Modeling Mackenzie Basin Surface Water Balance during CAGES with the Canadian Regional Climate Model

M. D. MacKay Climate Research Branch, Meteorological Service of Canada, Toronto, Ontario, Canada

Search for other papers by M. D. MacKay in
Current site
Google Scholar
PubMed
Close
,
F. Seglenieks Department of Civil Engineering, University of Waterloo, Waterloo, Ontario, Canada

Search for other papers by F. Seglenieks in
Current site
Google Scholar
PubMed
Close
,
D. Verseghy Climate Research Branch, Meteorological Service of Canada, Toronto, Ontario, Canada

Search for other papers by D. Verseghy in
Current site
Google Scholar
PubMed
Close
,
E. D. Soulis Department of Civil Engineering, University of Waterloo, Waterloo, Ontario, Canada

Search for other papers by E. D. Soulis in
Current site
Google Scholar
PubMed
Close
,
K. R. Snelgrove Department of Civil Engineering, University of Manitoba, Winnipeg, Manitoba, Canada

Search for other papers by K. R. Snelgrove in
Current site
Google Scholar
PubMed
Close
,
A. Walker Climate Research Branch, Meteorological Service of Canada, Toronto, Ontario, Canada

Search for other papers by A. Walker in
Current site
Google Scholar
PubMed
Close
, and
K. Szeto Climate Research Branch, Meteorological Service of Canada, Toronto, Ontario, Canada

Search for other papers by K. Szeto in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The Canadian Regional Climate Model has been used to estimate surface water balance over the Mackenzie River basin during the water year 1998–99 in support of the Canadian Global Energy and Water Cycle Experiment (GEWEX) Enhanced Study (CAGES). The model makes use of a developmental third-generation physics parameterization package from the Canadian Centre for Climate Modelling and Analysis GCM, as well as a high-resolution land surface dataset. The surface water balance is simulated reasonably well, though Mackenzie basin annual mean daily maximum and minimum temperatures were both colder than observed by 1.7°C. The cold bias contributed to a longer snow-covered season and larger peak snow water equivalent than was observed, though snow accumulated realistically compared with two independently observed estimates after 1 November. Mackenzie basin annual precipitation was simulated as 496 mm, about 9% larger than observed, and PE was 225 mm. Net soil moisture change during this water year was found to be −26 mm, though because of a spinup problem in the Liard subbasin, the value is more likely closer to −14 mm.

The simulation was used to drive offline two different hydrologic models in order to simulate streamflow hydrographs at key stations within the Mackenzie basin. Results suggest that when subgrid-scale routing and interflow are included, streamflow timing is improved. This study highlights the importance of orographic processes and land surface initialization for climate modeling within the Mackenzie GEWEX Study.

Corresponding author address: Murray D. MacKay, Climate Research Branch, Meteorological Service of Canada, 4905 Dufferin St., Toronto, ON M3H 5T4, Canada. Email: murray.mackay@ec.gc.ca

Abstract

The Canadian Regional Climate Model has been used to estimate surface water balance over the Mackenzie River basin during the water year 1998–99 in support of the Canadian Global Energy and Water Cycle Experiment (GEWEX) Enhanced Study (CAGES). The model makes use of a developmental third-generation physics parameterization package from the Canadian Centre for Climate Modelling and Analysis GCM, as well as a high-resolution land surface dataset. The surface water balance is simulated reasonably well, though Mackenzie basin annual mean daily maximum and minimum temperatures were both colder than observed by 1.7°C. The cold bias contributed to a longer snow-covered season and larger peak snow water equivalent than was observed, though snow accumulated realistically compared with two independently observed estimates after 1 November. Mackenzie basin annual precipitation was simulated as 496 mm, about 9% larger than observed, and PE was 225 mm. Net soil moisture change during this water year was found to be −26 mm, though because of a spinup problem in the Liard subbasin, the value is more likely closer to −14 mm.

The simulation was used to drive offline two different hydrologic models in order to simulate streamflow hydrographs at key stations within the Mackenzie basin. Results suggest that when subgrid-scale routing and interflow are included, streamflow timing is improved. This study highlights the importance of orographic processes and land surface initialization for climate modeling within the Mackenzie GEWEX Study.

Corresponding author address: Murray D. MacKay, Climate Research Branch, Meteorological Service of Canada, 4905 Dufferin St., Toronto, ON M3H 5T4, Canada. Email: murray.mackay@ec.gc.ca

1. Introduction

The Canadian Global Energy and Water Cycle Experiment (GEWEX) Enhanced Study (CAGES) is a research program centered on a 14-month-long field campaign, running from the summer of 1998 to the fall of 1999, over the Mackenzie River basin. As a component of the Mackenzie GEWEX Study (MAGS; Stewart et al. 1998), its purpose is to improve our understanding of water and energy fluxes and reservoirs in the region during this water year (WY) through enhanced observing periods of targeted variables, as well as comprehensive modeling studies. In support of this, a developmental version of the Canadian Regional Climate Model (CRCM), coupled with a third-generation physics parameterization package of the Canadian Centre for Climate Modelling and Analysis (CCCma) GCM has simulated the entire period. The purpose of these simulations is twofold. First we take advantage of a variety of specialized MAGS-observed datasets to evaluate the performance of the model in a traditionally data-sparse area. Second, given a reasonably well-validated simulation we use the model to estimate the surface water budget during the CAGES water year (1 October 1998–30 September 1999). This approach allows us to estimate budget components (e.g., soil moisture) that are currently not observable at the Mackenzie basin scale.

High-latitude regional climate modeling is hampered both by a relative paucity of observed data (with which to evaluate simulations), as well as an incomplete understanding of physical processes relevant to this region—especially those related to snow cover (e.g., turbulent exchange over snow, local advection over patchy snow, etc.) and frozen soils. In this study we critically examine observed estimates of precipitation (P), screen-level (i.e., 2 m) temperature, and snow cover, before comparing with our simulated results. A companion paper (Feng et al. 2003, in this issue, hereafter FEN), examines solar and longwave (LW) radiation fluxes at the top of the atmosphere (TOA) and surface. Each of the observed estimates has uncertainty but, as discussed below, when taken in concert they begin to provide a consistent basis for model evaluation. For example, the absolute error in our regional estimates of snow water equivalent (SWE) based on remotely sensed passive microwave data is unknown. However, observed cumulative precipitation estimates, and changes in both observed and modeled SWE all agree remarkably well during the early winter period, lending credibility to all three.

The following section outlines our experimental setup and model configuration. After a careful evaluation of our simulation in section 3, section 4 presents an estimate of the Mackenzie basin water balance for the CAGES water year. Streamflow hydrographs were generated offline by driving two hydrological models with output from our simulation, and were compared with observations to gain further insight into our simulated water balance. Conclusions about our ability to simulate the Mackenzie basin surface water balance are presented in section 5.

2. Model configuration

The primary climate model of use within MAGS is the CRCM (Caya and Laprise 1999). This limited-area model has been driven at the lateral boundaries by both a GCM (Laprise et al. 1998; Laprise et al. 2003), and by operational analysis from a numerical weather prediction model (MacKay et al. 1998) over western Canada. The model in this study differs from that in these previous works in the physical parameterization package that is used. Recently, a developmental third-generation physical parameterization package of the CCCma has been introduced into the CRCM. MacKay et al. (2003) used this model in an examination of mesoscale circulations forced by heterogeneities in land cover, but they did not evaluate their simulations with respect to observations. In the present study we take advantage of CAGES and other MAGS-observed datasets in order to more carefully evaluate the performance of the model. In this study the CRCM is run at a horizontal resolution of 51 km (true at 60°N) with 29 levels in the vertical, 10 of which are below 850 hPa. A 15-min time step was employed. The entire domain (100 points × 90 points) is indicated in Fig. 1, along with elevation, the outline of the Mackenzie basin, and a nine-point sponge zone used to nest the model. Note that there are numerous lakes evident in Fig. 1, though, apart from Lake Superior, lakes are not represented in the CRCM simulations described here. The impact of lakes on regional climate is currently an active area of research within MAGS.

The new physical parameterization package incorporates version 2.7 of the Canadian Land Surface Scheme (CLASS). A detailed description of CLASS up to version 2.0 was presented in Verseghy (1991) and Verseghy et al. (1993). Version 2.7, finalized in 1997, contains a number of additional features, mainly a new set of surface stability functions (Abdella and McFarlane 1996), modifications to allow inhomogeneity between soil layers, and the incorporation of variable soil permeable depth. Briefly, CLASS models the energy and moisture balances of the soil, vegetation, and snow cover (if any) of the land surface. The soil is divided into three layers of thickness: 0.10, 0.25, and 3.75 m. Separate prognostic values of temperature and liquid and frozen soil moisture are carried for each layer. The vegetation and snow are treated as thermally and hydrologically separate from the soil, and energy and moisture exchanges are calculated based on physical principles. CLASS has participated in numerous international land surface scheme validation experiments, notably the Project for Intercomparison of Land Surface Schemes (PILPS) and a project comparing and validating the modeling of snow [Snow Models Intercomparison Project (SnowMIP)], and has been shown to generate physically reasonable land surface fluxes and budgets under a variety of conditions.

