Solar influence on the spatial structure of the NAO during the winter 1900–1999



[1] The solar influence on the North Atlantic Oscillation (NAO), reported in a previous study based on 39 years of data, was confirmed by using 100 years of historical data. During the low solar activity winters, the NAO signal in the sea level pressure is confined in the Atlantic sector, while, during the high solar activity winters, NAO-related anomalies extend over the northern hemisphere, in particular over the polar region and Eurasian continent. It was also found that the solar influence on the NAO originates from the Eurasian continent in early winter. In addition, two prolonged periods of high NAO index in the early and late 1900s showed spatial structures that were quite different and similar to those that occur during the low and high solar activity winters, respectively.

1. Introduction

[2] The North Atlantic Oscillation (NAO) has been regarded for a long time as a dominant mode of the atmospheric variability in the Atlantic sector [e.g., van Loon and Rogers, 1978]. In a previous study using 39 years of NCAR/NCEP reanalysis data [Kalnay et al., 1996], the spatial structure of the NAO was reported to differ significantly, according to the phase of the solar cycle [Kodera, 2002, hereinafterafter referred to as K02]. During the maximum phases of the solar cycle, the NAO has a hemispherical structure extending into the stratosphere, while, during the minimum phases, the NAO is confined to the eastern Atlantic sector in the troposphere. The analysis period is, however, relatively short and covers fewer than four complete solar cycles.

[3] In the past, the relationships between the solar activity and the surface climate detected in shorter datasets have usually disappeared after a few more solar cycles. A number of such examples are illustrated in a review paper by Pittock [1978]. It should be noted that the solar relationship investigated here is not a simple linear relationship as that documented in the paper mentioned above, but, rather, a kind of modulation effect that is similar in some ways to that found by Labitzke [1987] in the winter stratosphere. In any case, it is crucial to investigate whether the relationship found in K02 can be consistently found for a longer period. For this, historical datasets are used, and the period of analysis is extended for 100 years from 1900 to 1999.

[4] To describe an average feature during the cold season, the extended winter (DJFM) mean is usually used. The NAO index in early winter exhibits somewhat different characteristics; however, on the other hand, to reveal the evolution in time, the investigation was conducted based on two three-month periods: early (NDJ) and late (JFM) winters.

2. Data

[5] The NAO index was calculated by Hurrell [1995] as a difference in the normalized monthly mean sea level pressure (SLP) between Lisbon and Stykkisholmur. Surface temperature (combined land surface air temperature and sea surface temperature) data were compiled by Jones [1994]. These two datasets are the same as those used in K02. The monthly mean SLP dataset was compiled by Trenberth and Paolino [1980]. Sunspot numbers (SSPs) are used as an index of solar activity.

3. Results

[6] Figure 1 shows standardized time series of the (a) winter (JFM) mean NAO index and (b) winter mean (NDJF) SSPs from 1900 through 1999. One hundred years of data covers nine complete solar cycles. High solar activity (HS) and low solar activity (LS) years are defined whether the sunspot numbers are above or below the mean value. During the 20th Century, the amplitude of the solar cycle varied quite significantly. For a small solar cycle, the maximum value did not reach the mean value of the largest solar cycles. Therefore, in the present study, the years are denoted as HS and LS, rather than as the maximum and minimum phases. The HS winters are marked with open circles in the figure. Of the total number of winters, 41 are HS and 59, LS. The different numbers between the HS and LS result from the fact that the solar cycle variation of the SSP is not sinusoidal but remains for a longer time in a low activity state.

Figure 1.

Standardized indices: (a) JFM-mean NAO index, (b) NDJF-mean relative sunspot numbers. Open circles indicate the high solar (HS) activity winters.

[7] Figure 2a shows the lagged correlation coefficients between the JFM-mean NAO index and NDJ-mean SLP at each grid point. The simultaneous correlations between the JFM-mean NAO index and the JFM-mean SLP at each grid point is displayed in Figure 2b. The left- and right-hand panels are for the LS and HS winters, respectively. The correlation coefficients corresponding to the 95%-significant levels are 0.28 and 0.3 for the HS and LS winters, respectively. The correlation pattern shows a compact north-south seesaw over the Atlantic Ocean for the LS winters, while, for the HS winters, the NAO-related variability extends over larger regions of the northern hemisphere (NH), in particular over the polar region and the Eurasian continent.

Figure 2.

(a) Lagged correlation coefficients between the JFM-mean NAO index and NDJ-mean SLP at each grid point for the period 1900–1999. (b) Same as in (a), but for the simultaneous correlation with JFM-mean SLP at each grid point. (c) and (d): Same as in (a) and (b), but for the correlation with the surface temperature at each grid point. The contour interval is 0.1, and absolute values below 0.4 are omitted. Negative values are indicated by dashed contours, and positive values are shaded.

