Studies have shown there has been a substantial increase in shrub and low cover vegetation densities within the Arctic Tundra. This
trend is forecast to continue and intensify over the coming decades. The increase has the potential to have many profound effects on
hydrologic processes in this region, including snow water equivalent at the end of winter, snow melt energy balance, spring runoff, and
the soil nutrient balance. Using observed data, including satellite images and air photos, the SnowModel (Liston) snow-evolution
modeling system was validated at a tundra site north of Inuvik, NWT in north-western Canada. Multiple model runs were utilized to
assess the effects of this shrub and low cover vegetation increase on end of winter SWE and melt.
Although an increase in vegetation has been documented, the hydrological effects, if any, have been largely undetermined. Hypotheses
that we consider include changes in snow accumulation (less blowing snow and therefore less sublimation leading to an increased end-of
winter snow pack) and changes in snowpack melt rates resulting from the increased shrub canopy density. After SnowModel validation
using observed data from the Trail Valley Creek research basin, the model was run with different shrub covers to examine the effect on snow
accumulation and melt.
This study is the first step in a series of studies using SnowModel as input for other hydrological models, such as TopoFlow, in an
effort to develop an improved coupled model for arctic regions, and as the basis for additional experimental model simulations.
The area of interest is the Trail Valley Creek basin, approximately 50km north of Inuvik N.W.T, east of the Mackenzie
delta. It is 63 km in area and underlain by continuous permafrost. The land cover of the area is dominated by open tundra
(78%) and shrub tundra (20%), with some forest (2%). The area is characterized by a fairly low relief with only some deeper
incised river valleys. A high resolution LIDAR-DEM (2m) was acquired for the basin in 2004. This was upscaled to 20m for
use with the hydrological model. Meteorological, snow accumulation, snow melt, and other relevant data have been
collected since 1992.
Study Site
An aggregation of four sub models, SnowModel is a spatially-distributed snow evolution numerical model. Applicable to areas conditions favoring snow formation, it can be run in gridded sections varying from 5m to 200m intervals with temporal intervals between
10minutes and 24 hours. SnowModel's aggregates allow for meteorological forcing (MicroMet), surface energy exchange estimation (EnBal), snow depth and snow water equivalent (SnowPack), and snow redistribution due to windy (SnowTran-3D), forest canopy
interception, unloading and sublimation, as well as snowpack melt.
Current Vegetation
Day number
10/1/1998 11/19/1998 1/8/1999 2/27/1999 4/18/1999 6/7/1999
Average wind velocity(m/s)
Precipitation (mm)
Snow water equivalent (m)
Day number vs Daily Average Wind velocity (m)
Precipitation (mm)
Day number vs Max swe(m)
Day number vs Average swe(m)
Impact of increased shrub density on snow accumulation and melt in the Arctic tundra
C. Marsh , S. Pohl , G.E. Liston
12 3
May 23, 1999
May 28, 1999
June 10, 1999
Fig. 2 Simulated SWE for entire model run
Fig. 1 Observed vs Modeled SWE for April 21
Fig. 3 Hourly wind speed and direction
Fig. 4 SCA from Satellite - Model - Air photo [channel network in pink & blue]
Model validation
Model Results
Our validation runs covered the period between October 1,1998 to June 30, 1999. At the end of winter (April 21, 1999), extended snow
surveys were conducted to determine snow water equivalent.
The model results compares very closely with the surveyed snow water equivalent for end of winter snowpack (Figure 1) as well as satellite
imagery and air photos collected through out the melt period (Figure 4) indicating that the model captures the spatially distributed snow
accumulation and ablation patterns very well. As indicated by Figure 3, north west winds dominated most of the winter storms. Consequently, the
model predicts most of the drifts to be on southeast facing slopes, which coincides with the observations.
The model seems to slightly underestimate the amount of drifting over the winter as indicated by the lower modeled drift SWE values compared
to snow surveys. This is the likely cause for the model's under prediction of snow covered area late in the melt period figure 5.
Figures 6,7,8 show the effect of vegetation on the snow accumulation.Figure 6 - all tundra run - produces the most
drifting, while the average SWE is fairly comparable to current day conditions. An all shrub vegetation cover (figure
7)would drastically reduce the redistribution of snow and lead to a much more uniform, and overall highers end of
winter snowpack. This would have considerable impact on a variety of factors such as snowmelt runoff, ground
insulation, and animal habitat. Figure 8 shows the average and peak snow water equivalents for current conditions,
predicted uniform tundra conditions, and full vegetation coverage conditions.
Fig 6. Modeled SWE for uniform tundra. April 21
Fig. 7 Modeled SWE for uniform shrub cover. April 21
Fig. 8 SWE comparison
(1) University of Saskatchewan, Saskatoon, Saskatchewan, Canada
(2) National Water Research Institute, Environment Canada, Saskatoon, Saskatchewan, Canada
(3) Cooperative Institute for Research in the Atmosphere (CIRA)
Colorado State University, Fort Collins, Colorado, USA
Fig. 5 Modeled SCA vs Observed SCA
Conclusion and Discussion
1. SnowModel predicts the distribution, magnitude and melt of the snowpack under existing conditions with reasonable
2. Mean SWE increases (due to decreased sublimation) as vegetation changes from tundra to shrub covered.
3. Maximum SWE in drifts decreases (due to a decline in drifting) as vegetation changes from tundra to shrub covered.
4. This suggests that as shrubs have invaded TVC over during the past decades, snowcover conditions are likely to have
also changed. As a result, any studies of past changes in runoff, must consider the effect of shrubs.
5. Previous studies (P.Marsh show that shrubs enhance melt in TVC, and Snowmodel does the same. However,
future shrub canopies may be denser, and therefore reduce melt rates (Pomeroy et al.). As a result, modeling of future
impact of shrubs requires information on future changes in shrub density.
6.Observations at TVC show that in certain years, shrubs are bent over and are fully covered by snow at the end of
winter. This complicates modeling shrub landscapes as shrubs behave like tundra for at least part of the winter and
during spring melt. These factors are not included in the current simulations.
The figure to the right
showing over prediction
of snow covered area
during the first half of
melt and excellent during
the last half