Seasonality of the northern hemisphere circumpolar vortex

Transcription

1 INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 27: (2007) Published online 14 November 2006 in Wiley InterScience ( Seasonality of the northern hemisphere circumpolar vortex Kalyn M. Wrona a, * and Robert V. Rohli a a Department of Geography and Anthropology, Louisiana State University, Louisiana, USA Abstract: In previous research, Rohli et al. (2005) identified long-term features of the northern hemispheric circumpolar vortex (NHCPV) in January. This research provides a seasonal analysis using December and February to augment the previously analyzed January data in representing winter, along with April, July, and October data to represent spring, summer, and autumn, respectively. A representative 500 hpa geopotential height contour was selected to delineate the NHCPV in each of the five months. The area, shape, and centroid of the monthly December, February, April, July, and October NHCPV are computed for to supplement the previously identified January properties. These geometrical features of the NHCPV reveal relationships between hemispheric-scale circulation and temperature anomalies throughout the year. A circularity ratio (Rohli et al., 2005) is used to characterize the shape of the hemispheric-scale circulation. Results suggest that only October exhibit long-term trends in either area or circularity, with July being the most variable month in area and October being the most variable month in circularity. s tend to be skewed toward the Pacific basin, except in spring, but few systematic temporal shifts in centroid position were noted for any month. The NHCPV is correlated with atmospheric teleconnection patterns in several months. For example, as was the case for January (Rohli et al., 2005), the Arctic Oscillation (AO) is associated with the area of the December, February, and April NHCPV, while in December the circularity is positively correlated to the AO Index. Also, the Pacific-North American index is correlated with the area of the December and February NHCPV and with the shape of the December and October NHCPV. Copyright 2006 Royal Meteorological Society KEY WORDS circumpolar vortex; atmospheric circulation variability; atmospheric teleconnections; climate variability and change Received 5 April 2006; Revised 25 August 2006; Accepted 25 August 2006 INTRODUCTION Variation in the hemispheric-scale, upper-level, prevailing atmospheric circulation results from variations in the input of energy across the earth s surface. One indicator of this prevailing circulation is the circumpolar vortex (CPV), a cyclonic circulation in the middle- to upper-troposphere circumnavigating the poles in each hemisphere. Because the CPV exists over the region of strongest thermal gradients, variation in area, shape, and position of the CPV characterizes the impacts of temperature changes on circulation. Furthermore, the CPV alsomayexplainvariationinother circulation-dependent atmospheric features, such as humidity, precipitation, and pollution advection. Rohli et al. (2005) suggested that in Januaries from 1959 to 2001, the Northern Hemisphere s CPV (NHCPV) exhibited no linear temporal trends in area or shape, and that the mean centroid of the NHCPV is displaced from the North Pole in the direction of the Pacific Ocean. Furthermore, Rohli et al. (2005) linked the Arctic Oscillation (AO) to variability in the area of the January NHCPV, and the Pacific-North American (PNA) teleconnection to Correspondence to: Kalyn M. Wrona, kwrona@hotmail.com the shape (but not area) of the January NHCPV over the same period. This research expands on previous research by examining variability and changes in the area, shape, and position of the NHCPV, and linkages to atmospheric teleconnections, throughout the year. Because the winter season is the time of strongest pressure gradients, and therefore, the most meaningful and identifiable NHCPV, December and February analyzes were added to corroborate the previous exploratory January results. Because of the labor-intensive nature of representing the NHCPV and calculating geometric features of it, April, July, and October were selected to represent the spring, summer, and autumn, respectively. PURPOSE The purpose of this research is to provide a more complete history and better understanding of the NHCPV and atmospheric circulation. The most fundamental question addressed in both this research and in Rohli et al. (2005) is whether the NHCPV exhibits long-term trends in geometrical and spatial features over the period. Therefore, this research includes several objectives to characterize the change in the NHCPV including: Copyright 2006 Royal Meteorological Society

