Results
The magnitude and timing of annual peak snowpack is sensitive to both
air temperature and precipitation changes in Wolf Creek as shown in
Figure 3. The interaction between air temperature and precipitation
affecting peak SWE is evident in the curvature and slope of the
contours; the interaction is complex in the alpine and shrub tundra
(curved contour lines, Figure 3a & c) but less so in the forest (Figure
3e). The sensitivity of peak SWE to precipitation is somewhat higher in
the high elevation alpine zone (contours have higher slope) and its
sensitivity to temperature is somewhat higher in the lower elevation
shrub tundra and forest zones (contours have lower slope). The peak SWE
in the shrub tundra zone is very sensitive to a decrease in
precipitation with warming due to precipitation phase change and
suppression of blowing snow redistribution from the alpine zone under
warmer air temperatures (Rasouli et al., 2014) and drops from 162 mm to
75 mm (87 mm reduction, Figure 3c) with 80% of precipitation and
+5oC of warming. The sensitivity of peak SWE to
increasing precipitation in the shrub tundra zone declines as the
temperature warms. The peak SWE in the forested zone is slightly less
sensitive to temperature than shrub tundra because unloading of
intercepted snow from the canopy, where it is prone to sublimation,
increases with winter air temperatures and moderates the impact of
declining snowfall with rising temperature. Whether 20% additional
precipitation can offset the effect of warming on snowpacks in Wolf
Creek is illustrated in Figure 3 by comparing the black dot, indicating
no change in air temperature or precipitation, to the white dot,
indicating the degree of warming that can be offset by a 20% increase
in precipitation. This is 3.5°C of warming for peak SWE in the alpine
zone (Figure 3a) and 2.7°C and 3°C of warming for peak SWE in the shrub
tundra (Figure 3c) and forest (Figure 3e).
There is no clear pattern to the small changes, less than six days, in
the timing of peak SWE in the Wolf Creek alpine zone with air
temperature and precipitation changes (Figure 3b). This is likely due to
the persistently colder temperatures during winter at high elevations in
the subarctic (Figure 3b). In the shrub tundra and forest, the mean
annual peak SWE occurs 25 and 20 days earlier respectively with 5°C of
warming and 20% reduced precipitation (Figure 3d and 3f).
In cold continental Marmot Creek, peak SWE in all zones is influenced by
changes in both air temperature and precipitation but responds more
strongly to temperature than in subarctic Wolf Creek (Figure 4). Peak
SWE is progressively more influenced by warming temperature with
declining elevation due to the influence of lapse rates on precipitation
phase and other factors. Because of reduced blowing snow inputs from the
alpine zone, the treeline forest zone loses the most snow (-422 mm under
5°C of warming and 20% less precipitation, Figure 4c), but because it
has the highest snow accumulation, snow is still deep and its
proportional change with temperature was not substantially different
from the other zones. In contrast, almost all snow is lost in the forest
zone, suggesting a high sensitivity of snow in Marmot Creek’s low
elevation forests to warming because of the large losses of snow. The
response of the peak SWE to warming and precipitation changes shows that
an increase in precipitation of 20%, slightly greater than the maximum
indicated by climate models, can offset the effect on peak SWE of
warming in the alpine of 2.9°C (Figure 4a), in the treeline forests of
2.1°C (Figure 4c) and in the forest and forest clearing of 1.8°C (Figure
4c, g). The peak snowpack in Marmot Creek is more sensitive to warming,
and so increased precipitation can offset less of a temperature increase
than in Wolf Creek.
The changes in the simulated timing of peak SWE in Marmot Creek are
substantial and complex. Timing responded much more to warming than to
precipitation change and precipitation increases could not compensate
for any degree of warming at any elevation (Figures 4b, 4d, 4f, and 4h).
In the alpine, forest, and forest clearing zones, peak SWE advanced
between 19 and 28 days for 2°C of warming, and between 60 and 70 days
for 5°C. In contrast, the treeline forest peak SWE timing advanced only
10 and 27 days for 2 and 5°C of warming, its lower sensitivity (range of
contours) due to the high snow accumulation in this zone associated with
continued redistribution of snow from the alpine (Figure 4d).
In Reynolds Mountain, annual peak SWE is very sensitive to increases in
air temperature and much less sensitive to changes in precipitation
(Figure 5a, 5c, 5e, and 5g). The slope and curvature of the annual peak
SWE contours show the sensitivity to precipitation change decreases as
temperature increases. This suggests that the effects of warming on SWE
cannot be easily offset by increased precipitation; a precipitation
increase of +20% can offset warming up to from 1.2 to 1.5°C depending
on location. The warmest and driest scenario (+5°C and -20%
precipitation) caused the peak SWE decline in all zones, e.g., from 570
mm to 58 mm in the sink (Figure 5a) and from 427 mm to 39 mm in the
interception zones (Figure 5e). The blowing snow sink zone lost more
snow with warming and drying than other zones due to the suppression of
blowing snow transport from the source zone (Figure 5a). An increase in
precipitation greater than 20% would be needed in Reynolds Mountain
than in Wolf Creek and Marmot Creek to offset the effect of the same
warming on peak SWE.
