Snow-cover extent and the process of snow melt are recognized as major components of atmospheric dynamics. Each also plays a dynamical role in the disposition of climate (Clark et al., 1999; Bamzai and Marx, 2000; Chen and Wu, 2000).
The areal extent of snow cover is thought to influence the atmospheric circulation throughout much of the troposphere by affecting the surface albedo and diabatic heating (Robinson, 1987, 1990; Leathers et al., 1995, 2002; Ellis and Leathers, 1998, 1999; Robinson et al., 1994; Serreze et al., 1998; Andrew and Hawkins 2001; Ellis and Hawkins 2001; Zhang et al., 2003). Zhang et al. (2003) showed that the accumulation of snow increases the albedo, lowers thermal conductivity, and absorbs latent heat energy. These impacts cool the continental and adjacent oceanic climates, which in turn modifies the atmospheric circulation.
The snow-melting process can also modify the atmospheric circulation, as the increase in water reduces the albedo through increased absorption and decreased reflection of solar energy. The snow-melt process can affect the circulation for days to months (Andrew and Hawkins, 2001; García-Herrera and Barriopedro, 2006), and the melting of an extensive snow cover can sometimes retard the shift from one season to the next (Groisman et al., 1994).
Groisman et al. (1994) suggested that impacts from snow melt are greatest in the spring, corresponding with the period of most rapid melt. Yasunari et al. (1991) showed that the albedo effect is dominant in spring at low latitudes, whereas the effect of snow melting is more important in the middle latitudes.
Other studies (Barnett et al., 1988, 1989; Yang 1993; Sobolowski and Frei 2006) have shown that snow cover can be associated with modulations in the atmospheric response to ENSO. Yang et al. (2001) demonstrated that North American surface climate anomalies related to ENSO are greatly enhanced by a local snow–albedo feedback, wherein the altered surface energy balance provides a feedback mechanism that can influence the direct response of surface temperature to ENSO.
Regionally, fluctuations in Eurasian snow cover extent have been related to anomalous conditions in the tropics during the following spring and summer. Pang et al. (2005) described the influence of Eurasian snow cover on the Indian Ocean dipole, while Bamzai and Marx (2000) showed model simulations establishing a relationship between spring-time snow cover and the upcoming Indian monsoon. Xie et al. (2005) discussed the relationship between snow-cover extent over the Qinghai-Xizang (Tibetan) Plateau and typhoon activity in western North Pacific.
In North America, the lateral snow extent can be associated with climate variability that extends well beyond the cool season. Hawkins et al. (2002) showed that snow cover is inversely related to precipitation totals and precipitation frequency associated with the North American monsoon. Ellis and Hawkins (2001) also discussed a teleconnection between snow cover and the atmospheric circulation over the tropical and subtropical Americas and the adjacent oceans during the North American monsoon.
In this study, we explore the relationship between wintertime snow cover extent over North America and Atlantic hurricane activity during the following hurricane season (June–November). To avoid extreme ENSO influence on hurricanes, we are especially interested in documenting the interannual aspects of this relationship in the absence of ENSO.
In Section 2, the inverse relationship between Northern Hemisphere (NH) wintertime snow cover extent and Atlantic hurricane activity is established, even after ENSO and trend have been removed from the data. In Section 3, composite analyses of the large-scale atmospheric anomalies are developed and examined based on the lightest snow-cover years (LSY) and heaviest snow-cover years (HSY). These analyses provide a more complete physical picture of the snow-cover/hurricane relationship than has appeared previously in the literature. In Section 4, the results of the study are discussed and important considerations are noted for future research into the snow cover/hurricane relationship.
2. Relationship between NH winter snow cover and subsequent Atlantic hurricane activity
Indices of monthly Northern Hemisphere snow cover extent (SCE) were derived from two sources. One source is from the National Centers for Environmental Prediction (NCEP) Climate Prediction Center (CPC) ranged from 1973 to 2009 (Armstrong and Brodzik, 2002; available at: ftp://ftp.cpc.ncep.noaa.gov/wd52dg/snow/snw_cvr_area/NH_AREA).
