1. Introduction
a. A case for snow cover
Surface boundary forcing exerts a strong influence on monthly and seasonal climate variability (Shukla 1984). Previous studies have emphasized sea surface temperatures (SSTs; e.g., Ropelewski and Halpert 1986; Bhatt et al. 1998) and soil moisture (e.g., Karl 1986; Wolfson et al. 1987; Wang and Kumar 1998) as boundary conditions; snow cover has only recently become a focus in investigations of seasonal atmospheric variability. Namias (1962) analyzed the severe 1960/61 winter and concluded that the anomalously high snow depth across North America initiated a positive feedback: deep snow cover altered local cyclogenesis patterns to support below-normal temperatures and further snowfall across the Northern Hemisphere. Later studies generalized these observations and extended them to western European and Eurasian snow cover (Namias 1978, 1985; Saito and Cohen 2003).
General circulation models (GCMs) have facilitated the most recent investigations of snow cover anomalies’ effects on Northern Hemisphere circulation. Walsh and Ross (1988) employed five two-run sets of the National Center for Atmospheric Research (NCAR) Community Forecast Model. Each set included one run with a forced initial snow cover anomaly across North America—drawn from 1976–84 data—and one run with an anomaly equal in magnitude, but opposite in sign. Downstream effects consisted of above-normal surface temperatures in northern Scandinavia, lower surface pressures over Iceland, and below-normal precipitation across western Europe and North Africa. Later studies confirmed these effects, but pointed to the importance of Eurasian, and particularly Siberian, snow cover in controlling Northern Hemisphere climate variability. Gong et al. (2003b) forced excess and deficient snow mass in North American and Siberia in the European Centre Hamburg (ECHAM3) model. Excess Siberian snow weakened the North Atlantic low and drove the jet stream southward into western Europe. Excess North American snow promoted the opposite pattern, strengthening both the oceanic low and a ridge across western Europe, but with a much weaker correlation between the snow cover and its downstream effects. Other observation–analysis and GCM studies have uncovered links between Eurasian snow cover and the propagation of springtime Rossby wave trains (Yasunari et al. 1991), downstream Eurasian surface temperatures (Watanabe and Nitta 1998; Cohen and Entekhabi 1999), and the Indian monsoon (Hahn and Shukla 1976; Dey and Kumar 1983; Dickson 1984; Vernekar et al. 1995).
These GCM and observation–analysis studies have suggested that snow cover can affect local and downstream atmospheric circulation. One finds little mention, however, of North American snow cover generating a significant transatlantic teleconnection. No substantial link has yet been found between North American snow cover and the North Atlantic Oscillation (NAO), the dominant atmospheric mode that explains much of the variability in the winter climate of western Eurasia and the eastern United States (Hurrell 1995a, 1996).
b. The North Atlantic Oscillation
Walker and Bliss (1932) defined the NAO as the propensity for Iceland’s sea level pressure (Psl) to be lower relative to that of the Azores, southwestern Europe, and northern Africa. The positive phase represents an intensification of this pattern, while the negative phase represents a weakening. The strengthened meridional pressure gradient of the positive phase advects warm, moisture-laden air from the Atlantic Ocean over northern and western Europe, resulting in warmer temperatures and increased precipitation (van Loon and Rogers 1978; Hurrell 1995a; Hurrell and van Loon 1997; Thompson and Wallace 1998). In the negative phase, the Icelandic low shifts south and the ridge over the Azores moves east, directing warm Atlantic air into southern Europe and the Mediterranean and leaving northern Europe cool and dry (Rogers and van Loon 1979; Hurrell and van Loon 1997; Ulbrich and Christoph 1999). The NAO is better defined in winter than in summer (Barnston and Livezey 1987). Strong correlations have been found between the NAO and winter snowfall in the northeastern United States (Hartley and Keables 1998) and the Appalachian Mountains (Hartley 1999), winter atmospheric circulation in the Middle East (Kutiel and Kay 1992), streamflow in the Tigris and Euphrates Rivers (Cullen and de Menocal 2000), rainfall in the Caribbean and Central America (Giannini et al. 2000, 2001), and Siberian precipitation and surface runoff (Peng and Mysak 1993), among other variables.
The NAO is the dominant mode of Northern Hemispheric, lower-tropospheric interannual variability (Ambaum et al. 2001). Itoh and Harada (2004) examined correlations between the tropospheric Pacific–North American (PNA) pattern, the NAO, and their stratospheric counterparts, a wavenumber-1 mode and the annular mode, respectively. They found that although the NAO displayed a slight correlation with the PNA, the NAO was largely independent of other tropospheric and stratospheric processes. Several studies have found that the NAO explained at least 35% of the variability in Atlantic 500-hPa heights during boreal winter (Esbensen 1984; Kushnir and Wallace 1989; Wallace 1996), much more than the PNA or the Arctic Oscillation (AO). Rogers (1990) demonstrated that the NAO explained the greatest percentage of interannual variability in North Atlantic Psl during eight months of the year.
c. Possible connections between snow cover and the NAO
Recent studies have implied that the NAO is a fundamental, internal atmospheric mode, and that external forcings cannot modify the pattern at a level significantly above background noise (e.g., Robertson 2001; Baldwin 2001). Other studies have challenged that claim, advancing stratospheric variability (Baldwin and Dunkerton 1999; Shindell et al. 1999), SSTs in the North Atlantic (Kushnir 1994; Marshall et al. 2001) and the Indian Ocean and the Pacific warm pool (e.g., Hoerling et al. 2001; Bader and Latif 2003, 2005; Hurrell 2004), sea ice extent and depth (Mysak and Venegas 1998), and snow cover (Walsh and Ross 1988; Cohen and Entekhabi 2001; Gong et al. 2002, 2003a, b) as potential mechanisms for modifying the NAO.
Gong et al. (2002) found that midautumn Siberian snow cover anomalies enhanced NAO variability, altering Psl over the North Atlantic and southern Europe. That study used two ensembles of ECHAM3 simulations, one with forced snow cover and the other with model-determined or “free” snow cover. Gong et al. concluded that a Psl anomaly over western Siberia showed a negative correlation with the NAO; excess (deficient) snow mass over Siberia led to higher (lower) pressures over Greenland and the Arctic and lower (higher) pressures over western Europe. Gong et al. (2003a) attributed this correlation to a positive feedback in the vertical propagation of stationary waves over Siberia. Siberian snow cover may not drive the NAO itself, but anomalous snow cover may create fluctuations in NAO phase or amplitude (Gong et al. 2003a, b).
Several studies have attempted to distinguish between the influences of Eurasian and North American snow cover on the NAO. Cohen and Entekhabi (2001) also used paired forced- and free-snow-cover GCM runs, in which the forced simulations added snow mass between 35° and 50°N across North America and Eurasia. The additional North American snow lowered Psl and 500-hPa heights in the North Atlantic, implying the positive NAO phase. Gong et al. (2003b) found that North American snow cover showed a much weaker correlation to NAO phase and intensity than Siberian snow cover. An excess of the former favored the positive NAO, while an excess of the latter favored the negative phase. Saito and Cohen (2003) explored lag correlations between excess Siberian and North American snow cover and the NAO. For the Siberian snow cover, a positive feedback pattern emerged: autumn Siberian snowfall led to a negative winter NAO, which in turn led to more intense spring snowstorms, thereby increasing springtime soil moisture content and driving heavier snowfalls the following autumn. There was no significant lag correlation between North American snow cover and the NAO. These studies imply that while there may be some tenuous connection between North American snow cover and the NAO, Eurasian snow cover has a much greater influence on the transatlantic circulation.
d. Purpose of the present study
The purpose of the present study is to investigate the relationship between persistent, regional North American snow cover anomalies and Eurasian winter climate. Previous studies have hinted at such a connection, but its existence remains uncertain. This study differs from recent research in this area (e.g., Gong et al. 2003b; Saito and Cohen 2003) because where those studies have investigated hemisphere- or continent-wide snow cover anomalies, we analyze the effects of snow cover in a comparatively limited region of the northern Great Plains on the wider transatlantic atmospheric circulation. As this spatial scale of snow anomaly has received little attention, our methods are comparatively simple, but consistent with past research in snow cover–atmosphere interaction: we employ an ensemble of atmosphere-only GCM (AGCM) simulations and prescribe an idealized snowpack that persists in the northern Great Plains throughout November–February. We then compare this ensemble to a second ensemble with no prescribed snow forcing. To investigate issues of intraseasonal variability in the atmospheric response to snow anomalies and the duration of our prescribed snow, we conduct a third ensemble in which the idealized snow anomaly is added in January. Detection of a teleconnection to the NAO in these simple experiments with idealized forcing would provide support for further studies using different GCMs, more complex or time-varying forcing, or studies using observational data that may provide additional evidence for a teleconnection. Section 2 describes our three ensembles and section 3 presents the differences between them; these results are discussed in further detail in section 4. Finally, the significant conclusions from this study are summarized in section 5.
