For Peer Review Only - University of Lethbridgepeople.uleth.ca/~stefan.kienzle/Documents/Nemeth_etal_2010_NSask... · For Peer Review Only Multi-variable verification of hydrological

For Peer Review O

nly

Multi-variable verification of hydrological processes in the upper North Saskatchewan River basin, Alberta, Canada.

Journal: Hydrological Sciences Journal

Manuscript ID: Draft

Manuscript Type: Original Article

Date Submitted by the Author:

n/a

Complete List of Authors: Kienzle, Stefan; University of Lethbridge, Geography Byrne, James; University of Lethbridge, Geography

Nemeth, Michael; University of Lethbridge, Geography

URL: http://mc.manuscriptcentral.com/hsj

Hydrological Sciences Journal

For Peer Review O

nly

Multi-variable verification of hydrological processes in the

upper North Saskatchewan River basin, Alberta, Canada. MICHAEL W. NEMETH, STEFAN W. KIENZLE & JAMES M. BYRNE

Department of Geography, University of Lethbridge, Lethbridge, Canada

[email protected] Received; accepted; open for discussion until Citation in reference format. After received dates, ranged left, same size font as received dates.

Abstract The ACRU agro-hydrological modelling system (ACRU) was set up to simulate the impacts of climate change on the hydrological behaviour of the 3,856 km2 Cline River watershed, the major headwater of the upper North Saskatchewan River basin in central Alberta, Canada. The physical-conceptual ACRU was set up to simulate all elements of the hydrological cycle on a daily time step for the 1961-1990 baseline period, to serve as the basis for later climate change impacts simulations. After parameterization, ACRU output was verified against temperature, snow water equivalent, glacier mass balance, potential evapotranspiration, and two sets of long term daily streamflow records. The multi-variable verification analyses were implemented to instill confidence that the simulated streamflow time series is based on a realistic combination of bio-physical parameters and represents the watershed behaviour in terms of annual water yields, seasonality, shape and magnitude of the hydrographs, and streamflow variability for the right reasons.

Key words hydrological simulation model; streamflow; snow; glacier; verification; ACRU; Canada

INTRODUCTION Throughout the province of Alberta there is increasing demand for water due to population growth, agriculture, industry, and the energy sector (Sauchyn & Kulshreshtha, 2008). However, availability of future water resources is uncertain due to the hydrological impacts of climate change. Recent studies have shown that declining streamflows are expected to reduce water availability for irrigation, industrial and domestic use, as well as hydroelectric power generation (Gan, 1998; Rood et al., 2005; Byrne & Kienzle, 2008; Sauchyn & Kulshreshtha, 2008). Expected impacts include decreased snow accumulations and associated earlier spring runoffs, increased winter and decreased summer and late fall streamflows, and decreased glacial melt contributions (Burn, 1994; Gan, 1998; Demuth & Pietroniro, 2003; Lapp et al., 2005; Sauchyn & Kulshreshtha, 2008). Therefore, the understanding of watershed balances is essential for current and future water resources planning. This research focuses on the estimation of the impacts of climate change on streamflows in the Cline River watershed, the headwater watershed in the upper North Saskatchewan River basin (UNSRB). The Cline River watershed contributes approximately 40% of the streamflow that flows through Edmonton in the North Saskatchewan River.

Hydrological models, coupled with climate change predictions, are the best tools available to determine hydrological effects of climate change in a region, as they provide a framework to conceptualize and examine the relationships between climate and the elements of the hydrological cycle within a watershed (Leavesley, 1994; Xu, 1999). Hydrological models use observed meteorological data to simulate streamflow, and are often verified using observed streamflow or other measurable hydrological variables (Rosso, 1994; Brooks et al., 2006, Liu et al., 2009). A model should be able to reasonably reproduce historical streamflow records, in order to be used in climate change studies, and can only be applied with confidence once the model output has

Page 1 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

been verified against observed data (Loukas et al., 2002; Jewitt et al., 2004). Models are simplifications of reality, but are an important tool in assessing future scenarios for water management strategies, if accurately parameterized and verified (Beven, 1989). Rigorous parameterization of a model is a crucial step to avoid problems during the model verification process (Refsgaard, 1997). Parameters should be based on field data, or should be estimated within a physically acceptable range based on available data and literature sources (Kienzle & Schmidt, 2008). The physical-conceptual ACRU agro-hydrological modelling system (Schulze, 1995) is capable of simulating all major elements of the hydrological cycle, and was chosen for this study. This paper presents the parameterization and verification of simulated against observed hydrological variables.

OBJECTIVES

The objective of this research was to estimate the impacts of climate change on water yield, streamflow extremes, and the streamflow regimes in the Cline River watershed, Alberta, Canada. This paper describes the ACRU agro-hydrological modelling system and the parameterization and verification analyses for the 1961-1990 base period, while a follow-up paper describes the setup of the climate scenarios as well as simulation results for the future periods 2010-2039, 2040-2069, and 2070-2099. In order to ensure that the simulated results replicate the hydrological behaviour of the Cline River watershed, observations within both the Cline River watershed and the larger, upper North Saskatchewan River basin (UNSRB) are used to compare simulated with observed data, including temperature, snow pillow and snow course time series, potential evapotranspiration, glacier melt and glacier equilibrium line, and two sets of daily streamflow time series. STUDY AREA

The Cline River watershed is situated southwest of Edmonton, Alberta, Canada (Fig. 1), constituting the main headwaters of the UNSRB. For the purpose of streamflow verification, the Cline River watershed was delineated into two nested subwatersheds, with Watershed 1 flowing into Watershed 2, which at its lower portion contains Lake Abraham, the reservoir for the Bighorn Dam (Fig. 1). The Cline River watershed was the focus of model parameterization and analysis, however, verification was preformed for the whole UNSRB to increase regional data availability and prepare for future regional analysis. The Cline River watershed has an area of a 3,856 km2 and consists of alpine, subalpine, and foothills landscapes located on the eastern slopes of Alberta’s Rocky Mountains. Dominant land cover includes bare rock (50.6%), coniferous forest (33.0%), and glaciers (6.8%). There are no towns or agricultural activities in the watershed. The watershed ranges in elevation from approximately 1,300 m at the outlet gauging station to just under 3,500 m at the Rocky Mountain peaks at the continental divide. Mean annual precipitation for the Cline river watershed is 1002 mm, ranging from 1408 mm in the upper reaches to 510 mm at the outlet. Mean annual temperature is -0.8°C, with mean temperatures of -6.2°C in the upper reaches and 2.9°C at the outlet. Streamflows are dominated by snowmelt and supplemented by glacial melt from the Columbia Ice Fields, Peyto, Athabasca, and Saskatchewan glaciers (NSWA, 2005). Mean annual glacier contributions to Cline River streamflow are 8%, with an annual range of 0.1 to 29.2%.

***** Please insert Figure 1 approximately here *****

Page 2 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

METHODS The ACRU agro-hydrological modelling system

The ACRU agro-hydrological modelling system has been developed at the School of Bioresources Engineering and Environmental Hydrology (formerly the Department of Agricultural Engineering) at the University of KwaZulu-Natal, Republic of South Africa since the late 1970s (ACRU, 2007). ACRU is a multi-purpose, multi-level, integrated physical-conceptual model that can simulate total evaporation, soil water and reservoir storages, land cover and abstraction impacts, snow water dynamics and streamflow at a daily time step. The ACRU model revolves around multi-layer soil water budgeting with specific variables governing the atmosphere-plant-soil water interfaces. Surface runoff and infiltration are simulated using a modified SCS equation (Schmidt & Schulze, 1987), where the daily runoff depth is proportional to the antecedent soil moisture content.