2 698 K. M. WRONA AND R. V. ROHLI examining the long-term variability and changes in the area, characterizing the temporal variability and changes in circularity, determining whether the mean geographic position has changed over time, and identifying correlations between the area and circularity of the hemispheric-scale flow (represented by the CPV) and sub-hemispheric-scale circulation features (represented by atmospheric teleconnections). First, it is important to document the change in area of the NHCPV. The size of the CPV serves as a proxy for the overall degree of equatorward displacement of polar air and the polar front jet stream at its leading edge. However, even if the NHCPV shows similar area over time, it may force differing impacts. For example, the NHCPV in two months may have the same area, but in one of the months, the flow could be characterized by intense ridges and troughs that give the NHCPV an meridional shape, while the other month could be characterized by a circular shape about the North Pole (and therefore zonal, or westerly, upper-level flow throughout the midlatitudes). Therefore it is important to examine changes in the circularity, or shape, of the NHCPV as well. This part of the analysis also addresses whether any correlation exists between the area and circularity. It is hypothesized that the global warming signal in surface temperatures would be associated with a reduction in area in the NHCPV over time as the cold pool of air over the poles shrinks. Furthermore, if the area decreases over the period, the NHCPV may become more meridional (i.e. less circular) because zonal winds may decrease when the temperature increases at the poles. However, because the January results failed to identify linear decreases in area as hypothesized (Rohli et al., 2005), further analysis of other months is needed to supplement the previous January analysis. A third component of the research analyzes whether any change in position of the NHCPV is evident over time. It is expected that the mean position of the NHCPV would be located near the North Pole. However, a shift from the pole to (say) a centroid over Siberia would suggest that the cold pool has shifted equatorward over Asia and concurrent surface environmental changes would likely occur. Furthermore, even if no changes in area, circularity, or position are observed in the NHCPV, the NHCPV could still show two very different patterns and have varying impacts because the shape may simply have rotated about the same centroid. Differentiation between two such patterns is performed through correlation of the NHCPV properties to atmospheric teleconnection indices. The teleconnections represent pressure oscillations around fixed points around the world. This part of the study is important because any association between variability in the area and/or circularity of the NHCPV and variability in teleconnection patterns would suggest that the regional-scale teleconnections fluctuate as the overall, holistic pattern varies. By contrast, the absence of such a correlation would suggest that regional changes could occur in the absence of (or independently of) variation in the overall holistic pattern. It is hypothesized that the teleconnections studied correlate with the geometric properties of the NHCPV. DATA AND METHODS Data ThesamedatasetfromRohliet al. (2005) is employed here. Specifically, mean monthly 500 hpa geopotential height data from 1959 to 2001 were obtained from the National Center for Atmospheric Research (NCAR, 2003). This data set was selected so that results would be comparable both to results from the January analysis and from previous work (Burnett, 1993). The daily observational data are projected to a 5 by 5 grid including the entire Northern Hemisphere and are then averaged to produce the monthly mean. Limitations of the data set are described in Rohli et al. (2005). The analysis of 5 months per year (February, April, July, October, and December) to augment the previously analyzed January climatology provides for analysis in the annual cycle, with more detail in the winter season when the NHCPV is most robust and influential on the hemispheric-scale climatology. To address the question of whether the geometrical attributes of the NHCPV are associated with atmospheric teleconnections, several standardized monthly indices were examined. These indices were obtained from the Climate Prediction Center (CPC, 2003) and the National Climatic Data Center (NCDC, 2003). Specifically, data were gathered for the Southern Oscillation Index (SOI) and Niño 3.4 Index to represent the El Niño/Southern Oscillation (ENSO) phenomenon, along with indices representing the AO, PNApattern, North Atlantic Oscillation (NAO), and Pacific Decadal Oscillation (PDO) (CPC, 2003; NCDC, 2003). Methodology The size of the data set permits statistical analysis of the NHCPV properties by month (e.g. examination of long-term changes in the January NHCPV, etc.). A specific isohypse is selected to represent the NHCPV for each month, as recommended by Frauenfeld and Davis, 2003, Table I. The 5460 hpa isohypse was selected to delineate the December and February NHCPV. Because the same isohypse was used for January, it is possible that the NHCPV may be over-represented in the core winter month of January compared with the more fringe months of December and February. Nevertheless, we use the months suggested by Frauenfeld and Davis (2003), who studied geopotential height change by analyzing both daily and monthly mean maps along a 5-degree meridian to identify the isohypse that fell within the primary baroclinic zone to best represent the NHCPV.