2. Methods
a. Motivation
The northern Great Plains of the United States experiences high interannual and intraseasonal variability in snow cover and depth (Robinson 1996; Frei and Robinson 1999; Robinson and Frei 2000). Frei and Robinson (1999) and Robinson and Frei (2000) used principle component analysis to identify regions in the Northern Hemisphere in which interannual variations in snow extent were correlated in time across the region; the northern Great Plains is one such region. Additionally, Robinson and Frei (2000) found that the northern Great Plains, the southern Canadian prairies, and the northeastern United States collectively account for at least 60% of the monthly variability in North American snow extent during November–March. Such interannual and intraseasonal variability, combined with the coherence of the variability in snow cover across the northern Plains, makes this region ideal for a sensitivity study of the effect of snow cover on the downstream circulation.
The land cover in the northern Great Plains consists mostly of low shrubs and grasses, which ensures that even a shallow prescribed snowpack will completely cover the ground and present a homogeneous surface to the atmosphere. Using 19 yr of albedo observations at the University of Minnesota, Baker et al. (1991) found that a snow depth of approximately 15 cm was sufficient for the albedo of a snowpack surface to be independent of the underlying surface. The impact of this finding on the design of our experiment will be discussed further in section 2b (2). The Great Plains’ location downstream of the Rocky Mountains promotes local cyclogenesis. Elguindi et al. (2005) found that increasing the snow cover extent in the Great Plains in the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MM5) weakened cyclones by an average of 4 hPa as they passed over the snow-covered surface. This weakening was primarily caused by a reduction in the near-surface thermal and moisture gradients in the warm sector of the cyclones. While the aim of the present study is to examine transatlantic teleconnections rather than the behavior of individual cyclones, Elguindi et al. (2005) indicated that anomalous Great Plains snow cover can systematically affect storm systems, causing effects that may propagate downstream.
b. Experiment design
1) The Community Climate System Model
This study employs two coupled components of the NCAR Community Climate System Model, version 2.0.1 (CCSM2; Kiehl and Gent 2004): the Community Atmosphere Model, version 2.0 (CAM2) and the Common Land Model, version 2.1 (CLM2). CAM2 has a 20-min time step, 26 vertical levels, and a triangular truncation at wavenumber 42 (T42). The grid cells are approximately 2.8° latitude by 2.8° longitude, equivalent to 300 km by 225 km in the middle latitudes. This study uses the Data Ocean version of CAM2, in which SSTs and sea ice are forced to a monthly climatology. A previous version of this model, the NCAR Community Climate Model, version 3 (CCM3) has been used to investigate global teleconnections resulting from changes in land cover in the tropics and in southeast Asia (Zhao et al. 2001). That study found that such land cover changes could alter the strength of the Hadley and Walker circulations and so modify tropospheric circulation patterns at higher latitudes.
The land surface parameterizations of CLM2 are of particular importance to our simulations. Each grid cell’s surface characteristics are determined by dividing the cell into as many as five subgrid ground-cover types—glacier, lake, wetland, urban, and vegetated—and specifying a weighting fraction for each type. The vegetated ground-cover type may itself be composed of up to 16 patches with distinct plant functional types (PFTs). CLM2 treats each subgrid area and vegetated patch as a separate column for energy and water calculations. The surface albedo of each land column is a function of the ground-cover type, vegetation, soil color and type, soil moisture, and snow cover (Oleson et al. 2004). The contribution of snow cover to the albedo depends on the snowpack age and soot content and the solar zenith angle (Wiscombe and Warren 1980; Oleson et al. 2003). CLM2 calculates the fraction of snow-covered ground as a function of the snow depth and the soil roughness length; the model then uses this fraction to weight the contributions of soil and snow to the total albedo. Snow cover also affects the albedo of vegetated patches by reducing the exposed leaf and stem area through ground accumulation and canopy interception (Dickinson et al. 1993; Oleson et al. 2003).
CLM2 operates on up to 5 layers within the snowpack and on 10 soil layers to facilitate more accurate snow–land surface interactions (e.g., heat and moisture exchange). Internal snowpack layers exchange heat and moisture with vertically adjacent layers, and the model allows for compression of an individual layer over time. The model varies snowpack parameters with ground-cover type and PFT. At each time step and for each grid point, CAM2 passes atmospheric data, including precipitation type and intensity, to CLM2. CLM2 then diagnoses snow mass as a function of canopy interception of snowfall, ground accumulation, snowmelt, evaporation, and sublimation.
2) The forced- and free-snow-cover experiments
Nine independent atmospheres—all based on data from 1 September—initialized one member apiece in three nine-member ensembles of six-month integrations. The first member in each ensemble (case 1) used the 1 September 2000 atmospheric dataset NCAR distributed with the CAM2 source code. For each of the eight subsequent members (cases 2–9), a random perturbation of up to ±1% was applied to temperature (°C), wind speed (m s−1), and absolute humidity (g kg−1) at all 26 vertical levels, as well as to surface pressure (Pa). Within five days, the Psl and 500-hPa height patterns of each member were not significantly spatially correlated to those of any other member. Ensemble simulations are necessary to distinguish impacts from prescribed forcings from those stemming from the model’s internal variability; past studies have emphasized the need for ensemble simulations to properly represent the NAO (Mehta et al. 2000). After two months of spinup, the ensembles diverged on 1 November into one free-snow-cover ensemble (i.e., a control) and two forced-snow-cover ensembles. All simulations were then integrated for four additional months, ending on 1 March. Of the two forced-snow-cover ensembles, the “early-season” ensemble was forced with anomalous snow cover beginning on 1 November, while the “late-season” ensemble commenced snow cover forcing on 1 January. Aside from the date on which the snow cover forcing began, the forced-snow-cover ensembles were identical. Most of our discussion will center on the early-season ensemble, as it allows us to investigate the effect of snow cover forcing throughout the winter. The late-season ensemble will determine whether the January and February results from the early-season ensemble are (i) the result of long-period snow cover forcing, or (ii) caused by changes in the atmospheric basic state during the season.