A number of ACRU input variables are estimated from the physical characteristics of the watershed. When not all required variables are available, they are estimated within physically meaningful ranges based either on the literature or local expert knowledge. Spatial variation of rainfall, soils and land cover is facilitated by operating the model in distributed mode, in which case the catchment is subdivided into either sub-watersheds or hydrological response units (HRUs), each representing a relatively homogenous area of hydrological response.

ACRU has been considerably expanded from its original version (Schulze, 1995) to include snow and glacier routines, as well as improve the spatial and temporal representations of numerous variables. Snow processes, such as canopy interception, sublimation, metamorphosis, or change in albedo and density, are simulated in a physically explicit manner. The snow melt simulation is based on either pre-determined monthly snowmelt factors, or a dynamic degree-day factor, which is determined by the model on a daily basis from incoming radiation and albedo estimates.

ACRU has been applied for water resource assessments (Everson, 2001; Kienzle et al., 1997; Schulze et al., 2004; Martinez et al., 2008) and climate change impacts (New & Schulze, 1996; New, 2002; Schulze et al., 2004; Forbes et al., 2010), and often requires extensive GIS pre-processing (Kienzle, 1993; 1996; Schulze et al., 1995). The conceptual structure of ACRU is shown in Fig. 2.


Creation of a digital elevation model (DEM)

At the time of the study, the only DEM available was from the Shuttle Radar Topography Mission (SRTM), with a grid cell size of approximately 75 m. This DEM had major errors and discontinuities within the alpine region of the study area, making it unusable for further GIS analyses. Therefore, a new DEM was created, using 1:50,000 topographic map sheets from the National Topographic System of Canada. The ‘Topo to Raster’ tool (ESRI, 2008) was used to create a 100 m resolution DEM of the study area from contour lines and elevation point data from each of the 36 map sheets. The topographic map sheets also contained river layers, which were used to

Page 3 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

further improve the DEM to ensure that the watershed delineation would be consistent with the actual stream network.

Delineation of watersheds and HRUs

The Cline River watershed was divided into two watersheds by honouring the location of the Water Survey of Canada (WSC) gauging stations. Delineating these watersheds allowed the model to be run on each of the watersheds individually for verification purposes. Each watershed was further divided into hydrological response units (HRUs), which are distributed, relatively homogeneous bio-physical units that are assumed to have similar hydrological characteristics and response (Flügel, 1995). The HRUs were delineated based on 100 m elevation bands, nine land cover classes, four mean annual radiation classes, and watershed boundary. GIS overlay analysis was performed using these four data layers, which resulted in 308 HRUs, with an average planimetric HRU area of 12.5 km2.

Each HRU was parameterized to have a unique combination of hydrological variables, most of which were derived by GIS overlay analysis. The area of each HRU was calculated based on its true, sloped area, as the planimetric area derived from ArcGIS is under-estimated in steeply sloped terrain. Due to the steep slope of the Cline River watershed, the area-underestimation by using the planimetric areas would have been 9.8%, which would have affected interception volumes, soil moisture storages, groundwater recharge rates, actual evapotranspiration volumes, and runoff coefficients (Kienzle, 2010).

Parameterization of ACRU

Climate data

In order to model impacts of climate change, a base period from 1961-1990 was used, as it is assumed that it has not been significantly impacted by climate change (Diaz-Nieto & Wilby, 2005). Climate data were accessed through the Environment Canada Meteorological Service (Environment Canada, 2008). Within the vicinity of the Cline River watershed, three stations had 30 or more years of mostly complete data, which were required in order to create complete daily time series of climate inputs to drive the model. As none of the stations had complete daily records for the 1961-1990 time period, infilling the missing station data was required for each of the three stations by means of regression analysis, using monthly normals for both precipitation and temperature time series.

Monthly normal climate surfaces are needed to calculate precipitation and temperature correction factors for each of the HRUs. As the climate data in the study area were too sparse and incomplete to be able to interpolate a meaningful climate surface, the Parameter-elevation Regressions on Independent Slopes Model (PRISM), developed by Daly (2006), was used. The 1971-2000 monthly normal PRISM surfaces of minimum and maximum temperatures and precipitation, at a 2 km resolution, were available. Temperature and precipitation values from the PRISM surfaces were verified against station values from long-term climate stations in the UNSRB. All stations correlated well with a coefficient of determination ranging from 0.996 to 0.904, with an average of 0.989. However, as all climate stations used to create the PRISM surfaces are situated in valleys, high elevation precipitation values exhibit a high level of uncertainty. Further uncertainties exist due to the inaccuracies, experienced by the authors, of the seasonality of PRISM precipitation surfaces along the eastern slopes of the continental divide, a result of the over-estimation of the

Page 4 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

influence of a more maritime dominated climate observed on the western slopes of the continental divide. Consequently, the PRISM derived precipitation values could only provide initial approximations, which potentially required subsequent seasonal adjustments to match the observed seasonality of streamflows.

Precipitation

In order to account for spatial differences between a climate station used to “drive” the model and each HRU, ACRU corrects daily precipitation values. Monthly precipitation correction factors are based on a ratio between the 1971-2000 climate normals observed at the base station and the mean PRISM precipitation value for each HRU, and are then internally transformed to daily values through a Fourier transformation. Precipitation inputs are further corrected based on size of the watershed, as precipitation measured at a point should not be transferred to an entire watershed without adjustment. This is especially true during the summer months, when convective precipitation events result in localized, high intensity precipitation events. The implementation of an areal precipitation reduction factor (ARF) reduces large, daily extreme events as a function of the precipitation event and the size of the watershed. All other daily precipitation values are then increased by a small proportion to balance the initial annual precipitation depths.

Temperature

Regional mean monthly lapse rates for maximum and minimum temperatures were initially calculated from the monthly maximum and minimum temperature PRISM surfaces and mean elevation for each HRU. Initial runs in the glaciated watershed showed that the magnitude of the mean daily lapse rate was too high, as only HRUs at very low elevation produced any glacier melt in Peyto Glacier, for which detailed glacier mass balance data were available (WGMS, 2009). Maximum temperature lapse rates were, therefore, lowered so that the elevations of simulated glacier melt matched the observed. The resulting lapse rates were in agreement with those reported by Shea et al. (2009).

To account for the differences in temperature between HRUs receiving different amounts of radiation due to exposition and shading effects, the daily lapse-rate derived temperatures were adjusted, resulting in increased temperatures on south-facing slopes and decreased temperatures on north-facing slopes (Kienzle, 2010). The rate of temperature change is a function of the daily temperature range. In addition, daily ground temperatures are adjusted according to the function of the leaf area index (LAI), as suggested by Glassy & Running (1994). As the LAI varies seasonally, the LAI is changed on a monthly basis. LAI data were obtained from MODIS (Moderate Resolution Imaging Spectroradiometer) satellite images that were downloaded from the National Snow and Ice Data Center (NSIDC, 2009). The result of these complex calculations is that each combination of elevation, slope, location, and land cover has a unique set of daily minimum and maximum temperatures.