This was based on snow cover reconstruction and NOAA satellite data. The reconstruction method used in situ snow depth and daily climate data from the United States, Canada and China to generate a monthly snow-cover index that was calibrated with NOAA satellite-derived estimates of SCE over the 1972–1992 period. The satellite data (1973–2009) is believed to have better quality, so the reconstruction data during the overlapped period is omitted. Adjustment is made to the reconstruction data by adding a constant, which is obtained by multiplying monthly average of satellite snow-cover index and divided by the monthly average of the reconstructed snow-cover index.
For the period 1950–2009, a statistically significant inverse relationship is evident at the 95% confidence level between Northern America snow cover extent in January (Figure 1(a), dotted blue line), and both the upcoming North Atlantic hurricane count (solid black line) and hurricane season strength as measured by the accumulated cyclone energy (ACE, Bell et al., 2000) index (dashed red line). The correlations for these relationships are − 0.29 and − 0.24, respectively, with corresponding P-values of 0.0169 and 0.0467. (Here, the North Atlantic hurricane count refers to those hurricanes originating over the North Atlantic Ocean only, and excludes hurricanes originating over the Gulf of Mexico and Caribbean Sea. The rationale is that North Atlantic snow extent influence on hurricanes by changes in large scale atmosphere circulation pattern in North Atlantic Ocean and even the North Hemisphere. It is not sensitive to those hurricanes largely dominated by local SST which may not be the part of large-scale circulation. More details will be discussed in Section 3).
The above snow-hurricane relationships are manifested partly in the long-term trend (Figure 1(b)), which reflects the combination of decreased snow cover and increased Atlantic hurricane activity (Bell et al., 1999, 2000, 2004, 2006, 2007; Goldenberg et al., 2001; Bell and Chelliah, 2006) in more recent years. After removing the trend, the resulting interannual time series' are still significant correlated at above 95% confidence level (Figure 1(c)).
ENSO is also known to strongly impact seasonal Atlantic hurricane activity (Gray, 1984; Bell and Chelliah 2006), and ENSO extremes can mask the snow cover- hurricane relationship. For example, the snow cover and Atlantic hurricane counts were both below average during 1969–1976, when either La Niña or El Niño was observed during the peak months (August–October) of each hurricane season.
To better isolate the show cover/hurricane signal, the trend and all ENSO events during ASO (as defined by NOAA's Climate Prediction Center) were removed. The most extreme (value ⩽− 0.25 SD) LSY and the most extreme (value ≥ + 0.25 SD) heavy snow-cover years (HSY) were then identified for further analysis. Nine years comprise the LSY: 1952, 1953, 1980, 1981, 1989, 1990, 1992, 2000 and 2003). Seven years comprise the HSY: 1959, 1960, 1968, 1978, 1979, 1984 and 1993). For both cases, approximately the same percentage of seasons occurred during the high-activity eras (1950–1970 and 1995–2008) and low-activity era (1971–1994).
Using NOAA's hurricane season classifications,† markedly different distributions of hurricane season strength are seen during the LSY and HSY (Figure 2(a)). For the LSY, approximately half of the hurricane seasons were above-normal hurricane seasons and none were below-normal. Conversely, for the HSY approximately half of the hurricane seasons were below-normal and none were above-normal.
The LSY and HSY also exhibit statistically significant differences exceeding the 95% confidence level (calculated using both the Wilcoxen Score and t test) in the distributions of total seasonal hurricane count (Figure 2(b)), total seasonal major hurricane count (Figure 2(c)), and the seasonal ACE index (Figure 2(d)). On average, the LSY show a 42% increase in the seasonal number of hurricanes compared to the HSY, and more than a 70% increase in both major hurricanes and ACE (Figure 2(e)). The average number of days in which a named storm was present (termed named storm days) increased by 56% during the LSY compared to the HSY, while the average numbers of days in which a hurricane and major hurricane were present (called hurricane days and major hurricane days) were higher during the LSY by 59 and 90%, respectively.
3. Atmospheric anomalies and large-scale evolution
We are interested in documenting the atmospheric anomalies and large-scale evolution that relates extremes in NH wintertime snow cover to the strength of the upcoming Atlantic hurricane season. A composite analysis shows that both the LSY (Figure 3(a)) and HSY (Figure 3(b)) are associated with hemispheric-scale patterns of 500 hPa height anomalies during January, which also persist into the peak of the Atlantic hurricane season (Figures 3(c, d)). The LSY feature below average heights in the polar region and above average heights in the middle latitudes, and the HSY feature an opposite anomaly pattern. These patterns reflect opposing phases of the Arctic Oscillation (AO), with the LSY (HSY) featuring a positive (negative) AO. These results are consistent with LeDrew et al. (1997), who demonstrated that a meridional circulation pattern (which is typical of the negative AO) is linked to periods of snow accumulation while a zonal circulation pattern (which is typical of the positive AO) corresponds to persistent snow ablation.
