Climate Dynamics

, Volume 29, Issue 1, pp 73–88

The role of terrestrial snow cover in the climate system

Authors

    • Center for Climatic ResearchUniversity of Wisconsin-Madison
Article

DOI: 10.1007/s00382-007-0226-0

Cite this article as:
Vavrus, S. Clim Dyn (2007) 29: 73. doi:10.1007/s00382-007-0226-0

Abstract

Snow cover is known to exert a strong influence on climate, but quantifying its impact is difficult. This study investigates the global impact of terrestrial snow cover through a pair of GCM simulations run with prognostic snow cover and with all snow cover on land eliminated (NOSNOWCOVER). In this experiment all snowfall over land was converted into its liquid–water equivalent upon reaching the surface. Compared with the control run, NOSNOWCOVER produces mean-annual surface air temperatures up to 5 K higher over northern North America and Eurasia and 8–10 K greater during winter. The globally averaged warming of 0.8 K is one-third as large as the model’s response to 2 × CO2 forcing. The pronounced surface heating propagates throughout the troposphere, causing changes in surface and upper-air circulation patterns. Despite the large atmospheric warming, the absence of an insulating snow pack causes soil temperatures in NOSNOWCOVER to fall throughout northern Asia and Canada, including extreme wintertime cooling of over 20 K in Siberia and a 70% increase in permafrost area. The absence of snow melt water also affects extratropical surface hydrology, causing significantly drier upper-layer soils and dramatic changes in the annual cycle of runoff. Removing snow cover also drastically affects extreme weather. Extreme cold-air outbreaks (CAOs)—defined relative to the control climatology—essentially disappear in NOSNOWCOVER. The loss of CAOs appears to stem from both the local effects of eliminating snow cover in mid-latitudes and a remote effect over source regions in the Arctic, where −40°C air masses are no longer able to form.

1 Introduction

The most extensive component of the cryosphere, terrestrial snow cover is also the most rapidly and seasonally changing cryospheric variable (ACIA 2005). A snow pack strongly influences the overlying air, the underlying ground, and the atmosphere downstream. Many of these first-order effects are described by Cohen and Rind (1991), who emphasize the major thermodynamic influences of snow cover: high albedo, high emissivity, low thermal conductivity, and latent heat sink. The effect of snow cover on the atmospheric circulation has been investigated for purposes of seasonal forecasting (Cohen and Entekhabi 1999, 2001). In addition, the ecological and hydrological impacts of snow cover are important for environmental and water-resource issues (Campbell et al. 2005; Mote et al. 2005). Furthermore, recent and anticipated reductions in snow cover due to future greenhouse warming are an important topic for the global change community (IPCC 2001).

Many prior studies have investigated the physical basis for snow cover effects on climate. Early observational work found that snow cover can decrease the lower-tropospheric temperature locally by several degrees over days to months (Namias 1962; Wagner 1973; Dewey 1977). Subsequent modeling studies by Walsh et al. (1982) and Walsh and Ross (1988) demonstrated that excessive snow cover is associated with near-surface cooling of 5–10 K confined to the lower troposphere (up to 500 hPa), related to large-scale atmospheric circulation modes such as the Pacific-North American (PNA) pattern, and more influential to the planetary circulation over Eurasia than North America. Similar conclusions regarding the especially strong influence of Eurasian snow cover were reported by Foster et al. (1983), Walland and Simmonds (1997), and Yasunari et al. (1991). Idealized simulations by Barnett et al. (1988) corroborated the empirical evidence of Blanford (1884) that the Asian summer monsoon is affected by antecedent continental snow cover and also showed that Eurasian snow cover induces a downstream pressure response over North America. The dynamical influence of modeled Eurasian snow cover anomalies was later examined by Watanabe and Nitta (1998) and Clark and Serreze (2000), who found that the largest upper-air height response occurs downstream over the North Pacific Ocean.

These linkages between snow cover and atmospheric circulation have motivated efforts to forecast seasonal climate on the basis of snow cover anomalies. Cohen and Entekhabi (1999) reported that excessive summer–autumn snow coverage over Eurasia favors unusually cold winters over Western Europe and the eastern US, due to forcing of the negative phase of the Arctic Oscillation (AO). Subsequent studies by Cohen and Entekhabi (2001), Saito et al. (2001), Gong et al. (2003a, b), and Cohen and Saito (2003) refined the physical basis of this relationship: early-season snow cover anomalies trigger vertically propagating planetary waves that quickly alter the stratospheric polar vortex, whose anomalous strength then propagates downward to affect the surface annular mode during the course of the winter.