Land cover data

Land cover information is based on the Circa 2000 Land Cover data set, a generalized land cover map developed by the National Land and Water Information Service from multi-spectral Landsat satellite imagery for the agricultural extent of Canada (NLWI, 2008). From this data set, seven land cover classes were used for this study: water/wetlands, bare rock, shrubland, grassland, coniferous forest, deciduous forest,

Page 5 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

and mixed forest. As glaciers and large lakes were not part of the original Circa 2000

Land Cover data set, polygons of glaciers and lake polygons were added to the land cover GIS database, resulting in a total of nine land cover classes. Each land cover class is associated with monthly hydrological variables, such as albedo, rooting depth, plant transpiration coefficient, and coefficient of initial abstractions.

As no observed albedo values were available, literature values for monthly albedo values for each land cover were used as model inputs (Brutsaert, 1981; Gerrard, 1990; Pomeroy & Dion, 1996; Ahrens, 2006). During winter months, albedo values are modified daily as a function of snow cover. After fresh snowfall, the albedo is set to 0.80, after which it decreases by 1.5% per day until it reaches 0.60. After the snow depth declines to 75mm, the albedo decreases rapidly, as an increasing proportion of the land surface is assumed to be snow free.

Model inputs for rooting depths for each land cover were based on published values (Strong and La Roi, 1983; Canadell et al., 1996; Jackson et al., 1996).

No reliable estimates of plant water use for vegetation present in the study area were available in the literature. Therefore, plant transpiration coefficients (PTCs) were calculated for the most prevalent land cover classes. Values for PTCs were calculated from observed meteorological and flux data from grassland, aspen forest, and coniferous forest sites from Fluxnet Canada (2010), and AmeriFlux (2010) flux towers in Alberta, Saskatchewan, and Colorado respectively. These stations were used as they were the most representative of the vegetation in the Cline River watershed. Measured latent heat flux data from each station were used to calculate the actual evapotranspiration (AET) for each site, using equation (1) (Hornberger et al. 1998):

ET = El / ρw * λv (1)

with ET = evapotranspiration rate [m s-1], El = latent heat flux [J m-2 s-1], ρw = density of water [kg m-3], and λv = latent heat of vaporization [2.45x106 J kg-1]. Meteorological data at each site were used in the Penman-Monteith equation

to calculate potential evapotranspiration (PET). Canopy height and latitude of the sites were included as part of each calculation. The stomatal resistance value was set to “1” for each of the three sites to essentially eliminate stomatal resistance and control on calculated PET values. The AET was then divided by the PET value for each month to give a monthly PTC for grasslands, deciduous forest, and coniferous forest. Shrubland was estimated by taking a ratio of grassland and deciduous forest, and mixed forest was based on the average values of deciduous and coniferous forest. Seasonality and magnitudes of calculated PTCs are shown in Fig. 3.

Percentages of forest canopy coverage were derived from both field measurements in the southern part of the watershed (Visscher et al., 2004) and from the Alberta Vegetation Inventory (AVI), which is a photo-based inventory supported with air and ground observations (Government of Alberta, 2009).

Seasonal changes of LAI are used in ACRU to simulate changes in plant evaporation and canopy interception. The latter is estimated using the Von Hoyningen-Huene method (Schulze, 1995). Observed MODIS LAI data were download (NSIDC, 2009) for the 2004 calendar year, as this year best represented the average climate when compared to all other available years.

Page 6 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly


Soil data

ACRU requires soil texture and depth information for both the A- and the B-horizons. If not otherwise available, the porosity, wilting point, field capacity, the portion of water redistributed from surface to subsurface horizons, and from subsurface to groundwater store can be estimated from the texture information, using internal look-up tables in ACRU (Schulze, 1995). There exists considerable uncertainty with soils information, especially in mountainous regions, due to very sparse soils data. A 1:50,000 soils map of a mountain park area within the UNSRB was obtained from Parks Canada, which included the dominant soil type for each land cover polygon. Published soil surveys and journal articles were also used as guides to facilitate the realistic estimation of soil types and soil depths, and their association with land cover (Peters & Bowser, 1960; Pettapiece, 1971; Peters, 1981). The acquired soils data were compiled in a GIS, where a regression analysis was performed between the available soil parameters and land cover. This resulted in GIS soil layers with representative, but generalized, soils data, including texture with percent sand, silt, and clay, and soil depth for both A- and B-horizons. Using the percent sand and clay values, soil textures were assigned to each land cover type. Finally, soil texture and depth for the A- and B-horizons were determined for each HRU. Based on the scarcity of soils information, the quality of these estimates remained uncertain.

Evapotranspiration

Evapotranspiration takes place from several principle water storages, including intercepted snow and rain, snowpack, the soil and rock surfaces, and transpiration by plants. Daily actual evapotranspiration is determined using the internally pre-determined reference evaporation and monthly PTCs for each land cover type. The amount of transpiration from the A- and B-horizons is controlled by the proportion of roots within the two horizons, which can vary seasonally. Soil water in excess of the permanent wilting point is considered available for transpiration by plants (Schulze, 1995).

In order to apply the Penman (1948) evaporation method, daily climate data are required, including daily minimum and maximum temperature, relative humidity, wind, incoming solar radiation, sunshine hours, and precipitation. The physically based Penman method for reference evaporation was used at a daily time step. The mass transfer component of the Penman equation describes the change of vapour from a vegetated surface (Schulze, 1995). Saturated vapour pressure is empirically related to observed mean air temperature using Tetens’ (1930) equation.

Daily shortwave radiation and monthly mean potential sunshine hours were calculated for each HRU by first calculating daily shortwave radiation for each base station using the Area Solar Radiation tool in ArcGIS (ESRI, 2008). The radiation calculation required monthly atmospheric transmittivity and diffuse radiation values, which were estimated using values from the nearest climate stations that recorded them. A 30-year time series of observed daily wind and relative humidity was compiled by infilling data from a nearby climate station with adequate data. Daily relative humidity data from six stations in the study area with short term data were used to interpolate monthly surfaces for relative humidity. Four seasonal wind maps (i.e. DJF, MAM, JJA, SON) of interpolated wind data from Environment Canada’s

Page 7 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Canadian Wind Energy Atlas (Government of Canada, 2003) were integrated in a GIS. Monthly mean wind and humidity values were derived for each HRU by GIS overlay analysis of the relevant surfaces.

Snow

In order to accurately model magnitudes and timing of streamflow in the study area, snow fall, snow pack and snow melt needed to be simulated dynamically and realistically. The threshold temperature that determines whether precipitation falls as rain or snow, as well as the temperature range within which a proportion of precipitation falls as rain, were calculated for each HRU, based on its average elevation and an empirical equation developed for the study area (Kienzle, 2008), which is based on a curvi-linear relationship between mean daily air temperature and the proportion of precipitation that falls as snow. Snow canopy interception and melt, snowpack evolution, rain on snow events, snow density and the area covered by snow during the last phase of snow are simulated in a physically explicit manner. The snow melt simulation is based on an empirically derived dynamic degree-day factor, which is determined on a daily basis from incoming radiation and albedo estimates.

In ACRU, snow water equivalent (SWE) is simulated differently in forested HRUs than in non-forested HRUs. When an HRU is predominantly forested, canopy interception and losses, changes in albedo values, net radiation, and temperature are calculated in a forest-specific way.