At 1000 hPa, the LSY composite feature warmer than average temperatures over much of Eurasia and North America (Figure 4(a)), while the HSY composite shows below average temperatures in these regions (Figure 4(b)). These anomaly patterns are consistent with the opposing phases of the AO and with the results of Leathers and Robinson (1993), and are also seen to persist into the peak of the Atlantic hurricane season (Figures 4(c, d)).
The 1000 hPa wind anomalies associated with the LSY and HSY also differ markedly, especially over Eurasia, North America, and the North Atlantic Ocean. For example, the HSY feature an anomalous anticyclonic circulation over North America during January. This circulation is associated with enhanced northerly flow extending from Canada into the central United States, along with anomalous easterly winds and a reduction in the normal onshore flow of marine air into western North America (Figure 4(b)). Both of these factors are consistent with the positive phase of the AO, increased NH snow cover, and below average surface temperatures.
The composite North Atlantic circulation anomalies during the peak of the Atlantic hurricane season are now examined. For the LSY, the 200-hPa streamfunction field during August–September shows positive anomalies across the subtropical North Atlantic and negative anomalies across the subtropical South Atlantic (Figure 5(a)). This pattern reflects enhanced ridges in the subtropics of both hemispheres, and is associated with anomalous upper-level easterly winds across the Main Development Region (MDR, which spans the tropical Atlantic and Caribbean Sea between 9.5°N and 21.5°N, Goldenberg et al., 2001). These conditions are also associated with a weaker-than-average Tropical Upper Troposphere Trough (TUTT) over the central extratropical North Atlantic, which importantly does not extend into the MDR.
At 1000 hPa, the LSY August–September (Figure 5(b)) feature an anomalous low-level anticyclonic circulation over the extratropical North Atlantic that persisted from January. An anomalous cyclonic circulation is also evident during August–September within the MDR, along with weaker easterly trade winds (westerly anomalies) and below average heights. At the same time, Atlantic SSTs are above average in both the tropics and at higher latitudes.
One of the most important parameters associated with seasonal fluctuations in Atlantic hurricane activity is the vertical wind shear during August–September between 850 and 200 hPa (DeMaria, 1996). For the LSY, the combination of anomalous upper-level easterlies and low-level westerlies results in decreased vertical wind shear across the MDR (Figure 5(c, d)). The LSY also feature anomalous ascending motion at 500 hPa along 10°N in the eastern Pacific and throughout the central and eastern MDR, indicating a stronger ITCZ that is shifted more north of normal (Figure 5(e)). These circulation and SST anomalies are typical of more active Atlantic hurricane seasons (Bell and Chelliah, 2006).
Conversely, the HSY composites show a weaker-than-average subtropical ridge at 200 hPa over the western MDR and eastern North Pacific (Figure 6(a)), and a well-defined TUTT extending into the heart of the MDR. These conditions are associated with upper-level westerly wind anomalies over the tropical North Atlantic. Also within the MDR, the HSY composites show near-average winds and heights at 1000 hPa (Figure 6(b)), below average SSTs (Figure 6(c)), and near-average vertical wind shear (Figure 6(d)). They also show anomalous descending motion along 10°N in the eastern Pacific and throughout the central and eastern MDR, indicating a weaker ITCZ that is more confined to the deep tropics (Figure 6(e)). These circulation and SST anomalies are typically associated with reduced Atlantic hurricane activity (Bell and Chelliah, 2006).
This study has focused on the nature of the inverse relationship between NH wintertime snow cover extent and the strength of the upcoming Atlantic hurricane season. Consistent with previous studies (Moore et al., 2009), this relationship is seen in the long-term trend and on interannual time scales. By removing the trend and discarding hurricane seasons in which ENSO was present, the present study provides a more complete picture of the snow-cover/hurricane relationship than has appeared previously in the literature.