The free-snow-cover ensemble included no externally prescribed snow cover forcing; the model could freely determine snow cover extent and depth. The forced-snow-cover ensemble prescribed an idealized snowpack in an approximately 925-km (north–south) by 900-km (east–west) “forcing area” centered over Nebraska (Fig. 1). Model grid cells inside the forcing area experienced a prescribed snowfall of 3 cm hr−1 for 24 h on 1 November (early-season experiments) or 1 January (late-season experiments). The 72-cm snowpack was fully built by the day after the forcing began and was then made to persist until the end of the simulation. Although the snowpack was fixed in depth and extent during the winter season, CLM2 could melt and evaporate snow at each time step according to its unmodified physics. An amount of snow equal to any that evaporated or melted during a given time step was immediately returned to the snowpack before the end of the time step to maintain a constant 72-cm depth. From the perspective of the model water-balance routines, the snow was added under the guise of precipitation, although no precipitation actually fell and no atmospheric variables were modified to induce snowfall. No water was artificially added to or removed from CAM2 or CLM2, and neither model was forced to freeze, melt, sublimate, or otherwise change the state of any water molecules. The snowpack was returned to its original depth before CLM2 could calculate surface moisture balances or energy fluxes, so that CLM2 “saw” a snowpack of the same depth and extent at each time step. Snow cover was the only boundary condition altered between the two ensembles. While our prescribed snowpack has a spatial extent consistent with events observed in the northern Great Plains, the snowpack’s depth and persistence throughout the winter season have been exaggerated from observations for the purposes of this sensitivity experiment. The late-season ensemble will examine the effects of the longer persistence; the impact of a deep snowpack is discussed below.
The idealized snowpack caused an increase of 0.46 in the ensemble-mean, time-mean surface albedo from the free-snow-cover ensemble (0.20) to the early-season ensemble (0.66). This albedo increase was spatially consistent across the forcing area and varied temporally only when the free-snow-cover ensemble generated substantial snow cover over the forcing area. Our prescribed snow cover covered most of the low-level vegetation in the northern Great Plains (section 2a; Baker et al. 1991). Increasing the depth of the snowpacks would have little further effect on the albedo perturbation, because the fraction of snow-covered ground in CLM2—used to determine the surface albedo—is already asymptoting toward unity at our prescribed depth. Any decrease would need to be sufficient to expose a large amount of vegetation or soil or to substantially change the fraction of snow-covered ground. Baker et al. (1991) demonstrated that the rate of increase in the albedo with increasing snow depth slowed considerably as the albedo approached 0.7, validating the response of the surface albedo in CLM2 to our snow cover forcing. As our prescribed snow depth is far greater than the 15-cm threshold required by Baker et al. (1991) to present a homogeneous surface to the atmosphere, our results will approximate the maximum impact that snow cover forcing in this region may have on the atmosphere. Although experiments using snowpack much thinner than the one considered here may demonstrate smaller impacts, or different impacts altogether, such experiments are left for future studies and will be discussed in section 5.
c. The bootstrap test
To determine the statistical significance of differences between the forced- and free-snow-cover ensemble means, we apply the bootstrap method described in Efron (1981). The advantage of the bootstrap method is that one need not assume that one’s data is normally distributed before applying the test. With our limited sample of nine ensemble members, this advantage is of great importance. In this method, one randomly selects a sample of N data fields from an initial dataset of size N. In our study, N is the number of ensemble members (9) and the data fields are the differences between a given forced-snow-cover member and the corresponding free-snow-cover member. The randomization process operates on the global data field (i.e., by randomly selecting one global map of differences) to preserve the spatial characteristics of the tested variable. This random selection is carried out “with replacement,” that is, each data field may appear in the sample once, more than once, or not at all. Once this sampling process is repeated enough to collect an arbitrarily large number of difference maps—Efron (1981) suggests 1000 repetitions—any problems owing to a limited set are virtually eliminated.
The set of 1000 difference maps are then sorted at each grid point from the most negative difference to the most positive difference. This rearranges the set so that the first difference map shows the most negative differences at each grid point and the 1000th shows the most positive differences at each grid point. In this rearranged set, the 25th and 975th maps give the 95% confidence interval: any signal appearing in both the 25th and 975th maps is significant at the 5% level. For example, if a positive temperature difference appears in the 25th difference map, it is significant at the 5% level since the 25th difference map gives the 25th most negative (i.e., coldest) differences from the randomization process. The same logic applies for a negative temperature difference in the 975th difference map. It is critical to note that the bootstrap method determines the statistical significance of individual grid points just as well as it determines the significance of larger spatial patterns.
3. Results
a. Local responses
Although the purpose of this study is to examine the downstream effects of North American snow cover anomalies, it is necessary to establish first that the CLM2 snowpack parameterization provides a physically reasonable local response in energy and moisture variables before analyzing the more complex downstream patterns.
1) Near-surface temperatures
The forced runs of each case were compared with the corresponding free run, beginning with the early-season runs. It is well established that snow cools the overlying air (e.g., Leathers and Robinson 1993) and influences local and downstream temperatures through dynamic and thermodynamic processes (e.g., Cohen 1994; Walland and Simmonds 1996). Variables were time averaged over the snow cover–forcing period, November 1–February 28 (hereafter NDJF). All cases exhibited cooler 2-m temperatures (T2m) over the snow-covered region in the forced runs, ranging from 2°C in cases 1 and 7 to more than 8°C in cases 3 and 6 (Fig. 2). All cases also showed some decrease in T2m downwind; cases 3 and 5–9 exhibited cooling of more than 2°C over Labrador. Although the local temperature differences may be enhanced by the effectively infinite sink of latent heat provided by our experiment design (i.e., CLM2 can melt the forced snow cover indefinitely), the amount of snowmelt was on the order of 0.1 cm day−1 during NDJF and so we believe this effect is limited.
The difference in T2m between the early-season forced- and free-ensemble means shows only the cooling over the forcing area and Quebec and Labrador (Fig. 3). To test the statistical significance of these differences, we conducted a 1000-sample bootstrap test (see section 2c). The cooling over the forcing area and Quebec and Labrador are the only negative T2m differences in the 975th sample (Fig. 4b) and are therefore the only two cold anomalies significant at the 5% level. As noted in section 2c, the bootstrap test indicates significance at individual grid points, such that each grid point with a negative difference in Fig. 4b is statistically significant at the 5% level. Since no positive differences appear in the 25th sample (Fig. 4a), the forced snow cover generates no statistically significant warm T2m anomalies. The 500th sample from the bootstrap test (Fig. 4c) is essentially the median and strongly resembles the mean (Fig. 3), which suggests that the differences between corresponding ensemble members are not strongly skewed and may be normally distributed.
2) Surface energy fluxes and moist static energy
The elevated albedo of the snow-covered ground causes a decrease of at least 60 W m−2 in net surface shortwave radiation over the forcing area (Fig. 5a). Several areas in which the forced snow cover caused T2m anomalies also show a corresponding change in sensible heat flux (SHF; defined as positive into the atmosphere; Fig. 5b). High-albedo snow cover cooled the surface and induced a negative SHF anomaly, thereby reducing the heat energy available to the lower atmosphere, as noted in Ellis and Leathers (1999). This stands in contrast to the downstream effect over Labrador, where a 2°C decrease in cooling results in a 20 W m−2 increase in SHF. Since all three ensembles were forced by the same SSTs, the positive SHF anomaly is likely the result of anomalously cool air in the early-season simulations overlying water of constant temperature between ensembles.
Ellis and Leathers (1999) also indicated that snow cover could weaken the surface latent heat flux to the atmosphere and dry the air immediately above the snow. Here, the forced snow cover removed both moisture and heat energy from the overlying atmosphere (Fig. 5c). Substantial anomalies in moist static energy extend to 700 hPa, indicating that local effects of the forced snow cover reach to 2–3 km. The weakening of the anomalies with height is consistent with the findings of Kumar and Yang (2003), who found a “bottom-up” vertical structure for the influence of snow cover on atmospheric variability. Some downward motion over the snow-covered region suggests weak local subsidence and an increase in Psl.
Although the forced snowpack drew moisture from the overlying air, total cloud cover increased over the area immediately downwind (Fig. 5d). The cool, dry air over the snowpack acted as a cold front permanently positioned along the snowpack’s outer edge. As the anomalously southerly lower-tropospheric winds (Fig. 5c) advected the air over the snowpack to the north and east, the cool air sank and forced the warmer, moister air off the snowpack to rise, enhancing convection. A 1000-sample bootstrap analysis indicated that the local responses in net shortwave radiation, sensible heat flux, and moist static energy were statistically significant at the 5% level, while the response in cloud cover was significant at the 10% level.
