In December 1999, Central Europe was hit by three violent storms, which claimed more than 130 lives and caused about 13 billion Euro of total economic losses. The first of these storms hit Denmark and the northernmost part of Germany on December 3, 1999 (Fig. 1a). It was associated with a cyclone named "Anatol" by the German Weather Service, DWD (the DWD gives a name to each North Atlantic cyclone). At the Danish and German North Sea coast, gales of more than 50 m s-1 and a record storm surge were observed. Extreme wind speeds also occurred in central and eastern Denmark, with gusts as strong as 43 m s-1 in Copenhagen. According to a report issued by the Danish Meteorological Institute (2000) this was the strongest storm of the century over Denmark.
The second storm event occurred on December 26. It was caused by a low named "Lothar", and left a trail of destruction from northwestern France to southern Germany and Switzerland (Fig. 1b). Major damages were reported from the industrial belt around Paris and from well-known tourist sites, like the church of Notre Dame in Paris and the park of Versailles. Gusts reaching wind speeds of almost 50 m s-1 were recorded at Paris-Orly airport. In Germany and Switzerland the persistently violent gusts (over 41 m s-1 in Karlsruhe/Germany and almost 37 m s-1 in Bern/Switzerland) uprooted countless trees.
Only one day later, on 27 December 1999, gales associated with a low named "Martin" hit southern France and northern Spain (Fig. 1c). Again extreme gusts over 40 m s-1 occurred, this time in a belt extending from Bordeaux to Clermont-Ferrand, causing numerous road blockages and a widespread electricity breakdown which lasted for days.
In this paper, we will first give a description of the development of each storm. The subsequent discussion of the large scale conditions in which the storms were embedded will reveal two common characteristics, namely a band of extremely high baroclinicity and high equivalent potential temperature. Evidence will be given that similar large scale anomalies can also be identified for previous disastrous storm events over Europe. Finally, we will compare the observed anomalies with trends in a greenhouse gas scenario run of a GCM. Our paper will not touch the issue of the predictability of the storms and of the timeliness of the warnings issued by the different weather services.
Station data and operational ECMWF analyses will be used to investigate the recent three storms, while NCEP reanalyses are the data source for the historical storm events. The analysis/reanalysis data are expected to give a good representation of the observed large scale conditions. We would like to point out that they fail, however, in representing small scale features like the extreme pressure gradients associated with Lothar (cf. the operational manual surface analysis issued by the German Weather Service (Fig. 1d) to the ECMWF numerical analysis (Fig. 1b)). This is not solely due to the moderate (60 km) resolution of the ECMWF grid. Rather, several observed extreme surface pressure tendencies and winds have been rejected by the ECMWF`s automatic data check and assimilation scheme (E. Andersson, ECMWF, pers. communication, 2000), for example the observations at Rouen, France, at 0600 GMT on 26 December 1999 (see Fig. 5). Without going into details, one should recognize that operational analyses shall serve a high quality forecast, which does not necessarily mean that observed extremes are incorporated.
Two parameters, the maximum Eady growth rate, σΒΙ, and the equivalent potential temperature, Θe, are considered to study the large scale atmospheric conditions in which the storms were embedded. The Eady growth rate (see Eady, 1949; Lindzen and Farrell, 1980; Hoskins and Valdes, 1990) is a measure of baroclinicity, i.e., it quantifies the large scale conditions for the potential growth of extratropical cyclones. It is defined as σΒΙ=0.31(f/N)|δv/δz| where f is the Coriolis parameter, N is the Brunt-Väisälä frequency, and δv/δz is the vertical wind shear. Since the gridded fields of σΒΙ calculated from three day averages of the input parameters for the lower (850 mbar to 700 mbar) and for the upper troposphere (500 mbar to 300 mbar) proved to be similar in structure, we will only show the upper level baroclinicity. The other considered parameter Θe at 850 mbar serves to quantify the combined effect of latent and sensible heat on the cyclones. It was computed on a 6 hourly basis according to a formula given in Bolton (1980).
The three recent storms are investigated using the high resolution ECMWF operational analyses. When comparing the recent storms to historical storm events and estimating the extremeness of the associated large scale conditions, we will use the lower resolution NCEP reanalysis for forty winters (1958/59-1997/98) as observational reference data. We performed an intercomparison of the considered large scale quantities for storm events included in both ECMWF and NCEP data, and found no significant systematic differences between them.