[8] The largest difference between the HS and LS winter occurs over the Eurasian sector. The correlation coefficient between the Mediterranean region (20°E, 35°N) and Siberia (120°E, 65°N) is −0.39 for the total period. However when the data are partitioned into two parts, high negative correlation (−0.77) appears during the HS period, while no correlation (−0.07) is found during the LS period. A possibility that it occurs by chance is examined by randomly dividing the data into two parts. The result of 100,000 trials indicates that the chance is lower than 0.07%.

[9] The difference in the characteristics of the NAO between the LS and HS winters is not limited to spatial structure. The evolution over time is significantly different, too. The lagged correlation map (Figure 2a) between the JFM-mean NAO index and NDJ-mean SLP indicates that, prior to the formation of the NAO-dipole pattern over the Atlantic sector in late winter (Figure 2b), a dipole-type anomaly already exists over the Eurasian continent in early winter for the HS case. On the other hand, a dipole-type anomaly is formed and develops over the Atlantic Ocean for the LS case.

[10] Figures 2c and 2d show the same correlation maps as those in Figures 2a and 2b but for the surface temperature. During the late winter of the LS case (Figure 2d-left), positive correlations occur over Europe and the east coast of America, and negative ones occur along the offshore of West Africa. In spite of a large area of missing data, strong positive signals can still be seen across the Eurasian continent during the late winters in the HS case (Figure 2d-right). It should be noted that high-temperature anomalies are also found in Siberia in early winter prior to the formation of the NAO pattern in the SLP field over the Atlantic Ocean (Figure 2c-right). During the HS late winter, high SLP anomalies in the subtropics extend from the Atlantic Ocean to the Mediterranean Sea. Accordingly, while a low temperature signal is seen in the eastern Atlantic during the LS winters, it appears to be more eastward in West Asia during the HS winters. The present results based on the historical datasets are essentially the same as those found in K02.

[11] The lagged correlation in Figure 2 suggests that the differences in the spatial structure originate from the early winter. In particular, the NAO pattern in the HS winters develops from the Eurasian continent. Accordingly, to investigate this in more detail, a Eurasian index (EUI) is introduced to represent a seesaw pattern in the SLP between the Mediterranean and Russian regions. The EUI is tentatively defined as a difference in the normalized SLP at 30°E, 35°N and 60°E, 55°N. Figure 3 shows lagged correlation coefficients between the DJ-mean EUI and the two-monthly mean SLP at each grid point for, from top to bottom, ND, DJ, JF, and FM. The correlation maps are displayed for the longitudinal range from 120°W to 120°E.

Figure 3.

Similar to Figure 2b, but for the lagged correlation coefficients between the DJ-mean Eurasian index and two month-mean SLP at each grid point: from top to bottom, ND, DJ, JF, FM. The number on each panel indicates the lag in the month.

[12] During the LS winters, there were no anomalies observed outside the Eurasian continent. In HS winters, negative anomalies over Russia expanded toward the polar region and Canadian sector in the DJ from a pair of positive and negative anomalies over the Eurasian continent in the ND. Positive anomalies then develop over the Atlantic subtropics in JF. Finally, the dipole-type anomaly over the Eurasian continent declines and transforms into a regional NAO dipole pattern over the Iceland-Azores sector of the Atlantic.

[13] To examine whether the solar relationship is stable, the same analysis as in Figure 2 is repeated by dividing the dataset into two equal 39-year periods of 18 HS and 21 LS winters: (A) 1922–1960 and (B) 1961–1999 (see Figure 1b). The first period of 21 years of very low solar activity is excluded in order to keep a solar cycle condition similar in order to use comparable solar cycle conditions from both periods. This low solar activity period at the beginning of the 1900s will be discussed later.

[14] The correlation coefficient between the NAO index and the SLP during the late winter for the two above-mentioned periods (A) and (B) is shown in Figures 4A and 4B. The left- and right-hand panels display the LS and HS cases, as in Figure 2b. In spite of some differences in the spatial pattern between the two periods, a similar solar influence emerges in both periods. The high-latitude negative anomalies are larger and extend over the polar and Eurasian regions, and the subtropical positive anomalies extend further inside the continents during the HS winters than they do in the case of the LS. Thus, in the HS winters, the NAO pattern is more hemispherical. In the earlier period (A), however, the Atlantic anomalies are more connected to those in the American/Pacific sector, while they are more related with the Eurasian variability in later period (B).

Figure 4.

The same as in Figure 2b, but for the correlation coefficients between the NAO index during the two 39-year periods: (A) 1922–1960 and (B) 1961–1999.

