Hydrological sensitivity of a northern mountain … · 2014-06-17 · Hydrological sensitivity of a...

HYDROLOGICAL PROCESSESHydrol. Process. (2014)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/hyp.10244

Hydrological sensitivity of a northern mountain basin toclimate change

Kabir Rasouli,1 John W. Pomeroy,1* J. Richard Janowicz,2 Sean K. Carey3 and Tyler J. Williams11 Centre for Hydrology, University of Saskatchewan, Saskatoon, SK, Canada

2 Yukon Environment Water Resources Branch, Whitehorse, YT, Canada3 School of Geography and Earth Sciences, McMaster University, Hamilton, ON, Canada

*Cof SE-m

Co

Abstract:

The hydrological sensitivity of a northern Canadian mountain basin to change in temperature and precipitation was examined. Aphysically based hydrological model was created and included important snow and frozen soil infiltration processes. The model wasdiscretized into hydrological response units in order to simulate snow accumulation and melt regimes and basin discharge. Modelparameters were drawn from scientific studies in the basin except for calibration of routing and drainage. The model was able to simulatesnow surveys and discharge measurements with very good accuracy. The forcing inputs of the hourly air temperatures and dailyprecipitationwere scaled linearly to examine themodel sensitivity to conditions included in a range of climate change scenarios: warmingof up to 5 °C and change in precipitation of +/� 20%. The results show that peak seasonal snow accumulation, snow season length,evapotranspiration, runoff, peak runoff, and the timing of peak runoff have a pronounced sensitivity to both warming and precipitationchange, where the impact of warming is partly compensated for by increased precipitation and dramatically enhanced by decreasedprecipitation. The snow regime, including peak snow accumulation, snow-free period, intercepted snow sublimation, and blowing snowtransport, wasmost sensitive to temperature, and the impact of awarming of 5 °C could not be compensated for by a precipitation increaseof 20%. However, basin discharge was more sensitive to precipitation, and the impact of warming could be compensated for by a slightincrease in precipitation. The impacts of 5 °Cwarmingwith a +/�20% change in precipitation resulted in snow accumulation, runoff, andpeak streamflow decreasing by from one half to one fifth and the snow-free period lengthening by from 46 to 60 days; in both cases, thesmaller change is associated with increased precipitation and the larger change with decreased precipitation. These results show thatmountain hydrology in Northern Canada is extremely sensitive to warming, that snow regime is more sensitive to warming thanstreamflow and that changes in precipitation can partly modulate this response. Copyright © 2014 John Wiley & Sons, Ltd.

KEY WORDS climate change impact; Yukon; sensitivity analysis; snow hydrology; cold regions; hydrological modelling

Received 11 November 2013; Accepted 6 May 2014

INTRODUCTION

Air temperatures in northern latitudes in general andspecifically in the Yukon Territory, Canada have increasedsignificantly within the last decades, while changes inprecipitation are more variable with summer precipitationgenerally increasing throughout Yukon, but winterprecipitation decreasing in southern Yukon and increasingin central and northern regions (Janowicz, 2010). Thesetemperature and precipitation changes are associated withchanges in the hydrological cycle (Bunbury and Gajewski,2012), such as the river ice cover period shortening due tolater freeze-up in the fall, and earlier ice break-ups in thespring and a greater frequency of mid-winter break-upevents with subsequent flooding (Janowicz, 2008, 2010). Awarming climate can result in permafrost degradation,expansion of shrub tundra (Tape et al., 2006), and altered

orrespondence to: JohnW. Pomeroy, Centre for Hydrology, Universityaskatchewan, Saskatoon, SK, Canada.ail: [email protected]

pyright © 2014 John Wiley & Sons, Ltd.

snow processes (Pomeroy et al., 2006), each of whichimpacts streamflow (St Jacques and Sauchyn, 2009). Thechanging shrub extent, which might enhance or delaymelting of permafrost in tundra regions (Sturm et al., 2005)and its relation to the surface energy balance alters albedofeedbacks to the climate system (Pomeroy et al., 2006).Studies show that river flow in winter and April has alsoincreased in Yukon and regions of significant permafrostwith ground thaw (Walvoord and Striegl, 2007). In April, atransition period between streamflow dominated bybaseflow and snowmelt runoff, groundwater contributionto streamflow has increased (Brabets and Walvoord, 2009).The changes in the ice/snow cover, permafrost, streamflow,soil moisture, vegetation, and other defining properties ofecosystems include important feedback processestransforming the ecology, surface energy balance, andhydrological cycle of northern Canada at local to globalscales (Osterkamp et al., 2009; Rawlins et al., 2009).In northern latitudes, alpine and shrub tundra snow is

redistributed by wind; the blowing snow transport and

K. RASOULI ET AL.

sublimation processes are affected by the interaction oflocal topography and landscape vegetation cover withregional wind flow patterns (Pomeroy et al., 1999;MacDonald et al., 2009). Snow is intercepted in needle-leaf forest canopies from which large proportions ofwinter snowfall sublimate rather than unload to thesurface (Pomeroy et al., 1999). High surface runoffderives from spring snowmelt and occurs as a result oflimited infiltration into frozen mineral soils at the time ofmelt and a relatively rapid release of water from meltingsnowpacks (Janowicz et al., 2003). Snowmelt timing andmelt rate are primarily controlled by the net inputs of solarradiation, longwave radiation, energy advected fromrainfall, and turbulent transfer of sensible and latent heat(Pomeroy et al., 2003). The impact of these inputs onsnowmelt are moderated by the storage of internal energyin the initially cold snowpacks and the snow surfacealbedo, both of which change rapidly in the pre-melt andmelt period. Snowmelt in shrub tundra is complicated bythe emergence and spring-up of tall tundra shrubs whichform a canopy over the snowpack changing the energyinputs to the underlying snow (Pomeroy et al., 2006;Bewley et al., 2010; Ménard et al., 2012). Meltwaterinfiltration into frozen soils can be restricted, limited, orunlimited depending on the degree of soil saturation byliquid water and ice and the over-winter development ofice layers on top of the soil (Janowicz, 2001; Janowiczet al., 2003). Frozen mineral soils usually have limitedinfiltration characteristics. The degree of saturation can beestimated from soil porosity, and the volumetric moisturecontent measured the preceding fall if overwinter soilmoisture changes are minimal. Substantial mid-wintermelts can create ice layers, which restrict infiltration suchthat most snowmelt forms overland flow (Gray et al.,2001). During summer, most rainfall infiltrates the poroussoils and then is withdrawn by plant roots for evapo-transpiration associated with the growth of mountaintundra and forests (Granger, 1999). Evapotranspirationoccurs quickly from wet surfaces such as water bodies,wetted plant canopies, and wet soil surfaces and relativelyslowly fromunsaturated surfaces such as bare soils and plantstomata (Granger and Gray, 1989). Any hydrological modelto be used for climate change assessment in this regionmustcorrectly address these hydrological processes.To assess the impact of climate change and variability

on the cold regions hydrological cycle, models require afull set of physically based representations of cold regionshydrological processes including direct and diffuseradiation to slopes, longwave radiation in complexterrain, intercepted snow, blowing snow, sub-canopyturbulent and radiative transfer, sublimation, energybalance snowmelt, infiltration to frozen and unfrozensoils, rainfall interception, evapotranspiration, sub-surfaceflow, depressional storage fill and spill, saturation excess

Copyright © 2014 John Wiley & Sons, Ltd.

overland flow and routing of surface, sub-surface, andstreamflow. The Cold Regions HydrologicalModelling platform (CRHM) (Pomeroy et al., 2007) is aflexible model assembly system that is primarilyphysically based and represents all of the above-mentioned cold region hydrological processes. CRHMhas been widely applied to studies in the continentalclimate of Western Canada in British Columbia, Yukon,the Canadian Rockies, the Canadian Prairies, NorthwestTerritories, and other cold regions in the Tibetan-QinghaiPlateau, Patagonia, the Pyrenees, and the Alps (Ellis andPomeroy, 2007; Dornes et al., 2008; Ellis et al., 2010;López-Moreno et al., 2012; Fang et al., 2013; Zhou et al.,2014). CRHM was evaluated in the recent SnoMIP2snow model intercomparison and performed relativelywell in modelling forest snowmelt at sites in Switzerland,USA, Finland, and Japan (Rutter et al., 2009).The difference between the spatial scales of climatic

and hydrological models makes coupling the two sets ofmodels for climate change analyses very challenging(Kite and Haberlandt, 1999). Hydrological modellinggenerally requires spatial resolutions of 0.1° latitude andlongitude or smaller (Salathé, 2003). The resolution of aGeneral Circulation Model (GCM), for instance the T63version of the Environment Canada CGCM3.1 model, is2.8°, which is much lower than needed for directapplication in a semi-distributed hydrological model suchas CRHM without downscaling. The GCM outputs alsoare not capable of capturing effects of the sub-gridfeatures such as orography, cloud, convection, and land-cover. Bennett et al. (2012) assessed hydroclimatologicalmodelling uncertainties in northwestern Canada andshowed that uncertainties from GCMs, emissionsscenarios and hydrological parameterizations were 84%,58%, and 31%, respectively, for winter in the 2050s inheadwaters of the Peace, Fraser and Campbell Rivers innortheastern British Columbia. Wilby and Harris (2006)found that simulations of low streamflows are very sensitiveto uncertainties in GCMs. Wilby (2005) has recommendedthat sensitivity analyses can help in quantifying hydrolog-ical uncertainty from various sources in climate-changeimpact studies. One method for doing this is the deltamethod, which examines the change in a variable per degreeCelsius of global warming (Fowler et al., 2007). While thismethod is very clear for hydrological response studies, theassumption of linear scaling of impact with temperature fornon-linear variables such as precipitation and for extremesintroduces uncertainties. An alternative method to deltamethod scaling in impact studies is the application ofclimate models projections to simulate future hydrology.This method can take advantage of a large number ofsimulated future climates calculated using various atmo-spheric and feedback assumptions (scenarios). However,projected changes in regional precipitation are very

