AR4 Special Report on EmissionsScenarios (SRES) [Nakicenovic et al., 2000] climate scenarios. The precipitationchanges projected by the GCMs selected for this study are on the lower end of therange of GCMs used in the previous MISBA study.4.3.2 Streamflow projectionsThe projected change in mean annual streamflow is variable across the threeGCMs. The GFDL-ESM2G-driven simulations project an increase in streamflowfor all climate scenarios by the end-of century, while the IPSL-CM5A-LR-drivensimulations project a decrease in streamflow for three of the four climate scenarios(Figure 4.3a). By end-of-century in RCP8.5, the IBIS-THMB simulated mean an-nual streamflow increases by 53% in the GFDL-ESM2G-driven simulations anddecreases by 10% and 12% for the MIROC5- and IPSL-CM5A-LR- driven simu-lations respectively, relative to today.Streamflow did not show a linear dependence on temperature (r2 = 0.0; p =0.95), but did linearly increase with precipitation (r2 = 0.2, p < 0.01). (Fig-ure 4.4). This is in contrast to previous MISBA projections where runoff wasmore strongly correlated with changes in temperature [Kerkhoven and Gan, 2011].Kerkhoven and Gan [2011] projected a change in mean annual flow of -8 to -54%by 2040–2069, compared to the IBIS-THMB projections of change in mean an-nual flow of -6.5 to 19.0% by 2041–2060. The results are not directly compara-ble, however, since the mid-century time periods and the reference baseline years(1957–2007 in Kerkhoven and Gan [2011]) differ between the two studies.800 1 2 3 4 5 6 7−1001020∆ T (oC)∆ P (%)GFDL−ESM2G2041−2060MIROC5IPSL−CM5A−LR0 1 2 3 4 5 6 71.522.533.5Rain:Snow∆ T (oC)2081−2100(a)(b)Figure 4.2: Change in (a) annual precipitation (∆P), and (b) the ratio of rain to snow, relative to the change inannual temperature (∆T) projected by the three GCMs and four RCPs in mid-century (2041–2060) andend-of-century (2081–2100).811990 2010 2030 2050 2070 209040060080010001200flow (m3 /s)1990 2010 2030 2050 2070 2090MJJAmonth1990 2010 2030 2050 2070 2090JFMAyearmonth1990 2010 2030 2050 2070 2090SONDyearmonthRCP 4.5RCP 8.5(a) (b)(c) (d)Figure 4.3: Streamflow patterns for the Athabasca River at the location Below Ops: (a) annual mean stream-flow, (b) centroid of flow distribution, (c) timing of spring runoff, (d) persistence of flow. Dashedlines show the mean values across all GCMs and shaded areas show the range of values across GCMs.Years are the mid-point of running 20-year time windows over which results are averaged. RCP2.6 andRCP6.0 are omitted for clarity.820 1 2 3 4 5 6 7 84006008001000Flow (m3 /s)∆ T (oC)440 450 460 470 480 490 500 510 520 5304006008001000Flow (m3 /s)∆ P (%)GFDL−ESM2G2041−2060MIROC5IPSL−CM5A−LR2081−21002011−2030(a)(b)Figure 4.4: Simulated streamflow relative to the (a) change in annual temperature (∆T), and (b) change inannual precipitation (∆P) projected by the three GCMs and four RCPs in mid-century (2041–2060) andend-of-century (2081–2100).83Three metrics were used to evaluate shifts in the seasonal patterns of stream-flow (Figure 4.3b-d,Figure 4.5), following Burn [2008]. First, the timing of springrunoff was estimated as the date by which 10% of the annual streamflow volumehad occurred. Second, the centroid of flow distribution was calculated as the flow-weighted average time of discharge. Third, the persistence of runoff was estimatedas the date by which 95% of annual flow volume had occurred.The annual centroid and spring runoff occurs earlier in all IBIS-THMB sim-ulations by mid-century and end-of-century (Figure 4.3b-c,Figure 4.5a-d). Byend-of-century in RCP8.5, all three GCMs project the centroid of flow distribu-tion to occur a month or more earlier, shifting from an average of early July to anaverage of late-May (Figure 4.3b), as well as the average timing of spring runoffto shift from mid-March to early February (Figure 4.3c). Late season runoff isless persistent by end-of-century, and most of the annual flow occurs earlier inthe year, by over half a month, in all three GCMs (Figure 4.3d, Figure 4.5f). Thecentroid of flow distribution, timing of spring runoff, and flow persistence alsooccur progressively earlier as the projected proportion of rain to snow increases(Figure 4.6).84∆ days−40−30−20−10010RCP2.6 RCP4.5 RCP6.0 RCP8.5−40−30−20−10010−40−30−20−10010∆ days−40−30−20−10010GFDL−ESM2GMIROC5 IPSL−CM5A−LR−40−30−20−10010∆ daysMIROC5−40−30−20−10010GFDL−ESM2GIPSL−CM5A−LRmid-century (2041-2060) end-of-century (2081-2100)(a) (b)(c) (d)(e) (f)Figure 4.5: Changes in streamflow patterns by mid-century (2041–2060)and end-of-century (2081–2100) relative to today (1991–2010) for theAthabasca River at the location Below Ops: (a–b) shift in the centroidof flow distribution, (c–d) shift in the timing of spring runoff, (e–f)shift in the persistence of flow.85M J J A1.522.53rain/snow ratioJ F M A1.522.53A S O N1.522.53y = −0.4x+4.7p < 0.01y = −0.3x+2.9p < 0.01y = −0.6x+8.3p < 0.01(a) (b) (c)Figure 4.6: Linear relationships between the ratio of rain to snow and the (a) timing of the flow centroid (b)timing of spring runoff (c) persistence of flow.864.3.3 Frequency of low flowsThe frequency of statistical low flows was calculated as a measure of streamflowimpact on in-stream flow needs. The analysis focuses on the seasonality of lowflows, rather than the seasonality of mean flows or hydrographs, because the lowflows are of concern for water management. Low flows were defined for eachmonth relative to a threshold magnitude, computed based on the magnitude ofhistorical flows between 1981 and 2010 that were exceeded 80% of the time inthat month [Locke and Paul, 2011]. This corresponds to a threshold flow requiredto meet full environmental protection of the Athabasca River (i.e. that maintainsthe conditions of an unaltered natural flow regime).Patterns of decreasing and increasing low flow frequency occur in the first(January-June) and second (July-December) halves of the year, respectively (Fig-ure 4.7). All three GCMs project an increase in low flow frequency across allclimate scenarios (except in RCP2.6) from August–October by mid-century, andfrom July–November by end-of-century. By the end of the century in RCP8.5, lowflows are projected to occur 85% more frequently in August for IPSL-CM5A-LR-driven projections, and 75% more frequently in September for GFDL-ESM2G-driven projections.87J FMAM J J ASOND020406080100low flow frequency (%)J FMAM J J ASOND020406080100J FMAM J J ASOND020406080100RCP 4.5RCP 8.5(a) (b) (c)Figure 4.7: Low flow frequency for each month of the year, for three 20-year time windows: (a) today, 1991–2010 (b) mid-century, 2041–2060 (c) end-of-century, 2081–2100. Dashed lines show the mean valuesacross all GCMs and shaded areas show the range of values across GCMs. RCP2.6 and RCP6.0 areomitted for clarity.884.3.4 Water withdrawalsSimulated water withdrawals decreased projected streamflow by a fixed amountthat was generally small compared to the magnitude of projected changes in flowdue to climate change (Figure 4.8). For example, by mid-century, between April–June in RCP8.5, the projected low flow frequency decreases on average by 13%due to climate change and increases on average by 1% due to water withdrawals.From August–October in RCP8.5, the projected low flow frequency increases onaverage by 39% due to climate change and only by an additional 4% due to with-drawals by mid-century. By end-of-century, the relative contribution of waterwithdrawals to the low flow frequency becomes even smaller. In months (e.g.December) when the relative contribution of water withdrawals to the low flowfrequency are similar or greater than that due to climate change, the actual changein low flow frequency is generally small.The frequency of low flows indicates periods of low water availability that canpotentially halt oil sands water withdrawals and therefore bitumen production ifthe protection of in-stream flow needs is considered. Periods of low water avail-ability for oil sands mining operations were quantified as a change in the num-ber of months in which low flows occurred at mid-century and end-of-century,relative to today (Figure 4.9). By mid-century, all but three streamflow simula-tions (GFDL-ESM2G in RCP8.5, MIROC5 in RCP4.5, and IPSL-CM5A-LR inRCP2.6) project an increase in the number of months with low water availability(i.e. a decrease in water availability). By end-of-century, all but one simulation(GFDL-ESM2G in RCP2.6) projects a decrease in water availability. Projectedwater availability is also seasonal, increasing by end-of-century (relative to today)by up to 17% during spring (April–June), while decreasing by up to 75% dur-ing summer (July–October). The IPSL-CM5A-LR-driven simulation for RCP8.5projects the maximum decrease in water availability by mid-century, which trans-lates into a 22% increase in interruptions to oil sands operations relative to today,and equivalent to over two years of oil production per decade. By the end of thecentury, this rises to a 28% increase in interruptions.89J F M A M J J A S O N D−20020406080∆ low flow frequencyJ F M A M J J A S O N D−20020406080J F M A M J J A S O N D−20020406080∆ low flow frequencyJ F M A M J J A S O N D−20020406080climatewithdrawals(a) (b)(c) (d)Figure 4.8: Change in low flow frequency relative to today (1991–2010) for (a) RCP4.5 at mid-century (2041–2060), (b) RCP4.5 at end-of-century (2081–2100), (c) RCP8.5 at mid-century (2041–2060), (d) RCP8.5at end-of-century (2081–2100). Red shows the change due to climate change only and blue shows thechange due to water withdrawals. Dashed and dotted lines show the mean value across all GCMs andshaded areas show the range of values across GCMs.90MIROC5−10010203040506070# monthsMIROC5−10010203040506070(a) (b)GFDL−ESM2GIPSL−CM5A−LRGFDL−ESM2GIPSL−CM5A−LRRCP2.6 RCP4.5 RCP6.0 RCP8.5Figure 4.9: Change in the number of months during the mid-century (2041–2060) and end-of-century time(2081–2100) periods that flow falls below the low flow threshold, relative to today (1991–2010), foreach climate scenario.914.4 DiscussionThe three climate models used in this study generally agree on the projected fre-quency of low flows, the primary tool used here for impact assessments of cli-mate change and water withdrawals. Clear seasonal patterns in the frequency oflow flows of the Athabasca River are projected to emerge over time as climatewarming continues. The models employed in this study project that by end-of-century, low flows (defined based on historical low flow levels) will no longeroccur (0% frequency) in some winter months (November–March) and will alwaysoccur (100% frequency) in some summer months. In contrast, water withdrawalshave a small aggregate effect on low flow frequency; for example, under condi-tions of maximum water withdrawals and no climate change (an extreme, unlikelyscenario), low flows will occur with a maximum 40% frequency and only duringthe winter (see Chapter 3). Climate warming, however, is projected to increaseflow in the winter months and counter the small effect of water withdrawals. Withclimate change, frequent low summer and late season flows become a primaryconcern instead, with little contribution from water withdrawals.These projected shifts in the timing of spring runoff and the seasonality ofhigh and low flows could impact ecosystems such as the perched lakes in thePeace-Athabasca Delta, which are adapted to a historical frequency and timing ofrecharge [e.g., Timoney, 2002, Wolfe et al., 2005, Prowse et al., 2006]. Stream-flow timing can determine whether certain life-cycle requirements are met, andinfluence the degree of stress or mortality associated with extreme conditions[Richter et al., 1996]. Shifts in the distribution and timing of annual flow can alsoincrease the potential for drought by affecting the availability of water resourcesfor human use later in the year [Lapp et al., 2005].The projected climate-driven changes in streamflow may have consequencesfor the ability to continue water withdrawals for oil sands operations. A produc-tion stop of up to 58 months, projected by mid-century in one case, would beequivalent to the interruption of roughly 900 million barrels of oil production atSuncor’s Millenium and Steepbank mines, based on estimated future production92capacity [The Oil Sands Developers Group, 2013]. Athabasca oil sands miningoperations are forecasted to continue through much of the mid-century time pe-riod, given that the timeline for planned projects currently under regulatory reviewinclude the Teck Resources Ltd Frontier mine, with Phase 1 scheduled to begin in2021 and Phase 4 to begin in 2030, as well as