1. Introduction
Extreme cold air outbreaks (CAOs) have a large impact on heavily populated areas of the United States in terms of energy consumption, agricultural losses, property damage, and loss of life. The agricultural losses are particularly devastating to the Florida citrus crop (Miller and Downton 1993; Downton and Miller 1993; Rogers and Rohli 1991). According to a documentation of severe citrus freezes in Florida since 1880 by Rogers and Rohli (1991), the average frequency of these events in Florida is typically about one per decade except for the 1890s (three events) and the period from 1977–89 (six events). On the larger scale, the societal impacts of one of the most extreme CAOs (December 1983) were documented by Quiroz (1984). He cited the loss of several hundred lives nationwide, tens of thousands of cattle in Midwestern feedlots, and hundreds of thousands of fish in the waterways of the western Gulf of Mexico. Shreveport, Louisiana, had its worst ice jam of the past century during the 1983/84 winter.
CAO frequency has been linked to the strength of the stratospheric polar vortex and the phase of the Northern Hemisphere annular mode (Thompson et al. 2002; Thompson and Wallace 2001). Rogers and Rohli (1991) have linked the frequency of CAOs that reach Florida to the phase changes in the Pacific–North American (PNA) pattern. Both of these associations can be traced to the strength of the large-scale meridional circulation, which is important in transporting colder air from higher to lower latitudes. Diagnosis based on numerical model simulations of CAOs have proven to be challenging because 1) major CAOs differ from each other in their evolution, and 2) forcing mechanisms such as diabatic processes are difficult to model. CAOs often follow rapid anticyclogenesis over northwestern North America with an amplifying 500-hPa ridge/trough pattern across North America (Schultz et al. 1998; Colle and Mass 1995; Mecikalski and Tilley 1992; Colucci and Davenport 1987; Bodurtha 1952). Even though the major CAO of February 1989 was preceded by rapid anticyclogenesis, the 500-mb upper air field over North America was not the typical ridge/trough upper-air CAO pattern (Hartjenstein and Bleck 1991). Instead, a deep 500-mb ridge formed farther west creating an omega pattern whose zonal flow brought a cold, shallow air mass into the upper Midwest. The major CAO of January 1985 was unusual in that it intensified rather than modified during its southward migration (Colucci et al. 1999). There have also been attempts to address the relative importance of forcing mechanisms in CAO evolution, usually for a specific event. Tanaka and Milkovich (1990) found that advective rather than diabatic processes were responsible for the formation of an extreme cold air pool in late January 1989 over Alaska, which was the genesis of the February 1989 CAO. According to Tanaka and Milkovich, the heat energy of the continental Alaskan air mass is “maintained by a balance of warm advection, adiabatic warming and radiative cooling.” Alaska's location in a region east of the North Pacific characterized by pronounced blocking formations, high mountain ranges, and a diminished diurnal cycle in winter make it particularly vulnerable to changes in the meridional circulation. In January 1989 this balance was altered due to a significant reduction in warm horizontal advection near the surface and cooling in the lower troposphere due to persistent upward motion (Tanaka and Milkovich 1990). Relative to these advective processes, net radiative cooling had only a secondary effect. This finding is quite different from the modeling results of Curry (1987), who found that diabatic processes involving the radiative cooling of the condensate was a significant process in the evolution of a cold-core anticyclone. Colucci et al. (1999) showed a subtle balance between diabatic and adiabatic processes in CAO formation in their ensemble of numerical model simulations of the January 1985 CAO. Other studies have neglected the diabatic term in their studies of CAO evolution (Konrad and Colucci 1989; Colle and Mass 1995). Tan and Curry (1993), who determined the major forcings in the evolution of the February 1989 CAO with the complete height tendency equation, found that diabatic processes were relatively unimportant compared to vorticity advection and differential thermal advection in the evolution of the anticyclone. They also identified ageostrophic motions, which are often filtered by national weather prediction model initialization schemes, as a significant contributor toward anticyclone evolution associated with a CAO.
It is evident from the preceding summary that the relative importance of various mechanisms contributing to CAO development has not been conclusively documented. Moreover, the linkages among CAO occurrences, climate, and climate change are not well understood. Against this backdrop, the present paper has a twofold objective. First, we will establish the low-frequency variability of extreme North American CAOs (both frequency and intensity) over the last half century based on surface station data (section 2). This low-frequency variability becomes an important issue with recent studies reporting warming of the polar air masses in the 1970s–80s (Kalkstein et al. 1990) and increases in winter mean North American temperatures (Houghton et al. 2001). Second, we will investigate the spatial variability and the interaction of the thermodynamic and dynamic processes in the evolution of major CAOs (section 3). This will be done by performing a daily heat budget analysis over the area from northern Canada to the Gulf of Mexico coast. A composite of the daily anomalies of these heat budget terms will also be presented.