Glaciers

A simple glacier mass balance model was used for this study, which requires a dynamic glacier melt factor and an estimation of the initial mean glacier depth per HRU. The glacier melt factor is calculated from the dynamic snowmelt factor increased by a factor of 2.5. This value was calibrated based on that proportion of summer streamflow produced by Watershed 1 that could not be explained by rain events and baseflow contributions, and is in agreement with observed ratios between glacier melt and snow melt factors (Comeau et al., 2009; Shea et al., 2009). The initial glacier depth values were based on the reported depth of glaciers in this region of the Rocky Mountains (Ommanney, 2002). Depths were used that ranged from an estimated 30 m at elevations of 1947 m to 143 m at elevations of 2245 m. From the depth and the slope corrected area of the HRU, the initial glacier volume was calculated. On an annual basis, at the end of the hydrological year on October 31, the accumulated glacier melt is subtracted, and the remaining SWE is added to the glacier volume, and both glacier area and depth are adjusted.

Streamflow

ACRU determines daily streamflow from same day and delayed stormflow, baseflow, snow melt, and glacier melt volumes. Runoff generation is based on a modified Soil Conservation Service (SCS) procedure (United States Department of Agriculture, 1985), where runoff potential is conceptualized as a depth, and is a function of the soil’s relative wetness (Schulze, 1995), conceptually representing the antecedent watershed conditions. Monthly varying coefficients of initial abstractions are used in the SCS stormflow calculations and represent the amounts of precipitation that do not contribute to the generation of stormflow, because of the processes of interception and the temporary surface storage in hollows.

Groundwater recharge occurs when the soil water store exceeds field capacity (Schulze, 1995). Recessions curves from hydrographs of two observed streamflow

Page 8 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

time series are used to derive exponential response coefficients to represent the temporal response of both surface runoff and baseflows for each of the two sub-watersheds.

VERIFICATION PROCEDURES, DATA, AND RESULTS

Simulated results were verified against five observed data sources: air temperature, A-pan potential evapotranspiration and regional potential evapotranspiration maps, snowpack accumulation and melt from snow pillows and snow courses, Peyto Glacier glacial melt contributions and its equilibrium line altitude, and daily observed streamflow time series for two WSC stations (Fig. 1). In order to reveal if the simulated hydrological variables were representative of the observed ones, a number of simulation objectives were set to assess the success of the simulations. These included the mean annual water yields, seasonal fluctuations, daily and monthly statistics, including the coefficient of determination (r2), the regression coefficient (slope), the regression intercept, and the difference of the variances and means. In addition, the Nash-Sutcliff coefficient (E) was used to evaluate the model performance (Legates & McCabe, 1999).

Temperature

Observed daily maximum and minimum temperature data from ten short-term and seasonal climate stations (Environment Canada, 2008) within the UNSRB were used to verify that the derived monthly minimum and maximum temperature lapse rates were properly simulating temperatures. Between five and 42 years of observations were available, totaling 37,604 days.

Fig. 4 shows the seasonality of mean monthly temperatures, with good association of simulated and observed temperatures. Objective measures of modelling performance for monthly mean temperatures simulated by ACRU are presented in Table 1. Monthly temperatures are slightly over-simulated, with monthly mean temperatures being statistically not different, as indicated by the P-value of 0.46 from a two-tailed t-test. Monthly differences between variances are low (-3.79%), indicating that periods of extreme temperatures were realistically simulated. Correlation statistics were also highly acceptable (Table 1).


***** Please insert Table 1 approximately here *****

Potential evapotranspiration (PET)

Mean annual PET was estimated to be 500-600 mm in the eastern portion of the UNSRB, 600-800mm in the central portion, and 800-1000 mm in the western, mountainous region of the UNSRB. These values are in full agreement with the Hydrological Atlas of Canada (1978). Locally, mean annual PET values range from a low of about 200 mm, found in areas that are both in north-facing radiation shadow and that have long periods of high albedo, to a high of 1100 mm on south facing slopes with low albedo values associated with coniferous forest.

The seasonality of PET values was compared against three A-pan stations approximately 100 km northeast of the UNSRB. All three A-pan stations monitored

Page 9 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

very similar values for each of the six months of observed data (May-October). Thus, monthly A-pan values were averaged for all three stations. One HRU with a similar land cover and elevation, and closest to the A-pan stations, was selected for statistical comparison. With a coefficient of determination of 0.78 and a difference in variance of 2.5% it was confirmed that the estimated seasonality was in acceptable agreement with regional observed A-pan values. PET in the Cline River watershed peaks in June, with roughly 75% of mean annual PET occurring between May and September. Snow water equivalent (SWE)

Observed snow water equivalent (SWE) data from 15 snow courses and two snow pillows in the UNSRB were used to validate simulated snowfall depths and the timing of snow melt (Fig. 1). Snow course and snow pillow data were provided by Alberta Environment and from the Foothills Orographic Precipitation Experiment (FOPEX) by Environment Canada (Smith, 2009). Snow course measurements are carried out approximately once a month. Daily snow pillow time series were used to determine if the snow pack development, maximum annual SWE, and timing of spring melt were properly simulated.

Model output was visually and statistically compared to observed data from 15 snow course and two snow pillow sites. Fig. 5 presents a typical comparison of simulated and observed daily SWE in terms of the magnitude of peak SWE depth and the timing of snowpack melt from a snow pillow. Characteristic time series of simulated and observed SWE time series for a snow course are shown in Fig. 6. When compared to all observed SWE, overall SWE depths were simulated fairly well, considering the geographical extent of the watershed, and the few climate base stations available to drive the model. Peak SWE is sometimes severely over- and sometimes severely under-simulated, while some years are well simulated.



The 15 snow courses had a total of 882 observations, and the two snow pillows a total of 7,625 observations, for a grand total of 8,507 days with observed data. Table 2 shows the statistical analysis for all the daily observed and simulated SWE for the snow courses, the snow pillows, and all SWE data. Daily SWE depths are slightly under-simulated, and are not statistically different for the snow courses, but are considered statistically different for the snow pillows and when all the snow data are compared together (Table 2). Over the validation period, the simulation of high and low snowfall years is preserved, as the difference between variances is low (4.03%) when compared to all observed SWE data. Correlation statistics are good when all SWE data are compared. The regression coefficient is less than unity for all SWE data groups, indicating a systematic and overall under-simulation.


Page 10 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Glacial contributions to streamflow

Glacial mass balance data from Peyto Glacier were used to verify that both the average annual streamflow contributions from glacial melt and the equilibrium line altitude (ELA), the altitude where accumulation of snow is exactly balanced by ablation, were appropriately simulated. Peyto Glacier is located in the southwest corner of the Cline River watershed (Fig. 1), and has been extensively monitored for a number of decades (Loijens, 1974; Luckman, 1998; Demuth & Pietroniro, 2003; Comeau et al., 2009; Demuth & Keller, 2009; Shea et al., 2009). Observed ELA values were used to confirm the altitude of glacier melt, so that only the appropriate portion of a glacier would be simulated to melt out and contribute to streamflow. The average cumulative net balance for Peyto Glacier for the last 30 years was used to calibrate average annual glacial melt contributions from Peyto.

The average snowline, which is similar to the ELA, was simulated for the Cline River watershed to average 2,616 masl (1966-2007), ranging from 2,146 to 3,145 masl. The ELA for Peyto, and other glaciers in this area, fluctuates annually. The average ELA for Peyto is reported to be between 2,612 masl (WGMS, 2009) and 2,700 masl (Demuth & Keller, 2009), but can be as high as 3,190 masl in warm years. Annual glacial mass balance data from Peyto Glacier were used to validate that average annual streamflow contributions from glacial melt were appropriately simulated (Table 3). Based on the mass balance information (WGMS, 2009), the annual Peyto Glacial melt has been very steady between 1977 and 2004 at approximately 730 mm, with very few years significantly deviating from this value. As glaciers were not simulated individually, but were divided in glacier HRUs with specific characteristics in terms of radiation and elevation, only the mean glacier behaviour of those glacier contained in the Cline River watershed can be simulated. The simulated glacier mass balance for all glaciers resulted in an average annual decline of 518 mm, with a gain of 570 mm in cold and years, such as 1974, and a maximum decline of 4,604 mm in warm years, such as 1970.