The response of the timing of annual peak SWE is much more sensitive to
warming than to precipitation change in all zones in Reynolds Mountain
(Figures 5b, 5d, 5f, and 5h). The timing changes in Reynolds Mountain
are the largest of the three basins with the change in peak SWE date
being between 50 and 70 days earlier for the maximum
5oC warming. Additional precipitation of 20% can only
offset the effect of 0.5°C of warming on peak SWE date (Figures 5b, 5d,
5f, and 5h).
The rate of change in the simulated snowpacks can be estimated in
relation to temperature. Peak SWE reduction per degree of warming is 8%
in Wolf Creek, 10% in Marmot Creek, and 17% in Reynolds Mountain
(Table 1). The loss of snowpack with warming is reflected in the
reduction in the snowcover duration of 11 days in Wolf Creek, 18 days in
Marmot Creek, and 30 days in Reynolds Mountain per degree of warming
(Table 1). The duration of snowmelt declines between 0 and 9 days per
degree of warming in all basins, much less than the snowcover duration,
and smaller than the advance in the timing of snow disappearance which
ranges from 7 (Wolf Creek) to 13 (Marmot Creek) to 21 (Reynolds
Mountain) days per degree of warming. Snow melts more slowly as the melt
season advances in some of these simulations, which partly offsets the
impact of the decrease in peak snowpack on snowmelt period duration.
In Wolf Creek, as in the other basins, the distribution of hourly
simulations of SWE widens if precipitation increases and narrows if
precipitation decreases (Figure 6). In the alpine, the accompanying
warming shifts the distribution to the left and causes additional
narrowing (Figure 6a). If the precipitation increase is large (+20%); a
warming of up to 3.5°C can be offset for all hourly SWE simulations in
all three zones in Wolf Creek (mean annual temperature exceeds 1.4°C).
In contrast to higher sensitivity of peak SWE in the forest zones, the
distributions of different snowpack regimes show that the forest changes
the least under warming and changes in precipitation because of cold
sub-canopy winter temperatures and reduction in sublimation losses from
intercepted snow with warming (Figure 6c). An increase in precipitation
can offset the impacts of some warming and affects high and medium
values of SWE the most and low SWEs only slightly (Figure 6). The
warming impacts SWE in early winter during the initiation dates of snow
accumulation and early spring during snow depletion more than peak
snowpacks in all of the snow regimes (Table 1). The snowpack regime in
Wolf Creek is more sensitive to changes in precipitation than to warming
because of its consistently very cold winters. Each of the five
distributions for the warming and changed precipitation scenarios in
each zone are significantly different (p-value ≤ 0.05) to the snowpack
distribution in the base period (0°C warming, 100% precipitation) based
on the Kolmogorov-Smirnov (K–S) test.
In Marmot Creek, the distribution of hourly simulations of SWE is wider
if precipitation increases and much narrower if temperature warms by
more than 2°C (Figure 7). Precipitation increases of 20% can offset
effect of a warming up to 2°C on snowpack regime in each zone (Figure 7)
but cannot offset warming of more than 2°C. The distributions of
different snowpack regimes show that a very shallow snowpack (SWE
< 100 mm) is expected in the forest (Figure 7c) and forest
clearings (Figure 7d) under the extreme case of 5°C warming and 20%
decrease in precipitation. Warming impacts peak more than shallower
snowpacks in all zones. An increase in precipitation can offset the
impacts of warming by increasing SWE, especially high values of SWE
(Figure 7). In general, the snowpack regime in Marmot Creek is equally
sensitive to warming and changes in precipitation (Figure 4). Each of
the five distributions for the warming and changed precipitation
scenarios in each zone are significantly different (p-value ≤ 0.05) to
the modelled snowpack distribution in the base period based on the
Kolmogorov-Smirnov (K–S) test.
In Reynolds Mountain, the distribution of hourly simulations of SWE is
wider if precipitation increases, and much narrower if temperature warms
by more than 1°C (Figure 8). Increasing precipitation offsets less of
the warming impact in Reynolds Mountain than in Wolf Creek or in Marmot
Creek; a precipitation increase of 20% can only offset the impact of a
1°C warming (mean annual temperature exceeds 6°C). An additional 20%
precipitation cannot offset warming of 2°C or more. The different
snowpack regimes are sensitive to impacts of 5°C warming and 20%
decrease in precipitation (Figure 8) and simulated maximum SWE values
drop below 240 mm from the base case of over 800 mm in the source and
sink HRUs (Figure 8a), interception (Figure 8c), and sheltered regime
(Figure 8d). Of particular interest is that high values of SWE
(>500 mm) do not occur. Each of the five distributions for
the warming and changed precipitation scenarios in each zone are
significantly different (p-value ≤ 0.05) to the snowpack distribution in
the base period based on the Kolmogorov-Smirnov (K–S) test.