The late-season ensemble means show most of the same direct effects as the early-season experiments during the late-season perturbed period (January and February, hereafter JF). Minor variations between the two forced ensembles include the location and sign of the temperature anomaly near Labrador, which changed from a cooling to a weak warming (Fig. 6). Although the sensitivity of Labrador T2m to the timing of the snow-cover forcing is an interesting result of this experiment, its underlying causes are outside the scope of this study, which is focused on the response of Eurasian winter climate. We note the Labrador temperature anomalies merely as an aside. A bootstrap analysis found the T2m, SHF, moist static energy, and cloud cover differences between the free-snow-cover and late-season ensembles to be significant at the same levels as in the early-season ensemble above.
b. Transatlantic effects
1) Tropospheric circulation
Downstream of the early-season snow-cover forcing, a hemisphere-wide negative Psl anomaly stretched across the high latitudes, with one anomalous trough west of Iceland and a second over the Caspian (Fig. 7a). Several weak positive Psl anomalies appeared in the midlatitudes; the strongest of these was a 2–3-hPa anomalous ridge over western Europe. The 4–6-hPa Psl anomalous dipole between the North Atlantic and southwestern Europe was reminiscent of the positive phase of the NAO. Wallace and Gutzler (1981) determined that the Psl at 30°N, 20°W was anticorrelated with the Psl at roughly 65°N, 20°W (their Fig. 8). Oscillations in the NAO almost certainly forced this dipole. Wallace and Gutzler indicated, however, that a strong positive NAO pattern had a Psl dipole of 20–25 hPa for a single winter season (their Fig. 22b).
Anomalies in lower-tropospheric heights mirrored those in Psl. A negative 850-hPa height (Z850) anomaly covered the higher latitudes south to 45°N latitude above the North Atlantic and the Caspian Sea, while a positive anomaly appeared over western Europe (Fig. 7b). No change in Z850 occurred over the snow-covered region, indicating that the forced snow could not directly influence local circulation much above the atmospheric boundary layer.
Midtropospheric heights can predict surface and boundary layer features. Chan et al. (2003) linked a midtropospheric dipole across the north-central Atlantic in October with stronger, more frequent snowstorms the following winter. Similarly, Ratcliffe (1976) described an objective technique for using mean 500-hPa height anomalies to predict monthly rainfall in the British Isles. Our 500-hPa height (Z500) pattern is similar to those seen at sea level and 850 hPa: a positive Z500 anomaly over southwestern Europe anchored two negative anomalies, one over the North Atlantic and the other northeast of the Caspian Sea (Fig. 7c). Midtropospheric divergence over Iceland likely promoted surface convergence and upward vertical motion under the negative Z500 anomaly, which Hastenrath and Greischar (2001) found indicated a positive NAO. Similarly, midtropospheric convergence over southwestern Europe likely resulted in surface divergence and downward vertical motion under the positive Z500 anomaly. The strong meridional pressure gradient would intensify the surface westerlies, in turn causing warming in western Europe, above-normal rainfall across the United Kingdom and Scandinavia, and dry conditions on the Mediterranean coast. The wave pattern in the Z500 difference field suggests a teleconnection between forced Great Plains snow cover and European winter climate. In the upper troposphere, the 200-hPa height (Z200) dipole (not shown) resembled the Z500 dipole in magnitude, although the Z200 dipole was shifted roughly 600 km to the north and west. The response of the atmosphere was mostly barotropic in nature in both the early- and late-season ensembles, which can be typical of the atmosphere in the NAO–dipole region.
Neither the anomalous Z500 dipole nor any of the waves associated with it can be seen in the first two months of the early-season [November–December (ND)] simulations (Fig. 7f). The dipole appeared in force during JF (Fig. 7i), with a magnitude of up to 120 gpm. The Psl and Zsl anomalies (Figs. 7d and 7g and Figs. 7e and 7h, respectively) showed the same temporal pattern as the Z500 anomalies: virtually no change during ND, followed by an intense dipole between the North Atlantic and southern Europe and the western Atlantic during JF. The JF dipole was weaker and less coherent for the Psl anomalies than in the Z850 anomalies, but still appeared clearly.
To determine the degree to which the anomalous circulation patterns in Fig. 7 projected onto the North Atlantic Oscillation, we computed the difference in ensemble-mean and time-mean anomalies in Z500, Z850, and Psl between Iceland and the Azores, the NAO centers of action (Table 1). In other words, we quantified the magnitude of the anomalous dipoles displayed in Fig. 7 for the early- and late-season ensembles. (Figure 9a shows the model grid boxes used for this calculation.) Table 1 displays the coefficient of variation (CoV) for each anomalous dipole, defined as the standard deviation of the anomalous dipoles from each ensemble member divided by the ensemble mean. As noted above, the dipoles in the early-season ensemble were substantially stronger for JF means than for NDJF. The CoV was also lower in JF, which implies that the individual ensemble members were in closer agreement over the JF pattern. The late-season ensemble displayed similar anomalous JF dipoles to the early-season ensemble, both in their spatial pattern (not shown) and their magnitude. The late-season ensemble had lower CoV values than the early-season ensemble in JF, suggesting that the JF pattern was more consistent among the ensemble members. A closer examination of the circulation patterns from each member revealed that the early-season ensemble contained one member that favored the opposite pattern to the other eight [i.e., higher (lower) heights and pressures over Iceland (the Azores)] which markedly increased the CoV for that ensemble. All nine late-season ensemble members displayed a pattern resembling the positive NAO.
This analysis demonstrates that although the JF dipoles shown in Fig. 7 were not perfectly collocated with the NAO centers of action, the differences likely projected onto the NAO as time-mean, ensemble- mean pressures and heights decreased (increased) over Iceland (the Azores), with only moderate variations among the ensemble members. The dipoles’ magnitudes were substantial for time-mean, ensemble-mean values, given the limited extent of our snow-cover forcing. Furthermore, the fact that the JF anomalous circulation in the late season agreed with the JF pattern from the early-season ensemble (rather than with the negligible ND results) suggests that the JF basic state may be responsible for the sensitivity of the atmosphere to Great Plains snow cover. The implications of this agreement will be discussed further in section 4.
In a 1000-sample bootstrap analysis of JF differences between the early-season and free-snow-cover ensemble means, the full Psl dipole disappeared, but a portion of the Mediterranean positive anomaly in the 25th sample (Fig. 8a) remained, as did a negative anomaly over Iceland in the 975th (Fig. 8b). Thus, both signals are significant at the 5% level. At 850 hPa, neither the Mediterranean positive anomaly nor the North Atlantic negative anomaly were present in both the 25th and 975th sample (Figs. 8c and 8d), but both were visible in the 50th and 950th samples (not shown) and so are significant at the 10% level. The Mediterranean 500-hPa anomaly proved significant at the 5% level (Fig. 8e), while the North Atlantic anomaly was significant only at the 15% level. Similar significance levels were obtained at each vertical level by the bootstrap method for JF differences between the late-season and free-snow-cover ensemble means.
2) European near-surface temperatures
The difference in Z500 between the forced- and free-snow-cover ensembles in JF suggests a teleconnection between Great Plains snow cover and western European winter climate. To verify this, we considered the difference between the forced and free ensembles in European T2m. The early-season ensemble displays a marked warming across central and northern Europe and into Scandinavia (Fig. 9a). A 0.5°–1°C cool T2m anomaly in the forced runs appeared near the Black Sea, while a second extended across the Arctic Ocean north of 70°N. The former is associated with the anomalous Z500 trough across southwestern Russia; the broad, zonal, negative height anomalies over the high latitudes agree with the latter. The differences across central and northern Europe and the Black Sea are consistent with patterns that have been observed during positive NAO phases (e.g., Hurrell 1996). No substantial differences were found between the early-season and the free-snow-cover ensemble in ND (not shown), which agrees with the negligible changes to the tropospheric transatlantic circulation then (Fig. 7).