Hydrol. Process. (2014)

HYDROLOGICAL SENSITIVITY OF A NORTHERN MOUNTAIN BASIN TO CLIMATE CHANGE

uncertain in future climate scenarios due to coarsesimulation of the synoptic dynamics that are responsibleformost precipitation events (IPCC, 2014) and the problemsof scale mismatch between hydrological and climatemodelsas noted above.In this study, an attempt is carried out to estimate

components of the hydrological cycle using a physicallybased model created using CRHM that will in turn be ableto assess the sensitivity of the hydrological response toclimate change in a Yukon mountain basin. The modelcan project changes to snow dynamics, precipitation, andtiming and magnitude of runoff. Given the uncertaintiesand to some extent disagreement between the climatemodel outputs for enhanced greenhouse gas scenariosover time, this paper examines the sensitivity ofhydrological responses to changes in air temperatureand precipitation by perturbing a recent measured timeseries of air temperature and precipitation up to the rangeof changes that are predicted by most climate modelsunder late 21st century scenarios of global change. Thislinear sensitivity analysis is a step in exploring the possiblealteration of the cold regions hydrological cycle inYukon byclimate change. The main purpose of this study is thereforeto highlight the combination of changes in temperature andprecipitation that are necessary to induce important changesin basin hydrology. Knowingwhat combination of warmingand precipitation change can induce hydrological change inthe North can help in assessing the threat to future watermanagement from climate change.

METHODS

Study area and data sources

Wolf Creek Research Basin (WCRB) drains 179 km2

of mountainous terrain in the headwaters of the YukonRiver in Canada (Figure 1). The ecological diversity ofthe basin combined with the available long-term compre-hensive hydrometeorological data (18 years) in each ofthree distinct ecosystems (boreal forest, shrub tundra, andalpine tundra) makes WCRB a suitable subject for thistype of climate change study. The elevation of the basinranges from 660 to 2080m with a generally north-easterlyaspect. Studies have shown slope and aspect to beimportant controls on runoff processes (Woo and Carey,1999; Pomeroy et al., 2003). The hydrology of WCRB isstrongly influenced by long winters (usually eightmonths) with small groundwater contributions tostreamflow, spring melt of the seasonal snowpack causinghigh streamflow, and low streamflow in the relativelyshort, warm, and dry summer (McCartney et al., 2006).Streamflow at the outlet of Wolf Creek is typical of amountainous subarctic regime with snowmelt-dominatedpeak flows in late May or early June and low flows in


March (Janowicz, 1999). Along the main branch of WolfCreek is Coal Lake (area 1 km2), which controls winterbaseflows. At low elevations, jack pine, white and blackspruce, and trembling aspen forest stands predominate(Francis et al., 1999). Above the treeline, shrub tundrawith birch and willow shrub heights from 30 cm to 2moccupies the majority of the basin (58%). At the highestelevations is an alpine tundra ecozone of bare rock andshort tundra moss and grass vegetation.The near-surface geology of WCRB consists primarily

of sedimentary rocks such as limestone, sandstone,siltstone, and conglomerate capped by a mantle of till atleast 2 m thick, as well as glaciofluvial andglaciolacustrine deposits (Seguin et al., 1998). Soils aregenerally well drained due to the coarse parent materials.Lower elevation forest soils are generally of clay to graveltexture, while upland soils are primarily of sandy loam togravely sandy loam texture. Surface organics commonlyextend to depths of 0.1m and are deepest in riparian areasand north-facing slopes (Zhang et al., 2010). Estimationsby Lewkowicz and Ednie (2004) suggest that 43% of thebasin contains permafrost which restricts the movementof water beneath the surface, particularly in the case ofsaturated frozen soils (Carey and Woo, 2001; Quintonet al., 2009).Hydrometeorological data have been collected from the

water year of 1993–1994 to 2010–2011 from three mainmeteorological stations in WCRB, one in each primaryecozone, and from four streamflow gauges (Figure 1,Table I). Hourly air temperature, relative humidity, windspeed, incoming shortwave radiation, and daily precipi-tation observations from above-canopy meteorologicalstations and streamflow data from hydrometric stationsare used in this study. Precipitation was measured bytipping bucket rain gauges, unshielded ‘BC stylestandpipe’ precipitation gauges, and Nipher-shieldedMeteorological Service of Canada (MSC) snowfallgauges. Nipher gauge solid precipitation measurementswere corrected using a wind undercatch correctionequation (Goodison et al., 1998) with wind speedsmeasured from nearby gauge-height anemometers. Gapsin data were infilled by establishing regression equationsfor meteorological variables between each of the threemeteorological stations and the Whitehorse Airport MSCstation located 15 km from WCRB. Because the basin isrelatively small (≈179 km2) and stations are close to eachother, meteorological variation is largely controlled byorography and hence changes in elevation rather thandistance between stations. Measurements of snow depthand density along snow survey transects were collected atleast monthly by Yukon Environment and universityresearchers at each of the three meteorological stations.These measurements and snow water equivalent (SWE)measured at a snow pillow located at the subalpine site


Figure 1. Wolf Creek Research Basin, Yukon, showing (a) elevation bands, (b) land cover, (c) slope, and (d) aspect representations used for defining (e)hydrological response units, HRUs in the different sub-basins (numbered). Green circles denote streamflow gauging stations, and triangles denote the

location of the meteorological stations. The basin drains to the north

K. RASOULI ET AL.

provide model diagnostic information for each ecozone(Pomeroy and Gray, 1995; Pomeroy et al., 1999).

Cold regions hydrological model for WCRB

Models created using CRHM can have a wide varietyof structures with differing levels of process detail andrepresentation including structures that are suitable for anorthern basin such as WCRB (Pomeroy et al., 2007).The CRHM system is very flexible and creates ‘objected-oriented’ models for particular basins, environments, andpredictive needs. Recent developments include optionsfor redistribution of alpine blowing snow (MacDonald,2010), an improved physical basis to soil moisture


accounting, fill-and-spill depressional storage (Fanget al., 2010), and enhanced forest canopy interceptionand radiation modules (Ellis et al., 2010). For this study, aCRHM model that accurately characterized the surfaceand near-surface cold regions hydrological processes ofWCRB was needed.A set of physically based process modules describing

the major processes can be combined into a modelinformed by results from previous modelling experimentsin the research basins such as WCRB (Pomeroy andGranger, 1999; Pomeroy et al., 2003, 2006; McCartneyet al., 2006; Carey et al., 2007, Dornes et al., 2008;Quinton and Carey, 2008, MacDonald et al., 2009).


Table I. Description of the main meteorological and hydrometric stations within Wolf Creek

Station EcozoneElevation(m a.s.l.) Location Site details

Meteorological stations (recorded over 1993–2011)Shrub tundra Shrub tundra 1250 60° 31.34′N, 135° 11.84′W East facing moderate slope,

shrub alder/willow to ~2m tallAlpine Tundra 1615 60° 34.04′N, 135° 8.98′W Windswept ridge top plateau,

sparse vegetationForest Boreal forest 750 60° 35.76′N, 134° 57.17′W Gently undulating terrain,

dense white spruce to heights of 20mWhitehorse Airport Boreal forest 706.2 60° 42.57′ N, 135° 04′ W Located in Whitehorse airport,

~15 km from the basin

Hydrometric gauges across the Wolf Creek BasinStation ID Drainage area (km2); record period

Upper Wolf Creek Alpine tundra 29AB006 60° 29.45′N, 135° 17.50′W 14.4; 1994–2011Coal Lake Outlet Shrub tundra 29AB005 60° 30.61′N, 135° 9.74′W 70.5; 1994–2011Granger Creek Shrub tundra 29AB007 60° 32.79′N, 135° 11.08′W 7.6; 1998–2011Alaska Hwy Boreal forest 29AB002 60° 36′N, 134° 57′W 179.0; 1993–2011


Modules were selected that could be run to robustlysimulate the hydrological cycle of the region in aprimarily physically based manner. Figure 2 shows theschematic setup of these modules, which include:

1) Observation module: reads and adjusts the meteoro-logical data with hydrological lapse rate, elevation,and wind-induced undercatch.

2) Radiation module: computes the theoretical globalradiation, direct and diffuse solar radiation, andmaximum sunshine hours based on latitude, elevation,ground slope, and azimuth (Garnier and Ohmura,1970).

3) Sunshine hour module: estimates sunshine hours fromincoming short-wave radiation and maximum sunshinehours.