Streamflow

Observed daily streamflow time series from a Water Survey of Canada (WSC) gauging station were used for Watershed 1. In order to eliminate the effects of Lake Abraham on streamflow, the naturalized flow data provided by Alberta Environment were used, which corrected the streamflow time series for both reservoir storage and releases.

The WSC gauging station at the outlet of Watershed 1 provided observed data for 1970-2007. Consequently, the calibration of streamflow hydrographs in terms of flood peaks, recession curves, seasonality, and annual yields using the entire 1961-90 time period was not possible. A further complication is the fact that the climate base station used to provide daily climate time series to the model consistently has years which exhibit an obvious mismatch between the precipitation observed at the climate station and the streamflow observed at the watershed outlet. Therefore, unreliable years, which were defined here as years with runoff coefficients well outside the typical range found in the region, were eliminated from the calibration period (Beven, 2001). For these reasons it was decided to calibrate streamflow for a minimum of ten consecutive years where the magnitude of precipitations from the climate station

Page 11 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

matched the observed streamflow, considering a typical range of runoff coefficients observed for the watershed, and where the time period selected contained at least one of the lowest and one of the highest observed peak flows in the entire observed time series. Accordingly, the 1981-1992 time period was selected for Watersheds 1 and 2. Once Watershed 1 was satisfactorily simulated in terms of water yields, seasonality, and general shape of the hydrographs, streamflow for the whole 1961-1990 time period was simulated and compared against observed streamflows. After the ACRU setup for Watershed 1 was considered complete, Watershed 2 was simulated and calibrated in a similar fashion.

Table 4 presents the statistical analysis for daily and monthly simulated and observed streamflow data for 1961-90. Total daily water yields were over-simulated by 2.6%. Mean flows were simulated well, as none of the simulated means were statistically different (p>0.05) from the observed (Table 4). There are small differences in daily and monthly streamflow variances, indicating that simulation of peak and low flows were preserved. Both Watersheds 1 and 2 have high coefficients of determination for daily (r2 of 0.81 and 0.82) and monthly (r2 of 0.89 and 0.91) simulations. Fig. 7 shows scatter plots with the regression equations for both daily and monthly streamflow time series. The regression coefficient is less than unity for both daily and monthly time series, which indicates a slight under-simulation of high flows, more so on a daily than a monthly basis. The Nash-Sutcliff coefficients of efficiency for the two watersheds are 0.79 and 0.80 for daily, and 0.87 and 0.91 for monthly streamflow simulations, indicating a strong association between observed and simulated values.



In summary, the streamflow behaviour, as well as the timing and magnitudes

of peak and low flows were simulated very well for Watershed 2 (Figs. 8 and 9). When the whole 1961-90 time series is evaluated, peak flows are sometimes over-simulated, sometimes under-simulated, and most often match the observed flows well. This is to be expected when climate data are transferred from a base station, which has a distance of up to 75 km to the watershed HRUs. Timing and magnitudes of baseflows show a similar pattern of over- and under-simulation as seen in the peak flows, but, generally, the overall recession curves are simulated very well. Fig. 10 presents the mean seasonality of observed and simulated daily streamflow for the 1961-90 time series. The seasonality is well replicated, however, it was slightly over-simulated in August, October, and November.




.

Page 12 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

DISCUSSION

Many studies that require the simulation of streamflow use only observed streamflow to verify model output or to calibrate a hydrological model (Rosso, 1994; Jewitt & Schulze, 1999; Drogue et al., 2004; Forbes et al., 2010). With the ability of ACRU to output many key variables in the hydrological cycle, multi-variable verification was carried out. The purpose behind the verification of multiple ACRU outputs was to: (1) instill confidence that selected key variables other than streamflow, such as air temperature, snowpack, potential evapotranspiration, and glacier melt were appropriately simulated, (2) to reduce the uncertainty associated with sparse input data, especially climate and soils data, and (3) thus to ensure that the observed streamflow is simulated based on realistic selection of those bio-physical variables that represent and govern the hydrological processes in the watershed.

The objective of this research was not to simulate daily streamflow to perfectly fit the observed hydrographs, but rather to accurately simulate mean annual water yields, as well as variance, magnitude and seasonality of streamflows to replicate the hydrological behaviour of the study watersheds under historical conditions, thus enabling the realistic simulation of future streamflow behaviour.

The importance of multi-variable verification analysis is the interdependence of the variables validated. The first variable to be validated was temperature. Initial temperature estimations for each HRU were based on PRISM (Daly, 2006) surfaces and then calibrated to match the observed ELA of Peyto Glacier (Demuth & Keller, 2009; WGMS, 2009). Then the temperature time series were compared to ten independent climate observations within the larger UNSRB, resulting in satisfactory agreement of daily temperature magnitudes (on average within 0.37°C), seasonality (daily r2=0.88, monthly r2=0.98) and variance (difference of 4.8%).

Both the separation of precipitation into rain and snow and the timing of snow melt depend on daily air temperatures. The challenges of simulating snow pack is that local conditions, influenced by snow redistribution, sun shading effects, and especially the specific land cover composition and density at the monitoring site, cannot be simulated well when only very generalized data are available for model parameterization. The comparison of simulated SWE against observed snow pillow time series revealed that some years were heavily over- or under-simulated, while other years were well simulated. However, the means and variance of snow water equivalent (SWE) were preserved (Table 2), and comparison to streamflow time series indicated that, on a watershed basis, SWE simulation errors tended to balance each other out. The simulated SWE time series also matched the observed snow course data quite well. Neither the SWE nor the snow melt can be simulated to an acceptable level if the daily temperatures were not simulated correctly.

Simulated daily radiation values, based on GIS derived maximum potential incoming radiation and adjusted based on one observed daily time series, have resulted in realistic potential evapotranspiration time series, because the Penman (1948) derived potential evapotranspiration matched very well both estimates in mean annual magnitude (Hydrological Atlas of Canada, 1978) and seasonality, when compared to observed A-pan time series. As the Penman equation is commonly used

Page 13 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

to provide better estimates than pan data (Chiew & McMahon, 1991), available A-pan data were only used to confirm that the seasonality of potential evapotranspiration was simulated correctly. Radiation is also used, in combination with albedo, to derive the dynamic snow melt and glacier melt factors. Without the correct separation of precipitation into snow and rain, and without appropriate levels of daily potential evapotranspiration and melt factors, the timing and magnitude of streamflow could not have been simulated successfully for the right reason.

Statistical comparison of simulated against observed streamflows from Watersheds 1 and 2 are very acceptable for both daily and monthly time series for the 1961-90 baseline period. The Nash-Sutcliff coefficient is high for both watersheds, with daily values of 0.79 and 0.80 for Watersheds 1 and 2, and monthly values of 0.87 and 0.91, indicating a strong association with the observed streamflow time series. Mean annual water yields were either slightly under- or over-simulated for both watersheds and for all time series, but well within the initial simulation objective of being within 5% (Schulze et al., 2004). The simulated streamflow regime compares favourably with the observed for both watersheds. For the entire Cline River watershed, simulated mean monthly streamflows were simulated well when compared to the observed, with slight over-simulations in August, October, and November. Differences between simulated and observed variances were below 10% for Watershed 1 and below 4% for Watershed 2, indicating a realistic simulation of peak and low flows.

