Temperature information on the lower artic stratosphere is now available for 47 winters (since 1955/56), and information on the "main state" of the stratospheric winters is available since 1952 (51 winters) when Scherhag discovered the Stratospheric Mid-Winter Warmings. It is well known that the interannual variability of the stratospheric circulation during the arctic winters is very large, that very cold as well as very warm winters are observed and that it is very difficult to identify temperature trends.
In this report which accompanies the Berlin Data Series on this CD, we want to summarize the observations during the last 51 winters, based mainly on the historical daily analyses of radiosonde observations made at the Freie Universität Berlin (FUB) which is the longest data set available. To describe the general state of the Arctic since the winter of 1955/56, Table 5, we are using the monthly mean 30-hPa temperatures over the North Pole as a key parameter which is highly correlated with the Arctic Oscillation (AO) (Baldwin and Dunkerton, 1999) and with the EOF 1 (Hoffmann, 2001). The data of the last winter (2001/2002) are from the improved analyses of the ECMWF.
The clearest change in temperature took place in November and the trend for the whole Period is -1.0K/decade, with a probability of 99%, Fig. 1. A similar trend emerges if the data series starts in 1964, not shown. Starting with the winter 1981/82, almost all temperatures in Table 5 are marked with a C which is based on the 30-year mean, but the average of the last 23 years (since 1979) is with -70.9°C 2K colder than the 30-year mean, and the definition for C should therefore be lower, too. At the same time, the Canadian Warmings which in November/December took place in 18 of the 31 years until 1981/82, are missing during the last 20 winters, except for 4 cases. This hints to a change in the meteorological conditions of the troposphere, as Canadian Warmings are connected with the intensification of the stratospheric Aleutian high which in turn is connected with the Aleutian low in the troposphere.
It is of interest to note that the trend was much weaker during the last 23 years (1979-2001), black trend line in Fig. 1, see discussion for December.
In December the trends start to be impossible to define, because of the very large interannual variability, Fig. 2. Certainly, one cannot find a clear negative trend, and it even appears that the trend is changing to a positive trend during the last 23 Decembers (black trend line in Fig. 2). During this period there have been sometimes very cold months, but also 2 Major Mid-Winter Warmings took place so early in winter, a feature not observed before. In total for the 47 winters the monthly mean average for December remains the same as for the first 30 years and the overall trend is zero; see also Table 5.
Figure 1: Time series of the monthly mean 30-hPa North Pole temperatures in November, 1955-2001. Different trends are indicated: The first 25 years until 1979 in red; the last 23 years starting in 1979 in black, and the whole period in blue; the respective informations on the trends, probabilities, means and standard deviations are given on the left hand side of the graph. Data: Free University Berlin, 2001: ECMWF
Figure 2: Time series of the monthly mean 30-hPa North Pole temperatures in December, 1955-2001. Different trends are indicated: The first 25 years until 1979 in red; the last 23 years starting in 1979 in black, and the whole period in blue; the respective informations on the trends, probabilities, means and standard deviations are given on the left hand side of the graph. Data: Free University Berlin, 2001: ECMWF
Table 5: Monthly Mean 30-hPa North Pole Temperatures in °C. RJ: monthly mean of the sunspot number in January; QBO: the phase of the Quasi-Biennial Oscillation (determined using the wind between 50- and 40-hPa (45-hPa) in January and February); FW: indicates the timing of the Final Warmings. CW stands for Canadian Warmings; * indicates the occurrence of a Major Mid-Winter Warming; C stands for a cold monthly mean (about half a standard deviation below the average). After the winter 1984/85 the 30-year mean is given with the standard deviation. At the bottom of the table: the longterm mean, the standard deviation, the linear trend (K/decade) for the whole period (n=47) and the confidence level in %
In the 34 winters from 1952 to 1985, 6 Major Warmings took place in January and 8 in February; there were no major warmings in December. During the last 17 years, we observed only 1 Major Warming in January -- but 3 in December(!) and 5 in February. It is tempting to speculate that there were significantly fewer Major Warmings in mid-winter during the last 17 winters, but if one counts these events for the 3 winter months (Dec., Jan., and Feb.), the result shows that there was no change in the numbers of events: 14/34 years and 9/17 years. Really striking is the fact that there was a period of 7 winters without Major Warmings.
The time series of the monthly mean 30-hPa North Pole temperatures in January is given in Fig.3, again with the indication of three different linear trends. Different trends (positive and negative) can be derived, depending on the interval selected, but the variability is so large that no significant trend can be established, even with 47 years of data.