2. Low-frequency variability of CAO
a. Data and method
Many different criteria have been used to identify CAOs and cold surges. These criteria were dependent on the focus of the research, the data available, and the nature of the CAO/cold surge under study. Previous research on cold surges in eastern Asia (Lau and Lau 1984), Central America (Schultz et al. 1997), and the Front Range of the Rocky Mountains (Mecikalski and Tilley 1992), were focused on the smaller scale [down to the meso-α scale for the Mecikalski and Tilley (1992) study] and included surface parameters such as the temperature drop at a particular station, sustained northerly flow, and isallobaric patterns. Schultz et al. (1998) examined the planetary- and synoptic-scale patterns associated with cold surges that reached Central America and based their study on a compilation by Reding (1992) of 177 events over 11 winter seasons. Konrad (1996) developed a synoptic climatology of CAOs in the southeastern United States and used a more inclusive definition of a CAO based on temperature anomalies, which yielded 182 CAOs over 12 cold seasons. Colle and Mass (1995) in their composite study of cold surges east of the Rocky Mountains used a sea level pressure of at least 1035 over central Nebraska and the 12-h isallobaric pressure rise in north-central Texas. In our present study, we are focusing on record-setting CAOs that affect a large portion of the contiguous United States.
Our identification of these extreme CAOs is based on temperature station data from the Surface Airways Hourly (SAH) network that were obtained from the National Climatic Data Center (NCDC). The analysis included hourly data from a relatively evenly distributed network of 17 stations during 54 winter seasons (November–March) from 1948/49 through 2001/02. Table 1 lists the station codes, locations, and missing seasons (if any) for the selected stations.
The following method was used for identifying extreme CAOs over these 54 winter seasons. Temperatures were selected every 6 h: 0000, 0006, 1200, and 1800 UTC in the central time zone; 0005, 1100, 1700, and 2300 UTC in the eastern time zone. Temperature anomalies relative to daily means for these hours (1948–2002) were then calculated. The 20 coldest 1-, 3-, and 5-day periods were found for each location by summing the anomalies over the 1-, 3-, or 5-day time span. In the case of temporally proximate CAOs, a CAO was assigned to a certain duration category only if it was separated by a time period of above-average temperatures from other CAO events in the same duration category. Each of these station lists of the 20 coldest 5-day CAOs were combined and then ranked again based on the percentage of stations affected by a particular CAO, resulting in a master list of 30 intense, long-duration, and large-spatial-scale CAO events from 1948 to 2001. The events so identified provided the basis for an investigation of the low-frequency variability of extreme CAO events over our entire station network. We also examined the spatial and temporal variability of CAO intensity with a time series analysis of the mean temperature anomaly at each station for its 20 coldest 5-day CAOs.
b. Results
Table 2 is the compilation of the 30 most extreme 5-day CAOs since 1948 (see previous section) that is ranked according to spatial extent. Note that three of the top four events, which were extreme for at least 75% of the available stations in our network, have occurred since 1983.
Figure 1 depicts the decadal frequency of the long-duration events in Table 2. The highest frequency of events occurred during the 1980s and 1960s. The peak in the 1980s is consistent with the findings of Rogers and Rohli (1991). There is no apparent trend in extreme CAO frequency over this 54-yr period. One might expect that this lack of trend in CAO frequency would also be reflected in the mean temperature field. However, Fig. 2 shows an overall warming trend since 1948 of the 17-station mean winter (November–March) temperature from our Surface Hourly Network dataset. This warming trend (r = 0.33; n = 54) is significant at the 99% level with a one-tailed test. If the time series is divided into halves, there is an even greater warming trend since the mid-1970s (r = 0.48; n = 27; significant at the 95% level with a one-tailed test). The cooling trend in the first half of the time series has a correlation coefficient of 0.17, which is not significant. A more striking warming trend is found over Alaska and western Canada (Fig. 3), a region typically associated with the formation of the most extreme North American CAOs, with trend values approaching +0.80°C decade−1. These trends are for mean winter (November–March) temperatures over the 1948–2002 period obtained from the gridded dataset of Jones and Moberg (2003) and are consistent with the previously cited studies in the introduction (Kalkstein et al. 1990; Houghton et al. 2001) concerning recent North American warming trends. These warming trends in mean winter temperatures both in the high-latitude CAO formation regions and over our observational CAO reporting network, with no accompanying decreasing trend in CAO frequency, represent an intriguing climatic paradox. It suggests that large-scale circulation may outweigh the mean temperature in determining low-frequency variability of CAOs. The increased temperature pattern in Fig. 3, with its dome of warmer air over northwestern North America, is consistent with this hypothesis, since it is representative of a meridional circulation bringing relatively colder air from the higher latitudes into the middle latitudes. Both the negative phase of the Arctic Oscillation (AO) or North Atlantic Oscillation (NAO) (Thompson and Wallace 2001) and the positive phase of the PNA pattern (Rogers and Rohli 1991) have been positively correlated with cold-wave outbreaks over North America. Below the decadal histogram of extreme CAO events in Fig. 1 is a tabulation of mean November–March values for each of the five 11-yr periods for the NAO and the PNA pattern. The NAO values are from a mobile NAO index (NAOm) that better captures the seasonality of the NAO by following the seasonal migration of the centers of action (Portis et al. 2001). The PNA values are from the University of Washington (http://tao.atmos.washington.edu/data_sets/pna). The two periods of peak extreme CAO events (1959–69 and 1981–91) were characterized by favorable values of either one or the other large-scale circulation patterns, but not both of them. The most recent 11-yr period (1992–2002) did not have a large number of extreme CAO events despite having a relatively high positive PNA index.