Imperial Oil’s Phase 3 of the Kearlmine, to begin in 2020 [The Oil Sands Developers Group, 2013]. At a rate of threemillion barrels of oil production per day (both mined and in-situ recovery) Albertaoil sands reserves are expected to last for over 150 years [Alberta Environment,2009]. Assuming that surface mining continues to make up 58% of oil production[The Oil Sands Developers Group, 2013], and since 20% of reserves are recov-erable by surface mining [Alberta Environment, 2009], mining operations can beexpected to continue for at least 50 years. The mid-century time period is there-fore a realistic planning horizon for anticipated bitumen extraction and associatedwater withdrawals.The frequency of future water withdrawal restrictions and availability will de-pend in part on how an acceptable low flow threshold is quantified. A major aimof Phase Two of the Lower Athabasca Water Management Framework, currentlyunder development, is to include an ecosystem base flow which establishes a flowthreshold, such as the one defined in this study, below which it is recommendedthere be no further withdrawals of water [Ohlson et al., 2010]. This serves toprotect aquatic habitat and river biodiversity during the lowest flow periods. Onechallenge in establishing the ecosystem base flow or any low flow threshold isthat thresholds based on long-term historical flow are only valid under stationaryclimate conditions [Dettinger et al., 2004, Stewart et al., 2004]. Another is thatthe threshold must negotiate the competing needs of industry and aquatic ecosys-tems for water. Implementing a low flow threshold in the next phase of the watermanagement framework will therefore require that industrial water demand adaptto projected changes in streamflow due to climate change.Such projected patterns of streamflow and future low flow frequency, and theassociated impacts on water availability, are expected to be a product of changes93in precipitation amount and type, evapotranspiration, and snowpack accumulationand melt in large western Canadian river basins like the ARB [Schnorbus et al.,2011]. In this study, the projected patterns of flow, particularly the timing of fu-ture low flow occurrences, are broadly consistent with the results of previous mod-elling studies and general understanding of the response of snow-dominated riverbasins to climate warming [e.g., Sauchyn and Kulshreshtha, 2008, Kerkhoven andGan, 2011]. The timing of future low flows also demonstrate seasonal shifts inthe runoff response that will drive annual averages and extremes in runoff. Differ-ences in the range of projected change in annual mean, minimum and maximumrunoff, between this and previous studies like Kerkhoven and Gan [2011], are at-tributed to differences in the reference baseline years and the timestep of modelruns, which prevent direct comparisons. Projected precipitation is highly variable,but is found to generally increase with climate warming. The projected increasein the annual ratio of rain to snow, as temperature increases, is also consistent withthe expectation that winter precipitation will increasingly fall as rain [Schindlerand Donahue, 2006]. Projections that both the spring runoff and the centroidof flow distribution will occur earlier in the year are consistent with recent ob-served trends that show increasing temperatures driving a progressively earliersnowmelt, a decline in maximum snowpack depth and persistence, and more fre-quent periodic winter melting [Serreze et al., 2000, Zhang et al., 2001, Schindlerand Donahue, 2006]. Projected shifts in flow persistence were smaller than shiftsin the timing of spring runoff and the centroid of flow distribution, possibly result-ing from increased (summer) precipitation contrasted with an earlier spring runoffthat is expected to reduce future summer flows [Sauchyn and Kulshreshtha, 2008].Projected flow patterns are also sensitive to the temporal and spatial variabil-ity in temperature and precipitation patterns across different climate scenarios andGCMs [Prowse et al., 2006, Toth et al., 2006]. For example, a warmer and drierscenario could increase evaporation relative to precipitation and result in reducedrunoff. On the other hand, less warming in a wetter scenario could result in in-creased snowpack accumulation and runoff [Hinzman et al., 2005]. The selected94GCMs in this study project an annual precipitation increase of 2 to 5% for theARB in RCP4.5 by the end of the century. This represents the middle of the rangein precipitation change, -4 to 14%, projected by all CMIP5 models for West NorthAmerica (28.6◦N to 60◦N, 130◦W to 105◦W) [Christensen et al., 2013], whichcontains the Athabasca River Basin. Employing a wider selection of GCMs inthis study might broaden the range of future projected streamflow; however, suchanalysis was not possible for this study because current available output from theother CMIP5 models lack the complete set of daily climate variables needed toforce IBIS for all climate scenarios.4.5 ConclusionsClimate change in the Athabasca River Basin is projected to be the primary driverof future low flow patterns. Seasonal increases and decreases in future low flowfrequency during the respective historical summer and winter periods are pro-jected to affect the seasonal availability of water for oil sands water withdrawals.The frequency of low flows can be used to quantify the frequency of future inter-ruptions to water availability for oil sands production, assuming that restrictionswill exist on water withdrawals during low flow periods. As a result, a tradeoffarises between meeting industrial and ecological water demands. Future wateruse in the Athabasca oil sands may require operational decisions that adapt thetiming of water withdrawals to the timing of available flows. Projected changes instreamflow due to climate warming can inform such decisions by providing a toolto estimate the magnitude and uncertainty of change in future water availability.95Chapter 5Future Water Supply and DemandManagement Options in theAthabasca Oil Sands5.1 IntroductionWater management strategies in recent decades have undergone paradigm shiftsthat have focused attention first on the protection and restoration of the naturalflow regime [e.g., Poff et al., 1997, 2010, Gleick, 2000], and second on the im-portance of adapting to future climate change impacts on human water resourceuse [e.g., Vo¨ro¨smarty et al., 2000, Milly et al., 2008]. In the Athabasca RiverBasin (ARB), the ongoing development of a water management framework forthe Lower Athabasca River that recognizes these objectives has been challeng-ing [Alberta Environment, 2007, Ohlson et al., 2010]. Future development of oilsands bitumen production in the region, and its associated water use, is forecastedto continue on a high-growth trajectory [The Oil Sands Developers Group, 2013,ERCB, 2013]. At the same time, future climate change in the ARB is projected toshift the seasonal hydrograph, and may change the availability of winter and sum-mer flows for water withdrawals (see Chapter 4). These will be important con-96siderations for the future management of water withdrawals from the AthabascaRiver as tradeoffs are made between maintaining continuous bitumen productionand protecting in-stream flow needs.Restrictions on water use by oil sands companies are currently regulated ac-cording to the Alberta Government’s Phase One Water Management Framework,which specifies the amount of water that each company can withdraw from theAthabasca River throughout the year based on calculated threats to in-stream flowneeds, primarily fish life cycle and habitat needs [Alberta Environment, 2007].Phase Two of the Water Management Framework (P2F) has been in developmentsince 2007, and aims to balance long-term industry withdrawals with social, envi-ronmental, and economic interests. In developing the P2F, the multi-stakeholderP2F committee considered multiple water management alternatives which wereprojected to remain robust under future climate change in all but the most extremeclimate scenarios. These alternatives were evaluated only in the context of a high-growth scenario for oil sands bitumen production, and did not consider scenariosof more restricted growth [Ohlson et al., 2010]. Despite efforts to define a newframework, final consensus on a specific set of water management rules has notyet been achieved. The final, industry-preferred alternative proposed by the P2Fcommittee could not reach consensus over water use restrictions and exemptionsduring low flow periods and this has remained a roadblock to actual implementa-tion of a new water management framework.A broad analysis of future water management options that includes the fullrange of potential tradeoffs between oil sands industry growth and environmen-tal protection is still lacking. In this study, we applied streamflow simulationsfrom IBIS-THMB, a combined land surface process model and streamflow rout-ing algorithm (see details in Chapter 2), driven by recent CMIP5 GCM outputsusing the IPCC AR5 Representative Concentration Pathway (RCP) climate sce-narios [Moss et al., 2010], to develop two water use scenarios that bookend thepossible approaches to basin water management. One scenario prioritizes a high-growth trajectory for bitumen production and associated water withdrawals, while97the other prioritizes maximum environmental protection of the Athabasca River,which maintains the conditions of an unaltered natural flow regime. Together,these scenarios cover a range of both industry and environmental protection op-tions. For each scenario, we evaluated the water supply needed to meet the esti-mated average industry demand, and the amount of storage water, in addition todirect river withdrawals, that would be required to maintain constant bitumen pro-duction over the mid-century time period. Using this approach, we examined thewater tradeoffs that emerge when adapting water rules to projected climate changeimpacts on streamflow, and explored the range of management options availableto balance future water supply and demand in the Athabasca oil sands.5.2 Methods5.2.1 Water management scenariosTwo water management scenarios were defined to bookend the range of futurewater supply and demand options. The first scenario, labelled as the ‘industry-first’ scenario, is a high-growth oil sands development scenario that is defined tohave a production rate of 3.5 million barrels per day, requiring an average industrywater withdrawal rate of 16 m3/s and a maximum water withdrawal rate of 29m3/s based on planned pipe diameters for river water intake [Golder AssociatesLtd., 2009]. The P2F committee applied this high-growth assumption to all watermanagement alternatives that they considered. The industry-first scenario appliesthe same high-growth assumption under new climate change scenarios to providea direct comparison to the P2F analysis. Since the high-growth assumption isbased on a 2008 long-term forecast which includes both announced and potentialfuture projects [Ohlson et al., 2010], the average industry withdrawal rate is higherthan that calculated in Chapter 3, which estimates water use based on a bottom-upapproach that only includes announced future projects (as of 2013).The industry-first scenario adopts the water withdrawal rules and thresholdsoutlined in the P2F committee’s final recommendation, Option H. The water with-98drawal rules divide the year into five sets of different weeks (Table 5.1). BetweenNovember and mid-April, three flow threshold conditions determine the permittedwithdrawal amount. For the remainder of the year, the only flow threshold is theecosystem base flow (EBF), a flow threshold of 87 m3/s, based on a 1 in 100 yearwinter (January) low flow statistic [Ohlson et al., 2010]). Typically, an EBF is athreshold flow below which all water withdrawals must cease in order to avoid ir-reversible stress on aquatic ecosystems. However, the industry-first scenario (Op-tion H) rules permit a water withdrawal rate of 4.4 m3/s below the EBF for specificoil sands operators in order to prevent mining infrastructure from freezing duringcold winter months (Albian Muskeg River, Canadian Natural Horizon), as well asexemptions for the oldest operations (Suncor, Syncrude) which lack water storagecapabilities.The second scenario, labelled as the ‘environment-first’ scenario, is a scenariodescribing maximum environmental protection for the Athabasca River. The wa-ter withdrawal rules and thresholds are defined according to the Alberta DesktopMethod [Locke and Paul, 2011], which was developed as a means to