Lateral boundary and initial (atmospheric) conditions are specified from the operational global data assimilation system (i.e., operational analysis) of the Canadian Meteorological Centre, as in MacKay et al. (1998). In order to run CLASS, we also require land cover and soil data, gridded at the resolution of the model. The Climate Research Branch of the Meteorological Service of Canada (MSC) has assembled a high-resolution database over North America for this purpose, details of which can be found in MacKay et al. (2003). Briefly, vegetation cover is taken from the Canadian Centre for Remote Sensing (Cihlar et al. 1999) over Canada, as well as from the United States Geological Survey (USGS) (e.g., Loveland et al. 1995) and Olson global ecosystem (Olson 1994a,b) land cover classifications elsewhere. Soil texture profiles and depth to bedrock information are derived from the “Soil landscapes of Canada” (Centre for Land and Biological Resources Research 1996) for Canadian regions, and from the USGS soil data processed by the continental United States soil characteristics dataset, CONUS-SOIL (Miller and White 1998), for the U.S. regions. Elevation is specified from the GTOPO30 digital elevation model (Gesch et al. 1999), aggregated onto the model grid. Because a significant fraction of the domain is ocean covered, SST and ice cover are taken from the monthly 1°-resolution Hadley Centre Global Sea-Ice and Sea Surface Temperature (GISST) dataset GISST 2.3b (Parker et al. 1995), linearly interpolated in time to the model time step. Ice mass, which is also required by the model, is as used in the CCCma general circulation model (GCMII; McFarlane et al. 1992), reconciled with the observed ice cover fraction.

Apart from (fixed) geophysical information, CLASS requires initial values of soil moisture and temperature. This can be problematic given that very few observed estimates for these fields exist, especially over such large high-latitude domains. Yet a sufficiently unrealistic initial soil state could contaminate the simulation of surface climate for many years, while the land surface model “spins up” to reach its own equilibrium. Our strategy here was to begin our simulation on 1 April 1997 without snow cover, but with saturated soil moisture (based on porosity), and to spin up the model for 18 months prior to the CAGES water year. As is described below, this appears to have been appropriate for most of the Mackenzie basin, though a cluster of grid cells in the mountainous Liard subbasin clearly had not reached equilibrium, even by the end of the simulation, and this had an impact on the simulated hydrograph for this region. Temperature in the first soil layer was initialized to the mean atmospheric temperature near the surface from operational analysis (in which the model was nested) for the first day of the simulation. Initial temperature for the third (deepest) soil layer was taken from the annual average surface temperature of the Climatic Research Unit's half-degree monthly climate dataset (New et al. 2000). The second soil layer was initialized as the average of the first and third layers.

3. Surface climate evaluation

When attempting to validate a model simulation vis a vis “reality,” we must carefully assess the nature and quality of the available observational data. This is especially true in remote, data-sparse regions, such as the Mackenzie basin. Recently the MSC has produced a 50-yr gridded, monthly climate dataset of precipitation and screen temperature over the Mackenzie basin (Louie et al. 2002) for use within MAGS. Briefly, this dataset is based on operational climate station data, which have been adjusted for all known measurement errors. These adjustments, for example, result in an average increase of 20% in precipitation within the Mackenzie basin, eliminating the well-documented low bias of gauge measurements. The final gridded product is the sum of a gridded climate normal, based on the square grid technique of Solomon et al. (1968), and a gridded climate anomaly (Zhang et al. 2000). An updated version of the Louie et al. (2002) climate data, which now covers all of Canada and makes use of improved climate normal grids, has recently been produced. This dataset, known as CANGRID, is used here for comparison with the model.

Given the remoteness of the region, considerable effort has been expended on the development of satellite algorithms for the retrieval of surface information in MAGS. In this paper, snow cover extent [from the Advanced Very High Resolution Radiometer (AVHRR)] and snow water equivalent (SWE) [from the Special Sensor Microwave Imager (SSM/I)] have been estimated over the Mackenzie basin for comparison with the simulations. A detailed evaluation of modeled surface and top of the atmosphere radiation fluxes based on satellite estimates is presented in FEN.

a. Precipitation and temperature

The CANGRID surface climate datasets were produced on polar stereographic grids with a horizontal resolution of 50 km (true at 60°N). This is so close to the grid of our simulations (polar stereographic, 51-km resolutions, true at 60°N) that quantitative comparison—at least at the regional scale—is meaningful, and problems associated with comparing point observations (such as rain gauge measurements) with grid cell averages are avoided. There is, however, considerable uncertainty in these observed gridded estimates of surface climate, owing to the relative sparseness of station data and the complexity of the terrain in this remote, high-latitude region (e.g., Milewska et al. 2002). One measure of uncertainty is the range in average climate that is produced when using different gridding techniques. For example, in Table 1 we compare the 1961–90 climate normals of temperature and precipitation from four estimates of Mackenzie basin surface climate. Included are the CANGRID data, based on the technique described above, as well as estimates based on the inverse distance-weighting approach (Hopkinson 2000), and two estimates based on thin-plate spline interpolation [ANUSPLIN, McKenney et al. 1996; Climate Research Unit (CRU) New et al. 1998]. For the Mackenzie basin, mean annual temperature ranges from −3.4° to −4.7°C, a difference of well over 1°. However, the inverse distance-weighting approach does not take orography into account, unlike the other three datasets, and is known to perform poorly in mountainous regions. It is not surprising that this approach should yield a significantly warmer mean temperature for the Mackenzie basin—when excluded, the observed range drops to 0.2°C. Annual precipitation ranges from 375 to 463 mm—a range of around 20%—with CANGRID, indicating the largest value. This is not too surprising because only CANGRID attempted to correct for gauge undercatch—a known problem, especially in northern regions. Note that all of these climatologies are based on similar sparse-climate station networks, which may not adequately represent the regional climate over much of the Mackenzie basin (Milewska and Hogg 2001). It is hoped that by focusing on Mackenzie basinwide averages, rather than local or regional values, the uncertainty related to network sparseness is minimized.

Minimum and maximum screen-level temperatures (Tmin, Tmax) from the CAGES water year simulation and CANGRID are compared in Figs. 2 and 3. Both the simulated annual mean daily minimum temperature (−9.6°C) and simulated annual mean daily maximum temperature (0.4°C) were too cold by 1.7°C, though, as indicated in Figs. 2b and 3b, the seasonal pattern in the Mackenzie basin monthly averaged temperature biases differed. In particular, the simulated monthly average Tmax tended to be close to that observed from March to May, but too cold otherwise. On the other hand, Tmin was too cold throughout the year with the exception of December and January.

A detailed analysis of top-of-atmosphere (TOA) outgoing solar radiation, and net surface solar radiation (NSSR) from this simulation (FEN) helps shed some light on these temperature biases. FEN found a simulated outgoing solar flux excess of about 31 W m−2 compared with satellite estimates during summer over the Mackenzie basin, which they attributed to excessive cloudiness in the model. This excessive outgoing flux was seen to occur largely at the expense of net surface solar radiation, which was found to be deficient by about 33 W m−2 (the remaining 2 W m−2 being an atmospheric absorption bias), and which likely contributed to the summertime surface cold bias in Tmax (summertime average bias, −2.1°C). During the spring a small positive bias in simulated NSSR was found (4 W m−2), and the corresponding bias in Tmax was only +0.6°C. On the other hand, even though the bias in NSSR during October is positive, there is a significant cold bias in Tmax (−2.2°C).

Satellite estimates of TOA outgoing longwave radiation (November 1998–March 1999) were also compared with this simulation in FEN. While the nighttime bias remains more or less steady during this period, the daytime bias steadily improves after the winter solstice, with increasing sun angles over the basin. An examination of basin-averaged Tmax and Tmin during this period shows that while Tmin indicates a more or less constant cold bias during spring, Tmax improves dramatically from March until June. This pattern is consistent with the LW flux bias, and may be reflecting a known nocturnal surface decoupling problem in this version of CLASS (e.g., Delage et al. 2002).

Like temperature, precipitation appears to be reasonably well distributed when both spatially and temporally compared with CANGRID (Fig. 4), with a mean bias of less than 9% overall (496 mm, simulated; 457 mm, observed). As discussed below, a cold wet bias in October greatly affected the simulated snowpack evolution and spring freshet. Determination of whether or not this autumn bias resulted from a systematic misplacement of the Aleutian low and resulting storm tracks (e.g., MacKay et al. 1998) must await analysis of a much longer simulation.

b. Snow cover

Snow cover is an important variable to validate in climate models because it directly influences both the hydrological cycle, as short-term surface storage, as well as the surface energy cycle, through its effect on albedo. Observed estimates of snow cover fraction over the basin are compared with simulated estimates in Fig. 5. The observed estimates were derived from the National Oceanic and Atmospheric Administration (NOAA) National Environmental Satellite, Data, and Information Service (NESDIS) weekly Northern Hemisphere snow charts that are generated from manual interpretation of AVHRR, Geostationary Operational Environmental Satellite (GOES), and other visible-band satellite data (Robinson et al. 1993). For comparison, an estimate based on SSM/I passive microwave data (see below) is also shown, indicating a high degree of consistency between the observed estimates. The simulated estimates assume a snow cover threshold of 10 kg m−2 (representing about 10 cm of freshly fallen snow), approximately the sensitivity of the observed estimates. Figure 5 indicates that the Mackenzie basin is more widely snow covered during October in the simulation than observations would suggest (though it is known that NOAA snow charts can underestimate extent in autumn, e.g., Wiesnet et al. 1987). As can be seen in Figs. 24, and as discussed below, this resulted from a simulated cold, wet bias in October that produced snowfall when rain was expected. The excessive October precipitation also resulted in a longer snow-covered season for the Mackenzie basin.