LIMITATIONS AND RECOMMENDATIONS

Data needed to properly model hydrological processes were limited in the UNSRB. Key model inputs, including land cover and soils, are generalized, and some variables, such as plant transpiration coefficients (PTCs), were nonexistent and needed to be determined to provide initial simulation parameters. The derived PTC values need to confirmed by other studies. In general, mountain areas have poor representative climate data, typically captured at lower elevation sites, which is the case in this study. Some climate data do exist at higher elevations, but these records are short, often seasonal, and inadequate for long-term climate studies (Luckman, 1998), however, they were used to verify temperature time series for high elevation HRUs.

This research has highlighted several data gaps that need to be filled to produce better and more reliable results from hydrological models in the UNSRB. Improved results can be expected when precipitation time series measured at high elevations become available. Critical bio-physical data, including detailed soils data, particularly soil texture and depth, a broad array of climate variables at different elevations, and several land cover variables such as forest densities, should be measured to improve modelling capabilities in this region. Acknowledgements We thank our colleagues, Dr. Hester Jiskoot, for her support on glacier mass balances, and Dr. Lawrence Flanagan, for his help in developing plant transpiration coefficients. This research was funded jointly by EPCOR Water Supply and the Natural Sciences and Engineering Research Council of Canada (NSERC), Grant #40286. Part of this research was also funded by the Alberta Ingenuity Centre for Water Research (AICWR), Grant #42321.

REFERENCES Ahrens, C. D. (2006) Meteorology Today: An Introduction to Weather, Climate, and the

Environment. Eighth Edition. Thompson, Brooks/Cole, California, USA.

Page 14 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

AmeriFlux (2010) http://public.ornl.gov/ameriflux/ [Date of access: July 2009]

Beven, K.J. (1989) Changing ideas in hydrology- the case of physically based models. J.

Hydrol. 105, 157-172.

Beven K.J. (2001) Rainfall-Runoff Modelling: The Primer. Wiley, Chichester, UK, 1-360.

Brutsaert, W. (1981) Evaporation into the atmosphere: theory, history, and applications. D. Reidel Publishing Company, Dordrecht, Holland.

Brooks, E.S., Boll, J. & McDaniel, P. (2007) Distributed and integrated response of a geographic information system-based hydrological model in the eastern Palouse region, Idaho. Hydrol. Processes 21, 110–122.

Burn, D.H. (1994) Hydrologic effects of climatic-change in West-Central Canada. J. Hydrol. 160, 53-70.

Byrne, J. & Kienzle, S.W. (2008) The Prairies. In: Lemmon, D, Warren, F., Bush, E. and Lacroix, J. (Eds.): From Impacts to Adaptation: Canada in a Changing Climate 2007. Report to the Government of Canada, Chapter 7, Ontario, Canada.

Canadell, J., Jackson, R.B., Ehleringer, J.R., Mooney, H.A., Sala, O.E. & Schulze, E.D. (1996) Maximum rooting depth of vegetation types at the global scale. Oecologia 108, 583-595.

Chiew, F.H.S. & McMahon, T.A. (1991) The applicability of Morton's and Penman's evapotranspiration estimates in rainfall-runoff modelling. J. Am. Water Resour. As. 27(4), 611-620.

Comeau, L.E.L., Pietroniro, A. & Demuth, M.N. (2009) Glacier Contribution to the North and South Saskatchewan Rivers. Hydrol. Processes 23, 2640-2653.

Daly, C. (2006) Guidelines for assessing the suitability of spatial climate data sets. Int. J.

Climatol. 26, 707-721.

Drogue, G., Pfister, L., Lefiandier, T., El Idrissi, A., Iffly, J.F., Matgen, P., Humbert, J. & Hoffman, J. (2004) Simulating the spatio-temporal variability of streamflow response to climate change scenarios in a mesoscale basin. J. Hydrol. 293, 255-269.

Demuth, M.N. & Pietroniro, A. (2003) The impact of climate change on the glaciers of the Canadian Rocky Mountain eastern slopes and implications for water resource-related adaptation in the Canadian Prairies: Phase I- Headwaters for the North Saskatchewan River Basin. Report for the Climate Change Action Fund- Prairie Adaptation Research Collaborative, 1-111.

Demuth, M.N. & Keller, R. (2009) An assessment of the mass balance of Peyto Glacier (1966–1995) and its relation to recent and past century climate variability. In: Demuth, M.N., Munro, D.S., Young, G.J. (Eds.), Peyto Glacier: One Century of Science, National Hydrology Research Institute Science Report No. 8, 83–132.

Diaz-Nieto, J. & Wilby, R.L. (2005) A comparison of statistical downscaling and climate change factor methods: impacts on low flows in the River Thames, United Kingdom. Climatic

Change 69(2-3), 245-268.

Environment Canada (2008) National Climate Data and Information Archive http://www.climate.weatheroffice.gc.ca/advanceSearch/searchHistoricData_e.html [Date of access: 3 September 2008]

Page 15 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://public.ornl.gov/ameriflux/

For Peer Review O

nly

ESRI (Environmental Systems Research Institute, Inc) (2008) ArcInfo (Version 9.3) [Computer software].

Everson, C.S. (2001) The water balance of a first order catchment in the montane grasslands of South Africa. J. Hydrol. 241, 110-123.

Flügel, W.A. (1995) Delineating hydrological response units by geographical information-system analyses for regional hydrological modelling using PRMS/MMS in the drainage-basin of the River Brol, Germany. Hydrol. Processes 9, 423-436.

Fluxnet Canada (2010) http://www.fluxnet-canada.ca/home.php?page=jobs&setLang=en [Date of access: July 2009]

Forbes, K.A. Kienzle, S.W., Coburn, C.A., Byrne, J.M. and Rasmussen, J. (2010) Simulating the hydrological response to predicted climate change on a watershed in southern Alberta, Canada. Climatic Change. DOI 10.1007/s10584-010-9890-x

Gan, T.Y. (1998) Hydroclimatic trends and possible climatic warming in the Canadian Prairies. Water Resour. Res. 34(11), 3009–3015.

Gerrard, J. (1990) Mountain environments: an examination of the physical geography of

mountains, MIT Press, Cambridge, Massachusetts, USA

Glassy, J. M. & Running, S.W. (1994) Validating diurnal climatology logic of the MT-CLIM model across a climatic gradient in Oregon. Ecol. Appl. 4, 248–257.

Government of Alberta: Sustainable Resource Development (2009) Forest and Vegetation Inventories: Alberta Vegetation Inventory http://www.srd.alberta.ca/MapsFormsPublications/Maps/ResourceDataProductCatalogue/ForestVegetationInventories.aspx. [Date of access: 19 October 2009].

Government of Canada: Agriculture and Rural Development (2003) Agroclimatic Atlas of Alberta: Maps http://www1.agric.gov.ab.ca/$department/deptdocs.nsf/all/sag7019 [Date of access: 16 April 2009].

Hopkinson, C. & Young, G.J. (1998) The effect of glacier wastage on the flow of the Bow River at Banff, Alberta, 1951-1993. Hydrol. Processes 12, 1745-1762.

Hornberger, G.M., Raffensperger, J.P., Wiberg, P.L. & Eshleman, K.N. (1998) Elements of

Physical Hydrology. The John Hopkins University Press, Baltimore, USA.