February is the month with the largest variability and Fig. 4 shows as an example maps of the deviations of the monthly mean 30-hPa temperatures from the 30-year mean (1965-1994) for one of the coldest (1997) and one of the warmest (2001) Februarys.
The typical signature of the Arctic Oscillation (AO) or the EOF 1 is obvious with the opposite changes between high and middle latitudes.
Figure 3: Time series of the monthly mean 30-hPa North Pole temperatures in January, 1956-2002. Different trends are indicated: The first 25 years until 1979 in red; the last 23 years starting in 1979 in black, and the whole period in blue; the respective informations on the trends, probabilities, means and standard deviations are given on the left hand side of the graph. Data: Free University Berlin, 2002: ECMWF. (Labitzke and Naujokat, 2000, updated)
Figure 4: Deviations of the 30-hPa monthly mean temperatures from the 30-year mean (1965-1994) for one of the coldest (1997) and one of the warmest (2001) Februarys. Data: Free University Berlin; (Naujokat, 2001)
Another example of the large differences between the winters is given in Fig. 5. Here, we show for the 30-hPa level daily zonal means of zonal winds and temperatures, in a time/latitude section. The cold, undisturbed winter of 1999 at the top and the warm, disturbed winter of 2000/01 at the bottom.
Figure 5:Meridional time section of daily mean zonal winds (m/sec) and temperatures (°C) for two different winters: (top) Cold Winter: 1999/2000; (bottom)Warm Winter:2000/2001. Data: Free University Berlin
Factors influencing the state of the stratospheric circulation are the Southern Oscillation (SO), the Arctic Oscillation (AO), the QBO and the 11-year solar cycle; (a detailed discussion of these different influences is found in: Labitzke, 1998 and Labitzke and van Loon, 1999). An example of the influence of the Southern Oscillation is given in Fig. 6, where the 30-hPa height and temperature anomalies in February are shown, for 4 warm events and 4 cold events of the SO. Typically, the warm events are connected with Major Mid-Winter Warmings, compare with Table 5, and the cold events with an undisturbed cold polar vortex (van Loon and Labitzke, 1987). This is valid during solar minimum and without large volcanic eruption (Labitzke and van Loon, 1989).
Figure 6:Top: Deviations of 30-hPa monthly mean temperatures (K) in February. Below: Deviations of 30-hPa monthly mean heights, February (decameters). Both are deviations from the February mean of the years 1965-1974. Left: for the 4 Warm Events 1966,1973,1977, 1987. Right: for the 4 Cold Events 1965,1974,1976,1997. Data: Free University Berlin. (Labitzke, 1998 and Labitzke and van Loon, 1999)
The Arctic Oscillation is very closely related to the changes of the stratospheric arctic. This has been shown by comparing the AO, the Principal component (PC) of EOF 1 and the march of the daily 30-hPa heights at the North Pole (Hoffmann, 2001). Based on our experience as synoptic meteorologists the North Pole point was chosen by us since the 1970s as a characteristic parameter for describing the state of the arctic stratospheric circulation (Naujokat and Labitzke, 1993). Now we can use the same values but rename it Arctic Oscillation or EOF 1, to follow the more modern technologies and language.
The correlations between daily values of the PC 1 and of the 30-hPa heights at the North Pole are usually above 0.7, and in many winters near 0.9. In Fig. 7 this correlation is shown for 4 winters, with 3 disturbed and 1 undisturbed case.
Figure 7: Comparison between the time series of the first Principal Component (PC 1) and the 30-hPa height anomalies at the North Pole (FUB data) for 4 selected winters. Blue = time series of PC 1, inner scale; red = anomalies (geopot. m), outer scale. (Hoffmann, 2001, Abb. 4.37)
The QBO determines the character of the early winter, leading to a colder and more stable vortex in December and January during the west phase of the QBO and to a more disturbed and warmer Arctic during the east phase of the QBO. The solar cycle influences the latter part of the winters when a clear difference is observed between periods of high and low solar activity. During high solar activity the winters in the west phase of the QBO tend to be disturbed and are often connected with Major Mid-Winter Warmings. One example of these differences is shown in Fig. 8.