prescribe fullprotection of river environments in the absence of available site data. The waterrules of the environment-first scenario permit 15% of river flow to be withdrawnwhen flow is above the weekly or monthly 80% flow exceedance value (Q80), andno withdrawals below the Q80 threshold.99Table 5.1: Annual water withdrawal rules for the industry-first scenario. For each set of weeks, a water rule(R) defines the maximum permitted withdrawal rate when the river flow (F) meets a specified threshold(T) condition (adapted from the Option H rules in Ohlson et al. [2010]).R1 (m3/s) R2 (m3/s) R3 (m3/s) R4 (m3/s)If Flow in River If Flow in River If Flow in River If Flow in RiverF > T1 T1 T1>F>T2 T2 T2>F>T3 T3 T3>FWeek allow up to: (m3/s) allow up to: (m3/s) allow up to: (m3/s) allow up to:1–15 16 270 6% of flow 150 9 87 4.416–18 16 87 4.419–23 20 87 4.424–43 29 87 4.444–52 16 200 8% of flow 150 12 87 4.4100Unlike the scenarios considered by the P2F committee, this scenario is notconstrained to a high industry growth rate. Two demand-side options were consid-ered in the environment-first scenario to explore water supply needs under differ-ent demand options; the high-growth average industry withdrawal rate of 16 m3/s,and the 2010 average industry withdrawal rate of 6 m3/s. As with the industry-first scenario, the maximum water withdrawal rate in this scenario is also 29 m3/s,based on pipe infrastructure limitations.5.2.2 Climate scenarios and modelsClimate projections in this study used output from three CMIP5 GCMs (GFDL-ESM2G, MIROC5, and IPSL-CM5A-LR) driven by the most recent IPCC climatescenarios: RCP4.5, a moderate climate change mitigation scenario and RCP8.5,the highest IPCC emissions scenario which roughly corresponds with the cur-rent emissions trajectory. These projections are used to drive a combination ofa land surface process model, IBIS [Foley et al., 1996, Kucharik et al., 2000],and a hydrological routing algorithm, THMB [Coe et al., 2002] to simulate dailystreamflow. IBIS and THMB have been used together in dozens of global, large-scale studies, including simulations of continental-scale runoff in North America[Lenters et al., 2000, Coe and Foley, 2001] and Africa [Li et al., 2005], Ama-zonian flooding [Coe et al., 2002], and Mississippi nutrient flux [Donner et al.,2002]. For full model details, see Chapter 2, and for full simulation details, seeChapter 4.Streamflow simulations in this study are therefore based on the most recentIPCC climate scenarios, in contrast to the streamflow projections considered bythe P2F committee, which were developed using the Modified Interactions be-tween the Soil-Biosphere-Atmosphere (MISBA) hydrologic model [Kerkhovenand Gan, 2006] forced by seven major GCMs using the four IPCC AR4 SpecialReport on Emissions Scenarios (SRES) [Nakicenovic et al., 2000]. In the P2Fcommitee’s analysis of climate change impacts, the projected percent changes inminimum and mean flows were used as indicators of the percent change in win-101ter and summer flows respectively, and applied as percent modifiers on a 50-yeardata set of winter (December to March) and summer (June to August) flows. Incontrast, projected climate change impacts in this study were analyzed using thefull monthly time series of IBIS-THMB simulated streamflow. The IBIS-THMBstreamflow simulations project that climate change will advance the timing ofspring runoff and shorten the persistence of late-season flow in the AthabascaRiver by mid-century, leading to an increase in streamflow in the first half of theyear, and a decrease in streamflow in the last half of the year (see details in Chap-ter 4). Differences between the reference baseline years and the timestep of modelruns used in the IBIS-THMB and MISBA projections prevent a direct compari-son of the two model results, however, the IBIS-THMB projections of seasonalrunoff timing are broadly consistent with the response of a snow-dominated basinto climate warming.5.2.3 Simulating water supply and demandMid-century (2041–2060) streamflow was simulated by IBIS-THMB for the six(3 GCMs and 2 RCPs) climate change scenarios. For each water management sce-nario, the water withdrawal rules and thresholds were calculated and then appliedto the simulated streamflow. In the environment-first scenario, Q80 thresholds foreach calendar month were calculated based on 30 years of simulated historicalflow between 1981–2010 at the location of the Water Survey of Canada ‘BelowMcMurray’ hydrometric station. The Q80 thresholds were calculated indepen-dently for each GCM-driven simulation to account for the small differences insimulated flow over 1981–2010. The range of Q80 thresholds varied by approxi-mately 4% between the different simulations.The water withdrawal rules were applied to weekly streamflow to yield aweekly permitted river withdrawal rate. If this amount was less than the ex-pected average industry withdrawal rate, an additional water supply, drawn fromavailable stored water, was required to maintain bitumen production. If permittedwithdrawals exceeded the industry withdrawal rate, the excess amount could be102stored for later use. Therefore, as the water rules were applied to the time seriesof mid-century streamflow in each scenario, permitted withdrawals in excess ofthe industry withdrawal rate were used to fill storage, while deficits in river with-drawals were supplemented with storage water to meet the industry withdrawalrate.In the industry-first scenario, storage filling is possible during open water sea-son in weeks 19 through 43, where the prescribed water rules allow withdrawalsin excess of the industry demand. In the environment-first scenario, storage fill-ing can occur when the river flow is above the Q80 threshold and the averageindustry water withdrawal rate is met. In all cases however, additional water canonly be stored if storage reservoir space is available. Over the 20 year mid-centurytime period considered here, the capacity of water storage reservoirs must be largeenough to supply water when needed through multiple fill and use cycles duringmid-century. The cycle of storage fill and use was calculated for each manage-ment and climate scenario combination, along with the minimum storage volumeneeded to maintain the industry withdrawal rate over consecutive periods of stor-age use. The calculation of storage fill and use assumed that the storage volumewas initially filled to maximum capacity.5.3 ResultsApplication of the industry-first scenario water rules to mid-century IBIS-THMBsimulated streamflow showed that river flows can supply an average industry with-drawal rate of 16 m3/s in all climate scenarios between weeks 18–38 (May toSeptember) (Figure 5.1a). For January to mid-April, early November, and lateDecember (weeks 1–15, 44–46, and 52), the available river flow cannot supplythe average industry withdrawal rate in any climate scenario, and industry with-drawals will require an alternate supply of water from other sources. The mini-mum required storage capacity in RCP8.5 was similar to the P2F committee rec-ommended storage capacity of 104 Mm3 required for a 1 in 200 year low flowoccurrence, while the minimum required storage capacity in RCP4.5 was closer103Table 5.2: The minimum storage capacity (Mm3) in each management sce-nario that is required to maintain the indicated average industry waterwithdrawal rate across all GCM-driven streamflow projections for mid-century (2041–2060). The number of days that the storage volume cansupply demand at the average industry withdrawal rate, is also shown.Industry-first Environment-first16 m3/s 16 m3/s 6 m3/s(Mm3) (days) (Mm3) (days) (Mm3) (days)RCP4.5 87 63 424 307 120 231RCP8.5 103 75 939 679 113 218to the 91 Mm3 required for a 1 in 100 year low flow occurrence (Table 5.2).During the open water season (weeks 16-43), the calculated Q80 thresholds inthe environment-first scenario (Table 5.3) were much higher than the EBF thresh-old in the industry-first scenario (Table 5.1). In contrast, the Q80 thresholds ofthe environment-first scenario during the winter (weeks 1–15, 44–52) were con-sistently lower than the T1 thresholds in the industry-first scenario. In general,the water rules in the environment-first scenario were more restrictive in the sum-mer months and less restrictive in the winter months, relative to the industry-firstscenario.1041 5 9 13 17 21 25 29 33 37 41 45 49020406080100water shortage frequency (%)1 5 9 13 17 21 25 29 33 37 41 45 49020406080100weekwater shortage frequency (%)(a)(b)GFDL−ESM2GMIROC5IPSL−CM5A−LRRCP4.5RCP8.5Figure 5.1: The percentage of time for each week of the year during mid-century (2041–2060) that river flowwithdrawals cannot supply the full average industry withdrawal rate of 16 m3/s for (a) the industry-firstscenario and (b) the environment-first scenario.105Table 5.3: Example of the weekly water rules in the environment-first sce-nario for GFDL-ESM2G and RCP4.5. The water rule (R) defines themaximum permitted withdrawal rate when the weekly average riverflow (F) meets the Q80 threshold (T) condition. Weeks are groupedhere for brevity, but R is calculated separately for each week.R (m3/s) If Flow inRiver F>T allow upWeek to: T(m3/s)1 - 1515% of flow inriver OR 29 m3/s,whichever islower94 - 22816 - 18 120 - 29119 - 23 391 - 69024 - 43 240 - 85944 - 52 240 - 310Application of the water rules of the environment-first scenario to mid-centurystreamflow shows that in the latter half of the year (from early July, or week 27,forward), river withdrawals cannot supply the high-growth average industry with-drawal rate of 16 m3/s in any climate scenario (Figure 5.1b). Deficits in wateravailability also occur during weeks in January and February for all climate sce-narios. The volume of water storage required to maintain an average industrywithdrawal rate of 16 m3/s is four to nine times greater than the largest volume re-quired in the industry-first scenario (Table 5.2). Even at the 2010 average industrywithdrawal rate of 6 m3/s, the minimum storage capacity required would exceedthat of the industry-first scenario by 110–140%.5.4 DiscussionThe withdrawal rules and rates, and the water storage requirements of the industry-first and environment-first scenarios, bookend a spectrum of future options forbalancing water supply and demand in the Athabasca oil sands (Table 5.4). Re-strictions on growth will limit the economic potential of the oil sands industry.If the focus is on the protection of industry, it is unlikely that policy makers will106Table 5.4: Matrix showing management options for a range of prioritiesbased on the evaluation of the industry-first and environment-first sce-narios.Industry protection Environmental protectionLimitedgrowthIndustry accepts economiclosses and reduces waterdemand∼ 218−231 days of storagerequiredHighgrowth∼ 63−75 days of storagerequiredVery high storagerequirements, > 1 yeardevelop water management rules that limit the growth of operations (top left box,Table 5.4). Instead, it is more likely that storage volumes will be built to accom-modate the increased demand for water that follows high-growth in oil sands pro-duction (bottom left box, Table 5.4). When environmental protection is a priority,permitting high-growth in oil sands production may encounter potential physicallimitations of building sufficient storage volumes (bottom right box, Table 5.4),while reducing water demand will lead to more reasonable requirements for stor-age capacity (top right box, Table 5.4). The 2010 average industry withdrawal ratethat was considered in the environment-first scenario is lower than the projectedbase-growth demand (based on announced or approved projects in 2006) of 11.3m3/s through 2030 [Golder Associates Ltd., 2009].In both the industry-first and environment-first scenarios, the availability ofwater for oil sands operational use depends on several factors. First, water rulesthat define the EBF, low flow thresholds, and permitted withdrawals determine therate and frequency of storage water use (when flow is below the threshold), as wellas the frequency of storage filling (when flows are above the threshold). Second,107the water intake capacity or water pipe diameter, determines the maximum rateof storage filling. Third, the average industry withdrawal rate determines boththe rate of storage use (flow needed to meet the industry rate), and the rate ofstorage fill (available river flow in excess of the industry rate). Lastly, the max-imum storage capacity limits the frequency and rate of storage filling and there-fore determines the number of consecutive low flow periods