Changes in the distribution of SWE in each headwater basin show that the
zones in Wolf Creek (Figure 6) and the treeline forest in Marmot Creek
(Figure 7b) are the least sensitive to air temperature and precipitation
changes. Each of the regimes in Reynolds Mountain (Figure 8) are
sensitive in terms of the absolute magnitude of snow loss.
The sensitivity of five main characteristics of basin snow regimes to
warming and change in precipitation averaged over Wolf Creek shows that
both changes in precipitation and warming affect the magnitude of the
peak SWE (Figure 9a). Precipitation increases of 20% can offset a 3°C
temperature increase in Wolf Creek peak SWE. Delay in the initiation of
snow accumulation is sensitive to warming rates above 3°C regardless of
precipitation changes (Figure 9b). The snow-free date advances from
late-June (June 28) in the recent climate to early June (June 11) with a
warming of 2°C (Figure 9c, Table 2). The snow-free date is also
sensitive to warming and almost insensitive to precipitation changes
(Figure 9c). The snow season duration in Wolf Creek is also driven by
warming and not by precipitation changes (Figure 9d). The snowmelt
period, the timing difference between peak SWE and the snow-free date,
is sensitive to warming and almost insensitive to precipitation changes
(Figure 9e).
In Marmot Creek, the peak SWE drops from 220 mm to 92 mm under a warming
of 5°C and decreasing precipitation (20%), (Figure 9f, Table 2). The
start of snow accumulation is not affected to a large amount by either
warming or precipitation (Figure 9g), but increased temperatures have a
large effect on the end date (Figure 9h) and snow season duration
(Figure 9i) but this is reduced with increased precipitation. The
duration of the melt season also is not affected (Figure 9j). In
contrast to Wolf Creek, the initiation date of snow accumulation is
sensitive to precipitation changes and would advance if warming rates
are below 2°C and precipitation increases. The snow-free date advances
from early June in the recent climate to late May with a warming of 2°C
(Figure 9h). Similar to the ablation period, snow accumulation start
date is sensitive to precipitation changes and to a lesser extent to
warming. With concomitant warming (5°C) and decreasing precipitation,
the snow-free date across the basin advances by 77 days to late March
(Figure 9h). As shown in Figure 9, the snow-free date is sensitive to
warming and insensitive to precipitation changes in Marmot Creek and
snow season length is affected by both warming and precipitation
changes. Similar to Wolf Creek, the combination of air temperature
increasing by at least 2°C and precipitation increasing by less than
20% results in declining peak SWE and deviation from the historical
ranges of snowpack in Marmot Creek.
In Reynolds Mountain, warming of 5°C and decreasing precipitation of
20%, the mean annual peak SWE decreases from 390 mm to 47 mm (Figure
9k, Table 2), snow accumulation starts later (Figure 9l) and ends
earlier (Figure 9m). The duration of the snow season (Figure 9n) and
duration of the melt period snow season (Figure 9o) become much shorter
than in present climate (Table 2). A 1°C warming advances the timing of
peak SWE by approximately 15 days (Table 1). The magnitude of peak SWE
is more sensitive to temperature than precipitation (Figure 9k); the
timing of the snow regime sensitive to temperature and less so to
precipitation (Figure 9 l-o).
The peak SWE is 136 mm in Wolf Creek, 220 mm in Marmot Creek, and 390 mm
in Reynolds Mountain; Wolf Creek and Reynolds peak SWE occur in early
March, and in Marmot Creek it occurs in late April (Table 2). With a
20% decline in precipitation and a warming of 5°C peak SWE declines to
61 mm (55% decrease) in Wolf Creek, to 92 mm (58%) in Marmot Creek,
and to 47 mm (88% decrease) in Reynolds Mountain. With a 20% increase
in precipitation and no warming and peak SWE increases to 169 mm (24%)
in Wolf Creek, to 281 mm in Marmot Creek (28%), and to 486 mm (25%) in
Reynolds Mountain. With 5°C warming and no changes in precipitation, the
onset of snow accumulation is delayed 17 days in Wolf Creek, 23 days in
Marmot Creek, and 42 days in Reynolds Mountain and the end of winter
comes earlier by 37 days in Wolf Creek, 67 days in Marmot Creek, and 104
days in Reynolds Mountain. When compared to no changes (Table 2), a 20%
increase in precipitation would lengthen the snowcover duration by 5 to
20 days.