Figure 4 compares the low-frequency intensity (mean temperature anomaly) of the 20 coldest 5-day CAOs for each of the 17 stations for the 54 winter seasons in our study. Most striking is the spatial variability of the CAO intensity trends with a pattern of decreasing intensity along the eastern seaboard and increasing intensity over the Midwest. Table 3 lists the correlation coefficients for each of these trend lines. The trends at Minneapolis (MSP), Detroit (DTW), Cincinnati (CVG), and Atlanta (ATL) are significant at the 95% level. The trend toward decreasing intensity at Boston (BOS) does not change with the inclusion of the severe 2003/04 winter in the Northeast.
3. Evolution of seven major CAOs
a. Data and method
Another focus of this study was an investigation into the thermodynamic and dynamic forcings in the evolution of major CAOs. For this analysis we chose the seven most extreme CAOs in the Midwestern United States (see Table 4) based on the 1-day temperature anomaly criterion used by Walsh et al. (2001). The criterion used in the Walsh et al. (2001) study was designed to select the most extreme CAOs affecting the Midwestern region indicated in Fig. 5. These seven CAOs were used instead of the 5-day CAOs identified above, because 1) a 1-day criterion has a sharper CAO outbreak date so that processes involved in CAO evolution could be better resolved; 2) these 1-day CAO events were based on impact in a particular region (the Midwest; cf. Walsh et al.'s Fig. 1) and were more compatible with the regional budget framework used in this part of the analysis; and 3) these 1-day events were based on the same National Centers for Environmental Prediction (NCEP) reanalysis data used to calculate the heat budgets in this part of our analysis. Note that the three coldest 1-day CAO events based on the NCEP reanalysis (December 1989; December 1983; February 1996) were also among the four coldest 5-day station-based CAO events (Table 2). All seven events in Table 4 occurred from 1972 onward.
A regional heat budget approach was implemented to study the relative roles of dynamic and thermodynamic processes in the evolution of a major CAO impacting the Midwestern United States. Using gridded NCEP reanalysis data (Kalnay et al. 1996) with a resolution of 2.5° latitude × 2.5° longitude, we constructed a daily heat budget climatology [section 3b(1)] for each of eight regions (Fig. 5) over 30 winters (November–March 1971/72 through 2000/01). The Midwestern region used to identify the seven major CAO events is shaded in gray in Fig. 5. Within this budget analysis, a Midwestern CAO refers to one of those seven CAOs that reached its peak 1-day anomaly criterion on “day–0” (see Table 4 for dates) over that shaded Midwestern area in Fig. 5.
In our vertical advection calculation, we used omega at the 925-hPa level. The environmental lapse rate was a one-sided lapse rate upstream of the vertical wind at 925 hPa. If omega was upward (negative) at 925 hPa, then the lapse rate used the temperature difference between 925 and 1000 hPa. If omega was downward (positive) at 925 mb, then the lapse rate used the temperature difference between 925 and 850 mb. We performed sensitivity tests to see if our vertical advection patterns would change if we used omega at 850 hPa rather than the 925 hPa as our reference level. The patterns were sufficiently similar and the conclusions were unchanged. Because substantial portions of the western regions have surface pressures less than 925 hPa, the computed vertical advection is more open to question in these areas.
A composite [section 3b(2)] and correlation [section 3b(3)] analysis was then done on the seven coldest 1-day CAO events from the peak date of the Midwestern CAO (day −0) until 10 days prior (day −10) to this peak date. The focus of this part of the study was to investigate the similarities, relative importance, spatial variability, and interaction among the heat budget terms during the evolution of the cold air mass.
b. Results
1) Heat budget climatology
The daily time series of the advective and diabatic heat budget anomalies for each of the 30 winters (November–March) over selected northern and southern regions in our study area are shown in Figs. 6, 7 and 8. Each panel within these figures is a daily time series for 30 winter seasons with each winter delimited by blank spaces from the previous winter. Each winter time series is plotted in historical order, so that days near the mean and early in the season are mostly obscured by later days and outliers. The red triangles indicate the minimum anomalies from the 10-day periods prior to, and including, the Midwestern CAO events listed in Table 4.
Each of the climatology figures (Figs. 6 –8) have the same daily anomaly scale (°C day−1) to facilitate comparison. The horizontal advection anomaly (Fig. 6) has the least spatial variability with a fairly constant magnitude between −5° and +5°C day−1 among the four regions shown and also for the other four regions (not shown). However, this small HA magnitude may be attributable to the areal averaging of HA, which does not capture the locally large negative HA associated with the CAO cold-frontal passage. Note that the horizontal advection anomalies for the seven major CAOs in region 2 are extreme relative to the climatology. These historically extreme horizontal advection values in region 2 for the seven major CAOs also coincide with an extremely cold pool of air to the north in region 4. This is seen by examining the 11-day time series of HA and mean temperature (not shown) for regions 2 and 4 for all of the seven CAOs. This extremely cold air in region 4 is advected into region 2 in the scenario indicated by the daily anomalies.