3 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 699 Table I. Geopotential height contours (m) used to define the NHCPV. (Frauenfeld and Davis, 2003). Month 500 hpa isohypse (m) January 5460 February 5460 March 5520 April 5580 May 5640 June 5700 July 5700 August 5760 September 5640 October 5580 November 5520 December 5460 The area and circularity of the NHCPV were compared within a month (e.g. Januaries can be compared against Januaries) using the predefined isohypse to represent the equatorward margin of the NHCPV. For each of the 215 months (5 months 43 years) of the analysis, the corresponding height (Table I) was contoured, the area and perimeter of the CPV in each month were calculated using the XTools algorithm (Delaune, 2000), and the centroid of each monthly NHCPV was calculated using the center of mass criterion (Ciudad, 2002). An example of the analysis is shown in Figure 1. Regression analysis determined whether the area of the NHCPV exhibits linear change over time for each of the 5 months analyzed. The map projection is of great importance to this research, since results may vary depending on the projection used (Rohli et al., 2005). More specifically, a map projection that preserves shape but distorts area would portray the size of the CPV incorrectly. Likewise, if a projection that preserves area but distorts shape were used to identify the centroid of the CPV, the results would be incorrect. Therefore, this research uses the Lambert s equal-area projection to address the research questions that depend on equivalency (i.e. preservation of area) and the conformal (i.e. shape-preserving) polar stereographic projection when identifying the centroid (Robinson et al., 1984). Both projections were used to define the shape of the CPV. The perimeter of a given CPV, projected in the stereographic projection, was recorded and used as the circumference of a reference circle. The radius was then derived from that circumference and used to determine the area of that reference circle. Then, the area of a given CPV projected in the Lambert s equal-area projection was divided by the area of the reference circle to determine the circularity ratio (R c, Rohli et al., 2005). Therefore, R c is directly proportional to the circularity Figure m geopotential height contour for January The represents the North Pole and the dot represents the NHCPV centroid.

4 700 K. M. WRONA AND R. V. ROHLI of the NHCPV, with small values of R c suggestive of amplified and/or more longwave ridges and troughs. R c was used to address the hypothesis regarding NHCPV circularity. Regression analysis was again used to identify linear temporal trends in R c and Pearson correlations were computed between area and R c of the NHCPV. To address the third hypothesis, visualization techniques and CrimeStat II software (Levine, 2002) were used to show temporal changes in the position of the centroid of the NHCPV over the period. Changes in the centroid location would suggest that the CPV shifted poleward over one area and simultaneously equatorward over another area, and may explain why some places warmed and others cooled simultaneously during the period of overall warming. After the centroid of each NHCPV was calculated and recorded, several analyses were performed using CrimeStat II software. These analyses included identifying nearest neighbor indices (NNI) and hotspots using spatial clustering, and computing the directional mean and circular variance, where 0 represents no variability and 1 represents very high variability (Levine, 2002). Next, Pearson correlations were used to determine whether the NHCPV area and R c correlate significantly with any of the teleconnection indices. Each of the months analyzed was included in this analysis, with one exception. July was not correlated with the PNA pattern, because the PNA index is not prominent in the summer months (Barnston and Livezey, 1987). RESULTS The results for each of the months analyzed can be found in the Appendix. The NHCPV area, circularity, and position Winter. Both December and February results corroborate the results found for January by Rohli et al. (2005). The mean NHCPV tends to grow through the traditionally defined winter season, with the December NHCPV displaying a smaller mean area ( km 2 )than that for January, and February exhibiting the largest mean area ( km 2 ) of the winter months. This result coincides with what is expected given the delayed onset of seasons over maritime areas such as the North Pole. Figure 2 identifies the characteristics of the winter NHCPV, with the smallest area occurring in December 1990 ( km 2 ), the largest area occurring Figure 2. Representative examples of the winter NHCPV from 1959 to 2001: (a) December 1990 (smallest in area); (b) December 1959 (most circular shape); (c) February 1986 (largest in area); (d) January 1963 (least circular shape).