We hypothesize that the stronger westerlies brought on by the positive NAO advected warm, moist air from the Atlantic over England, northern Europe, and Scandinavia. (This study also found a slight increase in winter precipitation across northern Europe.) A positive NAO would intensify the polar vortex and reduce Arctic surface temperatures. The T2m anomalies over western Eurasia agree with the Z500 pattern and further suggest that Great Plains snow cover may promote a positive NAO.
The early-season European T2m differences in JF did not pass a reasonable bootstrapping test using this small number of simulations. We must rely on the logical reasonableness of the results (i.e., their physical connection to the pattern of Psl and tropospheric-height differences) to think of them as an unproven tendency that could be used to motivate further simulations. In other words, we believe that the European T2m may be a significant signal, based on our statistically significant circulation patterns [section 3b(1)], and that a larger ensemble of simulations would be better able to determine the true nature of the results shown here.
In the late-season simulations, western Eurasian T2m displayed a more intense version of the anomalies seen in the early-season cases. Most of central and northern Europe warmed by 0.5°–2.0°C in the forced runs, slightly more than in the early-season ensemble (Fig. 9b). The cool anomaly over southeastern Europe in the early-season ensemble shifted to the northeast and expanded to encompass northern Scandinavia in the late-season runs because of a northeastward shift in the eastern European Z500 ridge (not shown). Analysis of a 1000-sample bootstrap dataset proved the central and northern European warming significant at the 85% confidence level. As for the transatlantic circulation patterns, the JF patterns from the late-season ensemble strongly resembled those from the early-season integrations. This suggests that the response of the atmosphere depends on the basic state rather than on the duration of the snow-cover forcing.
3) The North Atlantic Oscillation
The North Atlantic Oscillation index is traditionally calculated from the difference in normalized Psl anomalies (taken from climatological values) between observing stations near the NAO’s centers of action, the Azores and Iceland (Rogers 1984; Hurrell 1995b). This point-based measure cannot accurately account for interannual shifts in the centers of action or seasonal variations in the pattern (Barnston and Livezey 1987). To overcome the former difficulty and to fit the CAM2 map grid to the centers of action, we averaged four 2.8° × 2.8° grid boxes to obtain the nominal Iceland and Azores Psl (see boxes in Fig. 9a). Each daily mean Psl from the model was then compared against a monthly mean Psl from a 100-yr climatology, and the anomalous Psl (P′sl) was computed as the model Psl minus the climatological monthly mean. The Iceland P′sl was subtracted from the Azores P′sl to obtain the daily NAO index.
The ensemble-mean NAO indices for the early-season and free-snow-cover ensembles tracked each other through the first 60 days of the experiment (Fig. 10a), coinciding with the small differences seen in Z500 and European T2m for ND (Fig. 7f). Such a small difference in the NAO index indicates only a slight change in the meridional pressure gradient and little difference in the mean flow across the North Atlantic and western Europe. Around 10 January, 70 days after the snow-cover forcing began, the forced-snow-cover runs shifted toward the positive NAO, while the free runs favored the negative phase, with the difference reaching 30–35 hPa at several points (e.g., near 20 January and 15 February), indicating that either the Icelandic trough or the Mediterranean ridge—or more likely both, given Fig. 7g—was substantially stronger in the early-season ensemble.
A one-tailed Student’s t test on the time-mean difference in the NAO index after 10 January between corresponding individual ensemble members determined that the ensemble-mean, time-mean NAO index from the early-season ensemble was greater than that from the free-snow-cover ensemble at the 5% significance level (95% confidence). (The null hypothesis is that the 10 January–28 February mean NAO index from the early-season ensemble is not greater than the mean NAO index over the same period from the free-snow-cover ensemble.) We found no statistically significant difference in the NAO index prior to 10 January, confirming that the forced snow cover indeed affects the NAO only in JF. These results are consistent with the ND and JF pressure, height, and temperature anomalies noted earlier. Forced Great Plains snow cover promoted the positive phase of the NAO in JF, inducing lower (higher) Psl and tropospheric geopotential heights over Iceland (the Azores and southwestern Europe) and above-normal T2m over western Europe and Scandinavia.
The late-season ensemble also favored the positive phase of the NAO. When the snowpack was added on 1 January, the late-season ensemble-mean NAO index tended toward more positive values than the free runs within about 20 days (Fig. 10b). This period (21–31 January) is almost identical to the time that the early-season ensemble-mean NAO index began to display markedly more positive values (10–20 January). A similar Student’s t test to the one used for the early-season ensemble found that the late-season ensemble-mean NAO index was greater than the free-snow-cover ensemble-mean at the 5% significance level. Taken together with the similarities between the early- and late-season ensembles in JF in tropospheric circulation and European T2m, the positive shift in the NAO indicates that transatlantic circulation must be sensitive to the presence of substantial Great Plains snow cover during mid-to-late January and throughout February.
4. Discussion
Our ensembles have demonstrated that anomalous snow cover in the northern Great Plains can impact transatlantic atmospheric circulation. While our limited number of ensemble members has affected our ability to validate the differences in European T2m with a bootstrap test, the coherence among the statistically significant tropospheric circulation patterns and the NAO index indicates that anomalous snow cover in the Great Plains likely promotes the positive phase of the NAO.
Holland (2003) found that CCSM2 properly represented the NAO: a 550-yr control run of CCSM2 showed a statistically significant, positive spatial correlation to 48 yr of National Centers for Environmental Prediction (NCEP)–NCAR reanalysis data in Psl, surface temperature, precipitation, and sea ice extent over the North Atlantic and the Arctic Ocean. The control run included dynamic oceans and sea ice, which provided a more accurate representation of ocean–atmosphere interactions than the monthly SST climatology used in this study. Delworth (1996) compared a 1000-yr integration of the coupled GCM from the Geophysical Fluid Dynamics Laboratory to a 500-yr integration of the AGCM with forced seasonal SSTs, and concluded that the SST variability associated with the NAO was driven largely by surface energy fluxes. The lack of an interactive ocean did not substantially affect the model’s ability to reproduce a realistic NAO at 500 hPa. It is therefore quite likely that the observed flow patterns seen in the present study are related to the NAO, even though this study forces SSTs with a monthly climatology.
While previous studies have generally failed to find a connection between North American snow cover and the NAO (e.g., Walsh and Ross 1988; Gong et al. 2003a; Saito and Cohen 2003), our results suggest that such a connection could exist. The key difference between our study and past attempts is that here we have considered a limited, regional snow-cover forcing area. Most other investigations have employed hemisphere- or continent-wide snow-cover anomalies, an approach which implies that an entire landmass will produce a homogeneous response to anomalous snow. Yet there is no obvious reason why the atmosphere should display the same sensitivity to snow cover over such a broad forcing area. Some regions of a large forcing area may have no effect on the atmosphere, and others may counterbalance one another, producing a negligible or noisy response. Our results demonstrate that smaller areas of high inter- and intra-annual snow-cover variability—such as the northern Great Plains—can produce a response in the wider hemispheric circulation through the North Atlantic Oscillation.
Aside from the identification of the northern Great Plains as a critical region for snow-cover forcing, this study has also demonstrated that the response of the atmosphere can change markedly during the winter season. Differences between the free-snow-cover and early-season ensembles showed that snow cover had a much stronger and more statistically significant effect on the transatlantic flow during January and February. The late-season simulations confirmed that these stronger differences were caused by a changing mean state during the winter season, rather than by a long antecedent period of snow-cover forcing. The atmospheric response to snow-cover forcing beginning on 1 January was so similar to the response to snow-cover forcing beginning on 1 November that the persistence of the snow-cover forcing must have little effect on the response. In other words, it is the mean state of the atmosphere that primarily determines the atmosphere’s sensitivity to the anomalous snowpack, not the length of time for which the snowpack has existed. This suggests that the unrealistic duration of our forced snow cover—particularly in the early-season ensemble, when it persists for four months—is not much of a caveat. Taken separately from November and December, the January and February results from the early- and late-season simulations show that a small, appropriately placed snowpack of realistic spatial extent and depth [see section 2b(2)] can promote the positive phase of the NAO and thereby affect European winter climate.