Perhaps of more direct relevance to the surface hydrological cycle is the snow mass. SSM/I passive microwave satellite data have been used to derive regional SWE in the prairie region of western Canada for many years (Goodison and Walker 1995). Algorithms used to translate the SSM/I brightness temperatures into values of SWE (mm) have recently been developed for the boreal forest region in western Canada (Goita et al. 1997) and are being tested for the Mackenzie basin as a research investigation within MAGS (Walker and Silis 2002). For the Mackenzie basin area, four algorithms are used to derive a weighted SWE value for each grid point based on the fractions of (i) coniferous, (ii) deciduous, (iii) sparse forest, and (iv) open (nonvegetated) land cover types within the grid square. A detailed evaluation of this approach for the Mackenzie basin is not presented here. However, Fig. 6 compares individual pixels (25 km × 25 km resolution) of SSM/I-derived SWE with nearby snow course measurements at various locations throughtout the Mackenzie basin for the period January–May 1999. Figure 6a indicates bias (SSM/I − snow course) as a function of elevation. The dominant vegetation type for each pixel is also indicated. It is well known that increasingly rough terrain (generally associated with increasing elevation) increasingly degrades passive microwave estimates of SWE, partly the result of an apparent increase in microwave emission due to multiple reflections off of the rough surface. Figure 6a suggests that there is little useful SSM/I-derived SWE information above about 800 m, which we take as our cutoff elevation for the purposes of this study. For points below 800 m, Fig. 6b compares SSM/I-derived SWE with snow course measurements in a scatterplot, with dominant vegetation type again indicated. While the sample is quite small, and not necessarily representative of the Mackenzie basin as a whole (e.g., snow course measurements are not evenly distributed across the basin), it does suggest that the SSM/I-derived SWE tends to underestimate the observed on average by about 27 mm for this year. It is also evident that, while there is some scatter for all of the dominant vegetation types, pixels that are dominated by coniferous forest tend to yield the worst correlation, with few pixels estimating SWE larger than 50 mm, even for observations 3 times as large. This is perhaps not too surprising, given that snow lying beneath a dense, vegetated canopy would be extremely difficult to detect via passive microwave techniques. It does, however, represent a challenge for MAGS, because a significant fraction of the Mackenzie basin is covered by coniferous vegetation.

Time series of three selected clusters of snow course data are compared with nearby pixels of SSM/I-derived SWE, as well as nearby simulated grid cells in Fig. 7. Table 2 describes the basic surface features of these three regions, including the number of snow courses, mean elevation, and mean vegetation coverage. Region 1 (Fig. 7a) is dominated by deciduous vegetation; thus, the SSM/I-derived SWE estimate is dominated by the “deciduous” algorithm. Figure 7a suggests that the SSM/I-estimated, -observed, and -simulated SWE are in reasonably good agreement throughout the year, though it is of course difficult to meaningfully compare point measurements with large grid cell averages. Region 2 (Fig. 7b) is largely “open.” In this case the SSM/I estimate tends to underestimate the snow course data in late winter, but appears to be quite realistic by April. Again, the simulation appears to follow the snow course data quite well, though it may be starting the melt period a little early here. Finally, the third region (Fig. 7c) is dominated by coniferous vegetation. As was suggested above, the SSM/I estimates for areas dominated by coniferous vegetation tend to saturate near 50 mm, though the observations indicate about twice this amount (and the simulation 4 times this amount).

Though the above analysis is indeed suggestive, given the very limited (both spatially and temporally) snow course data available for the Mackenzie basin it is difficult to evaluate SSM/I-derived estimates of SWE over such a large region. On the other hand, we do have gridded, monthly estimates of precipitation for the region, and another approach is to compare wintertime-accumulated precipitation with the growth in SWE during the same period. We begin our accumulation period for the Mackenzie basin lowlands (below 800 m) on 1 November, because temperatures will be sufficiently cold to preclude any significant rainfall or snowmelt after this date until the spring melt period begins. Figure 8 compares accumulated precipitation for November and December 1998 from CANGRID, with SWE accumulation estimated from SSM/I for this period. The total accumulations of observed precipitation and SWE agree quite well for both months, though the spatial distributions can differ. For example, by 1 January the southwestern region of the Mackenzie basin lowlands has a large accumulation of precipitation but, apparently, a low accumulation of SWE. Because this region is dominated by coniferous vegetation, this further supports our hypothesis that our coniferous algorithm is underestimating the actual SWE.

Figure 9 shows normalized frequency histograms of accumulated precipitation (CANGRID) and accumulated SWE (SSM/I). Accumulation begins on 1 November and ends 1 January (Fig. 9a), 1 February (Fig. 9b), and 1 March (Fig. 9c). In each case the histograms are normalized by the number of pixels in each grid (about 1850 pixels for the SSM/I grid and 491 pixels for CANGRID cover the Mackenzie basin lowlands) so that the area under each curve in Fig. 9 is unity. Figure 9a shows that on 1 January the accumulated CANGRID precipitation and SSM/I-accumulated SWE have nearly the same mean (43 mm for CANGRID, 47 mm for SSM/I) and very similar distributions. However, over the next 2 months the distributions increasingly diverge. It is clear that the SSM/I algorithms saturate at about 100 mm, about the level found in previous studies (e.g., Walker and Silis 2002; De Sève et al. 1997; Armstrong et al. 1993), which distorts the shape and reduces the mean of the SSM/I-derived SWE, compared to the CANGRID precipitation, distribution after January. Also evident, especially on 1 March (Fig. 9c), is an excessive area of SWE less than 50 mm compared with the accumulated precipitation, which is presumably the result of the “coniferous” algorithm saturating near 50 mm, as suggested above. This will also contribute to an SSM/I underestimation of SWE after January.

After this brief examination of the SSM/I SWE retrieval for the Mackenzie basin lowlands, we are now in a position to evaluate the evolution of our simulated snowpack over this region. Figure 10a presents a time series of simulated (solid curve) and biweekly SSM/I-derived SWE (asterisks), averaged over the Mackenzie basin below 800 m. Also indicated is the cumulative (commencing 1 October) precipitation from the simulation (dotted curve) and CANGRID (dashed curve). A number of features are readily apparent, the first and foremost being that the model appears to overestimate SWE throughout the year. This is largely due to the model simulating excessive precipitation (nearly all of which is falling as snow) in October. For the Mackenzie basin below 800 m, the model produced 47 mm of precipitation in October, compared with 30 mm indicated by CANGRID over the same region, though, the two datasets agreed relatively well after this month. The model also simulated a mean October surface temperature of −3.7°C compared with the CANGRID estimate of 0.3°C. In the simulation by the end of October, either a small amount of rain has fallen, or an episode of melting has occurred, because the accumulated precipitation exceeds the SWE. However, after this, the simulated snowpack follows very closely the accumulated precipitation, indicating that the model is not producing any noticeable midwinter snowpack ablation due to sublimation processes.

Such a result is not new; in a study of snowpack processes in 21 land surface schemes under the PILPS 2(d) experiment, Slater et al. (2001) found that midseason ablation varied widely but was consistently underestimated in the models. They did not identify the underlying reasons for this, though they did cite the interplay between albedo and fractional snow coverage, as well as the formulation of aerodynamic transfer coefficients as candidates for further study. The actual extent of midwinter sublimation in the Mackenzie basin is unclear, though some observational studies in the Canadian boreal forest suggest that it can be significant at least locally (Harding and Pomeroy 1996; Pomeroy et al. 1998).

Returning our attention to the observations in Fig. 10a, the October-accumulated precipitation from CANGRID (30 mm) greatly exceeds the SSM/I estimate of SWE (about 8 mm). If both are reasonably accurate then this indicates that much of the precipitation in October fell as rain (not unreasonable given a mean observed temperature of 0.3°C), as opposed to snow, as was found in the simulation. However, after October the snowpack and precipitation for both the simulation and the observations appear to accumulate all at the same rate (at least until January or so). This is more clear in Fig. 10b, where we have reproduced Fig. 10a, except we now show accumulations of all fields commencing on 1 November. Figure 10b suggests a very high degree of consistency between the simulated accumulation of SWE, and two entirely independent observed datasets, from 1 November to perhaps as late as 1 February. After this point the SSM/I estimates appear to have saturated, while the simulated SWE continues to accumulate more or less at the rate of observed precipitation accumulation until about 15 March when the melt season begins in earnest. Note that this result holds for the Mackenzie basin lowlands on average: local differences between simulated and observed snow cover can of course exist (e.g., Fig. 7c). It is also clear from Fig. 10b just how closely the simulated SWE accumulation follows the simulated precipitation accumulation (i.e., after 1 November), the two diverging only at the onset of melt. However, the simulated SWE (accumulation) exceeds the SSM/I estimates during May by about 20 mm. If these observed estimates are correct, and the melt period is simulated reasonably well (note that simulated and observed temperatures for this region are virtually identical from March to June), then this may be indicating a midwinter snowpack ablation of as much as 20 mm (about 20% of the peak SWE). Given that SSM/I estimates during the melt period are particularly problematic and tend to underestimate the actual SWE value, this should be taken as an absolute upper limit.