Before the cyclone Anatol rapidly developed over the central North Atlantic south of Greenland, two precursors could be identified. At the surface level, a small cyclonic vorticity perturbation moved northeastward from Newfoundland, meeting an upper air geopotential height disturbance approaching from Labrador (not shown). A closed isobar at the surface could be identified in the ECMWF analysis at 1800 GMT on 2 December 1999 when the system was located west of Ireland (for the cyclone paths see Fig. 2). The system moved on eastward along the southern flank of a steering cold mid-tropospheric trough centred over Iceland (Fig. 2a). A tongue of warm air was located at the southern flank of the track (Fig. 2d). During its displacement within a highly baroclinic zone (Fig. 2g), Anatol deepened rapidly over the eastern North Atlantic. According to a ship observation, it reached a central pressure of about 990 mbar when it approached the British Isles at 0000 GMT on December 3. Its rapid intensification continued over the subsequent 12 hours, fed by the large meridional temperature contrasts of the air integrated in the system. Extremely cold tropospheric air located in the area between Greenland and Norway (e.g., -39oC at the 500 mbar level over the Faroer Islands on 3 December 1999, 0000 GMT) was advected to the North Sea in the wake of the system, while warm air prevailed in the south (-20oC over central England and Ireland at the same pressure level, see Fig. 2a). Later, Anatol`s central pressure remained almost constant while it swept from the central North Sea (957 mbar, 1200 GMT) over the Danish North Sea coast (955 mbar, 1500 GMT) to Sweden (954 mbar, 2100 GMT). The lowest pressure was recorded at 1800 GMT (953 mbar) near the east coast of Jutland. Anatol filled slowly after crossing the Baltic Sea and reaching the Latvian coastline (4 December 1999, 0600 GMT, 970 mbar).
Fig. 1a shows that the largest wind speeds occurred south of the cyclone core in a belt reaching from the German and southern Danish North Sea coast to Copenhagen and the Polish Baltic Sea coast (cf. also Table 1). Notably, gales exceeding 25 m s-1 were observed over Denmark during the eight hour period from 1400 to 2200 GMT, with the most violent winds between 1600 and 1800 GMT at the North Sea coast (cf. Sylt, Table 1). Wind direction and temperature recordings suggest that the peak gusts occurred with the passage of the cold front.
According to the classical concept of cyclone development, the high baroclinicity indicated by the strong vertical windshear (Fig. 2h) is susceptible to the formation of an intense cyclone starting from the small initial disturbance. An upper air anomaly would form, and the coupled upper and lower tropospheric disturbance would be expected to grow (type A development after Pettersen and Smebye, 1971). However, such growth did not occur until Lothar had crossed the Atlantic with a propagation speed of about 120 km/h and reached Europe on December 26, 0000 GMT. The intense jet over the Atlantic remained almost unaltered during Lothar`s passage over the Atlantic Ocean.
For a better understanding of the suppressed development of Lothar, it is instructive to compare the vertical structure of the disturbance shortly before and after its explosive growth. The non-growth phase is characterized by a geopotential height anomaly which is largest close to the surface (Fig. 3a). The low level warm air advection upstream of the low (evident, for example in Fig. 2e) results in a positive temperature anomaly. In contrast to the idealized situation of a growing baroclinic wave (e.g., Holton, 1992, his Fig. 6.6), however, the vertical trough axis and the axis of the warmest air coincide and there is no backward vertical tilt. As a consequence, the amplitude of the geopotential height anomaly decreases with height, vanishing at the 500 mbar level. Correspondingly, the upward directed vertical winds given in the ECMWF analysis with a maximum at 700 mbar do not extend to the upper troposphere (Fig. 3a). Only six hours later, the "lid on top of the system" is removed and the system`s structure develops into the classical picture of a rapidly developing cyclone (Figs. 3b and 3c).