5 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 701 in February 1986 ( km 2 ), the most circular NHCPV in February 1990 (0.786), and the least circular in January 1963 (0.497, as in Figure 2(d) of Rohli et al., 2005). Furthermore, the circularity of the NHCPV also decreases throughout the winter, with December being most circular on average, with a mean R c of 0.706, and February being least circular (mean R c of 0.683). Like January, December and February s NHCPV areas were distributed normally around the mean, with no significant skewness or kurtosis. However, the December R c distribution was significantly negatively skewed (z skewness = 2.869, p-value = 0.003), suggesting that most December NHCPVs were more circular than the mean R c. This result is consistent with that found for the January R c as reported by Rohli et al. (2005). Unlike January, which showed a leptokurtic distribution of R c (Rohli et al., 2005), no kurtosis was identified for R c in December or February. Therefore, the January NHCPV seems to be very consistent in circularity, while Decembers and Februaries are more likely to have anomalously large or small R c. The regression analysis did not identify any significant linear temporal trends in the December (p-value = 0.901) or February (p-value = 0.122) NHCPV area. Again, these results support those reported previously for January (Rohli et al., 2005). Similarly, trends in R c for December (p-value = 0.091) and February (p-value = 0.105) were not statistically significant, similar to the results from Rohli et al. (2005), which suggest no trend in January circularity (p-value = 0.453). Therefore, no statistically significant temporal changes in polar-tropical exchange in energy are apparent, but because of the implications of this result for global circulation changes in winter, this phenomenon should be investigated further. No significant correlation was identified between area and R c for December or February, corroborating the previous results for January. Thus, it appears that changes in the winter NHCPV area are independent of circularity. To examine the position of the NHCPV, the centroids for each December and (in a separate analysis) February were plotted. This analysis resulted in an almost identical pattern to that observed for January (Rohli et al., 2005). Figure 3(a) shows that most December centroids fell in the western hemisphere, with a directional mean at N, W, and falling km from the North Pole. Likewise, the February centroids showed a mean position at N, W, km from the North Pole (Figure 3(b)). The position of the winter NHCPV seems to be displaced toward northwestern North America during most of the period. The circular variance for each winter month was low, ranging from 0.09 in December to 0.11 in February. Rose diagrams further depict this displacement, with an average of 77% of the points falling at longitudes between Alaska and Siberia in December (Figure 4(a)) and January (Figure 4(b)). However, the February rose diagram shows a slight increase in variance over the period (Figure 4(c)). Perhaps this result is not unexpected as February may display more characteristics of a transition month, with slightly higher interannual variability. Further examination for clusters in the overall point pattern revealed significant NNIs for each of the winter months, ranging from in December to in January (p-value < 0.001). Therefore, the overall point pattern distribution for the NHCPV centroids was more clustered than random in each winter month. However, significant (p-value < 0.05) clusters, or hot spots, were only found for the January and February NHCPV (Figure 5), if the minimum number of points required for a cluster is set to three. April. April was chosen to represent spring and the 5580 hpa isohypse was selected to represent the April NHCPV (Frauenfeld and Davis, 2003). The largest area was found in 1978 (Figure 6(a)) and the smallest area was in 1962 (Figure 6(b)). Likewise, the most circular NHCPV occurred in 1985 (Figure 6(c)) and the most meridional April was 1983 (Figure 6(d)). Furthermore, the April NHCPV area data were evenly distributed around a mean of km 2 (z skewness = 1.834, Figure 3. s of the NHCPV for (a) December and (b) February. The dot represents the directional mean centroid. Values represent the position of the NHCPV in each December (February) chronologically (1959 = 1, 1960 = 2, etc.).

6 702 K. M. WRONA AND R. V. ROHLI (a) 165E 135W (b) 165E 135W 105E 40% 30% 20% 10% 75W 105E 50% 40% 30% 20% 10 % 45E 15W 45E 15W 0.005). Therefore, most April NHCPVs were more circular than the mean R c of Examination for trends in the area and R c of the April NHCPV produced similar results to that of the winter months. Regression analysis revealed no linear temporal trends in area (p-value = 0.604) or R c (p-value = 0.769). The April NHCPV, like the winter months, found only short and weak temporal changes over the period. Finally, as for the winter months, no correlation between the area and R c was identified (r = 0.002, p-value = 0.992). On the other hand, results for the April NHCPV did display very different results than those found in winter for centroid position throughout (Figure 7). The mean centroid was located over the Atlantic side of the Arctic Ocean at N, W, falling only km from the North Pole, and the circular variance was much greater at The rose diagram (Figure 8) shows the wide distribution of the April centroids around the North Pole, with the majority (16%) located between 60 W and 85 W. The distribution was much wider in April than in any other month analyzed. The April results also identified a NNI of 1.02, or a centroid distribution that is random. Not surprisingly, the variability in the NHCPV movement precluded the identification of clusters. Interestingly, April is the only month with most of the centroids located near the North Pole. These results are not surprising because of the transitional nature of the spring hemispheric circulation. (c) 165E 135W 75W 105E 40% 75W 30% 20% 10% 45E 15W Figure 4. Average wind rose diagrams. (a) December; (b) January; (c) February. z kurtosis = 0.372). Although no kurtosis was found in the R c data (z kurtosis = 1.519), the R c distribution was significantly negatively skewed (z skewness = 2.785, p-value = July. The summer NHCPV is represented by July, and the 5700 hpa isohypse (Frauenfeld and Davis, 2003) was contoured and used for the analysis of the area, circularity, and centroid. The largest NHCPV occurred in 1992 (Figure 9(a)), the smallest in 1998 (Figure 9(b)), the most circular in 1995 (Figure 9(c)), and the least circular in 2001 (Figure 9(d)). It should be noted that 2 years (1965 and 1968) were eliminated because the data were deemed to be spurious. Neither skewness (z skewness = for area and z skewness = for R c ) nor kurtosis (z kurtosis = for area and z kurtosis = for R c )was found in the area or R c. Therefore, the data were normally distributed around the mean area of km 2 and around the mean R c of Like the winter months, regression analysis did not identify a significant decreasing linear temporal trend in area (p-value = 0.083). Furthermore, no significant temporal trend in R c was observed (p-value = 0.925). A significant positive relationship between the July NHCPV area and circularity (Pearson r = 0.323, p-value = 0.039) was observed. Thus, a large (small) area of the July NHCPV is associated with a more (less) circular July NHCPV. This result suggests that Julies with anomalously small NHCPVs may at least partially offset their inability to advect cooler air southward by having more and/or larger troughs. This result should be investigated further.