Figure 8: Left: Correlations between the 10.7cm solar flux (the 11-year solar cycle) and 30-hPa heights in February, shaded for emphasis where the correlations are above 0.5; upper panel: years in the east phase of the QBO; lower panel: years in the west phase of the QBO. Right: Respectively, height differences (geopot. m) between solar maxima and minima (1958-2001). Data: Free University Berlin. (Labitzke, 2002)
Table 6: RJ: monthly mean of the sunspot number in January; FW: indicates the timing of the Final Warmings. CW stands for Canadian Warmings; * indicates the occurrence of a Major Mid-Winter Warming; C stands for a cold monthly mean (about half a standard deviation below the average)
The different events given in Table 5 are shown again in Table 6, but grouped according to the phase of the QBO. This makes a comparison between the different groups easier. The [C] marks a cold month due to the late winter cooling after a Major Warming.
If one compares the last 17 winters with the earlier years since 1956, one cannot find a convincing sign of mid-winter temperature change (cf. Fig.3). One feature is, however, worth mentioning: the late winters from 1993 to 1996, i.e. 4 winters in line belonged to the group of the QBO westerlies. This did not happen before when never more than 2 consecutive winters fell into the same group. In addition, during these winters the solar activity was low, and this combination agrees with the fact that we observed cold winters, all through February. During this period of the winter the interannual variability is, as discussed above, extremely large and therefore no significant trends can be determined, see Table 5, bottom line.
March is an important time of the winter, as a long lasting undisturbed cold vortex is a meteorological situation in which ozone can be destroyed over the Arctic, while earlier breakups of the vortex are connected with ozone transport into the Arctic. The spring of 1997 was the coldest within our series of 47 winters.
But one should not imply a clear trend here, as the overall trend in March in the Arctic is practically zero, Fig.9. This Figure may serve as a warning: if one uses the complete series since 1956 to calculate a linear trend, the trend is zero. If one uses the first 25 years, the trend is positive and statistically significant, whereas it is negative and, again, statistically significant from 1979 to 1997 (!). In other words, the trend depends on how one cuts the cake (Labitzke and van Loon 1999).
In accordance with the large variability of the North Pole temperatures in March, Fig.9, the timing of the spring change over shows a very large variability, Fig.10. Here, the dates are given when the 30-hPa temperature gradient between 80°N and 50°N has changed to summer conditions, i.e. high latitudes being warmer than middle latitudes.
In general, Major Warmings (*) in January or February lead to a "late winter cooling" [C] and a delayed transition into summer, while Major Final Warmings (*FWs) are a manifestation of an early transition, Tables 5 and 6. Again, the overall trend (1965-2001) in Fig.10 is not significant but indicates a delay of the timing of the spring change over of 3.5 days/decade. But one can easily find an advancement of the transition from the beginning of the data until 1989: During the first 9 years (1965-1973) no *FWs occurred and the transition was mostly late; but then a series of *FWs occurred (cf. Table 5), they took place in 10 years out of the 16 year period 1974 till 1989. After this winter no more *FWs have been observed until 2002 and the transition was delayed, particularly during the recent winters after major warmings (*). Note that in the 9 year period from 1956 till 1964 three *FWs took place and the transition was, of course, early then, cf. Tables 5 and Fig.9, and we had also a delay of the spring reversals from that period till the late 1960s.
This variability and the occurrence of the *FWs is connected with large planetary waves propagating from the troposphere into the stratosphere and especially with the intensification of the Aleutian High, and there is no explanation, so far, about its variability, similar to the occurrence or non-occurrence of the Canadian Warmings, as dicussed above; (see also Waugh et al., 1999 and Langematz et al., 2002).
Figure 9: Time series of the monthly mean 30-hPa North Pole temperatures in March, 1956-2002. Different trends are indicated. Data: Free University Berlin, 2002: ECMWF. (Labitzke and Naujokat, 2000, updated)
Figure 10: Time series of the days of the "spring change over" at the 30-hPa level, expressed by T(80°N) - T(50°N) >= 0 K. Different linear trends are indicated. Data: Free University Berlin, 2002: ECMWF
The coldness, the stability and size as well as the duration of the polar vortex, that is the timing of the "spring change over" are the key parameters which determine the formation and duration of the PSCs (Polar Stratospheric Clouds) and therefore the potential for and duration of ozone destruction over the Arctic. This problem was dealt with by us in several publications (first by Pawson et al., 1995) and 3 relevant figures are given here.