that can be suppliedwith stored water (for example, in the IPSL-CM5A-LR-driven RCP4.5 simulationof the environment-first scenario, storage water is depleted over two consecutiveyears in which storage use exceeds storage fill opportunities). The presence ofmultiple controls on the availability of water for withdrawals make it challengingto design specific water rules and select water supply and demand thresholds thatwill be flexible enough to adapt to different future climate change scenarios.Uncertainties in mid-century streamflow projections also complicate the de-sign of specific water rules. This is demonstrated in the industry-first scenario,where the difference between the P2F committee and IBIS-THMB projectionsof climate change impacts suggest different storage capacity requirements, de-pending on the climate scenario. Understanding the range of future streamflowvariability will also be important to future water use planning so that storage filland use cycles can take full advantage of seasonal water availability. Water rulesthat are defined based on an incorrect assumption of higher available flows duringcertain seasons or weeks may miss opportunities for filling storage otherwise. Forexample, if minimum (winter flows) are projected to decline as generally simu-lated in the P2F committee climate change analysis, then water rules would bedesigned to limit withdrawals in the winter. However, if minimum flows gener-ally increase as simulated by the IBIS-THMB climate change analysis, then waterwithdrawal rules may be relaxed to allow storage filling during these weeks, lead-ing to more efficient use of storage capacity. The calculated storage requirementsdepend upon the specific sequence of climate variability in these simulations. Thepossibility of different climate realities than forecasted is therefore also an argu-ment for the design of flexible water rules.108In order to maintain a specific industry average withdrawal rate in both theindustry-first and environment-first scenarios, there must be a continuous and ad-equate water supply from river withdrawals and other water sources. The P2Fcommittee considered several options for improving water access during periodsof low flow, along with the associated capital costs, timing and feasibility issues,operating costs, footprint, and reliability of each option [Golder Associates Ltd.,2010]. Options included advancements in water treatment, off-site water stor-age (Lesser Slave Lake and McMillan Lake, other lakes and dams on tributaries),on-site water storage (constructed fresh water ponds, tailings pond treatment ordelayed reclamation, delayed closure of pit lakes), and groundwater (Pleistoceneaquifer via Wiau Channel). Of these, groundwater was the least likely to sup-ply sufficient water, while water treatment was the least reliable and most costly($40/m3). Only two technologies were ultimately shortlisted as the preferred in-dustry options based on risk, reliability, complexity, and timing issues: on-sitefresh water ponds and on-site tailings ponds. Of these, on-site fresh water pondswere determined to be the more practical option, given that the ERCB Tailings Di-rective 074 requires that tailings ponds are decommissioned in a timely manner.These tailings ponds have a footprint that covers 22% of the total disturbed miningarea, and contain the waste water and residue from oil sands bitumen extractionthat can lead to the seepage of pollutants into surrounding soils and water, as wellas pose a danger to migratory birds [Alberta Environment, 2009].Oil sands operators will need to weigh the cost of building sufficient storagecapacity (Table 5.5) against the cost of lost bitumen production during periods ofwater shortage. Although plant shutdowns do occur periodically, they are gen-erally unplanned and any shutdown of water withdrawals may also lead to costsassociated with equipment damage [Ohlson et al., 2010]. The P2F committee con-cluded that industry would be more likely to build additional storage capacity thanto accept water supply shortfalls since the costs associated with a loss in produc-tion would exceed the cost of additional storage construction [Ohlson et al., 2010].Given a 2012 WTI crude oil price of $95 US per barrel [CAPP, 2014] along with109the P2F committee assumption of 3.5 million barrels of production per day bymid-century, a week of interrupted production would amount to $2.3 billion USin lost revenue, equivalent to the capital cost of building approximately 144 Mm3of on-site storage capacity. This storage volume would be sufficient to satisfy theindustry-first scenario in each climate scenario (Table 5.5). In the environment-first scenario, however, constructing sufficient storage to prevent a single weekof lost production during the mid-century time period could come at up to sixtimes the cost of a week of lost revenue. In addition to capital costs, the annualoperating costs and land area needed for storage will also influence decisions tobuild storage. Storage footprints in the industry-first scenario range from 44 to 52km2, while storage footprints in the environment-first scenario range from 212 to470 km2 and would exceed the mining area footprint of most oil sands operations(Table 5.5).110Table 5.5: Costs and footprint of freshwater pond storage per unit meter [Ohlson et al., 2010] and the cal-culated costs and footprint of storage requirements associated with the different management optionsconsidered.AnnualCapital cost operating Footprint Water lossStorage option ($/m3) cost ($/m3) (km2/Mm3) (Mm3/Mm3)freshwater pond storage 16 0.88 0.5 0.04AnnualWater Climate Withdrawal Capital cost operating Footprint Water lossscenario scenario rate (M$) cost (M$) (km2) (Mm3)industry-first RCP4.5 16 m3/s 1,392 76 44 3industry-first RCP8.5 16 m3/s 1,648 90 52 4environment-first RCP4.5 16 m3/s 6,784 371 212 16environment-first RCP8.5 16 m3/s 15,024 822 470 35environment-first RCP4.5 6 m3/s 1,920 105 60 5environment-first RCP8.5 6 m3/s 1,808 99 57 4111The design and planning of future oil sands operations now include the con-struction of on-site freshwater storage facilities in anticipation of periods withlow water availability that could interrupt bitumen production. The number ofconsecutive low flow periods that can be supplemented by stored water use islimited by the maximum built storage capacity. Imperial’s Kearl mine site hasa 30-day storage capacity intended to sustain production during winter months(∼ 2.8 Mm3 volume based on an estimated 1.07 m3/s withdrawal rate from oilsands project data compiled in Chapter 3 [Imperial Oil Limited, 2013]), whileTotal E&P Canada’s new Joslyn North Mine Project, scheduled to commenceproduction in 2020, incorporates a 90-day water storage capacity (∼ 2.9 Mm3volume based on an estimated 0.368 m3/s withdrawal rate) [Total E&P Canada,2014]. However, older operations without water storage capabilities, such as Sun-cor, have stated that implementing water storage facilities for their aging miningoperations would produce a net negative impact on the environment due to addi-tional land disturbance [Healing, 2010]. During the development process of theP2F, these companies have argued for a total 4.4 m3/s exemption below the EBFdue to plant designs that require continuous water withdrawals from the river andthe absence of appropriate on-site water storage facilities. Using the per unit stor-age cost estimates, the construction of 30 days of storage capacity in order tosupply Suncor’s portion of the exemption withdrawal rate (2 m3/s of the total 4.4m3/s) would require a∼ 3 km2 of freshwater pond storage area (Table 5.5), whichis less than 2% of Suncor’s mining footprint in 2010 [Suncor Energy Inc., 2011].The environment-first scenario shows that water rules that provide maximumenvironmental protection cannot also supply a high-growth industry withdrawalrate of 16 m3/s without prohibitively expensive storage volumes. If the currentemissions trajectory is maintained (represented by RCP8.5), then maximum en-vironmental protection will not be compatible with climate change and high in-dustry growth due to the implausibly high storage requirements. To avoid highstorage demands, limits could be placed on either bitumen production or the av-erage industry withdrawal rate. A reduction in the average industry withdrawal112rate without reducing production levels would require significant advancementsin water mitigation technologies. In recent years, oil sands mining operationshave taken successful measures to reduce their water consumption intensity; theamount of water needed to produce one barrel of oil. Suncor reports a currentwater consumption intensity of 2.06:1 (water:oil), a 10% reduction since 2007[Suncor Energy Inc., 2013a]. Syncrude, in turn, reports a 60% reduction in wateruse since the 1980’s [Syncrude Canada Ltd., 2012] Most oil sands mining op-erators have been exploring water use mitigation options both in retrofitting oldoperations and in the design and construction of future operations. For example,Suncors wastewater treatment plant, opening in 2014, is expected to reduce waterconsumption intensity by 65–75% relative to 2007 [Suncor Energy Inc., 2013b].In 2012, 41.4% of water withdrawn for Suncor’s operations was treated and re-turned to the Athabasca River, while for other operations such as Shell, no waterfrom mining and extraction operations is currently returned to the Athabasca River[Shell Canada, 2014]. Canada’s Oil Sands Innovation Alliance is also focused onaccelerating the development and commercialization of water treatment technolo-gies and managing salt accumulation in water streams on mine sites [Canada’s OilSands Innovation Alliance, 2014].Although the growth trajectory of oil sands mining operations is projected tocontinue rising, supply and demand forecasts don’t generally extend into the mid-century time period considered in this study yet. There are a wide range of in-teracting factors that control future bitumen production, including energy prices,technology improvements, operational costs, crude oil demand, and remainingbitumen reserves [ERCB, 2013, Dobson et al., 2013]. Fluctuations in these con-ditions will control the pace of development of the oil sands industry and whethergrowth or decline in production and associated water use will occur.5.5 ConclusionsA spectrum of future water management options for the Athabasca oil sands re-gion was considered in this study. At one extreme, maintaining both maximum113environmental protection and a high growth rate in water withdrawals is implausi-ble, since the water storage requirements would not be cost-effective compared tothe potential loss in production revenue and/or feasible with respect to availableland area. At the opposite extreme, minimizing environmental protection and re-ducing current bitumen production output, is unlikely to find agreement with anystakeholders. Future water use in the Athabasca oil sands will require tradeoffsin both water supply and demand that consider the range of options in betweenthese extremes. For example, water supply can be increased by relaxing the ruleson seasonal water withdrawals and/or building greater water storage capacity, ifenvironmental protection is reduced and/or additional capital and operating costsare incurred. Water demand for withdrawals, in turn, can be decreased by reduc-ing bitumen production and/or increasing water use efficiency, but would resultin lost revenue and/or increased research and development costs. The scale andcosts of these actions will depend, in part, on the degree to which environmen-tal protection and industry growth are each prioritized. In addition, there willbe some risk associated with making these tradeoffs, since uncertainty (some ofwhich is irreducible) in climate change projections introduces further uncertain-ties in estimating the future frequency and severity of low flow periods. The rangeof impacts and responses considered in this study can serve to inform future watermanagement planning for the Lower Athabasca River, and also serve as a gen-eral example of the type of emerging tradeoffs between industrial water needs andin-stream flow needs in a changing climate.114Chapter 6Conclusions6.1 Key insights and findingsThis dissertation describes the application of a land-surface model and a hydro-logic model to the analysis of climate change and water use in the ecologicallyand economically important Athabasca River Basin. Collectively, the results ofthis research find that both climate change and industry growth will drive the fu-ture availability of freshwater, a