7 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 703 Figure 5. The hierarchical NNA clusters identified at α Note that no significant clusters were found for December. Figure 6. Representative examples of the April NHCPV from 1959 to 2001: (a) 1978 (largest in area); (b) 1962 (smallest in area); (c) 1985 (most circular shape); (d) 1983 (least circular shape).

8 704 K. M. WRONA AND R. V. ROHLI Figure 7. s of the April NHCPV. 165E 135W 75W 105E 20% 10% Figure 8. Rose diagram for the April NHCPV. 45E 15W The July pattern resumed the winter tendency to have centroid displacement toward the Pacific Basin. All except one of the centroids (July 1988) fell within the western hemisphere (Figure 10). The directional mean was located at N, W, km from the North Pole. July displayed a very low circular variance of 0.06 and was the least variable of all the months analyzed. The variability is shown further in the rose diagram (Figure 11). The July centroids identified a more persistent pattern, with a significant NNI of (p-value < ). One first-order cluster was found at the 0.05 probability level, containing five points (Figure 12). Compared to the clusters found in the winter months (Figure 5), July s cluster position was closer to the North Pole and centered north of Canada and Greenland. October. The October NHCPV was analyzed to represent the autumn hemispheric-scale extratropical circulation and was represented by the 5580 hpa (Frauenfeld and Davis, 2003). The largest NHCPV occurred in 1976 (Figure 13(a)), while the smallest area was found in 1991 (Figure 13(b)). Similarly, the most circular NHCPV occurred in 1990 (Figure 13(c)) and the most meridional was in 1967 (Figure 13(d)). The mean NHCPV area of km 2 showed no significant skewness (z skewness = 1.578). However, the distribution of areas was found to be significantly leptokurtic (z kurtosis = 2.530, p-value = 0.011), suggesting that the data has smaller tails around the mean than that of a normal distribution. Furthermore, the October NHCPV R c was found to be negatively skewed (z skewness = 7.711, p-value < 0.001) and leptokurtic (z kurtosis = , p-value < 0.001), suggesting that most October NHCPVs were more circular than the mean R c of 0.724, with relatively few high- or low-circularity anomalies. Regression analysis did not identify a linear temporal change in the October NHCPV area (p-value = 0.999), but did suggest a significantly increasing linear trend in R c (p-value = 0.022). Therefore, the NHCPV became more circular over the period (Figure 14). Whether this increasing circularity is a cause or an effect of a global warming signal is a question for future research. Visually, it seems that 1976 (the largest October NHCPV) is an anomaly compared to the other October areas, and 1967 (the least circular shape) is an anomaly compared to the other October R c values. Finally, Spearman correlations did not find a significant relationship between the area and R c (r = 0.083, p-value = 0.600), suggesting that the two may not be influenced by the same factor.

9 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 705 Figure 9. Representative examples of the July NHCPV from 1959 to 2001: (a) 1992 (largest in area); (b) 1998 (smallest in area); (c) 1995 (most circular shape); (d) 2001 (least circular shape). Figure 10. s of the July NHCPV. Note that years 7 (1965) and 10 (1968) are missing because of spurious data.