Unlike the horizontal advection anomaly, the vertical advection anomaly (Fig. 7) exhibits large spatial variability with anomaly ranges increasing by a factor of 2–3 from the northeasternmost region (region 8; −5° and +5°C day−1 to the southeasternmost region (region 2; −15° and +15°C day−1). It must be remembered that our vertical advection estimates were limited by the coarse vertical resolution of the reanalysis output. Note also that the cold vertical advection anomalies for the major CAOs are not as extreme relative to the climatology as the cold horizontal advection anomalies were for region 2. This may be due to vertical advection having both cooling and warming roles in the evolution of a cold air outbreak. Descending cold air during a CAO event's peak over a region would result in an adiabatic warming component for VA.
The diabatic anomalies (Fig. 8) have the largest variability relative to HA and VA, particularly over north-central Canada (region 8). The diabatic anomaly distribution has more outliers than HA and VA over all of the 30 winter seasons in regions 7, 8, and 4. In Fig. 8, the outlier areas are highlighted (blue for <−10°C; red for >+10°C). The northern region outliers (regions 7 and 8) were evenly distributed among warm and cold anomalies. Over the southern region 2 (and to a lesser extent region 4), the distribution is skewed toward cold anomalies. As noted earlier, the diabatic anomalies may have the greatest uncertainty because they are calculated as residuals.
2) Composite analysis
We now investigate the cumulative contribution of each heat budget term toward CAO evolution over the 10 days prior to the peak date (day −0) of the Midwestern CAO for our study regions. The cumulative anomalies (shown in Fig. 9) are constructed by summing backward in time from day −1 each heat budget term's anomaly time series that was composited from the seven CAOs (see Table 4). In these cumulative plots, cooling periods are denoted by a positive slope and warming periods are denoted by a negative slope. We have also indicated cumulative cooling/warming periods that we discuss in the text by colored brackets where the bracket color corresponds to the budget term (see key). The brackets for the cooling (warming) periods are drawn below (above) the zero line.
The evolution of the composited CAO can be traced backward in time from region 4 and continuing upstream through regions 5 and 7. The accumulations are done from day 1. Over region 4, there is significant cumulative cooling, as evidenced from the positively sloping DT line, from day −1 through day 6. During this time, both the diabatic and horizontal advection terms contribute to the cumulative cooling, but the vertical advection term offsets this cumulative cooling with cumulative warming from subsiding cold air. Upstream in region 5, the cumulative cooling takes place from day −1 through day −8. Over this region, all three processes (DIA, HA, and VA) contribute to the cumulative cooling. Over region 7, cumulative cooling is evident from day −4 through day −9 and is dominated by the VA term. This is consistent with the work of Tan and Curry (1993). In their synoptic study of the evolution of the cold-wave event in late January–early February 1989 (which is one of our seven CAO events), there was a surface low located over southeastern Alaska and western British Columbia, Canada, during the initial stages of the evolution of the intense North American anticyclone (their Fig. 1). One can also trace the horizontal advection of cold air from region 5 downstream into regions 4 and 2. However, it is never the dominant term in the cumulative cooling composite.
The evolution of the CAO over regions 7, 5, and 4 is in line with the widely held belief that cold air surges often travel parallel to major mountain ranges (Colle and Mass 1995; McInnes and McBride 1993; McBride and McInnes 1993; Hartjenstein and Bleck 1991; Bell and Bosart 1988). The dashed line in Fig. 5 indicates the lee side of the mountain range within our study area. In fact, Colle and Mass (1995) studied the evolution of one of our CAO events (November 1986) when a cold air surge traveled along the Rocky Mountains. From Fig. 9, we see that eastern regions 8 and 6, which are located far from the lee side of the Rockies, do not have a strong signal of CAO evolution in their cumulative heat budget composite.
Region 3 has a different heat budget regime than its neighbor to the east (region 4). Region 4 is characterized by cumulative DIA and HA cooling, offset by cumulative VA warming. This VA warming is associated with subsiding cold air from regions 5 and 7, which leads to clear-sky conditions and increased radiational cooling (diabatic cooling). In contrast, region 3 is characterized by VA cooling and diabatic warming. This suggests orographic lifting of the air in the foothills, and the consequent cloud formation leads to diabatic warming (latent heat release and a decrease in radiational cooling). Note that the heat budget components of the southwesternmost region 1 show characteristics similar to mountainous region 3.
After the CAO has penetrated southward, diabatic warming occurs over the northern regions (region 6, 7, and in particular region 8).
3) Correlation analysis
The above composite analysis focused on cumulative cooling/warming contributions by heat budget terms during the evolution of seven extreme cold-wave events. We now document the linear relationship among the budget terms within individual events by correlating time series of budget terms concatenated over all seven CAOs. Each concatenated time series constructed from the individual events consisted of a total of 77 observations [7 events × 11 days (day −0 through day −10)]. Figure 10 depicts the correlation of the budget terms (and the mean temperature) with the 24-h change in temperature (DT). The correlation for a two-sided test at a 1% significance level is 0.29. For this discussion of the correlation results, note the location of the dashed line in Fig. 5 that denotes the lee side of the western mountain ranges.