In addition to its possible role as a climate predictor, the ecological and societal impacts of snow cover are of considerable interest because of how a snow pack affects wintertime herbivores, the temperature of the underlying soil, and the magnitude and timing of river runoff. Rain falling on snow cover can cause large ungulate mortality due to surface icing, which animals cannot penetrate, and these rain-on-snow events are projected to become much more common in a warming climate (Putkonen and Roe 2003). Deer mortality also tends to increase during years of heavy snow, due to the similar foraging difficulties it poses (Tatatsuki et al. 1994). The low thermal conductivity of snow makes it an excellent insulator of the ground, resulting in soil temperatures up to 15 K warmer below a snow pack (Mölders and Walsh 2004). Ground temperatures beneath snow cover can be influenced as much by the snow as the overlying air temperature, as demonstrated by Stieglitz et al. (2003), who attribute half of the rise in the 20-m soil temperature at Barrow, AK to locally increasing snow cover since the 1970s. The insulating ability of a snow pack allows a surprising amount of ecological activity to occur during winter, much of which happens in the soil, where temperatures can remain high enough to support a wide range of biotic activities (Campbell et al. 2005). Warmer winters may actually result in colder soils because of snow cover reductions, based on field studies in which snow is manually removed from test plots. Such experiments have been conducted in New England over several winters by Groffman et al. (2001) and Decker et al. (2003), who observed lower soil temperatures and increases in soil freezing where snow cover is absent. The associated consequences were reported to be greater root mortality, more nutrient losses, and decreased productivity of certain tree species. Water resources in snowy climates are directly affected by both the presence and timing of snow cover. Mote et al. (2005) report declining mountain snow packs in western North America since 1950 and warn that projected reductions in future snow cover “will have profound consequences for water use”. Similarly, Barnett et al. (2005) conclude that future water supplies may be hindered by a warmer climate producing less snowfall and earlier seasonal melting. Both of these changes would thwart efforts to capture runoff, because most the early meltwater would exceed reservoir storage capacities and would no longer be available during the higher-demand periods in summer–autumn.

Terrestrial snow cover is expected to decrease in concert with greenhouse warming, and there are signs that such trends have already begun. Groisman et al. (1994) noted a retreat of North American springtime snow cover associated with strong warming during the previous decades, consistent with subsequent observations of earlier springtime snow melt over western North America (Stewart et al. 2005) and longer-term declines in Northern Hemisphere snow cover extent since the early twentieth century (Brown 2000). The extent of boreal snow during spring and summer was lower during the 1990s than at any time in the past 100 years (IPCC 2001). Climate models project continued decreases during this century, ranging from 5 to 15% over North America during the twenty-first century (Frei and Gong 2005) and up to 26% over the entire Northern Hemisphere (Dery and Wood 2006).

The purpose of this paper is to estimate the climatic impact of terrestrial snow cover on a global scale. This effort assesses the role of snow cover in the present climate and provides insight into how altered snow cover patterns may affect future climate. Thus, the snow-removal experiment described here is not intended merely as a “what if” pursuit, but rather as a means to quantify the influence of an important component of the climate system. This approach is similar to the numerous sea-ice removal simulations intended to evaluate the impact of ice cover on climate (Fletcher et al. 1973; Royer et al. 1990; Simmonds and Budd 1991; Bromwich et al. 1998) and a forest-removal experiment to quantify the climatic role of trees (Renssen et al. 2003). Prior studies of snow cover have also attempted to quantify its role, using the empirical and modeling approaches described above, as well as through statistical methods (Klein 1983, 1985; Klein and Walsh 1983; Walsh et al. 1985). While useful in shedding light on the role of snow cover, this earlier work has been limited in spatial and temporal extent. Model simulations of prescribed enhancements or reductions of snow cover have been done over specified regions based on observed fluctuations and over short (monthly-seasonal) time periods (Walsh and Ross 1988; Barnett et al. 1988; Cohen and Rind 1991; Walland and Simmonds 1997; Watanabe and Nitta 1998; Gong et al. 2004). The present study is unique in its radical approach of suppressing terrestrial snow cover at all locations and all times, so that the total climatological impact of snow cover can be evaluated. While obviously idealized, this tack is useful for assessing the global and time-mean influence, rather than the smaller-scale and shorter-term effects of snow cover anomalies examined in previous studies. This study also provides an upper-bound estimate of the impact of snow cover reductions to occur from anthropogenic warming, in which the positive snow-albedo feedback has been identified as a key process in amplifying the warming signal (ACIA 2005; Hall 2004).

A description of the model and experimental design is given in the following section. Section 3 includes an evaluation of the simulated snow cover in the control run and the major climatic changes that ensue from the removal of snow cover in the experiment. An interpretation of the results is presented in Sect. 4, along with caveats and implications for climate change. Conclusions are found in Sect. 5, which summarizes what the experiment implies for the role of snow cover in the present-day climate system.

2 Model description and simulations

NCAR’s Community Climate System Model, version 3 (CCSM3) is a fully coupled global climate model with components representing the atmosphere, dynamical ocean and sea ice, and land surface (Collins et al. 2005). The configuration used here employs a T42 atmospheric and land grid (approximately 2.8°× 2.8°), with approximately 1° resolution for the ocean and sea ice. The atmospheric component contains 26 levels in a hybrid-sigma pressure coordinate system. The land model, which contains ten sub-surface soil layers, exchanges energy, mass, and momentum with the atmosphere but does not allow changes in vegetation composition [although an optional dynamical global vegetation model does exist (Bonan and Levis 2005)].