Our conclusions have relied upon results from a single AGCM and have not been validated against observations. We believe that our results, although not a complete justification for a teleconnection, provide a basis from which studies employing other models, varying snow forcing, or observational data could be undertaken. To effectively use observational data to identify teleconnections between regional-scale snow anomalies and a large-scale circulation pattern such as the NAO requires a priori knowledge of a suitable region to examine; we have argued here that the northern Great Plains should be considered such a region of interest. Furthermore, the use of observations mandates careful attribution of cause and effect, because the NAO responds to numerous forcings and many of these forcings interact with one another and produce opposing responses in the NAO. For example, past studies (e.g., Gong et al. 2002) have suggested that excess Siberian snow produces a negative signal in the NAO, whereas this study has implied a positive response from excess North American snow. Our use of an AGCM allowed us to control for all forcings other than Great Plains snow cover, such that the difference between our ensembles quantifies only the effect of our idealized, regional-scale snow forcing on the wider atmospheric circulation. While we certainly believe that our results should be confirmed in observations and in other GCMs before being accepted as concrete proof of the existence of a teleconnection between Great Plains snow and the NAO, the methods and analysis required to isolate such a signal in observations are outside the stated scope of this study: to use idealized snow forcing in an AGCM to identify a potential region of interest that may show teleconnections to the NAO. Accordingly, validation against observations is left for a future study.
5. Conclusions
We conducted three nine-member ensembles of simulations of the Data Ocean version of the NCAR CCSM2, comprising a control ensemble in which the model was free to determine snow cover and two ensembles with forced snow cover. In the latter ensembles, we forced a 72-cm snowpack in the northern Great Plains of the United States but allowed the model to freely determine snow cover elsewhere. The snow-cover forcing began on 1 November (1 January) in the early- (late-) season ensemble. Ensemble-mean differences between the free- and forced-snow-cover simulations yielded the following results:
(i) The early- and late-season ensemble members demonstrated local cooling over the snowpack of 2°C to 8°C (Fig. 2). In the early-season ensemble, this cooling extended to Quebec and Labrador. A 1000-sample bootstrap analysis (Efron 1981) showed that the cooling was significant at the 5% level in both ensembles.
(ii) The forced snow cover cooled and dried the overlying air, in agreement with Ellis and Leathers (1999) and Leathers and Robinson (1993). Sensible heat flux over the snow cover was more than 30 W m−2 lower and available moist static energy decreased by as much as 12 kJ kg−1 (Fig. 5). Late-season responses were similar to their early-season counterparts, with some spatial displacement. These results together with those in (i) validated the response of CAM2 to a deep snowpack.
(iii) The forced snow cover modified downstream tropospheric circulation, most substantially at 850 and 500 hPa (Figs. 7b and 7c). The snowpack intensified the typical winter wave pattern across the North Atlantic and western Eurasia. This barotropic anomalous circulation suggested that Great Plains snow might favor the positive NAO. In the early-season simulations, the anomalies proved much stronger in JF than in ND at all levels examined (Fig. 7). JF differences in the late-season ensemble were very similar to those in the early-season ensemble. Bootstrap analysis showed that the Z500, Z850, and Psl differences in JF were significant at the 15%, 10%, and 5% levels, respectively (Fig. 8).
(iv) Both ensembles showed JF anomalies in western Eurasian T2m comprising warmer temperatures across western Europe and Scandinavia, a signal which was stronger and more significant in the late-season ensemble (Fig. 9). The early-season features failed to pass a reasonable bootstrapping test, which is likely due to the limited number of runs in our ensembles. The T2m anomalies logically fit with the anomalous tropospheric circulation and the positive shift in the NAO index, which led us to accept the T2m as an unproven tendency that could motivate larger ensembles of forced and free runs.
(v) In JF, the early- and late-season runs consistently exhibited a more positive NAO index than the free-snow-cover ensemble, a result confirmed at the 5% level by a Student’s t test. This confirmed that Great Plains snow cover was capable of modifying the NAO.
(vi) Our study differs from previous efforts in that we have forced a limited area in the northern Great Plains, as opposed to an entire hemisphere or continent. In doing so, we have shown that a regional snow-cover anomaly can have a larger impact on the transatlantic circulation than a much broader anomaly. The region we modified was an area of high snow-cover variability and an area in which snow-cover anomalies had been shown to affect local cyclogenesis patterns (Elguindi et al. 2005).
(vii) The similarities between the early- and late-season results in JF indicated that the stronger JF results seen in the early-season ensemble (compared to ND) were due to changes in the mean state in January and February. These similarities also suggested that the duration of the snow-cover anomaly was secondary to the mean state of the atmosphere in determining the sensitivity of the atmosphere to our snow-cover forcing.
This study has shown that the possibility exists for a stronger teleconnection between snow cover in the Great Plains and European winter climate than has been previously suggested, particularly for a regional snow-cover anomaly. Some questions remain as to the mechanism behind this teleconnection and why the snow cover has a much stronger effect on European climate in JF than in ND. Also, it is unclear what effect the spatial location, depth, and temporal duration of the regional snow-cover anomaly has on European winter climate. Future studies should adjust these parameters and compare the results of an ensemble of forced runs to the forced-snow-cover ensembles presented here.
Acknowledgments
NPK was supported by a grant from the Eugene duPont Distinguished Memorial Fund at the University of Delaware. NPK would like to thank Dr. Hilary Weller for several useful discussions on this work. The authors are grateful to three anonymous reviewers for their helpful comments on a previous version of this manuscript.
REFERENCES
Ambaum, M., B. Hoskins, and D. Stephenson, 2001: Arctic Oscillation or North Atlantic Oscillation? J. Climate, 14:3495–3507.
Bader, J. and M. Latif, 2003: The impact of decadal-scale Indian Ocean sea surface temperature anomalies on Sahelian rainfall and the North Atlantic Oscillation. Geophys. Res. Lett., 30.2169, doi:10.1029/2003GL018426.
Bader, J. and M. Latif, 2005: North Atlantic oscillation response to anomalous Indian Ocean SST in a coupled GCM. J. Climate, 18:5382–5389.
Baker, D. G., R. H. Skaggs, and D. L. Ruschy, 1991: Snow depth required to mask the underlying surface. J. Appl. Meteor., 30:387–392.
Baldwin, M. P., 2001: Annular modes in global daily surface pressure. Geophys. Res. Lett., 28:4115–4118.
Baldwin, M. P. and T. J. Dunkerton, 1999: Propagation of the Arctic Oscillation from the stratosphere to the troposphere. J. Geophys. Res., 104:30937–30946.
Barnston, A. G. and R. E. Livezey, 1987: Classification, seasonality, and persistence of low-frequency atmospheric wave patterns. Mon. Wea. Rev., 115:1083–1126.
Bhatt, U. S., M. A. Alexander, D. S. Battisti, D. D. Houghton, and L. M. Keller, 1998: Atmosphere–ocean interaction in the North Atlantic: Near-surface climate variability. J. Climate, 11:1615–1632.
Chan, A. C., S. J. Colucci, and D. T. DeGaetano, 2003: Predicting East Coast winter storm frequencies from midtropospheric geopotential height patterns. Wea. Forecasting, 18:1177–1191.
Cohen, J., 1994: Snow cover and climate. Weather, 49:150–156.
Cohen, J. and D. Entekhabi, 1999: Eurasian snow cover variability and Northern Hemisphere climate predictability. Geophys. Res. Lett., 26:345–348.