critical resource for oil sands mining operations,as well as for people and ecosystems in the basin. In turn, the availability of fresh-water to supply industry needs could influence the scale of future development inthe Athabasca oil sands.The key model results show that climate change is projected to be the primarydriver of streamflow alterations that will directly affect seasonal water supply inthe Athabasca River Basin by lowering summer flows and increasing winter flows(Chapter 4). Oil sands industry water withdrawals are projected to have a com-paratively small impact on river flow (Chapter 3, Chapter 4). Since concern hasconventionally focused on mitigating the environmental impact of oil sands waterwithdrawals on the Athabasca River, these findings suggest a new and additionalmotive for the careful management of water withdrawals - to mitigate the impactsof climate change-driven water shortages on bitumen production.115Calculations using a range of future water supply and demand trajectoriesindicate that sufficient water storage will be needed in all scenarios to preventwater shortages throughout the year (Chapter 5). The design of adaptive waterrules should therefore optimize water storage by recognizing any seasonal shiftsin the hydroclimatological regime. For example, the availability of flows is pro-jected to increase during winter months, the historical low flow season when waterwithdrawals are typically minimized, and is projected to decrease during summermonths, when water withdrawal restrictions are typically relaxed. Based on theseprojections, water rules should then be optimized to supply water during peri-ods of low river flow by maximizing opportunities to fill water storage reservoirsduring periods of high river flow.The volume of storage needed to supplement river withdrawals depends notonly on changes in the magnitude and timing of the freshwater river supply, butalso on whether water demand for oil sands operations continues to rise, remainsstatic, or declines. While the oil sands operators and the province of Alberta,although prominent contributors to greenhouse gas emissions, cannot alone mit-igate the impacts of climate change on the Athabasca River Basin, there are op-tions to address the water demand for bitumen production. Making decisions onthe magnitude of water demand, and therefore the scale of bitumen production,will depend in part on how environmental protection versus industry growth isprioritized based on the tradeoffs between environmental and economic costs.The management of future water resources for industry use is complex be-cause it requires a scientific understanding of the regional water supply to informmanagement and policy options. For the Athabasca oil sands industry, this meansthat an understanding of the river basin hydroclimatology, the economics of bi-tumen projection, as well as the uncertainties in both future climate change andenergy demand trajectories, is needed to inform tradeoff decisions.1166.2 ContributionThis research contributes a scientific basis for the future adaptive management ofwater resources in the Athabasca River Basin and develops a range of possiblewater management options for policy makers to consider. Although the resultsare specific to the Athabasca River Basin, the methods developed are also relevantto other river basins that face challenges in balancing energy and water demandsunder a changing climate.Each chapter in this dissertation forms a part of the overall contribution. Exist-ing large-scale, process-based models were adapted for application in the AthabascaRiver Basin and a historical baseline for streamflow was developed through themodel parameterization and validation process (Chapter 2). This involved thecollection and organization of climatic and hydrologic data to drive model simu-lations, and the results provide an essential reference point from which to evaluatefuture flow and projected impacts on water resources. This modelling frameworkhas potential broader applicability to studies of the Mackenzie River Basin, aswell as other northern Canadian river basins.For the first time, a comprehensive set of oil sands water use estimates wassynthesized from sparse data records and sources that are currently limited underthe existing system of voluntary water use reporting (Chapter 3). This new data setprovides a spatially explicit representation of water withdrawals in the AthabascaRiver Basin, which was used to construct a range of future water use scenarios.The application of these scenarios to simulate streamflow impacts is the first at-tempt to model oil sands water withdrawals within a process-based hydrologicalmodelling framework.This research is also the first attempt to quantify the range of climate changeimpacts on future streamflow in the Athabasca River Basin using the most recentglobal climate projections and scenarios (Chapter 4). The results of this climatechange analysis constitute an assessment of the potential risks and vulnerabilitiesin future oil sands water supply, and highlight the dominant role of climate changein altering future streamflow availability. This discovery suggests that water man-117agers, industry, and policy makers may wish to also consider the risks to futurebitumen production, alongside the environmental risks of bitumen production.Finally, the synthesis of climate change and water withdrawal impacts con-ducted in this study identifies a full range of water management options with dif-ferent priorities and tradeoffs in environmental protection and industry growth,which can help to inform future water policy decisions (Chapter 5). This researchexpands on the body of knowledge that has been recently developed to draft anew regional water management framework. It considers different scenarios ofindustry growth and decline that have not been previously addressed. The meth-ods for examining water storage options, while specific to the oil sands industry, isbroadly relevant to any sector, such as agriculture, where freshwater withdrawalsand storage is required.The intersection of energy and water demands, along with the dependencyof energy production on water, leads to increasingly common tradeoffs that arenot unique to the oil sands industry [e.g., Richter et al., 2003, Chapagain andOrr, 2009, Do¨ll et al., 2009, Harma et al., 2012]. The modelling approach ofthis study provides a tool to identify the science behind, and therefore inform thefacilitation of, these tradeoffs that water managers may encounter in adapting awater management framework to future climatic and industry conditions.6.3 Strengths and limitations of the researchThe development of a scientific basis to inform water policy and management re-quires the integration of knowledge from different disciplines. These disciplinescan include climate and hydrological science, ecology, and natural resource eco-nomics and management. Given the breadth of fields involved, the scope of thisdissertation was necessarily focused on a subset of these topics. The approach forthis research was to develop the specific linkages between hydrological scienceand water use management that quantify the timing of water supply and demand,in order to inform a framework for adaptation of future water use. The examina-tion of water use is restricted to withdrawals by the oil sands mining industry; this118is justified given that river water withdrawals for in situ oil sands projects and forother economic sectors in the Athabasca River Basin are negligible in comparison.A large-scale process-based modelling approach was chosen for this study be-cause it was important to simulate the hydrologic response of the entire river basinin order to capture future climate change impacts. Although the key water use is-sues in this study are primarily contained in the lower reaches of the basin, theyreflect the upstream dynamics that drive downstream flow. A sacrifice in pursuinga large-scale approach is that small-scale processes, such as the river ice cycleand its impact on flow, are difficult to parameterize. However, there are inevitabletradeoffs between predictability and model complexity. For example, while alter-native models, such as those calibrated to a river reach, may describe historicalflows more accurately, a process-based model can be better able to capture thesensitivity of the model to changing drivers like climate.Another challenge with large-scale, process-based models is the accurate sim-ulation of all atmosphere, soil and vegetation exchanges across a large basin. Thecharacterization of these physical processes requires a spatially and temporallyextensive observational data set to validate each component of the modelled waterbalance, and such information was limited for the Athabasca River Basin. Al-though streamflow validation captured the timing of flows well, disagreementsbetween the magnitude of observed and modelled discharge persisted, particularlyin the winter months. An analysis of observed precipitation data also showed thatthe overprediction of streamflow in the last decade of the historical time periodwas likely due to inaccurately large reanalysis precipitation inputs to the modelsystem. To control for these biases in the simulated streamflow, the common prac-tice of studying the change in future projections relative to a baseline was adopted,rather than the use of raw future projection output. This approach still allows fora full range of water withdrawal and climate change impacts on streamflow to becaptured. Due to the differences between global climate models, irreducible un-certainties between climate change projections, especially regional ones, will alsooccur.1196.4 Potential future research directionsMany avenues exist for future research that expands and draws on the work inthis dissertation. Further development of the modelling system would improvethe accounting of various hydrological processes important to the Athabasca oilsands region. For example, river and lake ice dynamics could be simulated byparameterizing the basic sensitivity of freeze/melt timing and ice thickness totemperature. Another example is the model parameterization of wetland envi-ronments that would have implications for flow storage and pathways that affectthe downstream timing of flows. Modelling the spatial and temporal distributionof wetlands could involve a parameterization based on soil saturation and/or a pa-rameterization based on inundated areas. Also of note is that the current oil sandsmining landscape is not specifically captured in current model simulations. Theparameterization of land-surface exchanges over the oil sands mining area may beimportant in determining whether the water demand for future land reclamationof mined sites can be satisfied.Improvements to the modelling system would advance the accuracy of sim-ulated seasonal streamflow, however, any new model developments must still besupported by adequate observational data in order to validate the physical pro-cesses that are represented. Continued development of a comprehensive networkof observations in the Athabasca River Basin is needed to better validate modelpredictions of future water availability and therefore minimize the risk and uncer-tainty in management decisions. Observations are also needed to bridge the gapbetween experimental work at the river reach-scale, and the large-scale averagesthat are required for basin-scale climate modelling. In 2012, the provincial andfederal governments announced a three-year Joint Canada-Alberta Implementa-tion Plan for Oil Sands Monitoring which aims to increase monitoring efforts inthe oil sands region. The plan will examine the long-term cumulative impactsof the oil sands industry using an expanded network of monitoring sites, includ-ing increased water quantity monitoring, and improved methodologies for datacollection. Industry-funded initiatives such as the Regional Aquatics Monitoring120Program have also contributed to observations of discharge in the Athabasca Riverin recent years, although the majority of water quantity monitoring stations in thisnetwork are located along tributaries of the Athabasca River.In addition to further model development, existing features of the modellingsystem can also be applied to expand the scope of this research. For example, thehydrological routing algorithm, THMB, has the capability to examine basin sedi-ment flow and dissolved constituents [Donner et al., 2004, Donner and Kucharik,2008]. These algorithms could potentially be applied to investigate water qualityissues in the Athabasca oil sands, a key