This paper describes two parallel simulations to assess the global impact of terrestrial snow cover, using recent (1990) concentrations of greenhouse gases. The control run uses model-predicted snow cover over land and sea ice, whereas terrestrial snow cover is suppressed except on glaciers and ice sheets in the experiment (NOSNOWCOVER). Because this study is focused on the role of terrestrial snow cover, sea ice regions are not subjected to snow cover suppression. The climatic impact of snow cover on sea ice is probably quite different and more complex than that over land, because of competing thermodynamic influences of snow albedo and insulation on snow-covered ice packs (Maykut and Untersteiner 1971).

In NOSNOWCOVER, the snowfall generated within the atmospheric model was not altered, but all frozen precipitation was immediately converted into liquid–water equivalent upon reaching the land surface. This procedure avoids artificial latent heat effects in the atmosphere, compared with the alternative method of converting snowfall into rainfall before it reaches the ground, and ensures conservation of water. After the liquid–water-equivalent snowfall is applied to the land surface, the model predicts whether the water runs off, percolates into the soil, or ponds on the surface (details of the model’s land hydrology are found in Oleson et al. 2004). The model does not keep track of the temperature of melt water, but rather assigns the temperature of soil water to be that of the surrounding soil layer. Thus, melt water can percolate into the ground and refreeze, releasing latent heat as it does so. This process warms the soil, introducing an artificial heating term in NOSNOWCOVER, but the alternative treatment of warming snowfall to the melting point upon reaching the surface also fails to conserve energy. Fortunately, the effect of this artificial latent heating appears to be small, for it will be shown that soils cool dramatically in the absence of overlying snow cover (Sect. 3.2). The only other coding change in NOSNOWCOVER was to override part of the model’s canopy radiation module, which assumes that any standing water on leaves is snow when the canopy temperature is below freezing. In this experiment, such standing water is assumed to be liquid regardless of the temperature of the canopy.

The NOSNOWCOVER simulation was run 35 years, the first 15 of which were a transition to a new equilibrium and the final 20 were used for calculating the new climate state. The rapid adjustment time of the atmosphere and upper ocean is attributable to the (snow cover) forcing occurring only during certain times of the year and over a relatively small fraction of the globe that has a low heat capacity. A corresponding 20-year interval from the end of a long (1,000 year) control simulation was used for comparison to quantify the differences caused by the absence of snow cover.

3 Results

3.1 CCSM3 snow coverage in control simulation

The simulated and observed fractional snow coverage during mid-winter (January) and the transitional months (October and April) are displayed in Fig. 1 for the control run. Only the Northern Hemisphere is shown, because nearly all seasonal snow cover occurs in this hemisphere, except for a strip along the Antarctic coast. In each month the model underestimates the extent and concentration of the snow pack on both continents, particularly in Europe, southwestern Asia, and the western US. Quantifying this underestimate of total snow area is difficult, however, partly because the model assigns more land cover than observed (especially over the Canadian archipelago) and also because the choice of a snow-cover threshold is somewhat arbitrary in both the model and observations. The former problem is discussed in Frei et al. (2003) and Frei and Gong (2005), who opt instead to evaluate simulated snow cover as a fraction of the land area used by a particular GCM. By this measure, CCSM3’s Northern Hemisphere terrestrial snow coverage during mid-winter (around 0.55) is considerably smaller than the 0.75 observed (Brown 2000) and is the least among the 11 GCMs analyzed by Frei and Gong (2005). The exact reason(s) for this bias is beyond the scope of this paper, but it may stem from the model’s wintertime warm bias over extratropical land (Collins et al. 2005), which would favor relatively too much rainfall over snowfall in middle latitudes. By contrast, the simulated snow depth in high latitudes is too high, because of excessive precipitation that may also be attributable to the wintertime warm bias (Dickinson et al. 2006). Assuming that snow cover area is more important for atmospheric interactions than snow depth, the implication for this study is that the simulated climatic impact of terrestrial snow cover in CCSM3 is likely to be a conservative estimate compared with other models.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig1_HTML.gif
Fig. 1

Simulated (left) and observed (right) snow cover fraction during October (a, b), January (c, d), and April (e, f). The model produces too little Northern Hemisphere snow cover in each month (observations are from the Rutgers Snow Lab, http://www.climate.rutgers.edu/snowcover)

3.2 Results of NOSNOWCOVER simulation

The elimination of snow cover causes large annual near-surface air temperature increases of up to 6 K in northern Siberia and 5 K in northern Canada (Fig. 2). The global average warming (0.84 K) is one-third of the model’s 2.47 K response to a doubling of CO2 (Kiehl et al. 2006). As expected, the heating is enhanced in the Northern Hemisphere, which warms by 1.3 K in NOSNOWCOVER and by considerably more over extratropical land (Table 1). Much of the statistically significant warming (95% confidence limit, based on a Student’s t test) occurs well outside of the regions where snow cover is removed, even over interior oceans and well into the tropics. Approximately 75% of the world experiences a significant temperature rise, including 87% of the Northern Hemisphere and 98% of the boreal extratropics. The warming is not confined to the lower troposphere, unlike many previous prescribed snow cover simulations (Walsh and Ross 1988; Walland and Simmonds 1997; Kumar and Yang 2003), but instead extends to the tropopause throughout the world (Fig. 3). The strongest heating occurs in the lower levels of the Arctic, due to both the direct effect of enhanced heating of the land and the indirect response of sea ice reductions over the Arctic Ocean. Notice, too, the pronounced vertical heating around 30°N, caused by the elevated terrain of the Himalayas and Tibetan Plateau, which contributes to a strong-wave train forcing described below.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig2_HTML.gif
Fig. 2