Figure 11 shows in the upper panel the lowest 30-hPa temperatures observed over the Northern Hemisphere during winter since 1965, together with the march of the daily minimum temperatures during the last 3 winters (1998/99, 1999/2000 and 2000/2001) of the Berlin Data Series. The temperatures relevant for the formation of PSCs, Tnat and Tice are indicated. In the lower panel the area (Amax in % of the N.H.) covered with temperatures lower than 192K ( i.e., Tnat: -81°C) is shown, again for the whole period since 1965 and with the daily data of the last 3 winters. Obviously, the potential for the formation of PSCs was very weak during the winter 1998/99 when 2 Major Warmings took place (cf. Table 5), weak during the winter 2000/01 with 1 Major Warming, but very large during the very cold winter of 1999/2000, Fig.2 and 3. The size of the area covered by the relevant low temperatures (here 50-hPa temperatures below 195K, in % of the N.H.) is shown in Fig.12 for 10 winters. Again, the large variability is striking.
Figure 11: Time series of daily (top) Tmin and (bottom) Amax at 30 hPa obtained from the FUB analyses for 1998/99 (red), 1999/2000 (green), and 2000/01 (black); the blue shading shows the lowest Tmin or highest Amax reached in the period 1964/65 to 1997/98. (Labitzke and Naujokat, 2001, update of Fig.1 in Pawson and Naujokat, 1999)
Figure 12: Time series of the area (in % of the Northern Hemisphere) covered with temperatures lower than Tnat ( 195K at 50 hPa) between November 1 and March 31 for 10 winters. (Naujokat and Kunze, 2000, update of Fig.3 in Pawson and Naujokat, 1999)
Figure 13 shows the same variability for the whole period, 1965/66 till 2000/2001, for the 50- and 30-hPa level. Here, the number of days with temperatures below the PSC formation threshold and the integral of the size of the areas for their formation are given.
Figure 13: Number of days with temperatures lower than Tnat (white) or Tice (red) at (a) 50 hPa and (b) 30 hPa. Integral for each season of the area (in % of the Northern Hemisphere) which had temperatures low enough for the formation of PSCs, (c) at 50 hPa and (d) at 30 hPa. ( Fig.2 in Pawson and Naujokat, 1999, updated)
The temperature trend described above for the 30-hPa level is similar for the lower stratosphere, e.g., the 50-hPa level. But if one goes about 7 km higher, to the 10-hPa level where FUB-data are available since the winter 1964/65, the trend appears to change, and February is shown as an example, Fig.14. The trend is now positive, +2.1 K/decade. This agrees with positive temperature trends during winter reported from rocketsonde data (Chanin and Ramaswamy, 1999) and with model results which indicate only weak cooling in the middle stratosphere above a cooling in the lower stratosphere in connection with the observed ozone decrease (Langematz, 2000). But caution is recommended, as the variability is, as stressed before, very large and any trend can be derived, depending on the beginning and the end of the time series.
The 10-hPa North Pole temperatures are given in Table 7. The values in brackets are preliminary, as the daily analyses are not finished and not digitized and therefore they are not included in the data series of the CD. But they are interpolated from daily analyses and can be considered correct for the purpose of analysing trends. (E indicates that the data are taken from the analyses of the ECMWF).
Table 7: Monthly Mean 10-hPa North Pole Temperatures in °C. RJ: monthly mean of the sunspot number in January. At the bottom: the longterm mean, the standard deviation, the linear trend (K/decade) and the confidence level in %
Major Mid-Winter Warmings occur mostly in January-February. In addition to warming of the north polar region and reversal of the meridional temperature gradient, they are also associated with a breakdown of the polar vortex, which is replaced by a high. That is, the definition of a Major Warming requires not only the warming but also a total change of circulation. The definition of a breakdown of the polar vortex is that the usual westerlies in the Arctic at 10 hPa are replaced by easterlies so that the centre of the vortex moves south of 60 - 65°N. The vortex is either displaced entirely or split into two. This type of warming has not been observed in the Antarctic.
Minor Warmings can indeed be intense and sometimes also reverse the temperature gradient, but they do not result in a reversal of the circulation at the 10-hPa level. They are found in the Antarctic as well.
Canadian Warmings happen in early winter. They take place when the Aleutian stratospheric high intensifies and moves poleward. The Canadian Warmings can reverse the meridional temperature gradient and sometimes briefly change the zonal wind direction over the polar cap, but nevertheless they do not lead to a breakdown of the cyclonic polar vortex.
Final Warmings mark the transition from the cold cyclonic vortex in winter to the warm anticyclone centered on the pole in summer. Their intensity varies much and they can be divided into major and minor Final Warmings. The time when the Final Warmings take place - when the westerlies of winter are replaced by the easterlies of summer - also varies a good deal, so they are further divided into early and late Final Warmings. Naturally, there are also Final Warmings in the Southern Hemisphere (Labitzke and van Loon, 1999).
Stratospheric Research Group Last modified: Wed Sep 11 22:15:26 MST 2002