environmental concern.Many other potential research directions exist beyond the scope of hydrologi-cal modelling and address the breadth of disciplines related to water management.One key example is the exploration of environmental risk. A direct evaluation ofpotential threats to river ecology and in-stream flow needs is needed to determineif environmental tradeoffs are legitimate, and whether water use allocations ad-dress environmental risk as intended. The vulnerabilities of multiple ecosystemsneed to be considered, in addition to the conventional focus on fish habitat andlife cycles. Another dimension of environmental risk that can be examined is thecontribution of greenhouse gas emissions from the oil sands industry to climatechange, which in turn drives streamflow alterations that can interrupt bitumen pro-duction.Water use beyond the scope of the oil sands industry, could also be consid-ered in future research. For approximately the next decade, water use in othersectors is projected to remain small relative to water use for the oil sands miningindustry. Climate warming by mid-century and beyond could, however, lead toan increase in competition for water resources in the Athabasca River Basin bytransforming regions currently too cool and remote to sustain agriculture into vi-able sectors that require water for irrigation [Brklacich et al., 1997, Ramankuttyet al., 2002]. While water withdrawal needs for oil sands operations are relativelyconstant throughout the year, water withdrawal needs for irrigated agriculture fol-low a more seasonal pattern. Different patterns of water use could further compli-121cate the timing of available flows for either industry and make the design of waterwithdrawal rules across the basin more challenging.While the research in this dissertation addresses hydrological and industry is-sues specific to the Athabasca River Basin, the scientific methods and conceptsdeveloped can also be applied more broadly in future work. The models used toanalyze future water availability under climate change, for example, were adaptedto the Athabasca River Basin, a cold, northern river basin, and can potentiallybe applied now to a study of the entire Mackenzie River Basin and other cold-region river basins. Water use concepts addressed in this study, such as adaptingto freshwater supply and demand constraints, and tradeoffs between energy andwater supply, have wide applicability for water resource management in otherriver basins as well. Some regions may face tradeoffs in future energy and wa-ter security with respect to different industries such as hydroelectricity and shalegas exploration [POLIS Project on Ecological Governance, 2012]. Studies of theevolving interaction between climate and human streamflow perturbations in anybasin will help build an informed approach to the general climate change adapta-tion of future water use.122BibliographyAlberta Energy. Alberta’s leased oil sands area.http://www.energy.alberta.ca/LandAccess/pdfs/OSAagreeStats.pdf, January2013. Retrieved March 2013. → pages 55Alberta Energy Regulator. Oil sands projects production history graph.http://osip.alberta.ca/library/Dataset/Details/45, October 2010. RetrievedMarch 2013. → pages 57Alberta Environment. Water management framework: Instream flow needs andwater management system for the Lower Athabasca River. Technical report,Alberta Environment and Fisheries and Oceans Canada, February 2007. →pages 5, 9, 10, 11, 13, 50, 65, 66, 74, 96, 97Alberta Environment. Environmental management of Alberta’s oil sandsresource. Responsible.http://www.environment.gov.ab.ca/info/library/8042.pdf, January 2009.Catalogue #ENV-748-O. → pages 8, 9, 93, 109Alberta Environment and Sustainable Resource Development. Oil sandsoperators: Water use history.http://environment.alberta.ca/apps/OSIPDL/Dataset/Details/56?ProjectId=20,February 2013. Retrieved March 2013. → pages 57Alberta Environment and Sustainable Resource Development. Sectoral waterallocations. http://esrd.alberta.ca/focus/state-of-the-environment/water/surface-water/pressure-indicators/sectoral-water-allocations/, April 2014.Retrieved January 2011. → pages 10Alberta Utilities Commission. Timeline extension to complete the Jackpinepower plant. Decision 2010-577, Shell Canada Ltd., Calgary, Alberta,December 2010. → pages 69123AMEC Earth & Environmental. Athabasca River Basin. In Current and FutureWater Use in Alberta, chapter 11, page 64. Alberta Environment, Edmonton,Alberta, 2007. → pages 9, 10, 11, 49, 69D. D. Andres. The breakup process and the documentation of the 1978 ice jamson the Athabasca River at Fort McMurray. In Proceedings of the workshop onhydraulic resistance of river ice, pages 143–158, Burlington, Ontario, 1980.National Research Council of Canada. → pages 46T. Andrews, J. M. Gregory, M. J. Webb, and K. E. Taylor. Forcing, feedbacks,and climate sensitivity in CMIP5 coupled atmosphere-ocean climate models.Geophysical Research Letters, 39(9):1–7, 2012. → pages x, 77N. W. Arnell and N. S. Reynard. The effects of climate change due to globalwarming on river flows in Great Britain. Journal of Hydrology, 183:397–424,1996. → pages 77AWRI. Groundwater in Alberta: An assessment of source, use and change.Technical report, Alberta Water Research Institute, Edmonton, Alberta,February 2011. → pages 9N. H. Batjes. Global soil profile data. Available on-line athttp://www.daac.ornl.gov from ORNL Distributed Active Archive Center, OakRidge National Laboratory, Oak Ridge, Tennessee, U.S.A., 2000. → pages 21,52E. G. Bekele and H. V. Knapp. Watershed modeling to assessing impacts ofpotential climate change on water supply availability. Water ResourcesManagement, 24:3299–3320, 2010. → pages 39S. Beltaos. River ice breakup processes: recent advances and future directions.Canadian Journal of Civil Engineering, 34:703–716, 2007. → pages 47S. Beltaos. Hydrodynamic characteristics and effects of river waves cause by icejam releases. Cold Regions Science and Technology, 85:42–55, 2013. →pages 46A. Botta and J. A. Foley. Effects of climate variability and disturbances on theAmazonian terrestrial ecosystems dynamics. Global Biogeochemical Cycles,16(4):1070–1080, 2002. → pages 18, 51124A. Botta, N. Ramankutty, and J. A. Foley. Long-term variations of climate andcarbon fluxes over the Amazon Basin. Geophysical Research Letters, 29(9):1319–1322, 2002. → pages 18, 51M. Brklacich, P. Curran, D. Brunt, and D. McNabb. CO2-induced climaticchange and agricultural potential in the Mackenzie Basin case. In S. Cohen,editor, Mackenzie Basin Impact Study, final report 7. People: EconomicActivities and Communities, pages 242–246. Environment Canada, Canada,1997. → pages 121J. Bruce. Oil and water - will they mix in a changing climate? The AthabascaRiver story. Policy research initiative, Government of Canada, October 2006.→ pages 5, 11, 74D. H. Burn. Climatic influences on streamflow timing in the headwaters of theMackenzie River Basin. Journal of Hydrology, 352:225–238, 2008. → pages84D. H. Burn, O. I. Abdul Aziz, and A. Pietroniro. A comparison of trends inhydrological variables for two watersheds in the Mackenzie River Basin.Canadian Water Resources Journal, 29(4):283–298, 2004. → pages 4, 46, 74Canada’s Oil Sands Innovation Alliance. Water.http://www.cosia.ca/initiatives/water, 2014. Retrieved June 2014. → pages 113CAPP. Water use in Canada’s oil sands. Technical report, Canadian Associationof Petroleum Producers, July 2012. → pages 7CAPP. Basic statistics. Oil sands 2011 statistics.http://www.capp.ca/library/statistics/basic/Pages, 2014. Retrieved May 2014.→ pages 109A. K. Chapagain and S. Orr. An improved water footprint methodology linkingglobal consumption to local water resources: A case of Spanish tomatoes.Journal of Environmental Management, 90:1219–1228, 2009. → pages 118J. Choles. An overview of streamflows and lake levels for the Peace, Athabascaand Slave River Basins. Technical Report 146, Northern River Basins Study,1996. → pages 4, 5125J. H. Christensen, K. K. Kumar, E. Aldrian, S.-I. An, I. F. A. Cavalcanti, M. deCastro, W. Dong, P. Goswami, A. Hall, J. K. Kanyanga, A. Kitoh, J. Kossin,N.-C. Lau, J. Renwick, D. B. Stephenson, S.-P. Xie, and T. Zhou. Climatephenomena and their relevance for future regional climate change. In T. F.Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung,A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, editors, Climate Change 2013:The Physical Science Basis. Contribution of Working Group I to the FifthAssessment Report of the Intergovernmental Panel on Climate Change.Cambridge University Press, Cambridge, UK and New York, NY, 2013. →pages 6, 74, 95M. T. Coe. A linked global model of terrestrial hydrologic processes: Simulationof modern rivers, lakes, and wetlands. Journal of Geophysical Research, 103(D8):8885–8899, 1998. → pages 20, 55M. T. Coe. Modeling terrestrial hydrological systems at the continental scale:Testing the accuracy of an atmospheric GCM. Journal of Climate, 13(4):686–704, 2000. → pages 19, 22, 33, 53, 54M. T. Coe and J. A. Foley. Human and natural impacts on the water resources ofthe Lake Chad basin. Journal of Geophysical Research, 106(D4):3349–3356,2001. → pages 15, 18, 19, 51, 53, 75, 101M. T. Coe, M. H. Costa, A. Botta, and C. Birkett. Long-term simulations ofdischarge and floods in the Amazon Basin. Journal of Geophysical Research,107(D20):8044–8060, 2002. → pages 12, 15, 18, 19, 20, 51, 53, 55, 75, 101M. T. Coe, M. H. Costa, and E. A. Howard. Simulating the surface waters of theAmazon River basin: Impacts of new river geomorphic and flowparameterizations. Hydrological Processes, 22(14):2542–2553, 2008. →pages 12, 19, 20, 22, 34, 54, 55G. Coen and C. Wang. Estimating vertical saturated hydraulic conductivity fromsoil morphology in Alberta. Canadian Journal of Soil Science, 69(1):1–16,1989. → pages 30J. Conti and P. Holtberg. International energy outlook 2011. ReportDOE/EIA-0484(2011), U.S. Energy Information Administration, September2011. → pages 7126L. P. de Rham, T. D. Prowse, and B. R. Bonsal. Temporal variations in river-icebreak-up over the Mackenzie River Basin, Canada. Journal of Hydrology, 349:441–454, 2008. → pages 46C. Delire and J. A. Foley. Evaluating the performance of a landsurface/ecosystem model with biophysical measurements from contrastingenvironments. Journal of Geophysical Research, 104(D14):16,895–16,909,1999. → pages 18, 51M. D. Dettinger, D. R. Cayan, M. K. Meyer, and A. E. Jeton. Simulatedhydrologic responses to climate variations and change in the Merced, Carsonand American River Basins, Sierra Nevada, California, 1900-2099. ClimaticChange, 62:283–317, 2004. → pages 93S. Dobson, N. Lemphers, and S. Guilbeault. Booms, busts and bitumen: Theeconomic implications of Canadian oilsands development. Technical report,The Pembina Institute and E´quiterre, Calgary, AB and Montreal, QC,November 2013. → pages 113P. Do¨ll, K. Fiedler, and J. Zhang. Global-scale analysis of river flow alterationsdue to water withdrawals and reservoirs. Hydrol. Earth Syst. Sci., 13:2413–2432, 2009. → pages 13, 118S. D. Donner. The impact of climate and land use on nitrate export by theMississippi River. Ph.D. dissertation, University of Wisconsin - Madison,Madison, Wisconsin, 2002. → pages 33S. D. Donner and C. J. Kucharik. Evaluating the impacts of land managementand climate variability on crop production and nitrate export across the UpperMississippi Basin. Global Biogeochemical Cycles, 17(3):1085–1102, 2003. →pages 19, 53S. D. Donner and C. J. Kucharik. Corn-based ethanol production compromisesgoal of reducing nitrogen export by the Mississippi River. Proceedings of theNational Academy of Sciences, 105(11):4513–4518, 2008. → pages 121S. D. Donner, M. T. Coe, J. D. Lenters, T. E. Twine, and J. A. Foley. Modelingthe impact of hydrological changes on nitrate transport in the MississippiRiver Basin from 1955-1994. Global Biogeochemical Cycles, 16(3):1–19,2002. → pages 15, 19, 51, 53, 75, 101127S. D. Donner, C. J. Kucharik, and J. A. Foley. Impact of changing land usepractices on nitrate export by the Mississippi River. Global BiogeochemicalCycles, 18:1–21, 2004. → pages 19, 53, 121D. Droitsch. Watered down: Overcoming federal inaction on the impact of oilsands development to water resources. Report, Water Matters Society ofAlberta, Calgary, October 2009. → pages 8J. Dumanski, T. Macyk, C. Veauvy, and J. Lindsay. Soil survey and landevaluation of the Hinton - Edson area, Alberta. Report 31, Alberta Soil Survey,Alberta, 1972. → pages 28M. El Maayar, D. T. Price, C. Delire, J. A. Foley, T. A. Black, andP. Bessemoulin. Validation of the Integrated