4) Slope radiation module: adjusts the short-waveradiation on the slope with incoming short-waveradiation on a level surface.

5) Long-wave radiation module: calculates incominglong-wave radiation using short-wave radiation(Sicart et al., 2006).

6) Albedo module: estimates snow albedo throughoutthe winter and into the melt period and also indicatesthe beginning of melt (Verseghy, 1991).

7) Canopy module: estimates the snowfall and rainfallintercepted by and sublimated or evaporated from theforest canopy and unloaded or dripped from thecanopy, updates sub-canopy snowfall and rainfall,and calculates short-wave and long-wave sub-canopyradiation and turbulent transfer to snow. This modulehas options for open, small forest clearings, andforested environment (Ellis et al., 2010; 2013).

8) Blowing snowmodule: simulates the inter-hydrologicalresponse unit (HRU) wind redistribution of snow


transport and blowing snow sublimation losses through-out the winter period (Pomeroy and Li, 2000).

9) Energy balance snowmelt module: estimatessnowmelt by calculating the energy balance ofradiation, sensible heat, latent heat, ground heat,advection from rainfall, and change in internal energy(Marks et al., 1998).

10) All-wave radiation module: calculates the net all-waveradiation from short-wave radiation for snow-freeconditions (Granger and Gray, 1990).

11) Snowmelt infiltration module: This module has twoalgorithms for infiltration into frozen and unfrozen soils.One estimates snowmelt infiltration into frozen soilsusing a parametric equation describing the results of afinite difference heat and mass transfer model (Grayet al., 2001) and the other one estimates rainfallinfiltration into unfrozen soils with macropores usinga soil classification (Ayers, 1959). Surface runoffforms when snowmelt or rainfall exceeds the infiltra-tion rate.

12) Evaporation module: Granger’s evapotranspirationexpression (Granger and Gray, 1989; Granger andPomeroy, 1997) estimates actual evapotranspirationfrom unsaturated surfaces using an energy balanceand extension of Penman’s equation to unsaturatedconditions; the Priestley and Taylor evaporationexpression (Priestley and Taylor, 1972) estimatesevaporation from saturated surfaces such as streamchannels and lakes. Both evaporation algorithmsmodify moisture content in the canopy interceptionstore, ponded surface water store, and soil columnand by linking to the soil module are restricted by asoil water withdrawal relationship to ensure continu-ity of mass. The Priestley and Taylor evaporation alsoupdates moisture content for saturated surfaces.


Figure 2. Flowchart of physically based hydrological modules used in the CRHM model of WCRB

K. RASOULI ET AL.

13) Soil and hillslope module: estimates soil moisturebalance, depressional storage, surface/sub-surfaceflows in two soil layers and groundwater dischargein a groundwater layer, and interactions betweensurface flow and groundwater (Leavesley et al., 1983;Pomeroy et al., 2007; Dornes et al., 2008; Fang et al.,2010; 2013). The top layer is recharge layer, whichreceives infiltration from depressional storage, snow-melt, and rainfall. Evaporation takes water first fromcanopy interception and depressional storage and thenfrom the recharge layer or from both soil columnlayers via transpiration based on the plant availablesoil moisture and the rooting depth (Armstrong et al.,2010). Aquifer recharge occurs via percolation fromthe lower soil layer or directly from the surfacethrough macropores. Horizontal and vertical drainagefrom the soil and groundwater layers is regulated byDarcy’s flux, parameterized using Brooks andCorey’s (1964) relationship to estimate the actualhydraulic conductivity in unsaturated zone.

14) Routing module: Clark’s lag and route timing estima-tion method is used to route runoff (Clark, 1945).

The proposed physically based model operates on thespatially distributed control volumes of the HRU whichhave been found useful for modelling in basins wherethere is a good understanding of hydrological behaviour,but incomplete detailed sub-surface information to permita fully distributed fine scale modelling approach (Dorneset al., 2008). HRUs in a given basin are segregated based


on the land-cover, slope, aspect, and elevation utilizing adigital elevation model (DEM). The temporal resolutionof this CRHM model is hourly, while the spatialresolution is that of the HRU which vary from 0.92 km2

to 25.4 km2 in WCRB.

Model parameter estimation

Parameters were estimated from research basin resultsand those from other mountain modelling studies (SmokyRiver, Marmot Creek in Canada; Pomeroy et al., 2011;2013) with limited calibration for timing and routing ofthe streamflow. Blowing snow fetch distances of 500mwere found from the study of MacDonald et al. (2009)and used for the alpine and shrub tundra HRUs.Vegetation heights determined by field surveys wereused for the model simulations. Values of the blowingsnow redistribution factor were chosen as 2 and 5 for thealpine and shrub tundra HRUs, respectively (MacDonaldet al., 2009). Blowing snow was inhibited for the forestHRU. Initial soil saturation prior to the snowmeltinfiltration was estimated from both pre-melt volumetricsoil moisture content observations and soil porositymeasurements. The yearly antecedent soil moisture wasmeasured by water content reflectometer measurementsat each station. Soil types (organic and mineral soils) andtheir porosity values were measured by Carey and Woo(2005), Quinton et al. (2005) and Zhang et al. (2010).Initial soil temperature was measured by soil thermo-couples prior to the major snowmelt. The environmentcoefficient and surface saturation for frozen soil



infiltration were set up based on the recommended valuesfrom Gray et al. (2001). Infiltration opportunity time wascalculated by model simulation of snowmelt duration.Previous studies on soil properties in the Wolf Creek(Carey and Woo, 2001; Carey and Quinton, 2004; CareyandWoo, 2005; Quinton et al., 2005) found that the activesoil layer was composed of an upper organic soil layerover a mineral soil layer. Measured depth and porosityvalues for both organic and mineral soil layers were usedto approximate the capacity of the soil column layer. Thesoil recharge zone, a top layer of the soil column, wasassumed to be shallow and initially saturated.Modelling the complex processes controlling streamflow

generation in the model was relatively difficult due to thevarying magnitude and timing of the hydrologicalprocesses over short distances, aufeis formation, riverice formation and breakup, snow dammed channels, andthe presence of discontinuous permafrost slopes (Careyand Woo, 2001; McCartney et al., 2006; Boucher andCarey, 2010). Lag and route times in channels betweensub-basins were not measured by any study in WolfCreek. The capacity of surface depressional storage wasalso unknown and so both were calibrated. Subsurfacedrainage factors were thought to be affected by subsurfaceflow perched over thawing frozen soil which is not yetfully parameterized in CRHM and so were calibrated fromstreamflow. Table II shows the routing and drainageparameters calibrated from measured Wolf Creekstreamflow at the Alaska Highway using the dynamicallydimensioned search (DDS) approach (Tolson and Shoe-maker, 2007) and a calibration based upon 1000 modelruns over 3 years (1998–2001) for which measurementconfidence is relatively high. The DDS approach is a

Table II. List and range of the lag and route timing and drainageparameters calibrated using the DDS algorithm in

channels and HRUs

Parameter name UnitLowerbound

Upperbound

Basin-scale groundwaterlag time

h 0 100

Basin-scale runoff lag time h 0 100Groundwater storage constant day 0 50Groundwater lag time h 0 100Drainage rate fromgroundwater

mm/day 0 0.5

Aggregated storage constant day 0 50Aggregated lag time h 0 100Surface runoff storage constant day 0 50Surface runoff lag time h 0 100Drainage ratefrom depressionalstorage to groundwater

mm/day 0 0.5


heuristic search method developed to obtain an acceptableoptimal solution within a given number of modelevaluation runs. First, the DDS algorithm searches theparameter set globally, and then it looks for local solutionsas the number of model evaluations approaches thespecified try number. The transition from global to localsearch is carried out dynamically by reducing the size ofthe neighbourhood or dimensions of the model parame-ters tuned. The calibrated parameters are the lag and routetiming and storage constant, sub-surface drainage ratesand depressional storage of the HRUs. The range of theparameters was selected so that they were reasonable forthe study area based on the experience of the authors;the DDS search algorithm then optimized the parameterswithin this range. It is intended that more physicallyaccurate parameterisations of subsurface drainagewill become available, based on the studies by Quintonet al. (2005), but until that time, calibration of thesubsurface drainage is needed as per the approach ofDornes et al. (2008).

Model spatial configuration

A spatially distributed modeling structure (Figure 3)was developed with five sub-basins and 29 HRUs in WolfCreek basin, based on three elevation bands, fourgroupings of slope and aspect and three land coverecozones (Figure 1). A DEM with 30 m cell resolutionwas prepared by Dr. Lawrence Martz (Dept ofGeography, University of Saskatchewan). The extractedelevation, aspect, and slope were then intersected with thebasin land cover feature in ArcGIS, which generates theHRUs based on elevation, aspect, slope, and land cover(Figure 1). All the HRUs were categorized into one ofthree ecozone groups (Janowicz, 1999); alpine tundra,shrub tundra, or forest for parsimonious parameterization.Four streamflow gauges along Wolf Creek serve as theoutlet of sub-basins or the whole basin. An additionalsub-basin is Mid-Wolf Creek between Coal Lake and theforested ecozone in the lower basin. CRHM modellingstructures were replicated by ecozone to simulate thehydrological processes in the five sub-basins. A set ofphysically based modules were assembled for a number ofHRUs and the structure was replicated for each sub-basin.Each sub-basin then possessed the same module config-uration but was permitted varying parameter sets andvarying numbers of HRUs. Streamflow from each sub-basin was routed along the main channel of Wolf Creek.