Hydrological Atlas of Canada. (1978) Canadian National Committee for the International

Hydrological Decade. Fisheries and Environment Canada, Ottawa, Canada, 1-34.

Jackson, R.B., Canadell, J., Ehleringer, J.R., Mooney, H.A., Sala, O.E. & Schulze, E.D. (1996) A global analysis of root distributions for terrestrial biomes. Oecologia 108, 389-411.

Jewitt, G.P.W. & Schulze, R.E. (1999) Verification of the ACRU model for forest hydrology applications. Water SA. 25, 483-489.

Jewitt, G.P.W., Garratt, J.A., Calder, I.R. & Fuller, L. (2004) Water resources planning and modelling tools for the assessment of land use change in the Luvuvhu Catchment, South Africa. Phys. Chem. Earth 29, 1233-1241.

Kienzle, S.W. (1993) Application of a GIS for simulating hydrological responses in developing regions. In: HydroGIS 93: Application of Geographical Information Systems in Hydrology and Water Resources Management (Proc. of the Vienna Conference, Austria, April 1993). IAHS Publ. 211, 309-318.

Page 16 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://www.fluxnet-canada.ca/home.php?page=jobs&setLang=en

http://www.srd.alberta.ca/MapsFormsPublications/Maps/ResourceDataProductCatalogue/ForestVegetationInventories.aspx

http://www.srd.alberta.ca/MapsFormsPublications/Maps/ResourceDataProductCatalogue/ForestVegetationInventories.aspx

http://www1.agric.gov.ab.ca/$department/deptdocs.nsf/all/sag7019

For Peer Review O

nly

Kienzle, S.W. (1996) Using DTMs and GIS to define input variables for hydrological and geomorphological analysis. In: HydroGIS 96: Application of Geographical Information Systems in Hydrology and Water Resources Management (Proc. of the Vienna Conference, Austria, April 1996). IAHS Publ. 235, 183-190.

Kienzle, S.W., Lorenz, S.A. & Schulze, R.E. (1997) Hydrology and Water Quality of the

Mgeni Catchment. Water Resources Commission, Pretoria, Report TT87/97, 1-88.

Kienzle, S.W. (2008) A new temperature based method to separate rain and snow. Hydrol.

Processes 22, 5067-5085.

Kienzle, S.W. & Schmidt, J. (2008) Hydrological impacts of irrigated agriculture in the Manuherikia Catchment, Otago, New Zealand. J. Hydrol. (NZ). 47, 67-83.

Kienzle, S.W. (2010) Effects of area under-estimations of sloped mountain terrain on simulated hydrological behaviour. Hydrol. Processes (submitted Jan. 2010)

Lapp, S., Byrne, J., Townshend, I. & Kienzle, S. (2005) Climate warming impacts on snowpack accumulation in an Alpine watershed: a GIS based modeling approach. Int. J.

Climatol. 25(3), 521–536.

Leavesley, G.H. (1994) Modelling the effects of climate change on water resources- A review. Climatic Change 28, 159-179.

Legates, D.R. & McCabe, G.J. (1999) Evaluating the use of “goodness-of-fit” measures in hydrologic and hydroclimatic model validation. Water Resour. Res. 35(1), 233-241.

Liu, Z., Xu, Z., Huang, J., Charles, S.P. & Fu, G. (2009) Impacts of climate change on hydrological processes in the headwater catchment of the Tarim River basin, China. Hydrol.

Processes 24, 196-208.

Loijens, H.S. (1974) Streamflow formation in the Mistaya River Basin, Rocky Mountains. In: Proc. 42

rd Western Snow Conference, (Anchorage, Alaska, USA) 86-95.

Loukas, A., Vasiliades, L. & Dalezios, N.R. (2002) Potential climate change impacts on flood producing mechanisms in southern British Columbia, Canada using the CGCMA1 simulation results. J. Hydrol. 259, 163-188.

Luckman, B.H. (1998) Landscape and climate change in the central Canadian Rockies during the 20th century. Can. Geogr. 42(4), 319-337.

Martinez, C.J., Campbell, K.L., Annable, M.D. & Kiker, G.A. (2008) An object-oriented hydrologic model for humid, shallow water-table environments. J. Hydrol. 351, 368-381.

NSIDC (National Snow and Ice Data Center) (2009) http://nsidc.org/data/modis/ [Date of access: 12 May 2009].

NLWI (National Land and Water Information Service) (2008) Data http://www4.agr.gc.ca/AAFC-AAC/display-afficher.do?id=1226330737632&lang=eng [Date of access: 10 September 2008]

New, M.G. & Schulze, R.E. (1996) Hydrologic sensitivity to climate change in the Langriver catchment, Stellenbosch, South Africa and some implications for geomorphic processes. Z.

Geomorephol 107, 11-34.

New, M.G. (2002) Climate change and water resources in the southwestern Cape, South Africa. S. Afr. J. Sci. 98, 1-8.

Page 17 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

http://nsidc.org/data/modis/

http://www4.agr.gc.ca/AAFC-AAC/display-afficher.do?id=1226330737632&lang=eng

For Peer Review O

nly

North Saskatchewan Watershed Alliance (NSWA). (2005) State of the North Saskatchewan

Watershed Report 2005 - A Foundation for Collaborative Watershed Management. North Saskatchewan Watershed Alliance, Edmonton, Alberta, 202.

Ommanney, C.S.L. 2002. Glaciers of the Canadian Rockies. In Ferrigno, J. and R.S. Williams, Jr, eds. Satellite image atlas of glaciers of the world. Denver, CO, United States Geological Survey, J199–J289. (USGS Professional Paper 1386-J.)

Penman, H.L. (1948) Natural evaporation from open water, bare soil and grass. P. Roy. Soc.

Lond.Ser-A. 193(1032), 120-145.

Peters, T.W., & Bowser, W.E. (1960). Soil Survey Report No. 19: Soil Survey of the Rocky

Mountain House Sheet. Canada Department of Agriculture, University of Alberta, Alberta, 1-51.

Peters, T.W. (1981). Soil Survey Report No. 40: Reconnaissance soil survey of the Brazeau

Dam area. Agriculture Canada, Land Resource Research Institute Branch, Ottawa, Ontario, 1-68.

Pettapiece, W.W. (1971). Land Classification and soils in the Rocky Mountains of Alberta

along the North Saskatchewan River Valley. Alberta Institute of Pedology, University of Alberta, 1-40.

Pomeroy, J.W. & Dion, K. (1996) Winter radiation extinction and reflection in a boreal pine canopy: measurements and modelling. Hydrol. Processes 10, 1591-1608.

Refsgaard, J.C. (1997) Parameterisation, calibration and validation of distributed hydrological models. J. Hydrol. 198, 69-97.

Rood, S.B., Samuelson, G.M., Weber, J.K. & Wywrot, K.A. (2005) Twentieth-century decline in streamflows from the hydrographic apex of North America. J. Hydrol. 306, 215-233.

Rosso, R. (1994) An introduction to spatially distributed modelling of basin response. In: Rosso, R., Peano, A., Becchil, I., Bemporad, G.A. (Eds.), Advances in Distributed Hydrology. Water Resources Publications, 3-30.

Sauchyn, D. & Kulshreshtha, S. (2008) Climate change impacts on Canada’s Prairie Provinces: A summary of our state of knowledge, from “Prairies” in From Impacts to

Adaptation: Canada in a Changing Climate 2007, edited by D. Lemmen et al., Government of Canada, Ottawa.