10 706 K. M. WRONA AND R. V. ROHLI 165E 135W 105E 30% 75W 20% 10% Figure 11. Rose diagram for the July NHCPV. 45E 15W The October NHCPV centroid results resemble those in the winter months and July (Figure 15). The centroids were again displaced toward the Pacific with a mean centroid located at N, W, about km from the North Pole. The circular variance was relatively small at Furthermore, Figure 16 shows a high percentage (77%) of the centroids located between Alaska and Siberia. The cluster analysis reveals a significant NNI of (p-value < 0.001), suggesting that the overall distribution was significantly more clustered than random. Furthermore, two first-order clusters were found at the 0.05 probability level, with the first cluster including five points and the second cluster including four points (Figure 12). Both clusters were located in the same area as the clusters found in the winter months, leading to the conclusion that the autumn NHCPV acts similar to the winter NHCPV. The NHCPV and regional and hemispheric flow patterns To investigate the NHCPV further, associations between the NHCPV area and monthly atmospheric teleconnection indices, and in separate analyses, between R c and the same atmospheric teleconnection indices, were examined using Pearson correlation analysis. Any observed linkage between the NHCPV and regional-scale teleconnections would suggest that the geographical position of the longwave ridge trough configuration of the NHCPV tends to be anchored to certain locations. By contrast, an absence of linkage may imply that the NHCPV can rotate about a constant centroid (i.e. alter the ridge/trough configuration, as represented by teleconnection indices) without changing its area or shape. The Pearson correlation results between the area (and R c ) and each of the monthly atmospheric teleconnection indices are shown in Table II. Five significant correlations occur in December, more than any other month analyzed. First, the December NHCPV area was significantly positively correlated with the PNA pattern. Therefore, as the PNA teleconnection goes into a positive phase, with an amplified ridge trough pattern over North America (Leathers and Palecki, 1992), the overall area of the December NHCPV tends to increase. The December NHCPV area also correlated significantly with the AO, and actually had the strongest correlation of all variable pairs analyzed. Thus, the negative (positive) phase of the AO, with above- (below) normal sea level pressure (SLP) over the Arctic and anomalously low (high) SLP in lower latitudes and cold (warm) conditions particularly in northern Europe, is linked to an anomalously large (small) NHCPV. This result implies Figure 12. The hierarchical NNA clusters for July and October identified at 0.05.

11 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 707 Figure 13. Representative examples of the October NHCPV from 1959 to 2001: (a) 1976 (largest in area); (b) 1991 (smallest in area); (c) 1990 (most circular shape); (d) 1967 (least circular shape). Circularity Ratio y = x R 2 = p-value = Year Figure 14. October NHCPV R c scatterplot with regression line throughout the period. that when the December NHCPV is anomalously large (small), midlatitude temperatures in the Northern Hemisphere are anomalously low (high). It is noteworthy that Rohli et al. (2005) found a similar result for January. Not surprisingly, because of the similarity between the AO and its regional expression, the NAO (Marshall et al., 2001; Rogers and McHugh, 2002), a significant negative relationship also exists between the December NHCPV area and NAO. A positive phase of the NAO is associated with a strong Icelandic Low, strong westerlies across the North Atlantic, and high-pressure along 40 N (Wallace and Gutzler, 1981); during such periods, the December NHCPV decreases in area. By contrast, a large December NHCPV is linked to weakened westerlies across the north Atlantic basin typical of negative NAO Decembers.

12 708 K. M. WRONA AND R. V. ROHLI Figure 15. s of the October NHCPV. 165E 135W 105E 30% 75W 20% 10% 45E Figure 16. Rose diagram for the October NHCPV. 15W The December NHCPV R c correlated significantly with two teleconnections. First, as was found for January (Rohli et al., 2005) a significant negative relationship was found between R c and the PNA pattern. This result suggests that a decrease (increase) in circularity (i.e. an increase (decrease) in ridge and trough amplification over North America) is coincident with increased (decreased) ridging/troughing across the hemispheric Rossby wavetrain hemisphere as a whole. However, it should be noted that because the PNA-related ridging/troughing is a subset of the entire hemispheric NHCPV, a correlation between these two variables should not be surprising. Nevertheless, this result, when combined with results for other winter months (discussed in the following text), suggests that the winter PNA pattern appears to represent a subset of the hemispheric-scale circulation variability that is not offset by coincident opposite amplification tendencies elsewhere. The other significant December relationship is a positive association between the R c and the AO. Thus, when the AO was in a warm (cold), or positive (negative), phase, the NHCPV was more (less) circular. It is interesting that both the area and R c are tied to the AO, with the positive (negative) phase of the AO linked to a smaller and less circular (larger and more circular) December NHCPV. No other significant correlations were found between the December NHCPV area and R c. Interestingly, as was the case in January (Rohli et al., 2005) neither the SOI nor Niño 3.4 correlated significantly with either the area or R c of the NHCPV, suggesting that El Niño creates regional changes to the flow, but may not affect the NHCPV acting as a whole. Furthermore, no significant correlations between the latitude and longitude of the centroid and either of the ENSO indices were identified, for any of the months. Three teleconnections were found to correlate significantly with the area of the February NHCPV (Table II). First, the PNA pattern was found to be positively correlated with the area. Thus, as the PNA teleconnection moved into a positive (negative) phase, the area of the NHCPV increased (decreased). The correlation was found for December. Furthermore, even though the February pattern did not show a significant link between R c and the PNA pattern, the significant negative correlations for the other two winter months between the circularity and PNA teleconnection suggest that this teleconnection may be an important winter anchor for both the area and the circularity of the NHCPV. Only winter