The simulations show that changes in snow regime in these mountain
basins also result in moderated changes in mean annual runoff. Unlike
peak SWE, mean annual runoff is more sensitive to changes in
precipitation than air temperature (Figure 10). The near vertical lines
in Figure 10a & b indicate that changes in mean annual runoff are
driven predominately by precipitation in Wolf Creek and Marmot Creek
while in Reynolds Mountain temperature more strongly impacts runoff. A
1°C warming in Wolf Creek resulted in a 5% decrease in the annual
runoff (Table 1); total decreases rise to ~14% for a
5°C warming (171 to 147 mm, Table 2, Figure 10a). The most extreme
scenario of climate warming and decreased precipitation caused larger
declines in runoff, but if precipitation increases there is strong
compensation. For instance, if precipitation increases by 20% then
annual runoff increases by 35 mm (from 171 to 206 mm) with 5°C of
warming. Mean annual runoff is more sensitive than snow regime to
precipitation change in Wolf Creek. Similarly, in Marmot Creek, a 5°C
increase in air temperature results in a 4% decrease in the mean annual
runoff (402 to 384 mm Table 2, Figure 10b). The combination of 5°C of
warming and 20% decreased precipitation reduces mean annual runoff by
34% (135 mm from 402 to 267 mm, Table 2, Figure 10b). In Reynolds
Mountain, mean annual runoff has a stronger temperature sensitivity than
Wolf Creek or Marmot Creek (Figure 10). A 5°C increase in temperature
results in a 29% (371 to 263 mm, Table 2) decrease in the mean annual
runoff. The combination of 5°C of warming and 20% decrease in
precipitation reduces annual runoff by 43%, (371 to 161 mm, Table 2).
Changes in mean annual runoff (Figure 10) contrasts with the change in
mean annual peak SWE (Figures 3-5) in that mean annual runoff is more
sensitive to precipitation than temperature. The sensitivity of annual
runoff to temperature increase in Reynolds Mountain is because of the
longer snow-free season and an increased growing season and energy flux
for evapotranspiration with increasing temperature (Figure 5) whilst
runoff responds to both precipitation change and warming (Figure 10c).
In contrast to the sensitivity of snowpack to warming in Reynolds
Mountain, annual runoff is less sensitive, and the impact of warming on
annual runoff can be partly offset by an increase in precipitation in
Reynolds Mountain.
Annual runoff changes are given in Table 2 under different scenarios of
warming and changes in precipitation. Annual runoff responds strongly to
precipitation changes in Wolf Creek and Marmot Creek, and to both
warming and precipitation changes in Reynolds Mountain. The annual
runoff is the most resilient to warming in Marmot Creek and most
sensitive to warming in Reynolds Mountain. Under 5°C and a 20%
increased precipitation, annual runoff increases from 171 mm to 206 mm
(20%) in Wolf Creek and increases from 402 mm to 518 mm (29%) in
Marmot Creek and from 371 mm to 415 mm (12%) in Reynolds Mountain
(Table 1). This shows that increased precipitation with warming
increases the runoff in Marmot Creek more than the other two basins.
This is due to the very cold alpine snowpack at Marmot Creek which is
relatively unaffected by warming and the warm snowpacks at Reynolds
Mountain which become ephemeral with warming.
From this sensitivity analysis, the amount of additional precipitation
needed to offset the effect of increased temperature on peak SWE, and
annual runoff under future climate can be estimated. The largest
increase in precipitation projected by RCPs and NARCCAP RCM–GCMs is
34% for Wolf Creek, 18% for Marmot Creek, and 16% for Reynolds
Mountain. In Wolf Creek, when warming is limited to 1°C, increased
precipitation of 4% is able to offset the effect of warming on peak SWE
(Figure 11a); but, with warming of 5°C, an increase in precipitation of
34%, the amount expected from RCPs and NARCCAP, would be required to
offset the effect of warming. In Marmot Creek, the effect of a 1°C
warming on peak SWE can be offset by an 8% increase in precipitation;
however, the effect of a 5°C warming on peak SWE would require
precipitation increases that are greater than expected from RCP
scenarios and NARCCAP simulations. In Reynolds Mountain, the impact of a
1°C warming on peak SWE can be offset by a 16% increase in
precipitation, but the offset required for more than 2°C warming exceeds
the projected maximum precipitation increases.
Annual runoff is less sensitive than peak snowpack to warming and
smaller precipitation increases are required to offset the effects of
warming simulated here. These differences are due to differences in the
fraction of snow converted to rainfall in each basin under a warmer
climate. The additional precipitation needed to offset the impact of
warming on runoff varies with elevation range, precipitation regime and
latitude; offsetting the effect of warming of 5°C on annual runoff would
require precipitation increases of 8% in Wolf Creek (Figure 11a), 3%
in Marmot Creek (Figure 11b), and 14% in Reynolds Mountain (Figure
11c).