Biosphere Simulator overCanadian deciduous and coniferous boreal forest stands. Journal ofGeophysical Research, 106(D13):14,339–14,355, 2001. → pages 18, 52, 75Energy Resources Conservation Board. Oil sands operators: Water licenses.http://environment.alberta.ca/apps/OSIPDL/Dataset/Details/55, October2010. Retrieved March 2013. → pages 55D. Entekhabi, E. G. Njoku, P. E. O’Neill, K. H. Kellogg, W. T. Crow, W. N.Edelstein, J. K. Entin, S. D. Goodman, T. J. Jackson, J. Johnson, J. Kimball,J. R. Piepmeier, R. D. Koster, N. Martin, K. C. McDonald, M. Moghaddam,S. Moran, R. Reichle, J. C. Shi, M. W. Spencer, S. W. Thurman, L. Tsang, andJ. V. Zyl. The Soil Moisture Active and Passive (SMAP) mission. Proceedingsof the IEEE, 98:704–716, 2010. → pages 46Environment Canada. Water Survey of Canada hydrometric data.http://www.wsc.ec.gc.ca/applications/H2O/index-eng.cfm, April 2010.Retrieved March 2012. → pages 8, 24Environment Canada. Canada’s sixth national report on climate change. Report,Actions to meet commitments under the United Nations framework conventionon climate change, Ottawa, Ontario, April 2014. → pages 7ERCB. Coal mine atlas operating and abandoned coal mines in Alberta. SerialPublication ST45, Energy and Resources Conservation Board, January 212010. → pages 7128ERCB. Alberta’s energy reserves 2012 and supply/demand outlook 2013-2022.Serial Publication ST98-2013, Energy and Resources Conservation Board,May 2013. → pages 50, 96, 113T. Farr, P. Rosen, E. Caro, R. Crippen, R. Duren, S. Hensley, M. Kobrick,M. Paller, E. Rodriguez, L. Roth, D. Seal, S. Shaffer, J. Shimada, J. Umland,M. Werner, M. Oskin, D. Burbank, and D. Alsdorf. The Shuttle RadarTopography Mission. Reviews of Geophysics, 45(2):33, 2007. → pages 22, 53J. A. Foley, I. C. Prentic, N. Ramankutty, S. Levis, D. Pollard, S. Sitch, andA. Haxeltine. An integrated biosphere model of land surface processes,terrestrial carbon balance, and vegetation dynamics. Global BiogeochemicalCycles, 10(4):603–628, 1996. → pages 12, 15, 16, 19, 51, 53, 75, 101Forcorp Solutions Inc. Regional forest landscape assessment. upper athabascaregion. Technical report, Prepared for Forest Management Branch, AlbertaEnvironment and Sustainable Resource Development, Edmonton, Alberta,December 2012. → pages 47M. Friedl, A. Strahler, C. Schaaf, J. C. F. Hodges, and J. Salomon. BinaryMODIS MOD12C1 0.25 degree land cover climate modeler grid. Available athttp://duckwater.bu.edu/duckwater1/mod12c1 from the Department ofGeography, Boston University, Boston, Massachusetts, USA, 2001. → pages22, 52T. Y. Gan and E. Kerkhoven. Modeling the hydrology and possible impact ofclimate change of Athabasca River Basin by a Modified ISBA withGCM-scale data. In P. di Cenzo, editor, Proc. of 10th Annual ScientificMeeting of Mackenzie GEWEX Study (MAGS), pages 163–176, Vancouver,BC, Canada, 2004. → pages 5, 74GETECH. Global digital terrain model: Global DTM 5. Technical report,Department of Earth Sciences, University of Leeds, United Kingdom, 1995.→ pages 22, 53P. H. Gleick. A look at twenty-first century water resources development. WaterInternational, 25(1):127–138, 2000. → pages 96Golder Associates Ltd. Athabasca River water requirements for oil sandsoperations. Report 07-1345-0027, submitted to Oil Sands Developer Group,Calgary, AB, June 2009. → pages 98, 107129Golder Associates Ltd. Engineering mitigation options for meeting theAthabasca river water management framework. Report 07-1345-0027.5000,submitted to Oil Sands Developer Group, Environment and RegulatoryCommittee, Calgary, AB, February 2010. → pages 109P. Gosselin, S. E. Hrudey, M. A. Naeth, A. Plourde, R. Therrien, G. Van DerKraak, and Z. Xu. Environmental and health impacts of Canada’s oil sandsindustry. Technical report, The Royal Society of Canada Expert Panel, Ottawa,ON, December 2010. → pages 8, 9W. H. Green and G. A. Ampt. Studies on soil physics, 1. The flow of air andwater through soils. Journal of Agricultural Science, 4:1–24, 1911. → pages31M. Griffiths, A. Taylor, and D. Woynillowicz. Troubled waters, troubling trends:Technology and policy options to reduce water use in oil and oil sandsdevelopment in Alberta. Report 1st edition, The Pembina Institute, DraytonValley, AB, May 2006. → pages 57H. Hamilton, M. Thompson, and L. Corkum. Water quality overview ofAthabasca River Basin. Technical report, Alberta Environment PlanningDivision, Nanuk Engineering and Development, Cochrane, AB, August 231985. → pages 2, 4K. J. Harma, M. S. Johnson, and S. J. Cohen. Future water supply and demand inthe Okanagan Basin, British Columbia: A scenario-based analysis of multiple,interacting stressors. Water Resources Management, 26:667–689, 2012. →pages 118S. J. Hassol. Arctic climate impact assessment. Synthesis report, Arctic Counciland the International Arctic Science Committee, Canada, 2004. → pages 5L. E. Hay, R. L. Wilby, and G. H. Leavesley. A comparison of delta change anddownscaled GCM scenarios for three mountainous basins in the United States.Journal of the American Water Resources Association, 36(2):387–397, 2000.→ pages 78D. Healing. Oilsands reject halt to river water use. Calgary Herald, February2010. → pages 112130L. D. Hinzman, N. D. Bettez, W. R. Bolton, F. S. Chapin, M. B. Dyurgerov, C. L.Fastie, B. Griffith, R. D. Hollister, A. Hope, H. P. Huntington, A. M. Jensen,G. J. Jia, T. Jorgenson, D. L. Kane, D. R. Klein, G. Kofinas, A. H.Lynch, A. H.Lloyd, A. D. McGuire, F. E. Nelson, W. C. Oechel, T. E. Osterkamp, C. H.Racine, V. E. Romanovsky, R. S. Stone, D. A. Stow, M. Sturm, C. E. Tweedie,G. L. Vourlitis, M. D. Walker, D. A. Walker, P. J. Webber, J. M. Welker, K. S.Winker, and K. Yoshikawa. Evidence and implications of recent climatechange in northern Alaska and other arctic regions. Climatic Change, 72:251–298, 2005. → pages 94W. D. Holland and G. Coen. Ecological (biophysical) land classification of Banffand Jasper national parks. Report 44, Alberta Soil Survey, Alberta, 1983. →pages 28G. Holloway and S. Clare. Athabasca state of the watershed report: Phase 2.Report 1142, Prepared for the Athabasca Watershed Council by FieraBiological Consulting Ltd., Edmonton, AB, 2012. → pages 2, 7, 72C. Hopkinson and G. Young. The effect of glacier wastage on the flow of theBow River at Banff, Alberta, 1951-1993. Hydrological Processes, 12(10-11):1745–1762, 1998. → pages 6Imperial Oil Limited. Environmental responsibility. http://www.imperialoil.com/Canada-English/operations sands kearl environment.aspx, 2013. RetrievedJune 2014. → pages 112IPCC. Contribution of working group I to the fourth assessment report of theIntergovernmental Panel on Climate Change. In S. Solomon, D. Qin,M. Manning., Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, and H. L. Miller,editors, Climate Change 2007: The Physical Science Basis. CambridgeUniversity Press, Cambridge, United Kingdom and New York, NY, USA,2007. → pages 5R. Jordan. A one-dimensional temperature model for a snow cover. SpecialReport 16, U.S. Army Corps of Engineers, Cold Regions Research andEngineering Laboratory, 1991. → pages 30R. Kellerhals, C. R. Neill, and D. I. Bray. Hydraulic and geomorphiccharacteristics of rivers in Alberta, volume 72-1 of River engineering and131surface hydrology report. Alberta Cooperative Research Program in Highwayand River Engineering, Edmonton, AB, 1972. → pages ix, 23, 24, 32, 33, 54E. Kerkhoven and T. Y. Gan. A modified ISBA surface scheme for modeling thehydrology of Athabasca River Basin with GCM-scale data. Advances in WaterResources, 29:808–826, 2006. → pages 2, 4, 21, 36, 61, 101E. Kerkhoven and T. Y. Gan. Differences and sensitivities in potential hydrologicimpact of climate change to regional-scale Athabasca and Fraser River basinsof the leeward and windward sides of the Canadian Rocky Mountainsrespectively. Climatic Change, 106:583–607, 2011. → pages 5, 6, 72, 79, 80,94A. Kjearsgaard. Reconnaissance soil survey of the Tawatinaw map sheet. (83i).Report 29, Alberta Soil Survey, Alberta, 1973. → pages 28L. Knapik and J. Lindsay. Reconnaissance soil survey of the Iosegun Lake area,Alberta. Report 43, Alberta Soil Survey, Alberta, 1983. → pages 28J. Ko and W. F. Donahue. Drilling down: Groundwater risks imposed by in situoil sands development. Report, Water Matters Society of Alberta, Canmore,AB, July 2011. → pages 69T. Kowalcyk and F. Hicks. Observations of dynamic ice jam release on theAthabasca River at Fort McMurray, ab. In Proceedings of the 12th Workshopon River Ice, pages 369–392, Edmonton, Alberta, June 18-20 2003.Committee on River Ice Processes and the Environment, Hydrology Section,Canadian Geophysical Union. → pages 5C. J. Kucharik, J. A. Foley, C. Delire, V. A. Fisher, M. T. Coe, J. D. Lenters,C. Young-Molling, N. Ramankutty, J. M. Norman, and S. T. Gower. Testingthe performance of a dynamic global ecosystem model: Water balance, carbonbalance, and vegetation structure. Global Biogeochemical Cycles, 14(3):795–825, 2000. → pages 12, 15, 16, 18, 19, 21, 51, 52, 53, 75, 101S. Lapp, J. Byrne, I. Townshend, and S. Kienzle. Climate warming impacts onsnowpack accumulation in an alpine watershed. International Journal ofClimatology, 25:521–536, 2005. → pages 6, 92J. D. Lenters, M. T. Coe, and J. A. Foley. Surface water balance of thecontinental United States, 1963-1995: Regional evaluation of a terrestrial132biosphere model and the NCEP/NCAR reanalysis. Journal of GeophysicalResearch, 105(D17):22,393–22,425, 2000. → pages 15, 18, 51, 75, 101L. B. Leopold. A view of the river. Harvard University Press, Cambridge,Massachusetts, 1994. → pages 24, 54K. Y. Li, M. T. Coe, and N. Ramankutty. Investigation of hydrological variabilityin West Africa using land surface models. Journal of Climate, 18(16):2893–2908, 2005. → pages 15, 18, 31, 51, 75, 101Z. Li, W.-Z. Liu, X.-C. Zhang, and F.-L. Zheng. Impacts of land use change andclimate variability on hydrology in an agricultural catchment on the LoessPlateau of China. Journal of Hydrology, 377:35–42, 2009. → pages 39J. Lindsay, P. Heringa, S. Pawluk, and W. Odynsky. Exploratory soil survey ofAlberta map sheets 84C (east half), 84B, 84A and 74D. Report 58-1, AlbertaSoil Survey, Alberta, 1957. → pages 28J. Lindsay, A. Wynnyk, and W. Odynsky. Exploratory soil survey of Alberta mapsheets 83L, 83K, 83F and 83J. Report 64-2, Alberta Soil Survey, Alberta,1963. → pages 28J. Liu, D. T. Price, and J. Chen. Nitrogen controls on ecosystem carbonsequestration: a model implementation and application to Saskatchewan,Canada. Ecological Modelling, 86:178–195, 2005. → pages 18, 52, 75A. Locke and A. Paul. A desk-top method for establishing environmental flowsin Alberta rivers and streams. Technical report, Alberta Environment andAlberta Sustainable Resource Development, Edmonton, AB, April 2011. →pages 65, 87, 99R. Longley and B. Janz. The climatology of the Alberta oil sands environmentalresearch program study area. AOSERP Project ME 1.0, Alberta Oil SandsEnvironmental Research Program, Edmonton, AB, 1978. → pages 4M. Lynch-Stieglitz. The development and validation of a simple snow model forthe GISS GCM. Jounal of Climate, 7:1842–1855, 1994. → pages 30A. E. Mannix, C. Dridi, and W. L. Adamowicz. Water availability in the oil sandsunder projections of increasing demands and a changing climate: Anassessment of the Lower Athabasca water management framework (Phase 1).Canadian Water Resources Journal, 35(1):29–52, 2010. → pages 11, 50133D. Marks, J. Domingo, D. Susong, T. Link, and D. Garen. A spatially distributedenergy balance snowmelt model for application in mountain basins.Hydrological Processes, 13:1935–1959, 1999. → pages 30F. Mesinger, G. DiMego, E. Kalnay, K. Mitchell, P. C. Shafran, W. Ebisuzaki,D. Jovic´, J. Woollen, E. Rogers, E. H. Berbery, M. B. Ek, Y. Fan,R. Grumbine, W. Higgins, H. Li, Y. Lin, G. Manikin, D. Parrish, and W. Shi.North American Regional Reanalysis. Bulletin of the AmericanMeteorological Society, 87:343–360, 2006. → pages 43J. R. Miller, G. L. Russell, and G. Caliri. Continental-scale river flow in climatemodels. Journal of Climate, 7:914–928, 1994. → pages 33P. C. D. Milly, J. Betancourt, M. Falkenmark, R. M. Hirsch, Z. W. Kundzewicz,D. P. Lettenmaier, and R. J. Stouffer. Stationarity is dead: whither watermanagement? Science, 319:573–574, 2008. → pages 96D. N. Moriasi, J. G. Arnold, M. W. Van Liew, R. L. Bingner, R. D. Harmel, andT. L. Veith. Model evaluation guidelines for systematic quantification ofaccuracy in watershed simulations. Transactions of the ASABE, 50(3):885–900, 2007. → pages ix, x, 39, 43, 44, 45, 61, 62E. M. Morris. Snow and ice. In M. G. Anderson and T. P. Burt, editors,Hydrological Forecasting, pages 153–182. John Wiley, 1985. → pages 30R. H. Moss, J. A. Edmonds, K. A. Hibbard, M. R. Manning, S. K. Rose, D. P.van Vuuren, T. R. Carter, S. Emori, M. Kainuman, T. Kram, G. A. Meehl,J. F. B. Mitchell, N. Nakicenovic, K. Riahi, S. J. Smith, R. J. Stouffer, A. M.Thomson, J. P. Weyant, and T. J. Wilbanks. The next generation of scenariosfor climate change research and assessment. Nature, 463:747–756, 2010. →pages 76, 97MRRB. State of the aquatic ecosystem report 2003. Technical report, MackenzieRiver Basin Board, Fort Smith, NT, Canada, 2004. → pages 2, 7, 72N. Nakicenovic, J. Alcamo, G. Davis, B. de Vries, J. Fenhann, S. Gaffin,K. Gregory, A. Grbler, T. Y. Jung, T. Kram, E. Lebre La Rovere, L. Michaelis,S. Mori, T. Morita, W. Pepper, H. Pitcher, L. Price, K. Riahi, A. Roehrl, H.