Cohen, J. and D. Entekhabi, 2001: The influence of snow cover on Northern Hemisphere climate variability. Atmos.–Ocean, 39:35–53.
Cullen, H. M. and P. B. de Menocal, 2000: North Atlantic influence on Tigris–Euphrates streamflow. Int. J. Climatol., 20:853–863.
Delworth, T. L., 1996: North Atlantic interannual variability in a coupled ocean–atmosphere model. J. Climate, 9:2356–2375.
Dey, B. and A. Kumar, 1983: Himalayan winter snow cover area and summer monsoon rainfall in India. J. Geophys. Res., 88:5471–5474.
Dickinson, R. E., A. Henderson-Sellers, and P. J. Kennedy, 1993: Biosphere–Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model. NCAR Tech. Note NCAR/TN-387+STR, 72 pp.
Dickson, R. R., 1984: Eurasian snow cover versus Indian monsoon rainfall—An extension of the Hahn-Shukla results. J. Climate Appl. Meteor., 23:171–173.
Efron, B., 1981: Nonparametric estimates of standard error: The jackknife, the bootstrap, and other methods. Biometrika, 68:589–599.
Elguindi, N., B. Hanson, and D. J. Leathers, 2005: The effects of snow cover on midlatitude cyclones in the Great Plains. J. Hydrometeor., 6:263–279.
Ellis, A. and D. J. Leathers, 1999: Analysis of cold airmass temperature modification across the U.S. Great Plains as a consequence of snow depth and albedo. J. Appl. Meteor., 38:696–711.
Esbensen, S., 1984: A comparison of intermonthly and interannual teleconnections in the 700-mb geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 112:2016–2032.
Frei, A. and D. A. Robinson, 1999: Northern Hemisphere snow extent: Regional variability 1972–1994. Int. J. Climatol., 19:1535–1560.
Giannini, A., Y. Kushnir, and M. A. Cane, 2000: Interannual variability of Carribbean rainfall, ENSO, and the Atlantic Ocean. J. Climate, 13:297–311.
Giannini, A., M. A. Cane, and Y. Kushnir, 2001: Interdecadal changes in the ENSO teleconnection to the Carribbean region and the North Atlantic Oscillation. J. Climate, 14:2867–2879.
Gong, G., D. Entekhabi, and J. Cohen, 2002: A large-ensemble model study of the wintertime AO–NAO and the role of interannual snow perturbations. J. Climate, 15:3488–3499.
Gong, G., D. Entekhabi, and J. Cohen, 2003a: Modeled Northern Hemisphere winter climate response to realistic Siberian snow anomalies. J. Climate, 16:3917–3931.
Gong, G., D. Entekhabi, and J. Cohen, 2003b: Relative impacts of Siberian and North American snow anomalies on the winter Arctic Oscillation. Geophys. Res. Lett., 30.1848, doi:10.1029/2003GL017749.
Hahn, D. G. and J. Shukla, 1976: An apparent relationship between Eurasian snow cover and Indian monsoon rainfall. J. Atmos. Sci., 33:2461–2462.
Hartley, S., 1999: Winter Atlantic climate and snowfall in the south and central Appalachians. Phys. Geogr., 20:1–13.
Hartley, S. and M. J. Keables, 1998: Synoptic associations of winter climate and snowfall variability in New England, USA, 1950–1992. Int. J. Climatol., 18:281–298.
Hastenrath, S. and L. Greischar, 2001: The North Atlantic Oscillation in the NCEP–NCAR reanalysis. J. Climate, 14:2404–2413.
Hoerling, M. P., J. W. Hurrell, and T. Xu, 2001: Tropical origins for recent North Atlantic climate change. Science, 292:90–92.
Holland, M., 2003: The North Atlantic Oscillation–Arctic Oscillation in the CCSM2 and its influence on Arctic climate variability. J. Climate, 16:2767–2781.
Hurrell, J. W., 1995a: Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science, 269:676–679.
Hurrell, J. W., 1995b: Transient eddy forcing of the rotational flow during northern winter. J. Atmos. Sci., 52:2286–2301.
Hurrell, J. W., 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett., 23:665–668.
Hurrell, J. W., 2004: Twentieth century North Atlantic climate change. Part I: Assessing determinism. Climate Dyn., 23:371–389.
Hurrell, J. W. and H. van Loon, 1997: Decadal variations in climate associated with the North Atlantic oscillation. Climatic Change, 36:301–326.
Itoh, H. and K. Harada, 2004: Coupling between tropospheric and stratospheric leading modes. J. Climate, 17:320–336.
Karl, T. R., 1986: The relationship of soil moisture parameterizations to subsequent seasonal and monthly mean temperature in the United States. Mon. Wea. Rev., 114:675–686.
Kiehl, J. T. and P. R. Gent, 2004: The Community Climate System Model, version 2. J. Climate, 17:3666–3682.
Kumar, A. and F. Yang, 2003: Comparative influence of snow and SST variability on extratropical climate in northern winter. J. Climate, 16:2248–2261.
Kushnir, Y., 1994: Interdecadal variations in North Atlantic sea surface temperature and associated atmospheric conditions. J. Climate, 7:141–157.
Kushnir, Y. and J. M. Wallace, 1989: Low-frequency variability in the Northern Hemisphere winter: Geographic distribution, structure, and time-scale dependence. J. Atmos. Sci., 46:3122–3142.
Kutiel, H. and P. A. Kay, 1992: Recent variations in 700-hPa geopotential heights in summer over Europe and the Middle East, and their influence on other meteorological factors. Theor. Appl. Climatol., 46:99–108.
Leathers, D. and D. Robinson, 1993: The association between extremes in North American snow cover extent and United States temperatures. J. Climate, 6:1345–1355.
Marshall, J., Coauthors 2001: North Atlantic climate variability: Phenomena, impacts, and mechanisms. Int. J. Climatol., 21:1863–1898.
Mehta, V. M., M. J. Suarez, J. Manganello, and T. L. Delworth, 2000: Predictability of multiyear to decadal variations in the North Atlantic Oscillation and associated Northern Hemisphere climate variations. Geophys. Res. Lett., 27:121–124.
Mysak, L. and S. A. Venegas, 1998: Decadal climate oscillations in the Arctic: A new feedback loop for atmosphere–ice–ocean interactions. Geophys. Res. Lett., 25:3607–3610.
Namias, J., 1962: Influence of abnormal heat sources and sinks on atmospheric behavior. Proc. Int. Symp. on Numerical Weather Prediction, Tokyo, Japan, Meteorological Society of Japan, 615–627.
Namias, J., 1978: Multiple causes of the North American abnormal winter 1976/77. Mon. Wea. Rev., 106:279–295.
Namias, J., 1985: Some empirical evidence for the influence of snow cover on temperature and precipitation. Mon. Wea. Rev., 113:1542–1553.
Oleson, K., G. Bonan, C. Schaaf, F. Gao, Y. Jin, and A. Strahler, 2003: Assessment of global climate model land surface albedo using MODIS data. Geophys. Res. Lett., 30.1443, doi:10.1029/2002GL016749.
Oleson, K., G. Bonan, S. Levis, and M. Vertenstein, 2004: Effects of land use change on North American climate: Impact of surface datasets and model biogeophysics. Climate Dyn., 23:117–132.
Peng, S. and L. A. Mysak, 1993: A teleconnection study of interannual sea surface temperature fluctuations in the northern North Atlantic and precipitation and runoff over western Siberia. J. Climate, 6:876–885.
Ratcliffe, R. A. S., 1976: The use of 500-mb anomalies in long-range forecasting. Quart. J. Roy. Meteor. Soc., 102:265–266.
Robertson, A. W., 2001: Influence of ocean–atmosphere interaction on the Arctic Oscillation in two general circulation models. J. Climate, 14:3240–3254.