[15] The major difference in the previous study occurs over the mountainous region in China. In K02, a significant correlation occurs in the HS winters; however, in the present analysis of a similar 39-year period (1961–1999), no significant correlation was noted. The difference could be due to the extrapolation under the ground used for the calculation of the SLP. This discrepancy, however, does not affect the major conclusion of the present study.

4. Discussion

[16] The present study using 100 years of data confirms the results of the previous study by K02 based on 39 years of data. The stability of the relationship is also tested by dividing the data into two 39-year periods. During the LS winters, the NAO pattern fits more or less a classical image of the NAO, which is a seesaw of the SLP over the Atlantic Ocean between the regions of the Icelandic Low and the Azores High. On the other hand, during the HS winters, the NAO signals originate from the Eurasian sector and expand from the Siberian to the polar region to form a classical NAO pattern over the Atlantic Ocean at the end of the sequence (see Figure 3). Thus, when the winter mean is taken, the spatial feature of the NAO in HS winters exhibits a pattern that is more hemispherical.

[17] Because the upper air data are not available, the difference in vertical structure found in K02 cannot be verified. However, the evolution of SLP anomalies form the Eurasian to the Atlantic sector during the HS winter is quite consistent with the stratospheric involvement reported in K02, in particular, with the downward propagation of stratospheric zonal wind anomalies [e.g., Kodera and Koide, 1997; Kuroda and Kodera, 1999].

[18] The relationship between the NAO index and the warming of Eurasia has been noted [e.g., Hurrell, 1995]. This, however, does not necessarily mean that the NAO has an impact on Eurasia. A warming of Siberia can be seen prior to the formation of a dipole pattern over the Atlantic region (Figure 2a). In fact, the anomalous SLP pattern associated with the warming of Siberia [Rogers and Mosley-Thompson, 1995, Figure 2] more closely resembles the situation before the formation of the NAO dipole pattern over the Atlantic Ocean (Figure 3-right in DJ or JF). When such a precursor of the NAO is absent over the Eurasian continent, no warming is seen in the Asian sector in late winter, although warming appears in Europe (Figure 2-left).

[19] The end of the 20th Century is known as a period of prolonged high-positive NAO (H-NAO). In addition to this, another H-NAO epoch occurred at the beginning of the 20th Century (Figure 1a), during which Europe experienced mild winters [Rogers, 1985]. In contrast to the recent epoch, the solar activity was quite low in the early 1900s. Therefore, it is of interest to examine how the structure of the NAO differs between the two epochs.

[20] Three-month mean SLP for the two H-NAO epochs of 26 years, (a)1903–1928 and (b)1973–1998, are displayed in Figure 5 as anomalies from the climatology. The top and bottom panels show NDJ and JFM means, respectively. Shading indicates the region where the confidence level exceeds 95%. According to the H-NAO conditions, a pair of low- and high-pressure anomalies is found in late winter (JFM) over the Icelandic-Low and Azores-High regions in both epochs. In the more recent epoch, high-pressure anomalies extend over Europe and regions of the Mediterranean Sea, and low-pressure anomalies cover the polar region. However, during the early epoch, the positive anomalies did not extend inside the Eurasian continent, and the negative ones were confined to the Atlantic sector of the polar region. The differences are more distinct in the early winter. In (a), a weaker dipole pattern was noted over the Atlantic Ocean, whereas in (b), the dipole pattern was located over the Eurasian continent.

Figure 5.

Three-month mean SLP anomalies from the climatology averaged over a 26-year period: (a) 1903–1928 and (b) 1973–1998. Top and bottom panels are for early (NDJ) and late (JFM) winter means, respectively. The contour interval is 0.75 hPa, and negative values are indicated by dashed contours. Light and dark shading indicate the region where the confidence level of the negative and positive anomalies exceeds 95%, respectively.

[21] These differences in spatial structure and seasonal evolution are quite similar to those between the LS and HS years in Figure 2. In recent times, the solar activity was high, but, at the beginning of the 20th Century, it was quite low (Figure 1). Therefore, the difference in the spatial structure of the NAO between the two periods was consistent with the difference in the solar activity (Figure 2). It should be noted, however, that during the 20th Century, large changes occurred also in the Pacific sector [Trenberth and Hurrell, 1994]. Similarly, decadal changes in the equatorial SSTs in the Pacific-Indian Oceans could have had a large impact on the NAO [Hoerling et al., 2001]. Thus, other possibilities than a change in solar activity should also be examined. In any case, it would be important to investigate the different spatial structure and the seasonal evolution of the NAO.


[22] The author thanks the following persons and organizations for making the data available: original temperature and SLP datasets were produced by P. H. Jones and Trenberth and Paolino, respectively. They were compiled and distributed by T. Mitchell from JISAO. Sunspot numbers were obtained from the Solar Data Analysis Center at NASA/GSFC. NAO Index data were provided by the Climate Analysis Section, NCAR.