4. Surface water balance

The simulated annual Mackenzie basin surface water balance is summarized in Fig. 11. An average precipitation minus evaporation (Fig. 11a) of 225 mm was simulated, most of which occurred over the mountainous western part of the basin. Unfortunately, both our observed datasets and simulated results may be least reliable over this region, due to a general lack of observations on the one hand and absence of orographic process parameterizations (e.g., Leung and Ghan 1998) on the other. As noted above, the observed annual Mackenzie basin precipitation was estimated as 457 mm, but we have no reliable estimates for surface evapotranspiration (E) at this scale for this year. Louie et al. (2002) suggest a climatological (1961–90) evapotranspiration of 277 mm, which, if representative of the CAGES water year, suggests an observed PE of 180 mm, yielding a simulated bias of 45 mm (25%). As pointed out in Louie et al. (2002), however, the observed estimate of evapotranspiration (which is based on Morton's method) could be too high by as much as 10%, so the simulated bias here may be somewhat less. Also, it is of course not at all clear that the actual evapotranspiration for this year was close to the climatological estimate. All of this does suggest that our simulated PE is at least plausible.

Net change in soil moisture for this water year is indicated in Fig. 11b. For most of the Mackenzie basin soil moisture changes are modest, with a basin average of −26 mm. Notable exceptions occur in the Liard subbasin (central-western part of the Mackenzie basin), where several grid cells appear to be draining large amounts of water—in some cases more than 550 mm over the course of the year. Evidently these grid cells have not completely spun up with respect to their initial soil moisture, which, as noted above, was set to saturation. Figure 12a shows a time series of total soil moisture change from October 1997 to September 1999, averaged over the Liard subbasin (solid curve) and, for contrast, the Athabasca subbasin (dashed curve). The decreasing trend in total soil moisture in both regions is quite apparent, with both regions losing about 200 mm of soil moisture during this period (Liard: 187 mm, Athabasca: 213 mm), while the land surface spins up. However, the timing of this drainage varies significantly. Because the soil moisture content is so high (at least initially), significant drainage can only occur when the third (deepest)-layer soil temperature is above freezing. Figure 12b shows a time series of mean third-layer soil temperature for these two subbasins for this period. The Liard subbasin (third soil layer) is below 0°C for much of the 1997–98 water year, thus, it does not have the opportunity to drain all of its excess initial soil moisture until the 1998–99 water year. In particular, it drains 91 mm during WY 1997–98, and 96 mm during WY 1998–99. On the other hand, the Athabasca subbasin is above 0°C on average for longer periods, and manages to drain 201 mm of excess water during WY 1997–98, and only 12 mm during WY 1998–99. It is clear that the Liard subbasin on average has not equilibrated with respect to its land surface initial conditions before our water year of interest. On the other hand, the Athabasca subbasin is, if not already equilibrated, then at least much closer. Even though the Liard subbasin is only 16% of the total Mackenzie basin area, it accounts for more than half of the total volume of lost soil moisture. If we exclude this region, the average soil moisture change is only −14 mm, which may be a more representative estimate of the Mackenzie basin as a whole.

This analysis highlights the particular caution that must be taken when initializing land surface models in high latitudes. Any excessive initial soil moisture that was frozen would need to melt in order to drain away during the spinup phase of the simulation. This melt energy ultimately comes from net radiation, which can be very low indeed at high latitudes. Thus, high-latitude land surface model initialization and spinup requires careful consideration and evaluation. The problem of spinup was examined in the PILPS experiment (Yang et al. 1995) for 22 land surface models. This study found that spinup times depended on initial soil moisture and total soil moisture holding capacity, both of which are relatively uncertain throughout much of the Mackenzie basin. While initializing too wet was generally found to be better (i.e., require less spinup time) than initializing too dry, an accurate representation of precipitation and solar radiation was also important in the spinup process. Given excessive simulated precipitation and relatively low net radiation in the Liard basin, it is clear that initialization to saturation is not the most appropriate choice in this region for a short simulation, such as presented here.

Returning now to Fig. 11, surface and total runoff (surface runoff + deep drainage) for our CAGES simulation are shown in Figs. 11c and 11d, respectively. In the current implementation of CLASS, surface runoff occurs when PE exceeds the infiltration capacity of soil and a maximum surface pond depth is reached (Verseghy 1991). The vertical flux of water in the soil column is governed by Darcy's law, and deep drainage to the water table can occur if an impermeable soil layer has not been reached. Lateral movement of water, in particular interflow, is not represented in this model version, but is currently under development. The potential impact of the neglect of interflow is discussed below. As with PE, runoff is primarily generated in the mountainous region to the west, with total runoff reflecting the problematic grid cells in the Liard subbasin, as mentioned above. Annual totals of surface and total runoff are 132 and 246 mm, respectively. Note that the apparent water balance residual PE − ΔSR = 5 mm represents a 5-mm change in the basin average SWE during this water year. The Mackenzie basin surface water balance for the CAGES water year is summarized in Table 3.

We generally do not have at our disposal observed estimates of these components of the surface water balance over such a large scale. However, one can gain insight into this simulated water balance by coupling offline to hydrological models and comparing with hydrographs observed at key stations within the basin. In this study we consider two such models. First we simply route the simulated grid cell runoff with the University of Waterloo's Watroute channel-routing scheme (e.g., Arora et al. 2001), in which excess grid cell water is instantaneously put into stream channels and routed based on Manning's equation. Such an approach was taken by Arora and Boer (1999), who successfully routed runoff from the CCCma atmospheric GCM. In the second approach, we use the hydrologic model Watflood (Kouwen et al. 1993), forced with our simulated precipitation and temperature. In addition to using Watroute for the grid-scale routing component, Watflood also represents subgrid-scale routing and interflow. Watflood was calibrated for the Mackenzie basin by adjusting parameters to minimize the square of the difference between the measured and simulated flows over a subset of this basin for 2 yr (1995–96) prior to this experiment. The drainage network used in both approaches is presented in Fig. 13. Two large lakes in the basin (Great Bear and Great Slave Lakes) are modeled using a storage/outflow power function.

Two particular features of the Mackenzie basin require special mention. These include the W. A. C. Bennett dam in the southwest portion of the basin and the Peace–Athabasca Delta (PAD), located south of Great Slave Lake. The Bennett dam imposes significant regulation on the flows of the Peace River through the water storage potential of Williston Lake. Through this control, spring snowmelt from the Rocky Mountains is captured and used for electrical power production during the winter period. So as not to impose a penalty on the accuracy of the land surface runoff scheme in this area, measured outflows from the Bennett dam are provided to the Peace River at the dam site rather than the routed runoff amounts.

A second feature, which could potentially lead to spurious results in Watroute, is the Peace–Athabasca delta system. Under low- and normal-flow situations the Peace River and the outflow from Lake Athabasca converge at the Slave River, which in turn flows northward to Great Slave Lake. During high-flow events on the Peace River and/or ice restrictions on the Slave River, the Peace River flow may be diverted from its normal course and into Lake Athabasca and the surrounding delta area. This flow reversal has the effect of raising the water level of the lake and delta system, increasing local storage (Leconte et al. 2001) and decreasing outflow to the Slave River. These events are normally confined to short periods during the spring breakup with water stored temporarily in Lake Athabasca released once ice jamming on the Slave River has eased. Events of this type have had a reduced frequency since the construction of the Bennett dam and flow regulation on the Peace River, with only four flood events occurring in the 25 yr since water impoundment began in 1968 (Prowse and Conly 1998). Given that these events occur rarely, are difficult to predict, and are outside the scope of Watroute's hydraulic theory, their impact on Mackenzie River discharge is disregarded.