Robinson, D. A., 1996: Evaluating snow cover over Northern Hemisphere lands using satellite and in situ observations. Proc. 53rd Eastern Snow Conf., Williamsburg, VA, 13–19.
Robinson, D. A. and A. Frei, 2000: Seasonal variability of Northern Hemisphere snow extent using visible satellite data. Prof. Geogr., 52:307–315.
Rogers, J. C., 1984: The association between the North Atlantic Oscillation and the Southern Oscillation in the Northern Hemisphere. Mon. Wea. Rev., 112:1999–2015.
Rogers, J. C., 1990: Patterns of low-frequency monthly sea level pressure variability (1899–1986) and associated wave cyclone frequencies. J. Climate, 3:1364–1379.
Rogers, J. C. and H. van Loon, 1979: The seesaw in winter temperatures between Greenland and northern Europe. Part II: Some oceanic and atmospheric effects in middle and high latitudes. Mon. Wea. Rev., 107:509–519.
Ropelewski, C. F. and M. S. Halpert, 1986: North American precipitation and temperature patterns associated with the El Niño–Southern Oscillation (ENSO). Mon. Wea. Rev., 114:2352–2362.
Saito, K. and J. Cohen, 2003: The potential role of snow cover in forcing interannual variability of the major Northern Hemisphere mode. Geophys. Res. Lett., 30.1302, doi:10.1029/2002GL016341.
Shindell, D. T., R. L. Miller, G. A. Schmidt, and L. Pandolfo, 1999: Simulation of recent northern winter climate trends by greenhouse-gas forcing. Nature, 399:452–455.
Shukla, J., 1984: Predictability of time averages. Part II: The influence of boundary forcings. Problems and Prospects in Long and Medium Range Weather Forecasting, D. Burridge and E. Kallen, Eds., Springer-Verlag, 155–206.
Thompson, D. W. J. and J. M. Wallace, 1998: The Arctic oscillation signature in the wintertime geopotential height and temperature fields. Geophys. Res. Lett., 25:1297–1300.
Ulbrich, U. and M. Christoph, 1999: A shift of the NAO and increasing storm track activity over Europe due to anthropogenic greenhouse gas forcing. Climate Dyn., 15:551–559.
Van Loon, H. and J. C. Rogers, 1978: The seesaw in winter temperatures between Greenland and northern Europe. Part I: General description. Mon. Wea. Rev., 106:296–310.
Vernekar, A. D., J. Zhou, and J. Shukla, 1995: The effect of Eurasian snow cover on the Indian monsoon. J. Climate, 8:248–266.
Walker, G. T. and E. W. Bliss, 1932: World Weather V. Mem. Roy. Meteor. Soc., 4:53–84.
Wallace, J. M., 1996: Observed climatic variability; spatial structure. Decadal Climate Variability, D. Anderson and J. Willebrand, Eds., Springer, 31–81.
Wallace, J. M. and D. S. Gutzler, 1981: Teleconnections in the geopotential height field during the Northern Hemisphere winter. Mon. Wea. Rev., 109:784–812.
Walland, D. J. and I. Simmonds, 1996: Subgrid-scale topography and the simulation of the Northern Hemisphere snow cover. Int. J. Climatol., 16:961–982.
Walsh, J. and B. Ross, 1988: Sensitivity of 30-day dynamical forecasts to continental snow cover. J. Climate, 1:739–754.
Wang, W. and A. Kumar, 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103:28637–28646.
Watanabe, M. and T. Nitta, 1998: Relative impacts of snow and sea surface temperature anomalies on an extreme phase in the winter atmospheric circulation. J. Climate, 11:2837–2857.
Wiscombe, W. J. and S. G. Warren, 1980: A model for the spectral albedo of snow. I: Pure snow. J. Atmos. Sci., 37:2712–2733.
Wolfson, N., R. Atlas, and Y. C. Sud, 1987: Numerical experiments related to the summer 1980 U.S. heat wave. Mon. Wea. Rev., 115:1345–1357.
Yasunari, T., A. Kitoh, and T. Tokioka, 1991: Local and remote responses to excessive snow cover mass over Eurasia appearing in the northern spring and summer climate—A study with the MRI GCM. J. Meteor. Soc. Japan, 69:473–487.
Zhao, M., A. J. Pitman, and T. Chase, 2001: The impact of land cover change on the atmospheric circulation. Climate Dyn., 17:467–477.
The gray box represents the forcing region in the forced-snow-cover ensembles, in which 72 cm of snow cover were added on either 1 Nov or 1 Jan, then forced to persist through 1 Mar. The dashed lines inside the gray box indicate the CAM2 grid boxes.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The difference in T2m (°C) between corresponding early-season and free-snow-cover ensemble members (calculated as forced snow cover minus free snow cover, as in all other “difference” figures in this study), with means taken over NDJF. The contour interval is 1°C.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The NDJF mean difference in T2m (°C) between early- season and free-snow-cover ensemble means. This is the arithmetic mean of the nine panels in Fig. 2. The contour interval is 1°C.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The results of the 1000-sample bootstrap of T2m differences between the early-season and free-snow-cover ensemble means, using means over NDJF, for (a) the 25th most negative sample, (b) the 975th most negative sample, and (c) the 500th sample (essentially the median). The contour interval for (a)–(c) is 1°C.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The NDJF mean difference between the early-season and free-snow-cover ensemble means in (a) net shortwave radiation (W m−2), (b) sensible heat flux (W m−2; positive into the atmosphere), (c) moist static energy (contours; kJ kg−1) and meridional winds (vectors; m s−1) taken in a longitudinal cross section at 100°W, and (d) total cloud fraction. The contour intervals are (a) 5 W m−2, (b) 10 W m−2, (c) 1 kJ kg−1, and (d) 0.03 or 3% of the sky covered.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
As in Fig. 3, but for the JF mean difference in T2m (°C) between the late-season and free-snow-cover ensemble means.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The ensemble mean differences between the early-season and free-snow-cover ensemble means over (a)–(c) NDJF, (d)–(f) ND, and (g)–(i) JF for (a),(d),(g) Psl, (b),(e),(h) Z850, and (c),(f),(i) Z500. The contour interval is 10 gpm for Z500 and Z850 and 1 hPa for Psl; negative contours are dashed.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The 95% confidence interval from the 1000-sample bootstrap of differences between early-season and free-snow-cover ensemble means in (a)–(b) Psl, (c)–(d) Z850, and (e)–(f) Z500, using means over JF. In each row, (a),(c),(e) the 25th most negative sample and (b),(d),(f) the 975th most negative sample are displayed. The contour interval is 10 gpm for Z500 and Z850 and 1 hPa for Psl; negative contours are dashed.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
Differences in T2m during JF between forced- and free-snow-cover ensemble means for (a) the early-season ensemble and (b) the late-season ensemble. The contour interval is 0.5°C. The boxes in (a) labeled “Iceland” and “Azores” give the locations of the four model grid boxes averaged to quantify dipoles in transatlantic circulation [section 3b(1)] and to calculate the NAO index [section 3b(3)].
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The ensemble mean NAO index (mb) for (a) the early-season (solid) and free-snow-cover (dashed) ensembles during NDJF and (b) the late-season (solid) and free-snow-cover (dashed) ensembles during JF. On the horizontal axis, tick marks are spaced every (a) 5 and (b) 2 days.
Citation: Journal of Climate 21, 11; 10.1175/2007JCLI1672.1
The ensemble mean and CoV for the anomalous (forced snow cover minus free snow cover) dipoles in Z500, Z850, and Psl between Iceland and the Azores for the ensembles specified. The values given are for the mean magnitude of the anomalous dipole, with the mean taken over the months indicated. Units for the ensemble-mean dipole are geopotential meters (gpm) for Z500 and Z850 and hectopascals (hPa) for Psl; the coefficient of variation is dimensionless.