Sensitivity analysis

After defining the model structure, parameters, and thecontrol volumes of computation on the CRHM platform,a detailed sensitivity assessment of hydrological responseto climate warming at the given basin could be carried out


Figure 3. The spatially distributed modeling structure for the model of WCRB. The five sub-basins are composed of various HRUs, and each HRUcontains a set of physically based hydrological modules as its internal structure. Clark routing (shown by the dashed line) routes flow from non-channel

HRUs to valley bottom HRU in each sub-basin and then from all five sub-basins (solid lines) to the basin output

K. RASOULI ET AL.

with climate change impact model experiments assessing:the effect of temperature and/or precipitation forcingchange on hydrological response assuming that the basinvegetation and permafrost do not change. The purpose ofthis methodology is to look at first-order impacts of climatechange on hydrological processes including snow regimeand runoffmagnitude and timing. Thismethodology is a firststep examination of response to changing climate beforecomprehensive transient impacts due to concurrentlychanging vegetation and permafrost can be studied. Forinstance, Ménard et al. (2014) showed for snowmelt in asmall portion of the catchment thatmountain topography canreduce the impact of changing shrubs on snow redistribu-tion, spring energy balance, and meltwater generation. Infuture work, second-order impacts of changing climate suchas transient changes in vegetation and permafrost on thehydrology of WCRB will be investigated.The range of the changes in precipitation and warming

considered for this study is partly based on the changesestimated by Representative Concentration Pathways(RCPs) which are an alternative to the conventionalclimate change scenarios (Moss et al., 2010) and wereused in the recent Fifth Assessment Report (AR5) of theInter-governmental Panel on Climate Change (IPCC,2014). Four new atmospheric composition pathwayscorresponding to specific radiative forcing values of 2.6to 8.5W/m2 were used as a basis for long-term and near-term modeling experiments in climate change studies.Based on RCP2.6 and RCP8.5 pathways, respectively, a


warming of up to 2 °Cwith an increase in annual precipitationof less than 10% and a warming of up to 5 °C with a 20%increase in annual precipitation are expected for the southernYukon region around Wolf Creek. Most modelled scenariosproject the future climate to be wetter, but there are somescenarios used in AR5 that show a regional decrease inprecipitation up to 15% for the 2080s. As hydrologicalimpacts in a northern basin are largely driven by changes towinter and spring climate conditions and modelled scenariosdiffer in their assessment of seasonality of climate change,seasonal variations in temperature and precipitation wereconsidered unnecessary for an initial climate sensitivitystudy. Based on this consideration and the range of climatescenarios reviewed, a temperature increase of up to 5 °C and arange of�20% to +20% change in precipitation are analysedin this paper.

RESULTS

Snow regime simulation

As snowmelt is the major runoff-producing event of theyear and streamflow forecasts are derived from seasonalmaximum snow accumulation estimates, simulating snowdynamics correctly is critical for hydrological impactassessments in Yukon. The CRHM model of WCRB wasevaluated against snow surveys at the three meteorolo-gical stations (corresponding to the three ecozones) andagainst basin streamflow at the outlet. The purpose of this



evaluation was to ascertain whether the model hadsufficient physical realism in its predictions to be usedfor climate change assessment, not to assess it as astreamflow prediction tool. Modelled snow accumulationfor each HRU was aggregated by the ecozone-weightedaverage (alpine, shrub tundra, and forest) and thencompared to the snow survey measurements of SWE(Figure 4) made by Yukon Environment—WaterResources and its federal predecessors. The modelledsnow accumulation is shown as SWE and is a result ofsnowfall, rainfall, blowing snow transport, sublimation,snow interception, canopy unloading, evaporation, andsnowmelt calculations and so assesses a major part of themodel and over half of the hydrological year.Assessment of model performance for snow accumula-

tion used the root mean square error (RMSE) and the meanabsolute error (MAE)—both in mm SWE. RMSE rangedfrom 35mm in the alpine to 29mm in the forest and MAEranged from 28mm in the alpine to 23mm in the forest.These statistics indicate model errors of less than 25% inestimating snow dynamics. As shown in Figure 4, themodel often overestimated snow accumulation at allelevations compared to snow surveys. The representative-ness of the snow surveys and their accuracy must beconsidered in assessing this. Errors in weighted snow tubemeasurements of SWE of 10% are common (Pomeroy andGray, 1995) when compared to snow pit gravimetricmeasurements of density and depth. As well, the alpine andshrub tundra snow surveys did not include a driftaccumulation zone. Drift accumulation zones are veryimportant to the basin-scale snow mass balance of WolfCreek (Pomeroy et al., 1999; 2003). Hence, the alpine and

0

100

200

300

0

100

200

300

0

100

200

300

SW

E (

mm

)

20001995

Tim

Figure 4. Comparisons of simulated and observed snow water equival


shrub tundra modelled snow accumulation will likelyexceed the measured when the measurements do notinclude drift areas. Overestimation of snow accumulationin the forest may be contributed to by differences betweenthe aggregated forest HRU discontinuous canopy structureand that of a single snow survey transect in a continuouswhite spruce forest near the forest meteorological station.The snow survey line in the forest may represent a moredense forest canopy than is typical of the aggregated forestHRUs. Overall, the model seemed capable of simulatingthe SWE regime and the timing of peak snow accumulationand initiation of snowmelt well.

Streamflow simulation

The model was calibrated against the gauged flows ofWolf Creek at the Alaska Highway for the period of1998–2001 (Figure 5). To assess the multiple yearperformance of the model, flows were simulated from1993 to 2011 and compared to streamflow observations(Figure 5). The RMSE, MAE, and Nash–Sutcliff statistic(NS) for the calibration period are 0.72m3/s, 0.44m3/s,and 0.49 and for the full 18-year hourly simulation are0.88m3/s, 0.49m3/s, and 0.15, respectively. The errorrange and bias are similar between calibration andsimulation periods, but the NS statistic degrades, showingweaker hydrograph simulation over the simulation period.The acceptable performance of the model for bothecozone-scale SWE and basin-scale streamflow simula-tions given the minimal calibration of model parameters,along with the strong physical basis of its structure,suggests that this model can be used for climate changesensitivity simulations of snowpack and streamflow.

20102005

e (year)

ent in the three ecozones of WCRB: alpine, shrub tundra, and forest


20

15

10

5

0

1995 2000 2005 2010

Wol

f Cre

ek D

isch

arge

at

Ala

ska

Hig

hway

(m

3 /s)

Time (year)

Figure 5. Observed (points) and simulated (line) streamflow discharge ofWolf Creek at the Alaska Highway, Oct 1993 to Sep 2011. The period of

1998–2001 was used for calibration of parameters listed in Table II

K. RASOULI ET AL.

Impact of changing temperature and precipitation onsnow regime

The sensitivity analysis involved raising air tempera-tures by one-degree increments for each hourly intervalup to 5 °C and then altering daily precipitation eitherupwards or downwards by a multiplier for ±10% and±20%. The impact of raising hourly air temperatures by1 °C increments up to 5 °C on the seasonal snow regime isshown in Figure 6 for the alpine, shrub tundra, and forestecozones from October 1993 to September 2011. Climatewarming of up to 5 °C decreased the simulated annualpeak snowpacks from 30% to 45%. There are substantial

300

200

100

0300

200

100

0300

200

100

0

1995 2000

SW

E (

mm

)

Tim

Figure 6. Snow accumulation regimes for alpine, shrub tundra, and foresttemperatures b


declines in snow accumulation and duration of snowseason with increasing temperature over most winters,with some seasons (e.g. 2002–2003) showing largesensitivities and some (e.g. 1997–1998) almost nosensitivity to temperature increases. Warming affects thephase of precipitation, causing a shift from snowfall torainfall in the spring and fall transition seasons, and socauses the snow season to start later and end earlier. Italso can accelerate the initiation of snowmelt, which insome years slowed the melt rate as the melt period wasshifted forward into a lower solar irradiance period.Despite this effect, in most years snowmelt ended earlieras temperatures increased.The peak SWE was calculated as the maximum SWE

over the year; this generally occurs in April or May ofeach calendar year. The peak SWE and how it is affectedby climate warming is shown in Figure 7 for the shrubtundra ecozone. In some years, there are very strongdecreases in peak SWE with increasing temperature(e.g. 2000, 2005, 2006), whereas in other years thereis very little decrease (1998, 2002, 2007). Themost sensitiveyears are the wetter years as these often include latespring snowfall that occurs under relatively warmconditions and so is most vulnerable to phase changewith warming.A sensitivity analysis with warming air temperatures

(up to 5 °C) and both increasing (+10%, +20%) anddecreasing (�10%, �20%) precipitation was performedto assess whether changes in precipitation would dampen

2005 2010

e (year)

ecozones, under current climate and incremental warming of hourly airy up to 5 °C


300

200

100

0

1995 2000 2005 2010

Shr

ub tu

ndra

SW

E (

mm

)