The Mackenzie basin has a number of streamflow gauges at our disposal, and in this study we examine hydrographs from the Athabasca River at Athabasca, Alberta, the Smoky River at Watino, Alberta, the Liard River at Fort Liard, Northwest Territories, and, finally, the Mackenzie River at Arctic Red River, Northwest Territories, which represents outflow from about 95% of the entire Mackenzie basin (see Fig. 13). Simulated and observed hydrographs at these stations are shown in Fig. 14. Results for the Athabasca and Smoky Rivers are similar. For the Athabasca River (Fig. 14a), total annual simulated flow volume for both models was 2.2 × 1010 m3, compared with 1.3 × 1010 m3 observed. The excessive simulated flow may be indicating excessive precipitation locally. The simulated hydrographs, though producing the same net flow volume, have qualitatively different shapes. In the Watroute simulation the spring freshet begins and peaks too early, compared with what was observed. On the other hand, the Watflood result shows much better timing, with the excessive discharge occurring much later in the season. This may be the result of Watflood's representation of subgrid-scale routing and interflow, though there are other differences in the energy budgets of CLASS and Watflood as well. Note that surface flow and interflow are approximately one and two orders of magnitude slower than channel flow, respectively. In Watroute, on the other hand, both the overland runoff and deep drainage from CLASS are instantaneously put into stream channels. Results for the Smoky River are shown in Fig. 14b. Again, the simulations produced similar total flow volumes (1.7 × 1010 m3 for Watroute, 1.6 × 1010 m3 for Watflood), which in this case is about twice that of the observed (8.5 × 109 m3). Again, the timing and magnitude of the spring freshet is much better in the Watflood run compared with the Watroute run, with the excessive discharge occurring only later in the summer. The situation for the Liard River (Fig. 14c) is somewhat more complex. In this case the Watflood and Watroute simulations have very different flow volumes. The Watflood flow (7.9 × 1010 m3) exceeds the observed estimate (5.6 × 1010 m3) by 2.3 × 1010 m3, most of which can be explained by a CRCM precipitation bias for this subbasin (compared with CANGRID) of 65 mm, corresponding to a total volume of 1.8 × 1010 m3. The simulated freshet is late here, but the peak discharge is not grossly in error. On the other hand, the Watroute simulation produces a greatly exaggerated freshet, and a total flow volume more than twice that of the observed (12.3 × 1010 m3). In this case, the anomalous change in soil moisture for the Liard subbasin noted above (96 mm) contributes 2.6 × 1010 m3 to this. Because soil moisture is not directly used by Watflood (storage is parameterized based on flow), the problem of land surface pixels in the Liard subbasin spinning up by releasing a large volume of water does not occur in this case. Finally, hydrographs for the Mackenzie River at Arctic Red are shown in Fig. 14d. At this scale, total flow volumes of both simulations (Watroute, 2.70 × 1011 m3; Watflood, 2.65 × 1011 m3) are within about 1% of the observed value (2.68 × 1011 m3). In the context of Figs. 14a–c, this suggests that Mackenzie basin PE may be reasonably well simulated at the large scale, though poorly spatially distributed within the Mackenzie basin, and with a significant positive bias in mountainous regions. Note that errors in the Watroute case appear to be dominated by errors in the Liard subbasin; observations suggest that the Liard contributes 21% of the total Mackenzie River flow (at Arctic Red), while in the Watroute simulation it is 46%. This certainly highlights the Liard subbasin as a region requiring particular attention when attempting to simulate Mackenzie River hydrographs.

5. Conclusions

The Canadian Regional Climate Model has been used to estimate surface water balance over the Mackenzie basin during the water year 1998–99 in support of the Canadian GEWEX Enhanced Study. The model makes use of a developmental third-generation physics parameterization package from the Canadian Centre for Climate Modelling and Analysis. Among numerous enhancements to previous versions, this package also includes the Canadian Land Surface Scheme, and to fully exploit this, a high-resolution land surface geophysical dataset has been assembled for North America. Soil moisture and temperature were initialized based on saturation and climatology, respectively, and an 18-month model spinup was allowed prior to the CAGES period. Lateral atmospheric boundary conditions were specified by operational global analysis from the Canadian Meteorological Centre.

The simulation is first evaluated against several specialized datasets, developed under the Mackenzie GEWEX Study in this relatively data-sparse area. Simulated screen-level (i.e., 2 m) temperature is compared with gridded estimates from the CANGRID dataset of the MSC. Mackenzie basin annual mean daily minimum (−9.6°C) and daily maximum (0.4°C) temperatures were both found to be too cold by 1.7°C, compared with the observed estimates. A detailed examination of solar and longwave radiation fluxes from this simulation compared with satellite estimates in a companion paper (FEN) helps shed some light on these temperature biases. This study found that a positive summertime cloud cover bias produced excessive TOA outgoing solar radiation fluxes, resulting in a deficit of net surface solar radiation and cold surface temperatures during the day (i.e., Tmax). During spring, a weak positive bias in NSSR is consistent with a slight warm bias at the surface in Tmax. However, radiation cannot always explain the bias in Tmax: a significant positive bias in NSSR occurred during October when Tmax was more than 2°C too cold. Evidently spurious advection of cold air into the region took place during this month, perhaps related to a misplaced Aleutian low. Negative outgoing longwave flux biases at the TOA are also consistent with a cold surface temperature bias. During nighttime (November–March) longwave fluxes at the TOA were consistently underestimated, and Tmin was generally (though not always) too cold. On the other hand, with the onset of spring and higher sun angles over the basin, both the daytime longwave flux estimates (TOA) and Tmax improved. This suggests that the nocturnal surface decoupling problem found in Delage et al. (2002) may be active here.

Simulated Mackenzie basin average precipitation was 496 mm, compared with the 457 mm suggested by the CANGRID dataset, a bias of less than 9% overall. The spatial and temporal distribution of precipitation appear to reflect the CANGRID estimates reasonably well, though we note that both observed and simulated estimates of precipitation are less reliable in mountainous regions than elsewhere. Furthermore, a comparison of observed and simulated hydrographs at key locations suggest excessive orographic precipitation in the model.

Mackenzie basin snow cover extent was evaluated against a gridded snow extent dataset based on operational NOAA NESDIS weekly Northern Hemisphere snow charts. The longer simulated snow-covered season was found to result from a simulated cold, wet bias in October in which precipitation that fell as rain was simulated to fall as snow. Simulated SWE was also evaluated for the Mackenzie basin lowlands (below 800-m elevation) against observed estimates based on satellite passive microwave data. Because of the simulated cold, wet bias in October, the simulated SWE exceeds the observed estimates throughout the year, highlighting the importance of autumn temperature and precipitation simulation for the seasonal snowpack. When one considers the snowpack accumulation from 1 November onward, we find that the model simulates growth of the snowpack in very close agreement with observations of both accumulated precipitation and SWE until January or February, after which time the SSM/I-derived SWE estimates appear to saturate. After this time the model continues to accumulate SWE at the simulated precipitation rate; no midwinter ablation due to sublimation processes takes place. The importance of midwinter ablation in the Mackenzie basin is currently a matter of debate. If the simulated springmelt period and observed SSM/I-derived SWE estimates in May are reasonably accurate, then results here suggest that midwinter ablation can be no more than about 20% of the peak-simulated SWE (about 20 mm), and is probably considerably less, considering that passive microwave SWE estimates are questionable in spring, probably underestimating SWE in conditions of wet snow. The good agreement between the entirely independent observations of accumulated precipitation (from CANGRID) and accumulated SWE (based on SSM/I) suggests that early winter snowpack ablation is not significant at this scale, at least for this year.

The Mackenzie basin water balance was also estimated for the CAGES water year from this simulation. Mackenzie basin average PE was found to be 225 mm, with largest values occurring in the mountainous western region. There are no reliable observed estimates of evapotranspiration at this scale for this year, but a climatological estimate from Louie et al. (2002) suggests that our simulated PE is too large by no more than about 25%. Basin average soil moisture change was found to be −26 mm. Again, no observed estimates exist at this scale, but this is very likely excessively negative because numerous grid cells, particularly in the Liard subbasin, had evidently not yet equilibrated with respect to initial conditions (which were saturated), and were losing very large volumes of water as they continued to spin up. If one ignores the Liard subbasin and considers the remainder of the Mackenzie basin as representative of the whole Mackenzie basin, then a total soil moisture change of −14 mm is suggested.

Simulated streamflow is estimated based on two different hydrological models and is compared with observed streamflow measurements at key stations within the basin. Two approaches were taken: in the first, grid cell runoff is simply transferred instantaneously into stream channels and routed, based on Manning's equation with the University of Waterloo's Watroute channel-routing scheme. In the second approach, the hydrologic model Watflood (driven with CRCM temperature and precipitation) was used. In addition to using the grid-scale channel-routing scheme of Watroute, Watflood also estimates subgrid-scale surface and interflow. The two approaches were found to produce different results. For example, for the Athabasca and Smoky Rivers the subgrid-scale surface flow and interflow representation in Watflood may have improved the timing of the spring freshet due to the fact that surface and interflow generally occur much more slowly than channel flow. Hydrographs for the Liard River indicated that the surface storage spinup problem noted for this subbasin had a dramatic impact on the WATROUTE simulation. All three of these hydrographs suggest excessive total simulated flow volumes, compared with observed values. However, both simulated hydrographs for the Mackenzie River at Arctic Red, representing some 95% of the total discharge to the Arctic Ocean, show total flow volumes within about 1% of the observed estimate. This suggests that PE is reasonably well simulated at the large scale, but less well spatially distributed within the Mackenzie basin, with excessively large values in the mountainous western region.

Acknowledgments

This research was supported in part by the Climate Change Action Fund (CCAF), and by the Mackenzie GEWEX Study, which is funded by Environment Canada and the Natural Sciences and Engineering Research Council of Canada (NSERC). Thanks are due to Alanna Minogue who analyzed several observed gridded climate datasets over the Mackenzie basin, and to Chris Derksen for extracting the operational NOAA snow cover extent over this area. Useful comments on the manuscript were provided by Vivek Arora and Faye Hicks, as well as three anonymous reviewers.

REFERENCES

  • Abdella, K., and McFarlane N. A. , 1996: Parameterization of the surface-layer exchange coefficients for atmospheric models. Bound.-Layer Meteor., 80 , 223248.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Armstrong, R., Chang A. , Rango A. , and Josberger E. , 1993: Snow depths and grain-size relationships with relevance for passive microwave studies. Ann. Glaciol., 17 , 171176.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arora, V. A., and Boer G. J. , 1999: A variable velocity flow routing algorithm for GCMs. J. Geophys. Res., 104 (D24) 3096530979.