The dynamical reasons for the coincidence of temperature and height axis and the resulting "capping" shown in Fig. 3a are not clear. The effect is reminiscent of the mid-winter suppression of baroclinic wave growth over the Pacific (Nakamura, 1992). This reduction of mid-tropospheric baroclinic wave activity is associated with extremely high jet wind speeds. According to Christoph et al. (1997) the suppression is usually not active over the Atlantic, as zonally averaged jet wind speeds are usually below a threshold value (about 45 m/s for the monthly mean), but this is different for the present case. Another aspect might be that Lothar`s wavelength was too small for significant baroclinic growth.
Lothar's rapid development started when an additional factor set in at the time the storm reached the jet`s exit region at December 26, 0000 GMT (Fig. 4). A strong divergence centre associated with upper level diffluence was located between Brittany and Cornwall, north of the jet axis. It was already present at previous dates and thus was not generated by the approaching cyclone. Such areas of intense divergence north of the jet exit (induced by the ageostrophic winds associated with the wind speed reduction) are well known to induce rapid cyclone growth (e.g., Uccellini, 1990; Baehr et al., 1999). An additional amplification of the divergence in this region is expected from the secondary jet maximum (Fig. 4) reaching from England to the Alps (ageostrophic winds at the secondary jet`s right entrance region due to acceleration). When Lothar reached the French coast at 0600 GMT on December 26 its central pressure had fallen to 961 mbar, rising only slowly during its further movement over France to Germany (970 mbar at 0900 GMT) and Poland (980 mbar at 1800 GMT).
In general, the extreme pressure tendencies observed at stations along Lothar`s track were caused by the system`s high propagation speed (about 120 km/h over the Atlantic and 100 km/h over France and Germany) and its small diameter (Fig. 1d). The extraordinary pressure drop of almost 26 mbar/3 hours at Rouen/Normandy can be understood from the additional effect of the rapid deepening of the core pressure before 0600 GMT (Fig. 5). After Lothar`s passage, a pressure rise of the same order followed (28 mbar/3h) before 0900 GMT. Note that similar extreme pressure tendencies were also observed at other stations in northern France and the English Channel (Le Blancq and Searson, 2000). Gust wind speeds at Rouen reached 23 m s-1 (45 kn) during pressure falls and 39 m s-1 (76 kn) during pressure rise (Fig. 5). The occurrence of maximum wind speeds shortly after the time of minimum pressure is in accordance with a time series of 10 minute means at Karlsruhe, Germany. Widespread gales of more than 40 m s-1 hit France, Switzerland and Germany (cf. Fig. 1b and Table 1).
The next system, called "Martin", followed Lothar only one day later. As mentioned before, Lothar`s shallowness over the Atlantic had left the large scale situation nearly unchanged (Figs. 2f and 2i). Initially, Martin formed as a surface low on 25 December upstream of a longwave upper air trough over North America. The surface low moved northeastward along the trough`s flank, which itself was slowly progressing southeastward over the North American east coast. A transition phase followed on 26 December. The large scale trough over North America changed its structure. Its southerly part weakened while farther north a short upper air depression moved eastward from Newfoundland. During this transition phase Martin`s structure and track were ambiguous, but when it subsequently interacted with the eastward moving upper air depression (27 December, 0600 GMT), it moved rapidly across the Atlantic. In contrast to Lothar, Martin eventually modified the large scale airflow. Fig. 2c shows a situation when the change over the western Atlantic is evident, while the strong gradients still prevail in the east. In fact, the polar jet was extremely intense over the latter region. At 0000 GMT on 27 December the radiosonde at Brest (Brittany) measured a wind speed of 147 m s-1 at a height of 8138 m (Hontarrède, 2000), about 700 m below the 300 hPa level. Subsequently, Martin`s central pressure fell to 965 mbar by 27 December, 1500 GMT, just before it entered the European continent south of Brittany (Fig. 1c). Violent winds first hit the French Atlantic coast (Table 1), and, to a lesser extent, northern Spain (gust wind speeds for Spain have not been available to the authors). Westerly gales of more than 36 m s-1 were observed between 1500 and 2100 GMT at Bordeaux. The storm led to floodings at parts of the Atlantic coast. Subsequently, severe damage was reported along the track over southern France into the western Mediterranean.