Time (year)

Figure 7. Peak annual snow accumulation (SWE) for the shrub tundra ecozone; current and with warming of hourly air temperatures of up to 5 °C


or magnify the effect of warming climate on WCRB. Theoverall response of peak SWE magnitude and timing ineach ecozone to changing precipitation with warming isshown in Figure 8. Peak SWE declined markedly withincreasing temperature and decreasing precipitation in allecozones, with a slightly stronger response to precipita-tion in the alpine, a slightly stronger response totemperature in the shrub tundra and a fairly equalresponse to both temperature and precipitation in theforest ecozone. The response of peak SWE to changingprecipitation with moderate warming (2 °C) was non-linear suggesting sensitivity to the combination of

annual peak SWE (mm)

012345

change of annual pe

012345

012345

012345

Tem

pera

ture

incr

ease

(°C

)

80 90 100 110 120012345

Precipitation80 90 100

012345

Figure 8. Change in peak annual snow accumulation (SWE) magnitude, pewarming of hourly air temperatures and changes to daily precipitation for s

Negative signs in peak SWE date plot represent advance


warming and drying and relative insensitivity to warmingand wetting. For instance, with 2 °C warming, a 20%decrease in precipitation caused a 35% reduction in peakSWE, and a 10% decrease in precipitation caused a 26%reduction, but a 20% increase in precipitation caused onlya 9% increase in peak SWE. The timing of peak SWEadvanced dramatically with increases in temperature andwas only impacted by increases or decreases inprecipitation in the forest and shrub tundra ecozones.Besides a general advance in the date of peak SWE withcombined warming and drying, there was no clear patternin the alpine ecozone due to the complex interaction of

ak SWE (%)

(%)110 120

change of annual peak SWE date (day)

012345

012345

80 90 100 110 120012345

Alpine

Shrub tundra

Forest

rcentage, and date, at the alpine, shrub tundra, and forest ecozones underimulations using measured or perturbed meteorology from 1993 to 2011.in time and positives signs denote delay in peak SWEs


K. RASOULI ET AL.

blowing snow transport from the alpine snowpack withair temperature and precipitation, suggesting that thetiming of peak SWE in the alpine is insensitive to themost likely scenarios of warming and wetting climate.For increased precipitation without warming, the peak

SWE responded linearly in each ecozone with a 22%increase for a 20% precipitation increase. The shrubtundra snow regime showed the greatest sensitivity to thesimultaneous drying and warming; a 20% decrease inprecipitation along with 5 °C warming resulted in a 52%decline in SWE (from 162mm to 84mm) and a 25-dayadvance in timing. The high sensitivity of the shrubtundra snow regime to warming temperatures is contrib-uted to by the diminished redistribution of blowing snowfrom alpine to shrub tundra as the temperature warms(Figure 9). The relatively low sensitivity of the alpinepeak SWE date to temperature is due to the colder climateat high elevations. Warming temperatures in the relativelycold alpine zone cause only a small reduction in snowfallbut can cause a substantial reduction in blowing snowtransport from the alpine zone. Reduced snowfall andreduced snow erosion compensate for each other to somedegree, but compensation will vary with the meteorolog-ical characteristics of each snowstorm and snowfall event.The relatively lower sensitivity of the forest peak snowaccumulation to warming is due to the temperature-sensitive intercepted snow unloading processes in forestscounteracting reduced snowfall at higher temperatures;warmer temperatures lead to greater unloading ofintercepted snow and lesser sublimation loss (Gelfanet al., 2004; Figure 9). A fundamental question regardingthe impact of climate change on snowpack is whether themaximum likely increase in precipitation of 20% cancompensate for the reduced peak SWE caused by themaximum likely warming. This is explored in Figure 10

0

50

100

150

200

420

Temperature increase (°C)

Var

iabl

e (m

m/y

ear)

Figure 9. Changes in actual evapotranspiration for the basin, interceptedsnow sublimation in forest HRUs and blowing snow transport from alpineto shrub tundra ecozone with warming. Intercepted snow sublimation

decreases with warming of 5 °C from 21 to 15mm/year


where the cumulative distribution of peak SWE in eachecozone shows that the impact of a 20% increase inprecipitation can compensate for that of 5 °C of warmingon peak SWE only in the highest 1–2% of snowfall years—for lower snowfall years it is less likely that increases inprecipitation can compensate for the impacts of increasedwarming on peak SWE.Table III summarizes the basin scale magnitude and

timing of the peak SWE, snow season initiation, end andlength, peak flow, and total annual runoff for currentclimatic conditions and for scenarios of warming andchanges in precipitation. The average peak SWE in thebasin declined (55%) from 136mm to 61mm with 5 °Cwarming and a 20% decline in precipitation and increased(24%) to 169mm without warming and with 20%increase in precipitation. The most likely scenario of 5 °C warming and 20% increased precipitation resulted in18% decrease in annual peak SWE. With 5 °C warmingand no changes in precipitation winters in WCRB arepredicted to begin 17 days later and end 37 days earlierthan current winters; a concomitant 20% increase inprecipitation would lengthen the winter season by onlyone week. With warming of only 2 °C, increasedprecipitation of as little as 10% was able to compensatefor the effect of warming and further increases inprecipitation resulted in increased peak SWE. With 5 °Cwarming, peak snowpack decreased in all scenarios, evenwith increased precipitation of 20%, which is consistentwith the frequency distribution results from Figure 10.

Impact of changing temperature and precipitation onstreamflow

Streamflow simulations are sensitive to changes in thewhole range of hydrological processes and water balancecomponents as integrated by the basin. The daily runofffrequency curves for low and high runoffs of Wolf Creek(at the Alaska Highway) are shown in Figure 10.Streamflow exhibits a very different response to thechanges in precipitation under warming from that of thehistorical snowmelt driven regime. Under all temperatureincreases concomitant with precipitation increases of20%, medium and low flows in Wolf Creek are greaterthan those the current regime. This persists for high flowsas well, except with 5 °C of warming, for which low flowsincreased but high flows decreased compared to runoffunder the current climate; this dampening of hydrologicalvariability under a very likely climate scenario is of greatimportance to water resources management. The overallchange in annual runoff over the 18 years of simulation isshown in Figure 11. Change in runoff contrasts with thechange in peak SWE (Figure 8) in that runoff is muchmore sensitive to precipitation than to temperaturechange. A 1 °C increase in temperature resulted in a 4%


0 50 100 150 200 250 3000

0.2

0.4

0.6

0.8

1

SWE (mm)

0 50 100 150 200 250 300 350 4000

0.2

0.4

0.6

0.8

1

SWE (mm)

0 50 100 150 200 2500

0.2

0.4

0.6

0.8

1

SWE (mm)

cum

ulat

ive

dist

ribut

ion

func

tion

0 0.5 1 1.5 2 2.5 3 3.5 40

0.2

0.4

0.6

0.8

1

Runoff (mm/day)

P: 120% ; T: 0P: 120% ; T: 1P: 120% ; T: 2P: 120% ; T: 3P: 120% ; T: 4P: 120% ; T: 5P: 100% ; T: 0

Shrub tundra Alpine

Forest Basin

Figure 10. Cumulative distribution of peak SWE in each ecozone and daily runoff for Wolf Creek at the Alaska Highway under the current climate andwith precipitation increase of 20% and temperature warming of up to 5 °C

Table III. The area weighted average of SWE and discharge at the Wolf Creek outlet and their different timing in the hydrological yearstarting 1 October and ending 30 September for different scenarios of climate warming and changes in precipitation

Variable Index Unit

T:0 °C T:2 °C T:5 °C T:0 °C T:5 °C T:5 °C

P:100% P:100% P:100% P:120% P:80% P:120%

SWE Peak mm 135.9 117.3 84.8 169.4 61.0 107.4Accumulation start Day of year 5 7 22 4 27 18Timing of peak Day of year 167 163 154 172 148 155Timing of snow free Day of year 271 254 234 273 230 237Length of snow season Day 265 248 212 269 202 219

Discharge Annual runoff mm 171 160 147 235 95 205Annual peak rate m3/s 4.1 3.6 2.9 5.3 2.1 3.7Timing of rising flow Day of year 160 139 119 161 94 119Timing of peak flow Day of year 263 244 236 265 232 235Return to baseflow Day of year 353 337 343 357 340 336Summer flow length Day 193 197 225 196 246 217


decrease in annual runoff and grew to a 14% decrease fora 5 °C temperature increase. The combination of 5 °Cwarming and 20% decreased precipitation reduced annulrunoff to 96mm/year, 44% less than the current value(171mm/year, Figure 11); however, 5 °C with 20%increased precipitation increased runoff by 20% to205mm/year. The sensitivity of runoff to temperature isbecause of the longer snow-free season and increasedopportunity and energy for evapotranspiration withincreasing temperature. This is illustrated by the 45-mmincrease in annual actual evapotranspiration with 5 °C oftemperature rise (Figure 9). It is interesting to note that an