Diabatic processes have highly significant positive correlations with the 24-h change in temperature in the eastern part of the domain. This indicates the presence of a diabatic contribution to the day-to-day temperature changes. The correlations increase northward and also in the partially mountainous regions 7 and 5. Vertical advection has its highest correlation with the 24-h change in temperature over the mountainous regions 1 and 3 and to a lesser extent the two regions to the north whose topography does include flatter terrain (regions 7 and 5). The DT over the plains to the east only has a small negative correlation with VA. Horizontal advection has its strongest correlation with the 24-h change in temperature in the eastern plains south of region 8 and also in region 5, which is along the track of cold air surges along the Rocky Mountains. From the correlations in Fig. 10, HA explains about 36%–45% of the variance in the 24-h change in temperature over regions 6, 5, 4, and 2, but it was not a dominant, persistent term in the cumulative cooling of the air mass over those areas (Fig. 9). Note that region 2 was also at a disadvantage in Fig. 9 for resolving processes contributing to the CAO since the backward accumulation in time assumes that all processes at day −0 are zero and day −0 is time of peak cold over region 2. While these results depicting advective contributions are consistent with geographic variation of topography, they highlight the spatial variability of dynamic and thermodynamic processes in the formation of cold-wave events.
Figure 11a is a concatenated plot of the HA and DT anomalies for the individual CAOs over region 2; Fig. 11b is an average of these time series for the individual CAOs. Figure 11 illustrates the close association between HA and the 24-h change in temperature. Throughout the 11-day period, the HA anomalies correspond closely to the temperature change anomalies, but the CAO signal is only seen in the HA term about 2 days before the peak impact (day −0) on region 2. This implies that HA is a more time-specific or sudden impact, rather than a persistent process (e.g., VA and DIA) in CAO evolution.
As noted in previous sections, there is a degree of uncertainty in our VA term since we were limited in its calculation by the coarse resolution of the reanalysis output. If the correlation analysis of VA with the other budget variables is physically consistent, then we could have more confidence in the VA field. The following discussion focuses on physical consistency between VA and the other budget terms (DT, HA, and DIA).
We have already seen in Fig. 10 that VA has a positive correlation with the 24-h change of temperature (DT) over the western United States, but these correlations are weak and negative over our eastern study region. Cooling generated by rising vertical motion is mitigated by latent heat release and the opposing warming of horizontal advection during cyclogenesis. Both of these mitigating factors are more prevalent in our eastern study region, which would explain the decreased correlation between the temperature change and vertical advection. Within the United States, our eastern study region has a moister climate regime than our western region. A composite winter (November–March) precipitation climatology (available at http://www.cdc.noaa.gov/USclimate/USclimdivs.html) for 1950–95 from the National Oceanic and Atmospheric Administration's (NOAA) Climate Diagnostics Center shows the majority of the climate divisions within western regions 1 and 3 have less than 15.24 cm (6 in.) of rainfall in November–March, while the majority of climate divisions in eastern regions 2 and 4 have mean rainfall within the 30.48–76.2 cm (12–30 in.) range. Over the eastern United States, cyclogenesis and its associated warm air advection is the main cause of large-scale upward motion in winter. In contrast, large-scale upward motion in the western United States also has a strong orographic component. The negative correlation of the VA term with HA (Fig. 12) over the eastern United States supports the above discussion concerning the offsetting effect of HA opposing the VA term for regions where cyclogenesis is a dominant factor in the vertical advection field. However, there is also a significant negative correlation of HA and VA over southwestern region 1 where orography is also a component of the vertical advection field. A possible explanation for this is that the surface cold air is quite shallow by the time it reaches these southern latitudes, resulting in higher static stability and greater subsidence warming.
Another physically plausible result is the negative correlation of VA with DIA (Fig. 12) throughout the study region. Descending motion (+VA) is linked to clear skies and an increase in outgoing longwave radiation (−DIA). One can also make the reverse argument that a radiationally cooled air column will contract and is characterized by descending motion. Ascending motion (−VA) implies clouds and an increase in incoming longwave radiation at the surface (+DIA). Again the reverse argument can be made where a warming air column is expanding and is accompanied by ascending motion.
As stated above, the diabatic anomalies have the greatest uncertainty since they are calculated as a residual. Unfortunately, we cannot check the magnitude of the diabatic impacts using the other data since 1) cloud cover data do not include the diabatic processes of latent and sensible heat fluxes, and 2) model-derived radiative, latent, and sensible heat fluxes are in units of W m−2 and cannot be converted into °C day−1. However, we can verify that the diabatic term is consistent with extremes in cloud cover. Table 5 summarizes the average value of the diabatic term under two extremes in total cloud cover (>70% and <30%). These averages were taken for the seven major CAOs (day −10 through day −0) over three areas: the CAO track (regions 7, 5, 4, 2); northern (regions 8, 7, 6, 5); and southern (regions 1, 2, 3, 4). These three areas were selected to focus on the impact of clouds over the track of the CAO and to distinguish between the northern areas where minimal daylight would limit the clouds' daytime cooling impact and the southern areas where the clouds would have both a daytime cooling or nighttime warming impact. Our results show that there are physically consistent differences between the two extreme cloud cover categories and also among the three regions. All regions have more positive diabatic forcing with total cloud cover greater than 70%. Over the regions that follow the CAO track, the diabatic term changes sign from one extreme category to the other. Over the northern regions where daylight is minimal and the impact of clouds is not strongly affected by the diurnal cycle, the magnitude of the difference of the diabatic term between the categories is the greatest.