-H.Rogner, A. Sankovski, M. Schlesinger, P. Shukla, S. Smith, R. Swart, S. vanRooijen, N. Victor, and Z. Dadi. Emissions scenarios, a special report of134Working Group II of the Intergovernmental Panel on Climate Change. page570. Cambridge University Press, Cambridge, UK, 2000. → pages 80, 101National Energy Board. Canada’s oil sands: Opportunities and challenges to2015: An update. Energy market assessment, National Energy Board, Canada,Calgary, Alberta, June 2006. → pages 68Natural Resources Canada. Water use by the natural resources sector: Facts,March 2009. → pages 50, 59NOAA. National operational model archive and distribution system data access.http://nomads.ncdc.noaa.gov/data.php, December 2013. Retrieved September2011. → pages 21, 52D. Ohlson, G. Long, and T. Hatfield. Phase 2 framework committee report.Report, Cumulative Environmental Management Association (CEMA), FortMcMurray, Alberta, January 2010. → pages x, xi, 11, 50, 67, 93, 96, 97, 98,99, 100, 109, 111D. Panaˇ and R. Olson. Overview of uranium exploration work along the northernrim of the Athabasca Basin, northeastern Alberta. Open File Report 2009-19,Energy Resources Conservation Board, Alberta Geological Survey, Edmonton,Alberta, October 2009. → pages 7T. M. Pavelsky and L. C. Smith. Remote sensing of hydrologic recharge in thePeace-Athabasca Delta, Canada. Geophysical Research Letters, 35, 2008.doi:10.1029/2008GL033268. → pages 67D. L. Peters, T. D. Prowse, A. Pietroniron, and R. Leconte. Flood hydrology ofthe Peace-Athabasca Delta, northern Canada. Hydrological Processes, 20:4073–4096, 2006. → pages 5, 50D. L. Peters, W. A. Monk, and D. J. Baird. Cold-regions Hydrological Indicatorsof Change (CHIC) for ecological flow needs assessment. HydrologicalSciences Journal, 59(3-4):502–516, 2014. → pages 47A. Pietroniro, R. Leconte, B. Toght, D. L. Peters, N. Kouwen, F. M. Conly, andT. Prowse. Modelling climate change impacts in the Peace and Athabascacatchment and delta: III – integrated model assessment. HydrologicalProcesses, 20:4231–4245, 2006. → pages 6, 74135N. L. Poff, B. D. Richter, A. H. Arthington, S. E. Bunn, R. J. Naiman, E. Kendy,M. Acreman, C. Apse, B. P. Bledsoe, M. C. Freeman, J. Henriksen, R. B.Jacobson, J. G. Kennen, D. M. M. ad J. H. OKeeffe, J. D. Olden, K. Rogers,R. E. Tharme, and A. Warner. The ecological limits of hydrologic alteration(ELOHA): a new framework for developing regional environmental flowstandards. Freshwater Biology, 55:147–170, 2010. → pages 96N. L. P. Poff, J. D. Allan, M. B. Bain, J. R. Karr, K. L. Prestegaard, B. D. Richter,R. E. Sparks, and J. C. Stromberg. The natural flow regime: A paradigm forriver conservation and restoration. Bioscience, 47(11):769–784, 1997. →pages 96POLIS Project on Ecological Governance. The water-energy nexus: a westernCanadian perspective. Roundtable on water issues summary report, Universityof Victoria, Victoria, British Columbia, December 2012. → pages 122C. Prigent, E. Matthews, and F. A. ad W. B. Rossow. Remote sensing of globalwetland dynamics with multiple satellite data sets. Geophysical ResearchLetters, 28(24):4631–4634, 2001. → pages 46T. D. Prowse and S. Beltaos. Climatic control of river-ice hydrology.Hydrological Processes, 16:805–822, 2002. → pages 47T. D. Prowse, S. Beltaos, J. T. Gardner, J. J. Gibson, R. J. Granger, R. Leconte,D. L. Peters, A. Pietroniro, L. A. Romolo, and B. Toth. Climate change, flowregulation and landuse effects on the hydrology of the Peace-Athabasca-Slavesystem. Findings from the Northern Rivers Ecosystem Initiative.Environmental Monitoring and Assessment, 113(1-3):167–197, 2006. →pages 5, 6, 72, 74, 92, 94T. D. Prowse, B. R. Bonsal, C. R. Duguay, and M. P. Lacroix. River-icebreak-up/freeze-up: a review of climatic drivers, historical trends and futurepredictions. Annals of Glaciology, 46:443–451, 2007. → pages 47N. Ramankutty, J. A. Foley, J. Norman, and K. McSweeney. The globaldistribution of cultivable lands: current patterns and sensitivity to possibleclimate change. Global Ecology and Biogeography, 11:377–392, 2002. →pages 121136Regional Municipality of Wood Buffalo. Municipal census 2010. Technicalreport, Planning and Development Department, Fort McMurray, Alberta,2010. → pages 7B. D. Richter, J. V. Baumgartner, J. Powell, and D. P. Braun. A method forassessing hydrologic alteration within ecosystems. Conservation Biology, 10(4):1163–1174, 1996. → pages 13, 92B. D. Richter, R. Mathews, D. L. Harrison, and R. Wigington. Ecologicallysustainable water management: managing river flows for ecological integrity.Ecological Applications, 13(1):206–224, 2003. → pages 118D. Sauchyn and S. Kulshreshtha. Chapter 7 prairies. In D. S. Lemmen, F. J.Warren, J. Lacroix, and E. Bush, editors, From Impacts to Adapatation –Canada in a changing climate, pages 275–328, Ottawa, ON, 2008.Government of Canada. → pages 5, 6, 74, 94D. W. Schindler and W. F. Donahue. An impending water crisis in Canada’swestern prairie provinces. Proceedings of the National Academy of Sciences,103(19):7210–7216, 2006. → pages 5, 6, 74, 94D. W. Schindler, W. F. Donahue, and J. P. Thompson. Running out of steam? Oilsands development and water use in the Athabasca River-Watershed: Scienceand market based solutions. Report, University of Alberta, Edmonton, AB,May 2007. → pages 6, 8, 9, 11M. A. Schnorbus, K. E. Bennett, A. T. Werner, and A. J. Berland. Hydrologicimpacts of climate change in the Peace, Campbell, and Columbia watersheds,British Columbia, Canada: Hydrologic modelling project. Final Report 2,Pacific Climate Impacts Consortium, April 2011. → pages 94M. C. Serreze, J. E. Walsh, F. S. Chapin III, T. Osterkamp, M. Dyurgerove,V. Romanovsky, W. C. Oechel, J. Morison, T. Zhang, and R. G. Barry.Observation evidence of recent change in the northern high-latitudeenvironment. Climatic Change, 46:159–207, 2000. → pages 6, 94D. Shankar, V. Kotamraju, and S. R. Shetye. A quantitative framework forestimating water resources in India. Current Science, 86(4):543–552, 2004. →pages 19, 53137Shell Canada. Oil sands performance report.http://s00.static-shell.com/content/dam/shell-new/local/country/can/downloads/pdf/oil-sands/oil-sands-performance-report-2013.pdf, 2014.Retrieved June 2014. → pages 113A. J. Squires, C. H. Westbrook, and M. G. Dube. An approach for assessingcumulative effects in a model river, the Athabasca River basin. IntegratedEnvironmental Assessment and Management, 6:119–134, 2009. → pages 7, 74R. Srinivasan, X. Zhang, and J. Arnold. SWAT ungauged: hydrological budgetand crop yield predictionsin the Upper Mississippi river basin. Transactions ofthe ASABE, 53(5):1533–1546, 2010. → pages 39I. T. Stewart, D. R. Cayan, and M. D. Dettinger. Changes in snowmelt runofftiming in western North America under a ‘business as usual’ climate changescenario. Climatic Change, 62:217–232, 2004. → pages 93M. Stieglitz, A. Ducharne, R. Koster, and M. Suarez. The impact of detailedsnow physics on the simulation of snow cover and subsurface thermodynamicsat continental scales. Journal of Hydrometeorology, 2:228–242, 2001. →pages 30Suncor Energy Inc. Reclamation. In Suncor report on sustainability 2011.http://sustainability.suncor.com/2011/en/responsible/1794.aspx, 2011.Retrieved June 2014. → pages 112Suncor Energy Inc. Water. In Suncor report on sustainability 2013.http://sustainability.suncor.com/2013/en/environment/water.aspx, 2013a.Retrieved June 2014. → pages 113Suncor Energy Inc. Water management strategies. In Suncor report onsustainability 2013. http://sustainability.suncor.com/2013/en/environment/oil-sands-water-management-strategies.aspx, 2013b. Retrieved June 2014.→ pages 113B. Swainson. Rivers at risk: The status of environmental flows in Canada.Technical report, WWF Canada, 2009. → pages 6, 10, 50Syncrude Canada Ltd. Water management.http://www.syncrude.ca/users/folder.asp?FolderID=5911, 2012. RetrievedJune 2014. → pages 113138K. Taylor, R. Stouffer, and G. Meehl. An overview of CMIP5 and the experimentdesign. Bull. Amer. Meteor. Soc., 93:485–498, 2012. → pages 76M. Temimi, R. Leconte, N. Chaouch, P. Sukumal, R. Khanbilvardi, andF. Brissette. A combination of remote sensing data and topographic attributesfor the spatial and temporal monitoring of soil wetness. Journal of Hydrology,388:28–40, 2010. → pages 46The Oil Sands Developers Group. Oil sands project list.http://www.oilsandsdevelopers.ca/index.php/library/, April 2013. → pages 50,57, 93, 96K. Timoney. A dying delta? A case study of a wetland paradigm. Wetlands, 22(2):282–300, 2002. → pages 92Total E&P Canada. Water management.http://www.total-ep-canada.com/csr/water.asp, 2014. Retrieved June 2014. →pages 112B. Toth, A. Pietroniro, F. M. Conly, and N. Kouwen. Modelling climate changeimpacts in the Peace and Athabasca catchment and delta: I – hydrologicalmodel application. Hydrological Processes, 20:4197–4214, 2006. → pages 6,94L. Turchenek and J. Lindsay. Soils inventory of the Alberta oil sandsenvironmental research program study area. AOSERP report 122. Report 42,Alberta Soil Survey, Alberta, 1982. → pages 28J. A. Vano, J. A. Foley, C. J. Kucharik, and M. T. Coe. Evaluating the seasonaland interannual variations in water balance in northern Wisconsin using a landsurface model. Journal of Geophysical Research, 111(G02025):18, 2006. →pages 18, 31, 51C. J. Vo¨ro¨smarty, P. Green, J. Salisbury, and R. B. Lammers. Global waterresources: vulnerability from climate change and population growth. Science,289:284–288, 2000. → pages 96E. Watson and B. Luckman. Tree-ring based mass balance estimates for the past300 years at Peyto Glacier, Alberta, Canada. Quaternary Research, 62:9–18,2004. → pages 6139B. Weinhold. Alberta’s oil sands: Hard evidence, missing data, new promises.Environmental Health Perspectives, 119:126–131, 2011. → pages 8, 11B. B. Wolfe, T. L. Karst-Riddoch, S. R. Vardy, M. D. Falcone, R. I. Hall, andT. W. D. Edwards. Impacts of climate and river flooding on the hydro-ecologyof a floodplain basin, Peace-Athabasca Delta, Canada since A.D. 1700.Quaternary Research, 64:147–162, 2005. → pages 6, 74, 92B. B. Wolfe, R. I. Hall, W. D. Edwards, S. R. Vardy, M. D. Falcone,C. Sjunneskog, F. Sylvestre, S. McGowan, P. R. Leavitt, and P. van Driel.Hydroecological responses of the Athabasca Delta, Canada, to changes in riverflow and climate during the 20th century. Ecohydrology, 1:131–148, 2008. →pages 6, 74M. Woo and R. Thorne. Streamflow in the Mackenzie Basin. Arctic, 56(4):328–340, 2003. → pages 4, 36, 38, 50, 61D. Woynillowicz and C. Severson-Baker. Down to the last drop? The AthabascaRiver and oil sands. Oil Sands Issue Paper 1, The Pembina Institute, DraytonValley, AB, March 2006. → pages 11D. Woynillowicz, C. Severson-Baker, and M. Raynolds. Oil sands fever: theenvironmental implications of Canada’s oil sands rush. Technical report, ThePembina Institute, Drayton Valley, AB, November 2005. → pages 8, 9F. Wrona, J. Carey, B. Brownlee, and E. McCauley. Contaminant sources,distribution and fate in the Athabasca, Peace and Slave River Basins. J. Aquat.Ecosyst. Stress Recovery, 8:39–51, 2000. → pages 7A. Wynnyk, J. Lindsay, P. Heringa, and W. Odynsky. Exploratory soil survey ofAlberta map sheets 83O, 83P, and 73M. Report 64-1, Alberta Soil Survey,Alberta, 1963. → pages 28A. Wynnyk, J. Lindsay, and W. Odynsky. Soil survey of the Whitecourt andBarrhead area. Report 27, Alberta Soil Survey, Alberta, 1969. → pages 28W. Yuan, S. Liu, H. Liu, J. T. Randerson, G. Yu, and L. L. Tieszen. Impacts ofprecipitation seasonality and ecosystem types on evapotranspiration in theYukon River Basin, Alaska. Water Resources Research, 46(2), 2010.doi:10.1029/2009WR008119. → pages 18, 52140X. Zhang, K. Harvey, W. Hogg, and T. Yuzyk. Trends in Canadian streamflow.Water Resources Research, 37(4):987–998, 2001. → pages 6, 72, 74, 94141


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