8% increase in precipitation is necessary to compensate forthe impact of 5 °C warming on runoff from Wolf Creek.Figure 12 shows that peak runoff rates increase

proportionately greater than do changes in precipitation,demonstrating a high sensitivity of runoff to changes inprecipitation. Peak runoff timing advanced under mostcombinations of changed precipitation and warming(Figure 12). Peak runoff was more sensitive to temper-ature changes than was annual runoff. For the combina-tion of 2 °C warming and 20% increased precipitation,runoff timing advanced 4 days and peak rate increased by11%; however, for 5 °C warming and 20% increased


100

120

140

160

180

200

220

annual runoff (mm)

Tem

pera

ture

incr

ease

(°C

)

80 90 100 110 1200

1

2

3

4

5

−40

−30

−20

−10 0 10 20

30

change of annual runoff (%)

Precipitation (%)

80 90 100 110 1200

1

2

3

4

5

100 120 140 160 180 200 220 −40 −30 −20 −10 0 10 20 30

Figure 11. Change in mean annual runoff of Wolf Creek at the Alaska Highway with increases in hourly air temperature and changes in dailyprecipitation. Mean is taken from simulations using measured and perturbed meteorology from 1993 to 2011

1.2

1.4

1.6

1.8

2

2.2

2.4

annual peak runoff (mm/day)

Tem

pera

ture

incr

ease

(°C

)

80 90 100 110 1200

1

2

3

4

5

−40

−30

−20

−10

0

10

20change of annual peak

Precipitation (%)

80 90 100 110 1200

1

2

3

4

5

−40 −20 0 201.2 1.4 1.6 1.8 2 2.2 2.4

−30

−25

−20

−15

−10

−5

0

change of annual peak

80 90 100 110 1200

1

2

3

4

5

−30 −20 −10 0

runoff (%) runoff date (day)

Figure 12. Annual peak runoff rate, percentage, and date change for Wolf Creek at the Alaska Highway with increases in hourly air temperature andchanges to daily precipitation. Means are taken from simulations using measured or perturbed meteorology from 1993 to 2011. Negative signs in peak

timing represent an advance in timing and positive signs a delay in timing of peak runoff

K. RASOULI ET AL.

precipitation timing advanced 27 days and peak ratedeclined by 10%. This suggests that increased and earlierhigh flow potential is possible with moderate warmingand maximum wetting but not with maximum warmingwhich produces much earlier peak flow timing andreduced peak flows. Drying under 2 °C warming caused asubstantial reduction in peak flow; precipitation de-creases (10–20%) caused decreases in peak runoff rates(25–35%). Overall, increases in temperature tended toreduce peak flows by desynchronizing melt throughaccelerating the timing of spring flows (Table III) andreducing summer and fall flows as actual evapotranspi-


ration increases in the longer snow-free period. Theimpact on peak flow rates of warming by 2 °C and 4 °Ccan be compensated by increases in precipitation of 10%and 20%, respectively; however, even with a 20%increase in precipitation, peak flow rates decreased aswarming increased from 4° to 5 °C.

Synthesis

The average annual elevation-corrected precipitation ofWCRB is 334mm over 1993–2011, suggesting that runoffefficiency would only decline from 0.51 to 0.45 with 5 °C ofwarming. Conversely to thismodest temperature sensitivity,



runoff was more sensitive than the snow regime toprecipitation change and the impact of up to 5 °C ofwarming on annual runoff could be compensated for by anincrease in precipitation of less than 8%. The ambiguity inthe direction of change in precipitation causes greatuncertainty in the future hydrology of WCRB. If a 5 °Ctemperature increase is assumed, then this could cause a44% decrease in runoff if precipitation declines by 20%, a14% decrease if precipitation does not change, and a 20%increase in runoff if precipitation increases by 20%. Thecombination of climate warming and decreased precipita-tion could cause dramatic declines in runoff, but ifprecipitation increased as many scenarios suggest,then there could be some compensation. For instance, forthe not-unlikely scenario of precipitation increasing by 20%with 4 °C of warming, then runoff increases 23% and peakrunoff rates are unchanged, although the timing of peakrunoff is advanced by 20 days.The changes in Wolf Creek streamflow are largely

caused by the consequences of the decrease in depth andadvance in timing of peak snowpack with climate change.The combination of 5 °C warming and 20% decreasedprecipitation cause modelled declines in peak SWE of55% and the snow-covered period of 24%, resulting inincreased evapotranspiration of 30% and decreased inannual runoff and peak runoff rates of 42% and 44%,respectively. This suggests that WCRB hydrology has astrong sensitivity to a ‘loss of cold’ that is connected tolarge decreases in snowpack with warming temperatures.The greatest hydrological sensitivity and potential fordecline with increasing temperature is for peak snowaccumulation, peak runoff, and their timing. Theproportional drop in peak snow accumulation was closelyreflected in the proportional drop in annual runoff,highlighting the importance of correctly estimatingchanges in snowpack when assessing changing hydrologyin the region.

CONCLUSIONS

WCRB is Yukon’s most intensively studied headwaterbasin and has an 18-year archive of high altitude weather,snowpack, soils, and streamflow data that has beencollated and corrected. A comprehensive cold regionshydrological process model developed using the CRHMPlatform provides the basis for water balance estimatesand climate sensitivity studies in a characteristic Yukonheadwaters basin. Using parameters derived from theresults of local scientific research, the model was able toadequately predict snow accumulation and melt in alpine,shrub tundra and forest zones, and basin streamflow overthe simulation period. Perturbing the model withincreases in hourly air temperatures from 1 to 5 °C, andboth increases (to +20%) and decreases (to �20%) in


daily precipitation provided the basis for a hydrologicalsensitivity study to possible ranges of climate change.The simulated changes show that snowpacks are verysensitive to warming in wet years and less sensitive indrier years; this is mostly due to spring contributions towet year precipitation and the phase sensitivity ofrelatively warm spring snowfall (Harder and Pomeroy,2014). On average, a 5 °C warming with no change inprecipitation caused a 38% decline in maximum snowaccumulation. The impact of 2 °C warming on snowcould be fully compensated for by precipitation increas-ing by 20%, but greater warming (>3 °C) cannot becompensated with precipitation increases of thismagnitude. If precipitation decreases with warming,the impacts on snowpack multiply and a decline of 55%of peak SWE could occur with both 5 °C warmingand 20% decline in precipitation; though this is not alikely climate scenario, it would have catastrophicimplications for winter ecology and snow-basedtransportation. The smaller snowpacks and warmerweather would cause an increase in the snow-free periodby 52 days. These changes would also expand theevapotranspiration season, increasing annual evapo-transpiration loss by 30% and the importance ofrainfall–runoff mechanisms.The impacts from 5 °C warming with a 20% decrease

in precipitation on the hydrology of Wolf Creek would bevery severe and result in snow accumulation, streamflowvolume and peak runoff dropping approximately in half,peak flow timing advancing by one month and the snow-free period lengthening by 2months. This contrastsstrongly with the attenuated impact from the samewarming and an increase in precipitation of 20% inwhich snow accumulation decreases by 21%, streamflowvolume increases by 20% and peak runoff decreasesslightly whilst the snow-free period lengthens by 46 days.These results show that while the impacts of warming oncold regions hydrological processes are unequivocal—reduced snow contribution to streamflow, shorter snow-covered period, and greater evapotranspiration; themagnitude and direction of the impact of warming onstreamflow hydrology will depend on changes toprecipitation. Therefore, both warming and changes toprecipitation must be considered to evaluate futurehydrology in Yukon. The model developed here can beapplied to other Northern regions to assess the sensitivityto warming in environments with different climate, andvegetation conditions. It is considered most reliable for itstreatment of climate change impact on snow hydrology.However, improvements to the treatment of frozenground and inclusion of the transient response ofpermafrost and vegetation to changing climate wouldadd great insight to the impact of local feedbacks to theclimate system on hydrology. The extension of the


K. RASOULI ET AL.

sensitivity analysis to seasonal temperature changes andchanges in the intensity, duration, and frequency ofprecipitation as well as to its seasonality would reducethe uncertainty due to precipitation in the results. Inthe meantime, the results of this study may inform theearly development of adaptation strategies and optionswhich may include improving flood forecasting andwarning systems, infrastructure design modification,land zoning changes, and changes to policy, regulation,and legislation.

ACKNOWLEDGEMENTS

The authors wish to acknowledge two decades of WolfCreek Research Basin operation with substantialcontributions to installation, maintenance and operationof the core snow surveys, meteorological stations andhydrometric stations under inclement conditions fromGlenn Ford, Glen Carpenter, Kerry Paslowski, MartinJasek, Dell Bayne, Newell Hedstrom, Raoul Granger,and Richard Essery amongst many others. Modellingsupport from Tom Brown and Xing Fang are greatlyappreciated. Funding for the basin has come from YukonEnvironment, Environment Canada, DIAND, NSERC,CFCAS, NERC, NOAA, and other sources. Funding forthis study came from AAND Canada, Yukon Environment,and NSERC.

REFERENCES

Armstrong RN, Pomeroy JW, Martz LW. 2010. Estimating evaporation ina Prairie landscape under drought conditions. Canadian WaterResources Journal 35: 173–186.

Ayers HD. 1959. Influence of soil profile and vegetation characteristics onnet rainfall supply to Runoff. In Proceedings of Hydrology SymposiumNo.1: Spillway Design Floods. National Research Council Canada:Ottawa; 198–205.