4. Conclusions
Over the last five decades, there was no overall trend in CAO frequency, but there was an overall increase in mean wintertime temperature over the Midwest and especially upstream into the CAO formation regions of high-latitude North America. There were trends in the intensity of long-duration (5 day) CAOs, but these trends were not uniform across our network. Trends of intensity decreased along the East Coast, but increased in the Midwest. The CAO events of the most recent winter (2003/04) in the eastern United States may reduce the regional differences in these trends.
In our investigation of the evolution of extreme CAOs based on a daily heat budget analysis, the horizontal advection term has the smallest overall magnitude (daily anomaly values between −5° to +5°C day−1), while the daily anomaly values of the diabatic term have a range about twice as large. The vertical advection term is the most spatially variable, with the southeastern region having twice the magnitude as the northwestern region, where daily anomaly values generally range between −5° to +5°C day−1. However, the difficulties in estimating the vertical velocities using coarse-resolution reanalysis output make such conclusions tentative.
Despite the inherent difficulties in studying the evolution of extreme cold waves (limited sample; variability in both space and time; uncertainties in the diabatic heating term computed as a residual), our results show consistency and regional differences in the relative roles of the dynamic and thermodynamic processes in extreme cold-wave events. Specifically:
The cumulative cooling of the air mass can be traced in the heat budget analysis as it tracks southward along the Rocky Mountains into the Midwest. The earliest cumulative cooling starts in northwestern Canada nearly 10 days before the cold air mass reaches the Midwest. Downstream in southwestern Canada, both diabatic and advective (horizontal and vertical) processes contribute to the cumulative cooling of the air mass. At peak intensity over the Midwest, both the diabatic and horizontal advection processes cool the air mass, but subsiding cold air by vertical advection offsets this cooling with warming.
There are regional differences in the correlation of the 24-h temperature change with advective and diabatic processes. Vertical advection has the strongest association with temperature change over the mountainous regions in the Southwest, whereas horizontal advection explains more variance in the temperature change as the cold air mass tracks southward along the Rocky Mountains into the Midwest and the Southeast. Diabatic processes have strong positive correlations with temperature change over all regions (especially in central Canada) except for the mountainous regions in the United States that are to the west of the track of the cold air mass.
Correlations of VA with the other budget terms (HA and DIA) over the study domain are physically consistent and give credibility to the VA field. The sign of VA and HA correlations seem to be related to the region's topography and reflect the importance of orography and synoptic activity in the vertical advection field. The diabatic field is everywhere negatively correlated with vertical advection and is probably linked to the radiational differences due to cloud cover.
A contrasting heat budget regime exists in the regions to the west of the CAO track in the United States. Here vertical advection by orographic lifting cumulatively cools the air in the upslope flow regime associated with the low-level airflow around a cold air mass.
After a CAO has penetrated southward, diabatic warming occurs over the northern regions, particularly over north-central Canada.
The role of horizontal advection in cold surges along the eastern side of the Rocky Mountains has been examined by Colle and Mass (1995). In these surges of cold air, damming of cold air by the Rockies was found to play a role in the southward plunge of cold air along the mountains. The present study presents a much broader assessment of the role of cold air advection, both by including regions with and without substantial topography and by quantitatively comparing the horizontal advection's contribution to temperature changes with the contributions of other processes. In this respect, the study extends the earlier studies emphasizing synoptic and statistical approaches to a more complete diabatic–dynamical framework made possible by the suite of fields included in the output of an atmospheric reanalysis.
An intriguing issue with our study concerns the paradox between CAO frequency and warming of the polar air masses. Our study showed that the frequency of North American CAOs did not decrease as the polar air masses have warmed over recent decades (our Fig. 1; Kalkstein et al. 1990; Hassol 2004) and that over the Midwest, stations showed an upward trend in intensity. This suggests that CAOs are events that are influenced by large-scale circulation patterns. Therefore our ability to predict the occurrence of CAOs might be linked to understanding the phase dynamics of large-scale circulation patterns on different time scales. Thompson and Wallace (2001) demonstrated that the frequency of CAOs increases during the negative phase of the Arctic Oscillation and the easterly phase of the quasi-biennial oscillation (QBO). Rogers and Rohli (1991) show that polarity changes in the PNA pattern from 1947/48 to 1956/57 and 1976/77 to 1985/86 are associated with a stronger meridional flow and a cluster of citrus freezes in Florida during the later period. During the two peak periods in our histogram with extreme CAO events (1959–69 and 1981–91), either one or the other of these large-scale circulation patterns (but not both) was favorable to CAO formation. However over the 1992–2002 period when the mean PNA pattern was relatively high, CAO frequency was low. This might be attributable to the 11-yr periods not being optimum for capturing the low-frequency variability of our indices. In another observational study, Thompson et al. (2002) link the intensity of the stratospheric polar vortex to the phase of the Arctic Oscillation (AO) with an amplified stratospheric polar vortex forcing a positive phase of the AO or a period of stronger zonal flow. Song and Robinson (2004) explain this stratospheric influence on the tropospheric circulation in a modeling study. The relatively weak stratospheric forcing on the troposphere during periods of a highly amplified stratospheric polar vortex can be substantially amplified by transient eddies to excite existing tropospheric modes of variability such as the Arctic Oscillation. Greenhouse warming might also favor a more meridional large-scale circulation pattern that would lead to greater southward transport of the polar air mass. For example, Liang et al. (1996) used a coarse-resolution GCM to show that increased CO2 forces a more positive PNA phase that would favor polar air transport into North America. Community Climate System Model (CCSM3) experimental runs released in June 2004 (http://www.ccsm.ucar.edu/experiments/ccsm3.0/) also show a more amplified meridional flow in northwestern North America during December–February (DJF) when CO2 is doubled. The importance of these circulation changes vis-à-vis direct radiative warming from increased greenhouse gas concentrations will be a key determinant of future trends of CAOs in North America.