  • Arora, V. A., Seglenieks F. , Kouwen N. , and Soulis E. D. , 2001: Scaling aspects of river flow routing. Hydrol. Processes, 15 , 461477.

  • Caya, D., and Laprise R. , 1999: A semi-implicit, semi-Lagrangian regional climate model: The Canadian RCM. Mon. Wea. Rev., 127 , 341362.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Centre for Land and Biological Resources Research, 1996: Soil landscapes of Canada. Version 2.2. [Available online at http://sis.agr.gc.ca/cansis/nsdb/slc/intro.html.].

    • Search Google Scholar
    • Export Citation
  • Cihlar, J., Beaubien J. , Latifovic R. , and Simard G. , 1999: Land cover of Canada 1995. Version 1.1, Natural Resources Canada, Digital Dataset Documentation.

    • Search Google Scholar
    • Export Citation
  • Delage, Y., Bartlett P. , and McCaughey J. H. , 2002: Study of “soft” night-time surface-layer decoupling over forest canopies in a land surface model. Bound.-Layer Meteor., 103 , 253276.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • De Sève, D., Bernier M. , Fortin J-P. , and Walker A. , 1997: Preliminary analysis of snow microwave radiometry using the SSM/I passive-microwave data: The case of La Grande River watershed (Quebec). Ann. Glaciol., 25 , 353361.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Feng, J., Leighton H. , and MacKay M. , 2003: Radiation budgets in the Mackenzie River basin: Retrieval from satellite observations and an evaluation of the Canadian Regional Climate Model. J. Hydrometeor., 4 , 731747.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gesch, D. B., Verdin K. L. , and Greenlee S. K. , 1999: New land surface digital elevation model covers the Earth. Eos, Trans. Amer. Geophys. Union, 80 , 6970.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Goita, K., Walker A. E. , Goodison B. E. , and Chang A. T. C. , 1997: Estimation of snow water equivalent in the boreal forest using passive microwave data. Proc. GER'97 Int. Symp.: Geomatics in the Era of Radarsat, Ottawa, ON, Canada, Department of National Defence, CD-ROM.

    • Search Google Scholar
    • Export Citation
  • Goodison, B. E., and Walker A. E. , 1995: Canadian development and use of snow cover information from passive microwave satellite data. Passive Microwave Remote Sensing of Land–Atmosphere Interactions, B. J. Choudhury et al., Eds., VSP BV, 245–262.

    • Search Google Scholar
    • Export Citation
  • Harding, R. J., and Pomeroy J. W. , 1996: The energy balance of the winter boreal landscape. J. Climate, 9 , 27782787.

  • Hopkinson, R. F., cited 2000: Canadian gridded climate data. [Available online at http://www.cics.uvic.ca/climate/data.htm.].

  • Kouwen, N., Soulis E. D. , Pietroniro A. , Donald J. , and Harrington R. A. , 1993: Grouping response units for distributed hydrologic modelling. ASCE J. Water Resour. Manage. Planning, 119 , 289305.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laprise, R., Caya D. , Giguère M. , Bergeron G. , Côté H. , Blanchet J-P. , Boer G. , and McFarlane N. , 1998: Climate and climate change in western Canada as simulated by the Canadian Regional Climate Model. Atmos.–Ocean, 36 , 119167.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laprise, R., Caya D. , Frigon A. , and Paquin D. , 2003: Current and perturbed climate as simulated by the second generation Canadian Climate Model (CRCM II) over north-western North America. Climate Dyn., in press.

    • Search Google Scholar
    • Export Citation
  • Leconte, R., Pietroniro A. , Peters D. L. , and Prowse T. D. , 2001: Effect of flow regulation on the peace-athabasca delta 1996 summer flood. Regul. Rivers:. Res. Manage., 7 , 5165.

    • Search Google Scholar
    • Export Citation
  • Leung, L. R., and Ghan S. J. , 1998: Parameterizing subgrid orographic precipitation and surface cover in climate models. Mon. Wea. Rev., 126 , 32713291.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Louie, P. Y. T., Hogg W. D. , MacKay M. D. , Zhang X. , and Hopkinson R. F. , 2002: The water balance climatology of the Mackenzie basin with reference to the 1994/95 water year. Atmos.–Ocean, 40 , 159180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Loveland, T. R., Merchant J. W. , Brown J. F. , Ohlen D. O. , Reed B. C. , Olson P. , and Hutchinson J. , 1995: Seasonal landcover regions of the United States. Ann. Assoc. Amer. Geogr., 85 , 339355.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacKay, M., Stewart R. , and Bergeron G. , 1998: Downscaling the hydrological cycle in the Mackenzie Basin with the Canadian Regional Climate Model. Atmos.–Ocean, 36 , 179211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacKay, M., Szeto K. , Verseghy D. , Chan E. , and Bussières N. , 2003: Mesoscale circulations and surface energy balance during snowmelt in a regional climate model. Nord. Hydrol., 34 , 91106.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McFarlane, N., Boer G. , Blanchet J. P. , and Lazare M. , 1992: The Canadian Climate Centre second-generation general circulation model and its equilibrium climate. J. Climate, 5 , 10131044.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McKenney, D. W., Mackey B. G. , and Sims R. A. , 1996: Primary databases for forest ecosystem management—Examples from Ontario and possibilities for Canada: NatGRID. Environ. Monit. Assess., 39 , 399415.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Milewska, E., and Hogg W. D. , 2001: Spatial representativeness of a long-term climate network in Canada. Atmos.–Ocean, 39 , 145161.

  • Milewska, E., Hopkinson R. , and Niitsoo A. , 2002: Comparison of geo-referenced grids of 1961–1990 Canadian temperature and precipitation normals. Preprints, 13th Conf. on Applied Climatology, Portland, OR, Amer. Meteor. Soc., 191–196.

    • Search Google Scholar
    • Export Citation
  • Miller, D. A., and White R. A. , 1998: A conterminous United States multilayer soil characteristics dataset for regional climate and hydrology modeling. Earth Interactions,2. [Available online at http://EarthInteractions.org.].

    • Search Google Scholar
    • Export Citation
  • New, M., Hulme M. , and Jones P. , 1999: Representing twentieth-century space–time climate variability. Part I: Development of a 1961–90 mean monthly terrestrial climatology. J. Climate, 12 , 829856.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • New, M., Hulme M. , and Jones P. , 2000: Representing twentieth-century space–time climate variability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. J. Climate, 13 , 22172238.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Olson, J. S., 1994a: Global ecosystem framework—Definitions. USGS EROS Data Center Internal Report, 37 pp.

  • Olson, J. S., 1994b: Global ecosystem framework—Translation strategy. USGS EROS Data Center Internal Report, 39 pp.

  • Parker, D. E., Jackson M. , and Horton E. B. , 1995: The GISST 2.2 sea surface temperature and sea-ice climatology. Hadley Centre for Climate Prediction and Research, Met Office Tech. Rep. CRTN 63, 35 pp. [Available from Hadley Centre for Climate Prediction and Research, Met Office, London Rd., Bracknell, Berkshire RG122SY, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Pomeroy, J. W., Parviainen J. , Hedstrom N. , and Gray D. M. , 1998: Coupled modelling of forest snow interception and sublimation. Hydrol. Processes, 12 , 23172337.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Prowse, T. D., and Conly M. , 1998: Effects of climatic variability and flow regulation on ice-jam flooding of a northern delta. Hydrol. Processes, 12 , 15891610.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, D. A., Dewey K. F. , and Heim R. R. Jr., 1993: Global snow cover monitoring: An update. Bull. Amer. Meteor. Soc., 74 , 16891696.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Slater, A. G., and Coauthors. 2001: The representation of snow in land surface schemes: Results from PILPS 2(d). J. Hydrometeor., 2 , 725.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Solomon, S. I., Denouvilliez J. P. , Chart E. J. , Wooley J. A. , and Cadou C. F. , 1968: The use of a square grid system for computer estimation of precipitation, temperature, and runoff. Water Resour. Res., 4 , 919929.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stewart, R. E., and Coauthors. 1998: The Mackenzie GEWEX Study: The water and energy cycles of a major North American river basin. Bull. Amer. Meteor. Soc., 79 , 26652683.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Verseghy, D., 1991: CLASS—A Canadian land surface scheme for GCMs. I: Soil model. Int. J. Climatol., 11 , 111133.

  • Verseghy, D., McFarlane N. , and Lazare M. , 1993: CLASS—A Canadian land surface scheme for GCMs. II: Vegetation model and coupled runs. Int. J. Climatol., 13 , 347370.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Walker, A. E., and Silis A. , 2002: Snow-cover variations over the Mackenzie River basin, Canada, derived from SSM/I passive microwave satellite data. Ann. Glaciol., 34 , 814.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wiesnet, D., Ropelewski C. , Kukla G. , and Robinson D. , 1987: A discussion of the accuracy of NOAA satellite-derived global seasonal snow cover measurements. Large Scale Effects of Seasonal Snow Cover, IAHS Publication 166, 291–304.