Bennett, KE, Werner, AT, Schnorbus, M. 2012. Uncertainties inhydrologic and climate change impact analyses in headwater basinsof British Columbia Journal of Climate 25(17): 5711–5730.

Bewley D, Essery R, Pomeroy J, Ménard C. 2010. Measurements andmodelling of snowmelt and turbulent heat fluxes over shrub tundra.Hydrology and Earth System Sciences 14: 1331–1340.

Brabets TP, Walvoord MA. 2009. Trends in streamflow in the YukonRiver Basin from 1944 to 2005 and the influence of the Pacific DecadalOscillation. Journal of Hydrology 371: 108–119. DOI: 10.1016/j.jhydrol.2009.03.018

Brooks RH, Corey AT. 1964. Hydraulic properties of porous media.Colorado State University: Fort Collins, Colorado; 27.

Bunbury J, Gajewski K. 2012. Temperatures of the past 2000 yearsinferred from lake sediments, southwest Yukon Territory, Canada.Quaternary Research 77: 355–367.

Boucher JL, Carey SK. 2010. Exploring the use of chemical, isotopic andhydrometric data in a discontinuous permafrost catchment. HydrologyResearch 41: 508–519. DOI: 10.2166/nh.2010.146

Carey SK, Quinton WL, Goeller NT. 2007. Field and laboratory estimatesof pore size properties and hydraulic characteristics for subarcticorganic soils. Hydrological Processes 21: 2560–2571. DOI: 10.1002/hyp.6795

Carey SK, Woo MK. 2005. Freezing of subarctic hillslopes, Wolf CreekBasin, Yukon. Arctic, Antarctic, and Alpine Research 37: 1–10.


Carey SK, Quinton WL. 2004. Evaluating snowmelt runoff generation in adiscontinuous permafrost catchment using stable isotope,hydrochemical and hydrometric data. Nordic Hydrology 35: 309–324.

Carey SK, Woo MK. 2001. Slope runoff processes and flow generation ina subarctic, subalpine catchment. Journal of Hydrology 253: 110–119.

Clark CO. 1945. Storage and the unit hydrograph. Proceedings of theAmerican Society of Civil Engineering 69: 1419–1447.

Dornes PF, Pomeroy JW, Pietroniro A, Carey SK, Quinton WL. 2008.Influence of landscape aggregation in modelling snow-cover ablationand snowmelt runoff in a sub-arctic mountainous environment.Hydrological Sciences Journal 53: 725–740.

Ellis CR, Pomeroy JW, Link TE. 2013. Modeling increases in snowmeltyield and desynchronization resulting from forest gap-thinningtreatments in a northern mountain headwater basin. Water ResourcesResearch 49: 1–14. DOI: 10.1002/wrcr.20089

Ellis CR, Pomeroy JW, Brown T, MacDonald J. 2010. Simulation ofsnow accumulation and melt in needleleaf forest environments.Hydrology and Earth System Sciences 14: 925–940. DOI: 10.5194/hess-14-925-2010

Ellis CR, Pomeroy JW. 2007. Estimating sub-canopy shortwaveirradiance to melting snow on forested slopes. Hydrological Processes21: 2581–2593. DOI: 10.1002/hyp.6794

Fang X, Pomeroy JW, Ellis CR, MacDonald MK, DeBeer CM, BrownT. 2013. Multi-variable evaluation of hydrological model predictionsfor a headwater basin in the Canadian Rocky Mountains. Hydrologyand Earth System Sciences 17: 1635–1659. DOI: 10.5194/hess-17-1635-2013

Fang X, Pomeroy JW, Westbrook CJ, Guo X, Minke AG, Brown T. 2010.Prediction of snowmelt derived streamflow in a wetland dominatedprairie basin. Hydrology and Earth System Sciences 14: 991–1006.DOI: 10.5194/hess-14-991-2010

Fowler HJ, Blenkinsop S, Tebaldi C. 2007. Linking climate changemodelling to impacts studies: recent advances in downscalingtechniques for hydrological modelling. International Journal ofClimatology 27: 1547–1578.

Francis S, Smith S, Janowicz J. 1999. Data integration and ecologicalzonation of Wolf Creek watershed. In Wolf Creek Research Basin –Hydrology, Ecology, Environment – Proceedings of a workshop held inWhitehorse, Yukon, March 5–7, 1998. NHRI publication 37-121/1999E: Saskatoon, Saskatchewan; 93–100.

Garnier BJ, Ohmura A. 1970. The evaluation of surface variations in solarradiation income. Solar Energy 13: 21–34.

Gelfan A, Pomeroy JW, Kuchment LS. 2004. Modeling forest coverinfluences on snow accumulation, sublimation, and melt. Journal ofHydrometeorology 5: 785–803.

Goodison BE, Louie PYT, Yang D. 1998. WMO solid precipitationmeasurement intercomparison - Final Report. World MeteorologicalOrganization; 212.

Granger RJ. 1999. Partitioning of energy during the snow-free season inthe Wolf Creek research basin. In Wolf Creek Research Basin –Hydrology, Ecology, Environment – Proceedings of a workshop held inWhitehorse, Yukon, March 5–7, 1998. NHRI publication 37-121/1999E: Saskatoon, Saskatchewan; 121–130.

Granger RJ, Pomeroy JW. 1997. Sustainability of the western Canadianboreal forest under changing hydrological conditions - II- summerenergy and water use. In: Sustainability of Water Resources underIncreasing Uncertainty , Rosjberg D, Boutayeb N, Gustard A,Kundzewicz Z, Rasmussen P. (eds). IAHS Press: Wallingford, UnitedKingdom; 243–250.

Granger RJ, Gray DM. 1990. A new radiation model for calculating dailysnowmelt in open Environments. Nordic Hydrology 21: 217–234.

Granger RJ, Gray DM. 1989. Evaporation from natural non-saturatedsurfaces. Journal of Hydrology 111: 21–29.

Gray DM, Toth B, Zhao L, Pomeroy JW, Granger RJ. 2001. Estimatingareal snowmelt infiltration into frozen soils. Hydrological Processes15: 3095–3111. DOI: 10.1002/hyp.320

Harder P, Pomeroy, JW. 2014. Hydrological model uncertainty due toprecipitation phase partitioning methods. Hydrological Processes, thisissue. DOI: 10.1002/hyp.10214

Inter-governmental Panel on Climate Change (IPCC). 2014.Climate Change 2014: Impacts, Adaptation, and Vulnerability, SpecialIssue Canadian Geophysical Union 2014 IPCC WGII AR5.



Janowicz JR. 2010. Observed trends in the river ice regimes of northwestCanada. Hydrology Research 41: 462–470. DOI: 10.2166/nh.2010.145

Janowicz JR. 2008. Apparent recent trends in hydrologic responsein permafrost regions of northwest Canada. Hydrology Research39: 267–275. DOI: 10.2166/nh.2008.103

Janowicz JR, Gray DM, Pomeroy JW. 2003. Spatial variability of fallsoil moisture and spring snow water equivalent within a mountainoussub-arctic watershed. In Proceedings of the Eastern Snow ConferenceVol. 60. Sherbrooke, Quebec; 127–139.

Janowicz JR. 2001. Impact of recent climatic variability on peakstreamflow in northwestern Canada with implications for design ofthe proposed Alaska highway gas pipeline. In Proceedings, NorthernResearch Basins, 13th International Symposium and Workshop, 19-24August 2001, Saarisälkä, Finland / Murmansk, Russia, Kajander J,Kuusisto E (eds). Finnish Environmental Institute; 161–169.

Janowicz JR. 1999. Wolf Creek Research Basin – an overview. In WolfCreek Research Basin – Hydrology, Ecology, Environment – Pro-ceedings of a workshop held in Whitehorse, Yukon, March 5–7, 1998.NHRI publication 37-121/1999E: Saskatoon, Saskatchewan; 121–130.

Kite GW, Haberlandt U. 1999. Atmospheric model data for macroscalehydrology. Journal of Hydrology, 217(3): 303–313.

Leavesley GH, Lichty RW, Troutman BM, Saindon LG. 1983.Precipitation-runoff modelling system: user’s manual. US GeologicalSurvey: Reston, Virginia; 4238.

Lewkowicz AG, Ednie M. 2004. Probability mapping of mountainpermafrost using the BTS method, Wolf Creek, Yukon Territory,Canada. Permafrost and Periglacial Processes 15: 67–80. DOI:10.1002/ppp.480

López-Moreno JI, Pomeroy JW, Revuelto J, Vicente-Serrano SM. 2012.Response of snow processes to climate change: spatial variability ina small basin in the Spanish Pyrenees. Hydrological Processes27: 2637–2650. DOI: 10.1002/hyp.9408

MacDonald J. 2010. Unloading of intercepted snow in conifer forests. M.Sc. thesis, Department of Geography and Planning. University ofSaskatchewan: Saskatoon; 93.

MacDonald MK, Pomeroy JW, Pietroniro A. 2009. Parameterizingredistribution and sublimation of blowing snow for hydrologicalmodels: tests in a mountainous subarctic catchment. HydrologicalProcesses 23: 2570–2583.