The results of this study indicate that the LSY feature more Atlantic hurricanes, major hurricanes and seasonal ACE, while the HSY feature fewer hurricanes, major hurricanes and seasonal ACE. Accordingly, all of the LSY are associated with either a normal or above-normal Atlantic hurricane season, whereas all of the HSY are associated with either a normal or below-normal Atlantic hurricane season.
Composite analyses for the LSY and HSY show markedly different large-scale circulation anomalies during the winter, and also during the peak months of the subsequent Atlantic hurricane season. In both periods, the composite 500 hPa circulation for the LSY shows positive phase of the AO, which is a more zonal pattern with above average heights in the middle latitudes and below average heights in the polar region. Conversely, the HSY feature a more meridional circulation pattern captured by the negative phase of the AO. As such, the LSY feature above average surface temperatures over much of North America and Eurasia, while the HSY show cooler continental temperatures. At 1000 hPa, the LSY composites show an anomalous low-level anticyclonic circulation over the central North Atlantic during both winter and August–September, compared to an anomalous cyclonic circulation for the HSY.
The local factors influencing Atlantic tropical cyclogenesis in the MDR have been described in many studies (Gray, 1984; DeMaria, 1996; Goldenberg and Shapiro, 1996; Landsea et al., 1998; Bell and Chelliah, 1999, 2006; Bell et al., 1999, 2000, 2004). During August–September, the LSY and HSY are associated with notably different circulation anomalies across the main hurricane development region (MDR). For the LSY the anomalies are consistent with increased Atlantic hurricane activity, and include (1) an amplified subtropical ridge, a weaker TUTT and anomalous easterly winds at 200 hPa, (2) weaker low-level easterly trade winds, along with an anomalous cyclonic circulation and below average 1000 hPa heights, (3) warmer SSTs and anomalously weak vertical wind shear and (4) a strengthening and northward shift of the ITCZ across the eastern tropical Pacific and Atlantic Oceans.
Conversely, the composite conditions in the MDR for the HSY indicate (1) a near average strength of the TUTT and low-level easterly trade winds, (2) colder SSTs and near-average vertical wind shear and (3) an overall weaker ITCZ that is more confined to the deep tropics. None of these conditions are conducive to increased Atlantic hurricane activity.
This analysis raises two main questions: Why are the wintertime circulation anomalies associated with anomalous continental snow cover so persistent from winter to early Fall? and How does the anomalous extratropical circulation penetrate into the Tropics so as to affect the subsequent Atlantic hurricane season?
This study suggests that the snow-cover/hurricane relationships would not be present on interannual time scales were it not for the persistent AO conditions. However, the extent is unclear to which the persistent AO patterns are linked to internal atmospheric processes independent of the anomalous snow cover, and to what extent they are linked to the anomalous snow cover extent as suggested by Walland and Simmonds (1996) and Gong et al. (2003). Walland and Simmonds (1996) noted that the anomalous snow boundary condition can lead to significant changes in the meridional temperature gradients over the North Atlantic Ocean, thereby influencing the the baroclinic zones and hence the upper-level circulation anomalies associated with the AO. Gong et al. (2003) proposed an atmospheric teleconnection pathway to explain how the Siberian and North American snow anomaly influences the AO, stating that the combination of a large snow-forced local diabatic heating anomaly over a region of substantial stationary wave activity is required to produce strong upward wave activity flux anomalies which initiate the teleconnection pathway.
Also, several studies have shown that anomalous Northern Hemisphere snow cover can affect conditions in the tropics several months later (Bamzai and Marx, 2000; Chen and Wu, 2000; Ellis and Hawkins, 2001; Hawkins et al., 2002; Zhang et al., 2003; Pang et al., 2005). However, it remains unclear in this case how the persistent extratropical signal associated with the LSY and HSY penetrates into the MDR so as to ultimately affect the Atlantic hurricane activity by modulating the tropical easterly trade winds, SSTs, vertical wind shear, and Atlantic ITCZ. Detailed modelling studies are required to address this issue.
The authors acknowledge support for this study from the U.S. National Oceanic and Atmospheric Administration (NOAA) Grant #NAO3NES440015 through a cooperative Agreement (Climate & Weather Impacts on Society and the Environment) via the National Climatic Data Center, and the Charleston Coastal Services Center and on the CWISE extension, the NOAA INFORM project. The authors also acknowledge support from NOAA's Climate Prediction Center through the National Research Council of the United States.