    • Search Google Scholar
    • Export Citation
  • Yang, Z-L., Dickinson R. E. , Henderson-Sellers A. , and Pitman A. J. , 1995: Preliminary study of spin-up processes in land surface models with the first stage data of Project for Intercomparison of Land Surface Parameterization Schemes Phase 1(a). J. Geophys. Res., 100 , 1655316578.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, X., Vincent L. A. , Hogg W. D. , and Niitsoo A. , 2000: Temperature and precipitation trends in Canada during the 20th Century. Atmos.–Ocean, 38 , 395429.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

CRCM domain for the CAGES simulation. Elevation (shaded, contour interval 200 m), the Mackenzie basin outline, and the nine-point lateral sponge zone are also indicated

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 2.
Fig. 2.

Mackenzie basin screen level minimum temperature (°C) for water year 1998–99: (a) annual mean CANGRID; (b) annual mean CRCM; (c) monthly time series for CANGRID (dashed) and CRCM (solid)

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 3.
Fig. 3.

Same as in Fig. 2, but for Mackenzie basin screen level maximum temperature (°C)

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 4.
Fig. 4.

Same as in Fig. 2, but for accumulated precipitation (mm)

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 5.
Fig. 5.

Mackenzie basin average snow cover fraction from CRCM (solid) and, observed [NOAA NESDIS (crosses) SSM/I (diamonds)]. A simulated threshold of 10 kg m−2 is assumed.

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 6.
Fig. 6.

Selected SSM/I-derived SWE pixels and nearby snow course data within the Mackenzie basin for Jan–May 1999: (a) bias (SSM/I − snow course), as function of elevation; (b) scatterplot for locations below 800-m elevation. Also indicated is dominant vegetation class of SSM/I pixel

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 7.
Fig. 7.

Mean SWE for three regions (see text for details) within the Mackenzie basin during water year 1998–99 with dominant vegetation: (a) deciduous, (b) open (i.e., unvegetated), and (c) coniferous. CRCM (solid curves), SSM/I-derived estimates (asterisks), and snow course measurements (crosses) are indicated. Vertical lines through asterisks indicate range of SSM/I-derived values

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 8.
Fig. 8.

(a), (c) Cumulative observed (CANGRID) precipitation and (b), (d) SSM/I-derived SWE for the Mackenzie basin lowlands. Accumulation begins 1 Nov and ends (a), (b) 1 Dec and (c), (d) 1 Jan. Mean accumulations are (a) 18, (b) 15, (c) 43, and (d) 47 mm

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 9.
Fig. 9.

Normalized frequency distributions of cumulative observed (CANGRID) precipitation (dashed) and accumulated SSM/I-derived SWE (solid). Accumulation begins 1 Nov and ends (a) 1 Jan, (b) 1 Feb, and (c) 1 Mar. Mean accumulations are (a) CANGRID, 43 mm and SSM/I, 47 mm; (b) CANGRID, 72 mm and SSM/I, 57 mm; and (c) CANGRID, 89 mm and SSM/I 62 mm

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 10.
Fig. 10.

Mackenzie basin lowland mean SWE and cumulative precipitation for water year 1998–99: (a) CRCM (solid) and SSM/I-derived (asterisks) SWE, CRCM (since 1 Oct) (dotted) and CANGRID (since 1 Oct) precipitation (dashed); (b) as in (a) but all fields cumulative from 1 Nov

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 11.
Fig. 11.

Mean annual Mackenzie basin surface water balance (cm) for water year 1998–99: (a) PE, (b) net soil moisture change, (c) surface runoff, and (d) total runoff (surface runoff + deep drainage). Basin mean values are (a) 22.5, (b) −2.6, (c) 13.2, and (d) 24.6 cm. The locations of the Liard and Athabasca subbasins are indicated in (b)

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 12.
Fig. 12.

(a) Mean soil moisture change and (b) mean third soil layer temperature for Oct 1997–Sep 1999: Liard subbasin (solid) and Athabasca subbasin (dashed)

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 13.
Fig. 13.

Watroute drainage network

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Fig. 14.
Fig. 14.

Simulated and observed hydrographs at (a) Athabasca River at Athabasca, (b) Smoky River at Watino, (c) Liard River at Fort Liard, and (d) Mackenzie River at Arctic Red. Observed (solid), Watflood (dashed), and Watroute (dotted) hydrographs are shown. Both hydrological models are driven (offline) by the CRCM

Citation: Journal of Hydrometeorology 4, 4; 10.1175/1525-7541(2003)004<0748:MMBSWB>2.0.CO;2

Table 1.

Mackenzie basin 1961–90 climate normals for annual precipitation and mean screen level temperature based on four gridded datasets described in text

Table 1.
Table 2.

Mean surface characteristics of three clusters of snow course data described in text.

Table 2.
Table 3.

Mackenzie basin surface water balance for WY 1997–98. Units: mm.

Table 3.
Save
  • Abdella, K., and McFarlane N. A. , 1996: Parameterization of the surface-layer exchange coefficients for atmospheric models. Bound.-Layer Meteor., 80 , 223248.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Armstrong, R., Chang A. , Rango A. , and Josberger E. , 1993: Snow depths and grain-size relationships with relevance for passive microwave studies. Ann. Glaciol., 17 , 171176.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Arora, V. A., and Boer G. J. , 1999: A variable velocity flow routing algorithm for GCMs. J. Geophys. Res., 104 (D24) 3096530979.

  • Arora, V. A., Seglenieks F. , Kouwen N. , and Soulis E. D. , 2001: Scaling aspects of river flow routing. Hydrol. Processes, 15 , 461477.

  • Caya, D., and Laprise R. , 1999: A semi-implicit, semi-Lagrangian regional climate model: The Canadian RCM. Mon. Wea. Rev., 127 , 341362.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Centre for Land and Biological Resources Research, 1996: Soil landscapes of Canada. Version 2.2. [Available online at http://sis.agr.gc.ca/cansis/nsdb/slc/intro.html.].

    • Search Google Scholar
    • Export Citation
  • Cihlar, J., Beaubien J. , Latifovic R. , and Simard G. , 1999: Land cover of Canada 1995. Version 1.1, Natural Resources Canada, Digital Dataset Documentation.

    • Search Google Scholar
    • Export Citation
  • Delage, Y., Bartlett P. , and McCaughey J. H. , 2002: Study of “soft” night-time surface-layer decoupling over forest canopies in a land surface model. Bound.-Layer Meteor., 103 , 253276.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • De Sève, D., Bernier M. , Fortin J-P. , and Walker A. , 1997: Preliminary analysis of snow microwave radiometry using the SSM/I passive-microwave data: The case of La Grande River watershed (Quebec). Ann. Glaciol., 25 , 353361.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Feng, J., Leighton H. , and MacKay M. , 2003: Radiation budgets in the Mackenzie River basin: Retrieval from satellite observations and an evaluation of the Canadian Regional Climate Model. J. Hydrometeor., 4 , 731747.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gesch, D. B., Verdin K. L. , and Greenlee S. K. , 1999: New land surface digital elevation model covers the Earth. Eos, Trans. Amer. Geophys. Union, 80 , 6970.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Goita, K., Walker A. E. , Goodison B. E. , and Chang A. T. C. , 1997: Estimation of snow water equivalent in the boreal forest using passive microwave data. Proc. GER'97 Int. Symp.: Geomatics in the Era of Radarsat, Ottawa, ON, Canada, Department of National Defence, CD-ROM.

    • Search Google Scholar
    • Export Citation
  • Goodison, B. E., and Walker A. E. , 1995: Canadian development and use of snow cover information from passive microwave satellite data. Passive Microwave Remote Sensing of Land–Atmosphere Interactions, B. J. Choudhury et al., Eds., VSP BV, 245–262.

    • Search Google Scholar
    • Export Citation
  • Harding, R. J., and Pomeroy J. W. , 1996: The energy balance of the winter boreal landscape. J. Climate, 9 , 27782787.

  • Hopkinson, R. F., cited 2000: Canadian gridded climate data. [Available online at http://www.cics.uvic.ca/climate/data.htm.].

  • Kouwen, N., Soulis E. D. , Pietroniro A. , Donald J. , and Harrington R. A. , 1993: Grouping response units for distributed hydrologic modelling. ASCE J. Water Resour. Manage. Planning, 119 , 289305.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laprise, R., Caya D. , Giguère M. , Bergeron G. , Côté H. , Blanchet J-P. , Boer G. , and McFarlane N. , 1998: Climate and climate change in western Canada as simulated by the Canadian Regional Climate Model. Atmos.–Ocean, 36 , 119167.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Laprise, R., Caya D. , Frigon A. , and Paquin D. , 2003: Current and perturbed climate as simulated by the second generation Canadian Climate Model (CRCM II) over north-western North America. Climate Dyn., in press.

    • Search Google Scholar
    • Export Citation
  • Leconte, R., Pietroniro A. , Peters D. L. , and Prowse T. D. , 2001: Effect of flow regulation on the peace-athabasca delta 1996 summer flood. Regul. Rivers:. Res. Manage., 7 , 5165.