Marks D, Kimball J, Tingey D, Link T. 1998. The sensitivity of snowmeltprocesses to climate conditions and forest cover during rain-on-snow: acase study of the 1996 Pacific Northwest flood. Hydrological Processes12: 1569–1587.

McCartney SE, Carey SK, Pomeroy JW. 2006. Intra-basin variability ofsnowmelt water balance calculations in a subarctic catchment.Hydrological Processes 20: 1001–1016. DOI: 10.1002/hyp.6125

Ménard CB, Essery R, Pomeroy JW. 2014. Modelled sensitivity of thesnow regime to topography, shrub fraction and shrub height. Hydrologyand Earth System Sciences Discussions 11: 223–263.

Ménard CB, Essery R, Pomeroy JW, Marsh P, Clark DB. 2012. A shrubbending model to calculate the albedo of shrub-tundra. HydrologicalProcesses. DOI: 10.1002/hyp.9582

Moss RH, Edmonds JA, Hibbard KA, Manning MR, Rose SK, VanVuuren DP, … , Wilbanks TJ. 2010. The next generation ofscenarios for climate change research and assessment. Nature463: (7282): 747–756.

Osterkamp TE, Jorgenson MT, Schuur EAG, Shur YL, Kanevskiy MZ,Vogel JG, Tumskoy VE. 2009. Physical and ecological changesassociated with warming permafrost and thermokarst in interior Alaska.Permafrost and Periglacial Processes 20: 235–256.

Pomeroy JW, Shook K, Fang X, Brown T, Marsh C. 2013. Developmentof a snowmelt runoff model for the Lower Smoky River. Centrefor Hydrology Report: Report No. 13, University of Saskatchewan,Saskatoon, http://www.usask.ca/hydrology/reports/CHRpt13_Smoky_Snowmelt_Runoff_Model.pdf

Pomeroy JW, Fang X, Williams BT. 2011. Modelling snowwater conservation on the Canadian Prairies. Centre for HydrologyReport: Report No. 11, University of Saskatchewan, Saskatoon, http://www.usask.ca/hydrology/reports/CHRpt11_Prairie-Shelterbelt-Study_Apr11.pdf

Pomeroy JW, Gray DM, Brown T, Hedstrom NR, Quinton WL, GrangerRJ, Carey SK. 2007. The cold regions hydrological model, a platform


for basing process representation and model structure on physicalevidence. Hydrological Processes 21: 2650–2667. DOI: 10.1002/hyp.6787

Pomeroy JW, Bewley DS, Essery RLH, Hedstrom NR, Link T, GrangerRJ, Janowicz JR. 2006. Shrub tundra snowmelt. HydrologicalProcesses 20: 923–941. DOI: 10.1002/hyp.6124

Pomeroy JW, Toth B, Granger RJ, Hedstrom NR, Essery RLH. 2003.Variation in surface energetics during snowmelt in a subarctic mountaincatchment. Journal of Hydrometeorology 4(4): 702–719. DOI: 10.1002/hyp.6124

Pomeroy JW, Li L. 2000. Prairie and arctic areal snow cover mass balanceusing a blowing snow model. Journal of Geophysical Research105: 26619–26634. DOI: 10.1029/2000JD900149

Pomeroy JW, Hedstrom N, Parviainen J. 1999. The snow mass balance ofWolf Creek, Yukon: effects of snow sublimation and redistribution. InWolf Creek Research Basin – Hydrology, Ecology, Environment –Proceedings of a workshop held in Whitehorse, Yukon, March 5–7, 1998.NHRI publication 37-121/1999E: Saskatoon, Saskatchewan; 15–30.

Pomeroy JW, Granger RJ. 1999. Wolf Creek Research Basin: Hydrology,Ecology, Environment. National Water Research Institute. EnvironmentCanada; 160. http://www.taiga.net/wolfcreek/proceedings.html

Pomeroy JW, Gray DM. 1995. Snow accumulation, relocation andmanagement. In National Hydrology Research Institute Science ReportNo. 7. Environment Canada: Saskatoon; 144.

Priestley CHB, Taylor RJ. 1972. On the assessment of surface heat fluxand evaporation using large-scale parameters, Monthly Weather Review100: 81–92.

Quinton WL, Bemrose RK, Zhang Y, Carey SK. 2009. The influence ofspatial variability in snowmelt and active layer thaw on hillslopedrainage for an alpine tundra hillslope. Hydrological Processes23: 2628–2639. DOI: 10.1002/hyp.7327

Quinton WL, Carey SK. 2008. Towards an energy-based runoffgeneration theory for tundra landscapes. Hydrological Processes22: 4649–4653. DOI: 10.1002/hyp.7164

QuintonWL, Shirazi T,CareySK, Pomeroy JW. 2005. SoilWater Storage andactive layer development in a sub alpine tundra hillslope, southern YukonTerritory, Canada. Permafrost and Periglacial Processes 16: 369–382.

Rawlins MA, Steele M, Serreze MC, Vörösmarty CJ, Ermold W,Lammers RB, Zhang J. 2009. Tracing freshwater anomalies through theair-land-ocean system: A case study from the Mackenzie River Basinand the Beaufort Gyre. Atmosospher-Ocean 47: 79–97. DOI: 10.3137/OC301.2009

Rutter N, Essery RLH, Pomeroy JW, Altimir N, Andreadis K, Baker I,Barr A, Bartlett P, …, Yamazaki T. 2009. Evaluation of forest snowprocesses models (SnowMIP2). Journal of Geophysical Research114: 18. DOI: 10.1029/2008JD011063

Salathé EP. 2003. Comparison of various precipitation downscalingmethods for the simulation of streamflow in a rain shadow river basin.International Journal of Climatology 23: 887–901.

Seguin MK, Stein J, Nilo O, Jalbert C, Ding Y. 1998. HydrogeophysicalInvestigation of the Wolf Creek Watershed, Yukon Territory,Canada. In Wolf Creek Research Basin – Hydrology, Ecology,Environment – Proceedings of a workshop held in Whitehorse, Yukon,March 5–7, 1998. NHRI publication 37-121/1999E: Saskatoon,Saskatchewan; 55–78.

Sicart JE, Pomeroy JW, Essery RLH, Bewley D. 2006. Incominglongwave radiation to melting snow: observations, sensitivityand estimation in northern environments. Hydrological Processes20: 3697-3708. DOI: 10.1002/hyp.6383

St Jacques JM, Sauchyn DJ. 2009. Increasing winter baseflow and meanannual streamflow from possible permafrost thawing in the NorthwestTerritories, Canada. Geophysical Research Letters 36. DOI: 10.1029/2008GL035822

Sturm M, Schimel J, Michaelson G, Welker JM, Oberbauer SF, ListonGE, Romanovsky VE. 2005. Winter biological processes could helpconvert arctic tundra to shrubland. Bioscience 55: 17–26. DOI: 10.1175/1520-0442(2001)014<0336:SSIIAT>2.0.CO;2

Tape K, Sturm M, Racine C. 2006. The evidence for shrub expansionin Northern Alaska and the Pan-Arctic. Global Change Biology12: 686–702. DOI: 10.1111/j.1365-2486.2006.01128.x

Tolson BA, Shoemaker CA. 2007. The dynamically dimensioned search(DDS) algorithm for computationally efficient watershed model


http://www.taiga.net/wolfcreek/proceedings.html

K. RASOULI ET AL.

calibration. Water Resources Research 43: W12603. DOI: 10.1029/2005WR004723

Verseghy DL. 1991. CLASS -A Canadian land surface scheme for GCMs,I. Soilmodel. International Journal of Climatology 11: 111–133.

Walvoord MA, Striegl RG. 2007. Increased groundwater to streamdischarge from permafrost thawing in the Yukon River basin: Potentialimpacts on lateral export of carbon and nitrogen. Geophysical ResearchLetters 34: L12402. DOI: 10.1029/2007GL030216

Wilby RL. 2005. Uncertainty in water resource model parameters used forclimate change impact assessmentHydrological Processes19: 3201–3219.

Wilby RL, Harris I. 2006. A framework for assessing uncertainties inclimate change impacts: Low-flow scenarios for the River Thames, UK.Water Resources Research 42: W02419.


Woo M-K, Carey SK. 1999. Permafrost, seasonal frost and slopehydrology, central Wolf Creek basin, Yukon. In Wolf CreekResearch Basin – Hydrology, Ecology, Environment –Proceedings of a workshop held in Whitehorse, Yukon, March5–7, 1998. NHRI publication 37-121/1999E: Saskatoon, Saskatchewan;45–54.

Zhang Y, Carey SK, Quinton WL, Janowicz JR, Flerchinger GN. 2010.Comparison of algorithms and parameterizations for infiltration intoorganic-covered permafrost soils. Hydrology and Earth System Sciences14: 729–750. DOI: 10.5194/hess-14-729-2010.

Zhou J, Pomeroy JW, Zhang W, Cheng G, Wang G, Chen C. 2014.Simulating cold regions hydrological processes using a modular modelin the west of China. Journal of Hydrology 509: 13–24.


Hydrological sensitivity of a northern mountain … · 2014-06-17 · Hydrological sensitivity of a...

Documents

Transcript of Hydrological sensitivity of a northern mountain … · 2014-06-17 · Hydrological sensitivity of a...