Schmidt, E.J. & Schulze, R.E. (1987) Flood volume and peak discharge from small catchments in southern Africa, based on the SCS technique. Water Resource Commission, Pretoria, Technical Transfer Report TT/3/87, 1-164.

Schulze, R.E. (1995) Hydrology and Agrohydrology: a text to accompany the ACRU 3.00

agrohydrological modelling system. Report TT69/95. Water Research Commission, Pretoria, RSA.

Schulze, R.E., Smithers, J.C., Lynch, S.D. & Lecler, N.L. (1995) Obtaining and interpreting output from ACRU. In: Smithers, J.C. and Schulze, R.E. ACRU Agrohydrological Modelling

System: User Manual Version 3.00. Water Research Commission, Pretoria, Report TT70/95. Pp AM7-1 to AM7-21.

Schulze, R.E., Lorentz, S., Kienzle, S.W. & Perks, L. (2004) Modelling the impacts of land-use and climate change on hydrological responses in the mixed underdeveloped / developed Mgeni catchment, South Africa. In: Kabat, P. et al. (Eds.): Vegetation, Water, Humans and

Page 18 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

the Climate A New Perspective on an Interactive System. BAHC-IGBP Publication, Springer, 17pp, with 14 Figures and 2 tables.

Shea, J. M., Moore, R. D. & Stahl, K. (2009) Derivation of melt factors from glacier mass balance records in western Canada. J. Glaciol. 55(189), 123-130.

Smith, C. D. (2009) The relationship between monthly precipitation and elevation in the Canadian Rocky Mountain Foothills. In: Proc. 77

th Western Snow Conference, (Canmore,

Alberta, Canada), 37-46.

Strong, W.L. & La Roi, G.H. (1983) Root-system morphology of common boreal forest trees in Alberta, Canada. Can. J. Forest Res. 13, 1164-1173.

Tetens, O. (1930) Über einige meteorologische Begriffe. Z. Geophys. 6, 297-309.

United States Department of Agriculture. (1985) National Engineering Handbook, Section 4, Hydrology. USDA Soil Conservation Service, Washington DC, USA.

Visscher, D., Frair, J., Fortin, D., Beyer, H. & Merrill, E. (2004) Spatial-temporal dynamics of elk forages in the central east slopes elk study area. In: Beyer, H., Frair, J., Visscher, D., Fortin, D., Merrill, E., Boyce, M., Allen, J. eds. , Vegetation map and dynamics of elk forage for the central east slopes elk and wolf study, Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada, 15-60.

WGMS. (2009) Glacier Mass Balance Bulletin No. 10 (2006-2007). Haeberli, W., Gärtner-Roer, I., Hoelzle, M., Paul, F. and Zemp, M. (eds.), ICSU(WDS)/IUGG(IACS)/UNEP/UNESCO/WMO, World Glacier Monitoring Service, Zurich, 1-96.

Xu, C.Y. (1999) Operational testing of a water balance model for predicting climate change impacts. Agr. Forest Meteorol. 98, 295-304.

Young, G.J. (1991) Hydrological interactions in the Mistaya basin, Alberta, Canada. In:

Proc. Vienna Symposium, Snow, hydrology and Forests in High Alpine Areas, (Vienna, Austria), 237-244.

Page 19 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Cline River watershed and the Upper North Saskatchewan River Basin, including hydropower reservoirs and locations of observed data for model verification.

279x215mm (100 x 100 DPI)

Page 20 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Major components of the ACRU agro-hydrological modelling system 102x78mm (300 x 300 DPI)

Page 21 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Monthly plant transpiration coefficients. 172x100mm (300 x 300 DPI)

Page 22 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Mean monthly simulated and observed temperatures for ten climate stations in the UNSRB. 79x62mm (300 x 300 DPI)

Page 23 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Simulated and observed soil water equivalent for the Limestone Snow Pillow site. 154x82mm (300 x 300 DPI)

Page 24 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Simulated and observed soil water equivalent (SWE) for the Nigel Creek snow course. 140x84mm (300 x 300 DPI)

Page 25 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Regression scatter plots for simulated and observed daily and monthly streamflow for Watershed 1 (1970-1990) and Watershed 2 (1961-1990). The solid line is the 1:1 line, the dotted line is the

slope of the regression. 120x116mm (300 x 300 DPI)

Page 26 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Simulated and observed hydrograph for 1988. 120x72mm (600 x 600 DPI)

Page 27 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Simulated and observed hydrographs for entire 1960-90 verification period, showing comparisons in both normal (for peak flows) and logarithmic (for low flows) scales.

143x109mm (300 x 300 DPI)

Page 28 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Simulated and observed mean monthly streamflows for the outlet of the Cline River watershed. 110x68mm (600 x 600 DPI)

Page 29 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Observed Sample Size (Months) 499 Simulated Sample Size (Months) 499

Observed Mean (°C) 0.40

Simulated Mean (°C) 0.77

P(T<=t) two-tail 0.46

Observed Variance 67.04

Simulated Variance 64.59

% Difference -3.79

Coefficient of Determination (r2) 0.98

Regression Coefficient (Slope) 0.97

Regression Intercept 0.39

Page 30 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Snow Course

Snow

Pillow

All

SWE

Observed Sample Size (Days) 882 7,625 8,507

Simulated Sample Size (Days) 882 7,625 8,507

Observed Mean (mm) 190.6 74.1 86.2

Simulated Mean (mm) 197.0 67.8 81.2

% Difference 3.2 -9.3 -6.1

P(T<=t) two-tail 0.45 0.00 0.01

Observed Variance 34,342 5,414 9,670

Simulated Variance 28,013 6,275 10,076

% Difference -22.59 13.71 4.03

Regression

Coefficient of Determination

(r2) 0.61 0.57 0.63

Regression Coefficient (Slope) 0.70 0.81 0.81

Regression Intercept 62.9 7.7 11.3

Regression Through Origin

Coefficient of Determination

(r2) 0.54 0.56 0.62

Regression Coefficient (Slope) 0.87 0.86 0.87

Page 31 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

ELA

(masl) Cumulative Mass

Balance (mm) Comments

Observed 2,612 -25,000 Values are for Peyto Glacier

Simulated 2,648 -22,824 Average of all glaciers

Page 32 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

For Peer Review O

nly

Watershed 1 Watershed 2

Daily Monthly Daily Monthly

Observed Sample Size (Days/Months) 7,510 248 10,957 360 Simulated Sample Size (Days/Months) 7,510 248 10,957 360

Observed Mean (m3/s) 53.6 53.3 81.2 80.8

Simulated Mean (m3/s) 53.4 53.3 83.4 82.9

% Difference -0.37 -0.03 2.64 2.57

P(T<=t) two-tail 0.84 0.99 0.09 0.74

Observed Variance 4,569 3,874 8,756 7,419

Simulated Variance 4,793 4,265 8,862 7,718

% Difference 4.7 9.2 1.2 3.9

Observed Standard Deviation 67.6 62.2 93.6 86.1

Simulated Standard Deviation 69.2 65.3 94.1 87.9

% Difference 2.3 4.7 0.6 2.0

Coefficient of Determination (r2) 0.81 0.89 0.82 0.91

Regression Coefficient (Slope) 0.88 0.90 0.90 0.94

Regression Intercept 6.82 5.49 6.22 3.00

Nash-Sutcliff Coefficient 0.79 0.87 0.80 0.91

Page 33 of 33



123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960