13 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 709 Table II. Summary of significant Pearsons correlations between NHCPV area/circularity and teleconnection indices, by month. SOI Niño 3.4 PNA AO NAO PDO Area December p-value <0.001 <0.001 January 0.43 p-value (Rohli et al., 2005) February p-value <0.001 <0.001 April p-value < July 0.32 p-value October p-value R c December p-value <0.001 January 0.44 p-value (Rohli et al., 2005) February p-value April p-value July p-value October p-value months showed strong correlations between the PNA and the NHCPV. The AO and the NHCPV area are significantly negatively related in February, as in the other two winter months. Once again, the uniform response for the AO in all the three winter months suggests that this pattern was a dominant feature associated with variability in NHCPV area. Finally, the area was also significantly negatively correlated with the NAO. While the area of the February NHCPV correlated significantly with three teleconnections, the R c did not correlate significantly with any of the teleconnections studied. Two significant correlations were identified in April (Table II). A significant negative relationship was found between the area and AO, as was the case for all the three winter months. According to Wallace and Gutzler (1981), temperatures increase (decrease) around the North Pole when the AO is in a positive (negative) phase. Likewise, a significant negative relationship was found between the area and the closely related NAO. Once again, this tendency mimics the results found for most of the winter months. No other significant correlations were found in either the area or circularity. Only one significant correlation was found for July: a positive correlation between the area and Niño 3.4 (Table II). The positive phase of the Niño 3.4 index is associated with anomalously high SSTs in the region bounded by 5 N and5 S and W; when the index is in its positive phase, the July NHCPV area increases (Newman et al., 2003). July is the only month in which the Niño 3.4 or SOI seems to be an important factor associated with NHCPV geometric properties. This result is interesting because the effects of the ENSO phenomenon have been regarded to be more magnified in winter rather than summer (Newman et al., 2003). Perhaps the great distance between the pressure fluctuation and the NHCPV is associated with a greater time lag in the effects of ENSO on the NHCPV than in the tropics and midlatitudes. July also had the fewest significant correlations of any month analyzed. Two atmospheric teleconnections were found to be related to the October NHCPV (Table II). First, a significant negative relationship exists between R c and the PNA pattern. As the PNA index becomes positive (negative), the NHCPV becomes less (more) circular. This result suggests that, as was the case for December and January, the PNA-related ridge trough configuration was propagated around the hemisphere rather than compensated by opposite anomalies of amplification upstream or downstream. Perhaps the PNA pattern tends to become established as a part of the hemispheric Rossby wavetrain in autumn and persists into the early part of the winter. In addition, October is the only month analyzed to show significant (negative) correlations between the NHCPV and the PDO. Specifically, the R c became less