Change in annual surface air temperature (K) in NOSNOWCOVER. Shaded regions are statistically significant at the 95% confidence interval, based on a Student’s t test

Table 1

Global and hemispheric statistics for the control and NOSNOWCOVER simulations. CAO days are defined relative to the control run’s climatology

 

Control

NOSNOWCOVER

Difference

Global surface air temperature annual (K)

287.82

288.66

0.84

NH extratropical land surface air temperature DJF (K)

266.98

270.81

3.83

NH extratropical upper-layer soil temperature DJF (K)

275.23

271.71

−3.52

NH extratropical land surface air temperature JJA (K)

291.27

292.53

1.26

NH extratropical upper-layer soil temperature JJA (K)

291.07

292.44

1.37

NH sea ice cover (10km2)

11.17

8.30

−2.87 (−26%)

SH sea ice cover (10km2)

14.26

12.88

−1.38 (−10%)

NH extratropical permafrost fractional coverage

0.092

0.158

0.066 (72%)

NH extratropical upper-layer soil moisture (m3/m3)

0.35

0.26

−0.09 (−26%)

CAO days/year (Asia)

1.8

0.4

−1.4 (−78%)

CAO days/year (Europe)

2.9

0.2

−2.7 (−93%)

CAO days/year (N. America)

2.0

0.1

−1.9 (−95%)

https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig3_HTML.gif
Fig. 3

Change in the vertical cross section of air temperature (K) due to the elimination of terrestrial snow cover. Shaded regions are statistically significant at the 95% confidence interval

As expected, the large increase in surface air temperature annually is dominated by the wintertime response, but significant warming occurs throughout the Northern Hemisphere in all seasons, including summer (Fig. 4; Table 2). Wintertime temperature rises of up to 8–10 K occur over the interior of Eurasia and North America, while summertime heating of up to 3–5 K exists across northern Siberia and Canada. The cooling influence of snow cover stems from its high surface albedo, emissivity, and insulation (Cohen and Rind 1991), variables that decrease considerably in NOSNOWCOVER: winter-spring surface albedo, for example, falls by up to 0.5 over formerly snow-covered land. Despite the extreme mid-high latitude atmospheric warming during winter in the absence of snow cover, removing the insulating presence of snow cover causes extreme cooling of the ground. The upper soil temperature falls by up to 14 K in northern Canada and 22 K in Siberia, while significant wintertime cooling spreads well into the middle latitudes of both continents. The below-ground cooling extends throughout all ten of the model’s soil layers (approximately 3 m), but the most pronounced drops in temperature occur in the surface layer.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig4_HTML.gif
Fig. 4

Top Change in wintertime a surface air temperature and b upper soil temperature in NOSNOWCOVER. Bottom Change in summertime c surface air temperature and d upper soil temperature. Shaded regions are statistically significant at the 95% confidence interval

Table 2

Monthly surface air temperature change averaged over the Northern Hemisphere in NOSNOWCOVER

Month

Air temperature change (K)

January

1.90

February

1.73

March

1.51

April

1.29

May

1.17

June

1.02

July

0.72

August

0.61

September

0.66

October

1.22

November

1.53

December

1.78

The highest air temperature increases during summer of more than 2 K occur in regions of climatological June snow cover in the control simulation, but the warming is enhanced by a drying of the soil and thus a shift in the relative proportion of surface turbulent heat fluxes from latent (evaporation) to sensible energy (Bowen ratio increases of 20–200%). Despite the warmer soils during summer, the much more extreme wintertime cooling in NOSNOWCOVER causes a large expansion in simulated permafrost (defined as locations where the deepest soil layer remains at or below freezing in every month) (Fig. 5). Throughout Eurasia and North America, the permafrost margin extends equatorward fairly uniformly by about 5–10° latitude. Due to the earth’s curvature, this meridional expansion into middle latitudes translates into a surprisingly large areal increase of 70%, including the permanently frozen soil that emerges in the Himalayas when snow cover is removed. Although the snow-cover removal applied here is idealized, this result may be relevant for demonstrating the complexity in the effect on permafrost of atmospheric warming and projected snow cover loss during future greenhouse warming; simple associations of permafrost thawing driven by temperature increases may not be valid.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig5_HTML.gif
Fig. 5

Permafrost margins in the control simulation (thin line) and NOSNOWCOVER (thick line)