Acknowledgments
This work was supported by the National Science Foundation's Climate Dynamics Program through Grant ATM-03-32081.
REFERENCES
Bell, G. D., and L. F. Bosart, 1988: Appalachian cold-air damming. Mon. Wea. Rev, 116 , 137–161.
Bodurtha, F. T., 1952: An investigation of anticyclogenesis in Alaska. J. Meteor, 9 , 118–125.
Colle, B. A., and C. F. Mass, 1995: The structure and evolution of cold surges east of the Rocky Mountains. Mon. Wea. Rev, 123 , 2577–2610.
Colucci, S. J., and J. C. Davenport, 1987: Rapid surface anticyclogenesis: Surface climatology and attendant large-scale circulation changes. Mon. Wea. Rev, 115 , 822–836.
Colucci, S. J., D. P. Baumhefner, and C. E. Konrad II, 1999: Numerical prediction of a cold-air outbreak: A case study with ensemble forecasts. Mon. Wea. Rev, 127 , 1538–1550.
Curry, J., 1987: The contribution of radiative cooling to the formation of cold-core anticyclones. J. Atmos. Sci, 44 , 2575–2592.
Downton, M. W., and K. A. Miller, 1993: The freeze risk to Florida citrus. Part II: Temperature variability and circulation patterns. J. Climate, 6 , 364–372.
Hartjenstein, G., and R. Bleck, 1991: Factors affecting cold-air outbreaks east of the Rocky Mountains. Mon. Wea. Rev, 119 , 2280–2292.
Hassol, S. J., 2004: Impacts of a Warming Arctic: Arctic Climate Impact Assessment. Cambridge University Press, 146 pp.
Houghton, J. T., Y. Ding, D. C. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Mastell, and C. A. Johnson, 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 944 pp.
Jones, P. D., and A. Moberg, 2003: Hemispheric and large-scale surface air temperature variations: An extensive revision and an update to 2001. J. Climate, 16 , 206–223.
Kalkstein, L. S., P. C. Dunne, and R. S. Voss, 1990: Detection of climatic change in the western North American Arctic using a synoptic climatological approach. J. Climate, 3 , 1153–1167.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc, 77 , 437–471.
Konrad II, C. E., 1996: Relationships between the intensity of cold-air outbreaks and the evolution of synoptic and planetary-scale features over North America. Mon. Wea. Rev, 124 , 1067–1083.
Konrad II, C. E., and S. J. Colucci, 1989: An examination of extreme cold-air outbreaks over eastern North America. Mon. Wea. Rev, 117 , 2687–2700.
Lau, N-C., and K-A. Lau, 1984: The structure and energetics of midlatitude disturbances accompanying cold-air outbreaks over east Asia. Mon. Wea. Rev, 112 , 1309–1327.
Liang, X. Z., W. C. Wang, and M. P. Dudek, 1996: Northern Hemispheric interannual teleconnection patterns and their changes due to the greenhouse effect. J. Climate, 9 , 465–479.
McBride, J. L., and K. L. McInnes, 1993: Australian southerly busters. Part II: The dynamical structure of the orographically modified front. Mon. Wea. Rev, 121 , 1921–1935.
McInnes, K. L., and J. L. McBride, 1993: Australian southerly busters. Part I: Analysis of a numerically simulated case study. Mon. Wea. Rev, 121 , 1904–1920.
Mecikalski, J. R., and J. S. Tilley, 1992: Cold surges along the Front Range of the Rocky Mountains: Development of a classification scheme. Meteor. Atmos. Phys, 48 , 249–271.
Miller, K. A., and M. W. Downton, 1993: The freeze risk to Florida citrus. Part I: Investment decisions. J. Climate, 6 , 354–363.
Portis, D. H., J. E. Walsh, M. El Hamly, and P. J. Lamb, 2001: Seasonality of the North Atlantic Oscillation. J. Climate, 14 , 2069–2078.
Quiroz, R. S., 1984: The climate of the 1983–84 winter—A season of strong blocking and severe cold in North America. Mon. Wea. Rev, 112 , 1894–1912.
Reding, P. J., 1992: The Central American cold surge: An observational analysis of the deep southward penetration of North American cold fronts. M.S. thesis, Department of Meteorology, Texas A&M University, 177 pp. [Available from Dept. of Meteorology, Texas A&M University, College Station, TX 77843-3150.].