    • Search Google Scholar
    • Export Citation
  • Leung, L. R., and Ghan S. J. , 1998: Parameterizing subgrid orographic precipitation and surface cover in climate models. Mon. Wea. Rev., 126 , 32713291.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Louie, P. Y. T., Hogg W. D. , MacKay M. D. , Zhang X. , and Hopkinson R. F. , 2002: The water balance climatology of the Mackenzie basin with reference to the 1994/95 water year. Atmos.–Ocean, 40 , 159180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Loveland, T. R., Merchant J. W. , Brown J. F. , Ohlen D. O. , Reed B. C. , Olson P. , and Hutchinson J. , 1995: Seasonal landcover regions of the United States. Ann. Assoc. Amer. Geogr., 85 , 339355.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacKay, M., Stewart R. , and Bergeron G. , 1998: Downscaling the hydrological cycle in the Mackenzie Basin with the Canadian Regional Climate Model. Atmos.–Ocean, 36 , 179211.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • MacKay, M., Szeto K. , Verseghy D. , Chan E. , and Bussières N. , 2003: Mesoscale circulations and surface energy balance during snowmelt in a regional climate model. Nord. Hydrol., 34 , 91106.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McFarlane, N., Boer G. , Blanchet J. P. , and Lazare M. , 1992: The Canadian Climate Centre second-generation general circulation model and its equilibrium climate. J. Climate, 5 , 10131044.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McKenney, D. W., Mackey B. G. , and Sims R. A. , 1996: Primary databases for forest ecosystem management—Examples from Ontario and possibilities for Canada: NatGRID. Environ. Monit. Assess., 39 , 399415.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Milewska, E., and Hogg W. D. , 2001: Spatial representativeness of a long-term climate network in Canada. Atmos.–Ocean, 39 , 145161.

  • Milewska, E., Hopkinson R. , and Niitsoo A. , 2002: Comparison of geo-referenced grids of 1961–1990 Canadian temperature and precipitation normals. Preprints, 13th Conf. on Applied Climatology, Portland, OR, Amer. Meteor. Soc., 191–196.

    • Search Google Scholar
    • Export Citation
  • Miller, D. A., and White R. A. , 1998: A conterminous United States multilayer soil characteristics dataset for regional climate and hydrology modeling. Earth Interactions,2. [Available online at http://EarthInteractions.org.].

    • Search Google Scholar
    • Export Citation
  • New, M., Hulme M. , and Jones P. , 1999: Representing twentieth-century space–time climate variability. Part I: Development of a 1961–90 mean monthly terrestrial climatology. J. Climate, 12 , 829856.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • New, M., Hulme M. , and Jones P. , 2000: Representing twentieth-century space–time climate variability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. J. Climate, 13 , 22172238.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Olson, J. S., 1994a: Global ecosystem framework—Definitions. USGS EROS Data Center Internal Report, 37 pp.

  • Olson, J. S., 1994b: Global ecosystem framework—Translation strategy. USGS EROS Data Center Internal Report, 39 pp.

  • Parker, D. E., Jackson M. , and Horton E. B. , 1995: The GISST 2.2 sea surface temperature and sea-ice climatology. Hadley Centre for Climate Prediction and Research, Met Office Tech. Rep. CRTN 63, 35 pp. [Available from Hadley Centre for Climate Prediction and Research, Met Office, London Rd., Bracknell, Berkshire RG122SY, United Kingdom.].

    • Search Google Scholar
    • Export Citation
  • Pomeroy, J. W., Parviainen J. , Hedstrom N. , and Gray D. M. , 1998: Coupled modelling of forest snow interception and sublimation. Hydrol. Processes, 12 , 23172337.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Prowse, T. D., and Conly M. , 1998: Effects of climatic variability and flow regulation on ice-jam flooding of a northern delta. Hydrol. Processes, 12 , 15891610.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, D. A., Dewey K. F. , and Heim R. R. Jr., 1993: Global snow cover monitoring: An update. Bull. Amer. Meteor. Soc., 74 , 16891696.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Slater, A. G., and Coauthors. 2001: The representation of snow in land surface schemes: Results from PILPS 2(d). J. Hydrometeor., 2 , 725.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Solomon, S. I., Denouvilliez J. P. , Chart E. J. , Wooley J. A. , and Cadou C. F. , 1968: The use of a square grid system for computer estimation of precipitation, temperature, and runoff. Water Resour. Res., 4 , 919929.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stewart, R. E., and Coauthors. 1998: The Mackenzie GEWEX Study: The water and energy cycles of a major North American river basin. Bull. Amer. Meteor. Soc., 79 , 26652683.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Verseghy, D., 1991: CLASS—A Canadian land surface scheme for GCMs. I: Soil model. Int. J. Climatol., 11 , 111133.

  • Verseghy, D., McFarlane N. , and Lazare M. , 1993: CLASS—A Canadian land surface scheme for GCMs. II: Vegetation model and coupled runs. Int. J. Climatol., 13 , 347370.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Walker, A. E., and Silis A. , 2002: Snow-cover variations over the Mackenzie River basin, Canada, derived from SSM/I passive microwave satellite data. Ann. Glaciol., 34 , 814.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wiesnet, D., Ropelewski C. , Kukla G. , and Robinson D. , 1987: A discussion of the accuracy of NOAA satellite-derived global seasonal snow cover measurements. Large Scale Effects of Seasonal Snow Cover, IAHS Publication 166, 291–304.

    • Search Google Scholar
    • Export Citation
  • Yang, Z-L., Dickinson R. E. , Henderson-Sellers A. , and Pitman A. J. , 1995: Preliminary study of spin-up processes in land surface models with the first stage data of Project for Intercomparison of Land Surface Parameterization Schemes Phase 1(a). J. Geophys. Res., 100 , 1655316578.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zhang, X., Vincent L. A. , Hogg W. D. , and Niitsoo A. , 2000: Temperature and precipitation trends in Canada during the 20th Century. Atmos.–Ocean, 38 , 395429.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    CRCM domain for the CAGES simulation. Elevation (shaded, contour interval 200 m), the Mackenzie basin outline, and the nine-point lateral sponge zone are also indicated

  • Fig. 2.

    Mackenzie basin screen level minimum temperature (°C) for water year 1998–99: (a) annual mean CANGRID; (b) annual mean CRCM; (c) monthly time series for CANGRID (dashed) and CRCM (solid)

  • Fig. 3.

    Same as in Fig. 2, but for Mackenzie basin screen level maximum temperature (°C)

  • Fig. 4.

    Same as in Fig. 2, but for accumulated precipitation (mm)

  • Fig. 5.

    Mackenzie basin average snow cover fraction from CRCM (solid) and, observed [NOAA NESDIS (crosses) SSM/I (diamonds)]. A simulated threshold of 10 kg m−2 is assumed.

  • Fig. 6.

    Selected SSM/I-derived SWE pixels and nearby snow course data within the Mackenzie basin for Jan–May 1999: (a) bias (SSM/I − snow course), as function of elevation; (b) scatterplot for locations below 800-m elevation. Also indicated is dominant vegetation class of SSM/I pixel

  • Fig. 7.

    Mean SWE for three regions (see text for details) within the Mackenzie basin during water year 1998–99 with dominant vegetation: (a) deciduous, (b) open (i.e., unvegetated), and (c) coniferous. CRCM (solid curves), SSM/I-derived estimates (asterisks), and snow course measurements (crosses) are indicated. Vertical lines through asterisks indicate range of SSM/I-derived values

  • Fig. 8.

    (a), (c) Cumulative observed (CANGRID) precipitation and (b), (d) SSM/I-derived SWE for the Mackenzie basin lowlands. Accumulation begins 1 Nov and ends (a), (b) 1 Dec and (c), (d) 1 Jan. Mean accumulations are (a) 18, (b) 15, (c) 43, and (d) 47 mm

  • Fig. 9.

    Normalized frequency distributions of cumulative observed (CANGRID) precipitation (dashed) and accumulated SSM/I-derived SWE (solid). Accumulation begins 1 Nov and ends (a) 1 Jan, (b) 1 Feb, and (c) 1 Mar. Mean accumulations are (a) CANGRID, 43 mm and SSM/I, 47 mm; (b) CANGRID, 72 mm and SSM/I, 57 mm; and (c) CANGRID, 89 mm and SSM/I 62 mm

  • Fig. 10.

    Mackenzie basin lowland mean SWE and cumulative precipitation for water year 1998–99: (a) CRCM (solid) and SSM/I-derived (asterisks) SWE, CRCM (since 1 Oct) (dotted) and CANGRID (since 1 Oct) precipitation (dashed); (b) as in (a) but all fields cumulative from 1 Nov

  • Fig. 11.

    Mean annual Mackenzie basin surface water balance (cm) for water year 1998–99: (a) PE, (b) net soil moisture change, (c) surface runoff, and (d) total runoff (surface runoff + deep drainage). Basin mean values are (a) 22.5, (b) −2.6, (c) 13.2, and (d) 24.6 cm. The locations of the Liard and Athabasca subbasins are indicated in (b)

  • Fig. 12.

    (a) Mean soil moisture change and (b) mean third soil layer temperature for Oct 1997–Sep 1999: Liard subbasin (solid) and Athabasca subbasin (dashed)

  • Fig. 13.

    Watroute drainage network

  • Fig. 14.

    Simulated and observed hydrographs at (a) Athabasca River at Athabasca, (b) Smoky River at Watino, (c) Liard River at Fort Liard, and (d) Mackenzie River at Arctic Red. Observed (solid), Watflood (dashed), and Watroute (dotted) hydrographs are shown. Both hydrological models are driven (offline) by the CRCM