The drying of the ground surface occurs not only in summer but throughout the year, causing large annual decreases in moisture in the top soil layer (Fig. 6). Maximum reductions in both continents mostly range from 40 to 60%, although even greater drying happens over the Tibetan Plateau. The percentage decrease in near-surface soil moisture peaks during winter and spring (not shown), and its spatial variation generally follows that of snow cover in the control run. There are two reasons for this similarity: (1) the model assigns all snowmelt water to the top soil layer in the control run, and (2) infiltration of surface water into deeper layers is greater in NOSNOWCOVER because the upper soil layer is drier and thus more permeable. As a result, the column-integrated soil moisture actually increases over much of the formerly snow covered domain. This explanation is consistent with how little the upper-layer soil moisture anomalies resemble the pattern of precipitation changes (Fig. 7), which exhibits both wetter and drier conditions over mid-high latitude land, with the most coherent shift toward significantly greater precipitation (>10%) over the Arctic Ocean. This precipitation increase in polar regions resembles that in greenhouse forcing simulations (Meehl et al. 2005), suggesting that the increased moisture content of a warmer global atmosphere produces the same hydrological signature regardless of the exact forcing mechanism.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig6_HTML.gif
Fig. 6

Percentage change in annual moisture content of the upper-most soil layer in NOSNOWCOVER. Shaded regions are statistically significant at the 95% confidence interval

https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig7_HTML.gif
Fig. 7

Percentage change in annual precipitation in NOSNOWCOVER. Shaded regions are statistically significant at the 95% confidence interval

Another important hydrological effect of snow cover is its impact on the annual cycle of runoff, which is regarded here as the sum of the model’s surface runoff and sub-surface drainage. The simulated seasonal variation of runoff differs greatly among low-, mid-, and high latitudes in the Northern Hemisphere and displays a strong snowmelt signature in the extratropics (Fig. 8). As expected, the annual cycle of runoff in low latitudes is affected only weakly by snow cover and instead closely follows the annual cycle of precipitation minus evaporation (P − E) (not shown). By contrast, the control simulation shows a distinct spike in runoff, coincident with peak snow melt, during spring in middle latitudes and May–June in the Arctic. The corresponding timing of runoff in NOSNOWCOVER is dramatically different, consisting of only weak seasonal variations in middle latitudes and an annual cycle in polar regions that resembles that of P − E, including a minimum during spring and maxima in both variables during September–October.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig8_HTML.gif
Fig. 8

Annual cycle of terrestrial runoff rate (kmday−1) in the control simulation (open circles) and NOSNOWCOVER (solid circles) integrated over the tropics, middle latitudes, and Arctic

The pattern of surface pressure change (Fig. 9), whose wintertime structure consists of pressure falls over the Arctic, is fairly typical of greenhouse warming experiments (Teng et al. 2006) and in agreement with the greater atmospheric convergence expected from the precipitation rise over this region. Specifically, the pattern of extratropical pressure change takes the form of a positive phase of the North Atlantic Oscillation (NAO), with a large drop of 4 hPa over the Nordic Seas and a substantial rise of 3 hPa centered in the eastern Atlantic. An even more pronounced pressure anomaly dipole forms in the Pacific-Alaskan sector, consisting of a significant pressure fall (rise) over Alaska (the central Pacific Ocean). These mid-latitude pressure increases largely compensate for the loss of atmospheric mass in high latitudes, consistent with the hydrostatic response of the Arctic to the pronounced surface warming.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig9_HTML.gif
Fig. 9

Change in wintertime sea level pressure (hPa) in NOSNOWCOVER. Shaded regions are statistically significant at the 95% confidence interval

Given the general decrease in surface pressure in high latitudes, a surprising result is the behavior of the Siberian High. This center of action is supposedly regulated by the underlying surface, such that positive snow cover anomalies strengthen the anticyclone by enhancing radiative cooling (Lydolf 1977; Cohen and Entekhabi 2001). Yet the removal of snow cover and accompanying warming (Fig. 4a) produce a significant pressure increase of 2 hPa that is practically co-located with the core of the modern Siberian High near Lake Baikal (Fig. 9) (Panagiotopoulos et al. 2005). A likely dynamical explanation for this paradox is the change in upper-air circulation that emerges from the anomalous surface heating patterns when snow cover is suppressed. The deep tropospheric warming in NOSNOWCOVER (Fig. 3) results in large geopotential height increases over most of the Northern Hemisphere, including pronounced ridging over eastern North America and northern Eurasia (Fig. 10). The latter feature, in combination with an anomalous trough centered around 40°N, 80°E, appears to drive the surface pressure rise in the core of the Siberian High due to wind convergence/negative vorticity advection at upper levels. Other height changes aloft also appear to force the major pressure changes at the surface, such that the lower pressure over Alaska is supported by upper-air troughing centered near the date line, while the central Pacific pressure rise is controlled by the overlying ridging in a more equivalent barotopic structure typical of maritime anomalies (Blackmon et al. 1979). These large-scale circulation changes are impressive and indicative of planetary wave propagation forced by snow cover removal. In particular, the anomalous stationary wave train comprised of a central Pacific ridge, northern–eastern Pacific trough, and eastern North American ridge resembles the geopotential anomaly pattern during winters with low Eurasian snow cover in observations and model simulations, the cause of which is attributed to decreased transient eddy activity (Walland and Simmonds 1997; Watanabe and Nitta 1998; Clark and Serreze 2000). The extreme ridging over the central Pacific in NOSNOWCOVER is particularly noteworthy, because it represents the largest upper-air height response, yet it is far removed from the terrestrial forcing domain on either continent.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig10_HTML.gif
Fig. 10