Rogers, J. C., and R. V. Rohli, 1991: Florida citrus freezes and polar anticyclones in the Great Plains. J. Climate, 4 , 1103–1113.
Schultz, D. M., W. E. Bracken, L. F. Bosart, G. J. Hakim, M. A. Bedrick, M. J. Dickinson, and K. R. Tyle, 1997: The 1993 superstorm cold surge: Frontal structure, gap flow, and tropical impact. Mon. Wea. Rev, 125 , 5–39. Corrigendum. 125 , 662.
Schultz, D. M., W. E. Bracken, and L. F. Bosart, 1998: Planetary- and synoptic-scale signatures associated with Central American cold surges. Mon. Wea. Rev, 126 , 5–27.
Song, Y., and W. Robinson, 2004: Dynamical mechanisms for stratospheric influences on the troposphere. J. Atmos. Sci, 61 , 1711–1725.
Tan, Y. C., and J. A. Curry, 1993: A diagnostic study of the evolution of an intense North American anticyclone during winter 1989. Mon. Wea. Rev, 121 , 961–975.
Tanaka, H. L., and M. L. Milkovich, 1990: A heat budget analysis of the polar troposphere in and around Alaska during the abnormal winter of 1988/89. Mon. Wea. Rev, 118 , 1628–1639.
Thompson, D. W. J., and J. M. Wallace, 2001: Regional climate impacts of the Northern Hemisphere annular mode. Science, 293 , 85–89.
Thompson, D. W. J., M. P. Baldwin, and J. M. Wallace, 2002: Stratospheric connection to Northern Hemisphere wintertime weather: Implications for prediction. J. Climate, 15 , 1421–1428.
Walsh, J. E., A. S. Phillips, D. H. Portis, and W. L. Chapman, 2001: Extreme cold air outbreaks in the United States and Europe, 1948–99. J. Climate, 14 , 2642–2658.
Histogram of the extreme 5-day CAO events in Table 2. Listed below are the mean monthly (Nov–Mar) index values for the NAO and the PNA patterns for each of the 11-yr periods within the histogram. The year associated with the index is the year of Jan
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Time series of the mean winter (Nov–Mar) temperature anomaly for the stations in Table 1. The data are plotted in the year of Nov. The vertical gray line at 1974 divides the time line into half. The three regression lines are for the whole time period (dashed) and for each half of the time period (solid)
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Decadal trends (°C decade−1) of the surface mean winter (Nov–Mar) temperatures over the 1948–2002 period
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Scatterplots with regression lines of mean temperature anomalies (°C) during the 20 coldest 5-day CAOs at each of the stations listed in Table 1
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Locations of the eight regions used in the heat budget analysis. The dashed line indicates the lee side of the mountain ranges within those regions. Gray region denotes the Midwestern region used to identify the seven major CAO events
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Daily time series of the horizontal advection anomaly (°C day−1) for the 30 winters from 1971/72 through 2000/01 for northern regions 7 and 8 and southeastern regions 4 and 2 (see Fig. 5). The red triangles indicate the extreme values of horizontal advection during the day −10 to day −0 time series for each of the seven CAOs listed in Table 4
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
As in Fig. 6, except for the vertical advection anomaly (°C day−1)
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
As in Fig. 6, except for the diabatic anomaly (°C day−1). The outlier areas are highlighted (blue for <−10°C; red for >+10°C)
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Cumulative regional anomalies (°C day−1) of each of the four heat budget terms. Panels are arranged according to each region's geographic position (Fig. 5). The cumulative anomalies are constructed by summing backward in time each heat budget term's anomaly time series composited from the seven CAOs (see Table 4). Anomalies are color coded by budget term (see key). Brackets drawn below (above) the zero line indicate cooling (warming) periods that are discussed in the text
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Regional correlations of budget anomalies (VA, HA, DIA) and mean temperature anomalies (TMEAN) with the daily change in temperature anomaly (DT). Panels are arranged according to each region's geographic position (Fig. 5). Correlations are done with budget anomaly time series concatenated over each of the seven CAOs (see Table 4) from day −0 through day −10
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
Horizontal advection (HA) and 24-h change in temperature (DT) budget anomalies over region 2 for (a) each of the seven CAO events (see Table 4); (b) a composite of the seven CAOs. The individual CAO events are indicated below (a)
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
As in Fig. 10, except for the correlation of budget anomalies (HA, DIA) with the vertical advection anomaly, VA
Citation: Monthly Weather Review 134, 2; 10.1175/MWR3083.1
List of Surface Airways Hourly stations used in documenting cold air outbreaks for the winter season (Nov–Mar) from 1948 to 2002. Missing seasons are indicated in the last column
CAO events ranked by the percentage of available stations in Table 1 having that CAO event in their top twenty 5-day CAO list. The average rank of CAO is the mean CAO rank of that CAO among the top twenty 5-day CAO lists
Correlation coefficients for the CAO intensity trends
Dates of the seven CAOs used in the budget analysis
Mean diabatic forcing (°C day−1) under the extreme cloud cover categories (>70% cloud cover; <30% cloud cover)