14 710 K. M. WRONA AND R. V. ROHLI (more) circular as the northern Pacific SSTs demonstrate a positive (negative) warm ( cold ) phase of the PDO (Newman et al., 2003). This result should be interpreted with caution, however, because the period of the PDO is long approximately 20 to 30 years (Newman et al., 2003) while the period studied only consisted of 43 years. Therefore, the true nature of the PDO-induced variability may not appear to a sufficient degree during the study period. No other correlations were found between the teleconnections and either area or R c. CONCLUSIONS AND FUTURE RESEARCH The geometric properties of the monthly averaged NHCPV, including area, circularity, and centroid location, were examined seasonally over the period. Relationships between the area and circularity of the NHCPV and several atmospheric circulation indices were identified. For these analyses, the months of December and February were chosen to supplement the previously studied January to provide a clear picture of the NHCPV s wintertime properties. April, July, and October were chosen to represent spring, summer, and autumn, respectively. It was hypothesized that the area of the NHCPV would decrease over the period in association with the concurrent general hemispheric warming trend. However, no significant linear trends in area were found for any of the months analyzed over the period studied. Only October showed a significant linear temporal trend in NHCPV circularity (R c ), and that trend was toward increasing circularity over time. If global warming decreases the latitudinal temperature gradient, then decreased NHCPV circularity might be expected, but the presence of a linear trend in R c only in October leaves little convincing evidence for a global warming signal in the monthly 500 hpa NHCPV area and circularity. Interestingly, July was the only month to identify a correlation between the area and R c. The third research question stated that the NHCPV should be centered on the North Pole. However, April and July were the only two months to support this hypothesis. For most of the 215 months in the period (43 years, 5 months per year), the NHCPV position showed a strong displacement toward the Pacific basin, north of Alaska and Siberia. This result supports that which was found previously for January. Only April displayed a tendency for centroid displacement toward the Atlantic basin. Perhaps the greater increase in sea ice accumulation over the Atlantic as compared to the Pacific in late winter causes the baroclinic zone and NHCPV to shift equatorward over the Atlantic by April. By contrast, early in the winter, sea ice is already accumulated over the Pacific sector. Most of the months showed a low circular variability for the centroid location, suggesting that the position did not move drastically between 1959 and In winter, the intense cooling of the landmasses may possibly cause the zone of air mass contrast (and therefore also the midlatitude westerlies, polar front jet stream, and the NHCPV) to be displaced equatorward over this landdominated region. The PNA teleconnection pattern could be causing a shift in the NHCPV. Therefore, when the PNA teleconnection is in a positive or even in a neutral phase, the trough in the Alaska to Siberia region may cause the NHCPV centroid to migrate in that direction. The NNI analysis suggested that the point pattern of NHCPV centroid is random in April and October, while the centroid tends to exhibit a more preferred location in winter and summer. Identifiable spatial clusters of at least a few months within the period were found in most of the monthly analyses, with no identifiable spatial clusters in December and April. The winter and autumn clusters were located north of Alaska and Siberia, while the summer cluster was north of Greenland. The research illustrated that the winter and autumn NHCPV centroid seems to be positioned similarly, while the spring NHCPV position seems to differ from the other seasons, perhaps because of increases in accumulation of sea ice in late winter in the Atlantic sector. Relationships between teleconnections and NHCPV variability throughout the year were also analyzed. It was expected that correlations between the indices of the teleconnections studied and the area and R c of the NHCPV would be found. Results showed that the SO (as represented by the SOI) is the only teleconnection that did not correlate with area or R c in any of the months analyzed. Another indicator of ENSO, the Niño 3.4 index, was significantly (negatively) correlated only with the July area. By contrast, Pearson correlations identified several significant relationships between the NHCPV and the PNA pattern, including a positive correlation to NHCPV area in December and February, and a negative correlation with R c in December and October (along with the previously identified January). Likewise, the AO shows a strong relationship to the NHCPV in several months, including a negative correlation with the area in December, February, and April (along with January) and a positive correlation with the R c in December. A closely related teleconnection to the AO is the NAO, but the NAO was negatively correlated with the area only in December, February, and April. Finally, the PDO was significantly negatively correlated with the October R c. Thus, this hypothesis was partially confirmed; some links to the well-known modes of low-frequency variability in atmospheric flow were found. The results of this research provide further questions that should be investigated. For example, more months could be analyzed to represent the spring, summer, and autumn seasons more comprehensively than in this study. Also, daily analysis of the geometric properties of the NHCPV would provide more insight into the nature of spatial and temporal variability at submonthly timescales. Likewise, the addition of more vertical levels of analysis would provide further information about the degree to which the NHCPV acts as a barotropic system. Finally, additional analysis should be conducted

15 NORTHERN HEMISPHERE CIRCUMPOLAR VORTEX 711 to investigate possible forcing mechanisms for the longterm centroid migration of the NHCPV and associations to teleconnections. Because broad-scale atmospheric circulation impacts environmental surface features such as temperature and precipitation and a wide range of dependent environmental conditions, it is important to understand the characteristics of the circulation (Davis and Benkovic, 1994). Results from this research provide clues to understanding the puzzle of long-term atmospheric circulation variability. After the characteristics and processes behind broadscale atmospheric circulation are understood, the range of possible impacts on social systems, and communities impact on climate from macro- to local-scale, can begin to be understood. Eventually, such knowledge of this cascade of linkages (Terjung, 1976) can contribute to the development of physical-human-process-response systems the highest level of scholarship in geographical climatology (Terjung, 1976). APPENDIX Area, circularity ratio, and coordinates of centroids for monthly mean northern hemisphere circumpolar vortex, Year Month Area (km 2 ) R c latitude ( N) longitude W E W W W E E E E W E W E W E W E W W W W E W E W E W W W W W E Year Month Area (km 2 ) (Continued) R c latitude ( N) longitude W W W E W E W W E W E E W W E E W W E E E E W W W W E E E W W W E W W W E W E W W W E W W E W E W W E W E W W W W E (continued overleaf )

16 712 K. M. WRONA AND R. V. ROHLI (Continued) (Continued) Year Month Area (km 2 ) R c latitude ( N) longitude Year Month Area (km 2 ) R c latitude ( N) longitude E W E W W W W W E W E W E W W W W W E W E E E E W W W W W E W W W W E W E E E E E W W W W W W W W E W W W W E W E E W E E W W E W E E W W W E E W E E E E W W W E E W W W W W W W W W W W W W W W W W W W W W W W W W W W E W W W W W