As in Fig. 9 but for 250 hPa geopotential height (m) with the global mean increase subtracted

One of the most important characteristics of snow cover is its ability to regulate not only mean temperatures but extreme values as well. Especially under favorable synoptic conditions of clear skies, low humidity, and calm winds, a fresh snow pack’s high emissivity and low thermal conductivity can cause nocturnal temperatures to fall by several degrees (Baker et al 1992; Dewey 1977). In addition, snow cover can exert a regional influence by strengthening synoptic-scale cold-air outbreaks (Namias 1985; Konrad 1998). The NOSNOWCOVER simulation captures and quantifies these effects under the most extreme conditions of total global snow cover suppression (Fig. 11). The model simulates an increase of the minimum daily-mean surface temperature during the year of more than 10 K across most of the territory covered by snow in the control run and more than 20 K over widespread regions in eastern Europe, western Asia, and central Canada. Even areas outside the simulated climatological snow pack (Fig. 1) experience very large moderations in the coldest temperature of the year—relative to the mean wintertime temperature rise (Fig. 4a)—due to the absence of snow cover reinforcing cold air during synoptically favorable events, such as polar air mass incursions. The much larger increase in the extreme minimum temperature than the mean is similar to the response of the climate system under greenhouse warming, in which the lowest daily temperature of the year is projected to rise more than the wintertime mean in most terrestrial regions, due largely to reduced snow cover (Hegerl et al. 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig11_HTML.gif
Fig. 11

Change in the lowest daily mean temperature (K) of the year in NOSNOWCOVER

Another way to view the impact of snow cover on extreme cold is by comparing the frequency of bitterly cold days (e.g., ≤ −40°C mean temperature) with and without snow cover in the Arctic air mass source regions. In reality, these areas are virtually always covered with snow during winter and thus an observational comparison is impossible. However, the model indicates that such extreme cold would be virtually non-existent in the absence of a snow pack (Fig. 12). Whereas the frequency of daily mean temperatures ≤ −40°C exceeds three weeks per winter across the coldest regions of Siberia and Canada when snow cover is present, almost no such bitterly cold days occur in NOSNOWCOVER. The precise physical mechanism(s) responsible for this difference is beyond the scope of this paper, but it is likely attributable to a combination of effects, both thermodynamic (lower surface emissivity and greater ground heat flux without a snow pack) and dynamic (greater heat advection from lower latitudes).
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig12_HTML.gif
Fig. 12

Number of days per winter on which the daily mean temperature is ≤ −40°C in (top) control run and (bottom) NOSNOWCOVER

The weakening of Arctic air masses without snow cover would be expected to curtail the frequency and magnitude of extreme cold air outbreaks (CAOs) over middle latitudes. CAOs are defined here as in Vavrus et al. (2006): an event of at least two consecutive days during which the daily average surface air temperature at a grid point is at least two standard deviations below the simulated wintertime mean surface air temperature. Based on this definition, the preferred locations for CAOs are Europe-central Asia and northwestern North America (Fig. 13a), regions whose wintertime climates are relatively temperate but which are situated close enough to Arctic air mass source regions to sustain occasional incursions of bitterly cold air [these favored regions agree well with observations (Vavrus et al. 2006)]. The frequency of CAOs under a different climatic regime can be defined one of two ways: using an absolute temperature threshold based on the mean temperature and standard deviation of the control climate or using a different absolute temperature threshold based on the mean temperature and variability of the altered climate. Because each criterion provides useful information, the pattern of CAOs in NOSNOWCOVER is depicted using both methods (Fig. 13b, c). Using the stricter definition of CAOs (relative to the control climate), these extreme events virtually disappear in the absence of snow cover (Fig. 13b). Their frequency dwindles to less than a half day per winter over the US, Canada, and most of Eurasia, representing average decreases of well over 80% across the Northern Hemisphere (Table 1). Interestingly, CAOs actually become more common over parts of southern Asia, around India and Pakistan, due to the circulation changes that favor enhanced northerly flow in this region (Fig. 9). The alternative approach is to define a CAO relative to the warmer and less variable NOSNOWCOVER climate (Fig. 13c), which provides a sense of how extreme cold events might be perceived under a milder base state. Using this criterion, the changes in CAO days display an interesting dipole pattern of increased frequency in high latitudes and reductions of over 50% along the margins of typical polar air mass excursions in the US, southern Asia and Europe. This result shows that snow cover assists in expanding the spread of Arctic air masses into milder climatic regions; without this reinforcing boundary condition, bitterly cold air masses tend to remain more confined to high latitudes.
https://static-content.springer.com/image/art%3A10.1007%2Fs00382-007-0226-0/MediaObjects/382_2007_226_Fig13_HTML.gif
Fig. 13

Number of extreme cold-air outbreak days per winter in the (top) control run and (middle) NOSNOWCOVER, relative to the control climate (i.e., using the mean temperature and standard deviation from the control run to define a CAO). Bottom The change in CAO days per winter relative to the NOSNOWCOVER climate (using the mean temperature and standard deviation from the NOSNOWCOVER run to define an extreme cold-air outbreak)

In addition to this local thermodynamic role in accentuating Arctic air mass migrations, snow cover also seems to affect CAOs by altering the general circulation. In NOSNOWCOVER the anomalous upper-air flow over the western half of North America during winter becomes decidedly southerly (Fig. 10). By contrast, the equatorward migration of Arctic air masses into the continental US is known to be most likely when the upper-air circulation is precisely the opposite: a northerly flow pattern formed by a ridge (trough) over western (eastern) North America, such as the positive PNA pattern (Konrad 1998). In addition, the large reduction of average surface pressure during winter in NOSNOWCOVER over the Arctic air mass source region of North America (Alaska and the Yukon) (Fig. 9) is not conducive to the powerful polar anticyclones that drive extremely cold air into the contiguous US during CAOs (Dallavalle and Bosart 1975). The same kind of argument holds for Eurasia, where lower mean surface pressure suggests a reduction in very strong polar anticyclones that would drive Arctic air into Europe. Indeed, the frequency of the most powerful anticyclones that form over northern Siberia (>1,040 hPa) decreases by about 40%, while 40–80% declines occur over the favored source regions of North America (not shown).

4 Discussion

The results demonstrate that in addition to its very strong local influence, snow cover has far-reaching impacts in space and time from its direct terrestrial forcing during the cold months of the year. Polar climate is strongly affected by the snow cover removal, both by virtue of the region’s in situ snow cover and through polar amplification from sea ice feedbacks. In NOSNOWCOVER, the oceanic areas of the Arctic experience significant changes in sea ice cover (26% decrease), precipitation (>10% increase), and surface pressure (up to 4 hPa lower), even though the snow cover forcing is confined to land areas. In fact, 72% of the global ocean area shows significant temperature increases, including 63% of the tropical oceans (30°S–30°N). This widespread warming is not merely surface-trapped but extends throughout the troposphere virtually everywhere in the world (Fig. 3). The associated large changes in upper-air circulation (Fig. 10) are remarkable both in their planetary extent and extreme remote response that generates the largest geopotential height increase nearly as far away as possible from any continent. This PNA wave train is probably forced from Eurasia, based on similar behavior found in previous prescribed snow cover simulations (Barnett et al. 1988; Yasunari et al. 1991), and may have very important implications for past and future global climate change. Yasunari et al. (1991) simulated reduced pressure aloft over eastern North America during anomalously high Eurasian snow cover and suggested that this troughing would help initiate ice sheets over the presumed nucleation region of northeastern Canada (this pattern is analogous to the anomalous ridge formed over eastern North America due to snow cover removal in Eurasia in NOSNOWCOVER). Likewise, Lamb (1977) proposed a remote-forcing mechanism in which glaciation in Scandinavia is triggered by a North American ice sheet triggering downstream adjustment of stationary planetary waves. Conceivably, a similar kind of dynamical forcing could become important for glacial ablation in a warming future climate, if terrestrial snow packs retreat as expected.

Another point about future climate change and the large-scale impact of snow cover is the magnitude of snow cover forcing in this experiment relative to that of greenhouse gases. As noted earlier, the global surface air temperature increase (0.84 K) in NOSNOWCOVER is one-third as large as the model’s sensitivity to 2 × CO2, thus indicating that snow cover is an important component of the global climate system. On a per area basis, however, the response of global temperature is much more sensitive to snow cover than to the CO2 increase, given that greenhouse forcing occurs continuously over the entire surface area of the earth, whereas the snow cover removal only applies to ice-free land during the snow-covered portion of the year. In NOSNOWCOVER the elimination of snow cover across 3.7% of earth’s area producing a 0.84 K global warming represents a normalized temperature response 9 times larger than that from a doubling of CO2, in which the 2.47 K warming was driven by forcing across all of earth’s surface. Alternatively, the traditional definition of climate sensitivity—the global mean temperature change relative to the global top-of-atmosphere (TOA) radiative perturbation—also indicates a much stronger influence from snow cover. The 0.84 K warming from a 0.57 W m−2 TOA perturbation caused by snow cover removal (estimated using the method of Gregory et al. (2004)) represents a sensitivity twice as large as the model’s 2 × CO2 case of a 2.47 K warming from a 3.6 W m−2 TOA perturbation [as reported by Kiehl et al. (2006)].

Several caveats apply to the results of this study. First, the reader is reminded that the snow cover generated by CCSM3 is smaller than observed (Fig. 1) and thus the estimated impact of snow cover is probably underestimated compared with a more realistic model. Second, because terrestrial snow cover was suppressed globally, this experiment is not able to assess the influence of snow cover by region (likewise, the relative contributions of remote forcing vs. the local role of sea ice reductions in shaping the large Arctic Ocean response cannot be determined from the existing experimental design but would require supplemental simulations with fixed sea ice). However, follow-up sensitivity tests using regional snow cover masking are being conducted to evaluate the relative influence of snow cover in Eurasian versus North American snow cover and in high-latitudes versus middle-latitudes. The results of these experiments will be presented in a future paper. Third, the simulated changes in soil temperature and permafrost expansion probably are exaggerated due to the shallowness of the model’s soil depth (3 m). Such a thin layer of soil may respond too sensitively compared with a deeper soil module that has greater thermal inertia. Fourth, the simulated expansion of permafrost area through idealized snow cover removal may not be directly applicable to the effects of snow pack retreat under greenhouse forcing. Future changes will be driven by atmospheric radiative heating that will simultaneously favor both snow melt and warmer soils. In addition, future snow cover retreat will probably be concentrated along the margins of the snow pack, where permafrost is uncommon. Fifth, this study has not considered any superimposed influence of vegetation changes, even though these undoubtedly would occur under such extreme terrestrial-based climate changes as those induced by snow cover removal. Primarily through their effect on surface albedo, extratropical shifts in vegetation cover can strongly influence climate change, through such positive feedbacks as those triggered by the conversion of tundra to boreal forest (Foley et al. 1994; Levis et al. 1999) and vice versa (Gallimore and Kutzbach 1996). Given the competing effects on vegetation of the climate changes caused by snow cover suppression—atmospheric warming but large wintertime cooling of the ground, and drying of the upper soil but not deeper layers—it is not clear how vegetation would respond to and influence the climate. A follow-up CCSM3 simulation incorporating the optional dynamical global vegetation model will allow vegetation to interact with the atmosphere and thus address this potentially important aspect of the influence of snow cover on climate.

5 Conclusions

A GCM simulation with terrestrial snow cover suppressed is used to assess the role of snow cover in the global climate system. While idealized, this approach yields useful information about the importance of this climatic component under present conditions and suggests potential influences of snow cover during past and future climate regimes. The results support the following primary conclusions about the influence of snow cover under present-day conditions.
  • Snow cover significantly cools the air throughout the troposphere at most locations, causing near-surface air temperatures to decrease annually by up to 6 K over high-latitude land and as much as 10 K during winter.

  • The global cooling effect of snow cover on surface air temperature is one-third the magnitude of 2 × CO2 forcing and nine times as large on a per area basis (twice as large using the traditional definition of climate sensitivity).

  • The insulating effect of snow cover is dominant during winter, causing much warmer soils over the climatological snow pack. The insulating impact is large enough to counteract the effect of overlying air temperature changes on the ground temperature. This suggests that an interesting interplay could develop in the future between a warming climate acting to shrink permafrost extent but waning snow cover favoring permafrost expansion.

  • The existence of snow cover greatly reduces the amount of permanently frozen soil at present, resulting in permafrost margins being about 5–10° further poleward.

  • Near-surface soils are much wetter as a result of overlying snow packs, which moisten the near-surface soil by about 25% across the boreal extratropics.

  • Snow cover dramatically affects the annual cycle of surface runoff, which peaks due to snow melt during spring-early summer in the extratropics and would instead follow the annual cycle of P − E in polar regions in the absence of snow cover.

  • Polar sea ice is much more extensive due to terrestrial snow cover, expanding by 26 and 10% in the Arctic and Antarctic regions, respectively.

  • The resulting colder, icier conditions in high latitudes affect the hydrological cycle: annual precipitation over the Arctic Ocean is significantly reduced (>10%) by the existence of adjacent snow cover on land.

  • Snow cover strongly affects atmospheric circulation in mid-high latitudes. The presence of snow cover causes Arctic surface pressure to be higher and to favor a more negative NAO pattern. Snow cover also shapes the flow pattern aloft, resulting in local and downstream height changes throughout the Northern Hemisphere.

  • Extreme temperatures during winter are strongly regulated by snow cover. Snow packs depress the coldest wintertime daily mean temperature by up to 20 K and are essential for the formation of bitterly cold Arctic air masses (below −40°C).

  • Extreme cold air outbreaks are highly dependent on snow cover. Over most of Eurasia and North America, the chilling effect provided by the underlying snow pack is required for CAOs to occur. Without this reinforcing boundary condition, bitterly cold air masses tend to remain more confined to high latitudes.

Acknowledgments

This project is supported by collaborative National Science Foundation grants ATM-0332099, ATM-0332081, and OPP-0327664. Constructive input from and collaboration with John Walsh, Diane Portis, and Bill Chapman on extreme cold air outbreaks were valuable in strengthening the manuscript. Suggestions by Michael Notaro on the overall content were also very helpful in improving the paper. Sam Levis was instrumental in assisting on the coding changes within the CCSM land component model.

Copyright information

© Springer-Verlag 2007