Local wind speeds associated with the three storms were the largest on record at several synoptic stations at the affected parts of Europe. The synoptic developments of the storms are, however, different. Both Anatol and Martin were of typical horizontal scale and modified the large scale airflow in their wake. In contrast, Lothar was a small scale system (Fig. 1d) and did not have a strong effect on the large scale flow over the Atlantic. Its peculiarity was the "capping" over the Atlantic. Another difference regards the tracks of the systems. Lothar and Martin crossed the Atlantic at a rather southerly position, while Anatol developed at higher latitudes (about 54oN, Fig. 2a).
The storms revealed common characteristics regarding the large scale conditions leading to their development. The one aspect is the existence of a large band of very high baroclinicity (Figs. 2 g,h,i) at the path of the cyclones over the Atlantic, which extends to the European continent, the other is the existence of an airmass with high equivalent potential temperatures, Θe, south of the cyclone track (Figs. 2d,e,f) which could enhance cyclone growth, in particular due to the release of latent heat.
The extraordinariness of the large scale conditions associated with the three storms was quantified from the exceedance of the 95 and 99 percentiles for the reference period 1957/58- 1997/98. Figs. 2g,h,i indicate that for all three storms baroclinicity clearly exceeded the 95 percentile of the respective values for the winters (December, January, February) in the reference period. The two late December storms, Lothar and Martin, were even associated with a zonally elongated band of 99 percentile exceedance over the Atlantic. For computing the percentiles of Θe , its strong annual cycle has to be taken into account. Thus, for each storm a sample containing all 6-hourly values from the respective 40 weeks (centred on the storm date) in the period 1958-1997 was used. The Θe percentiles also reveals rather large positive anomalies for each of the three cyclones, with the source of the moist air apparently lying in the Sargasso Sea (Figs. 2 d,e,f). Exceedance of the 95 or 99 percentiles is, however, found only at a few grid points south of the track.
In order to put the December 1999 storms into a historical context, we have chosen three other intense storms that produced large damage over Europe (see, e.g., Lamb, 1991). Their tracks and the associated anomalies in baroclinicity and Θe were computed. The track of the cyclone which caused major damage in Denmark on 23 November 1981 is shown in Figs. 6a and 6d. As for the recent storms, the large scale fields reveal a zone of extraordinarily high baroclinicity located between Iceland and England (Fig. 6a), i.e. upstream of the region of largest wind speeds. High local Θe values also occurred in conjunction with the storm. They ranked as extraordinary over a rather large area around the British Isles (Fig. 6d).
The second storm considered for comparison hit northwestern France and England on 15-16 October 1987. The corresponding cyclone track was at a very southerly position before the low reached Europe (Fig. 6b). Again, a band of very high baroclinicity was present over the eastern North Atlantic. The track curved northward when reaching the European continent. Unusually high Θe values observed over a fairly large area (Fig. 6e) led to a dominance of latent heat release over the dry baroclinic instability process during the rapid intensification of the system (Shutts, 1990). According to Fig. 6b, high baroclinicity was not observed over soutwest England where the storm reached its maximum intensity.
The third storm considered is one of the series of European storms in February 1990. The damage occurred mainly in Germany and Switzerland on 1 March 1990. Like the other storms, this cyclone is associated with a band of high baroclinicity near the track of its core. However, local values do not exceed the 99 percentile. There is also an area of enhanced Θe, but only few local values are rated as extreme.
Summing up, we find that the existence of bands of extremely high baroclinicity and high values of Θe are typical for European storm events (see also Busch et al., 1998). In detail, there are differences in how extreme these bands are, differences in their relative importance, and in their location. The anomalies are, however, always found over the eastern North Atlantic, close to or extending into the European continent.
With respect to the scientific and public discussion on the anthropogenic climate change one should emphasize that the occurrence of the storms cannot be taken as evidence for an enhanced greenhouse gas effect. Firstly, such an attempt would be inappropriate from a statistical point of view (see Frei and Schär, 2001). Secondly, only the first of the three storms, Anatol, seems to fit the greenhouse gas signal recently described by Knippertz et al. (2000). Considering a particular greenhouse gas forcing experiment they found a rising number of extreme wind events in the northern parts of Europe, associated with an increasing number of deep lows and enhanced mean upper air baroclinicity over the northeastern Atlantic. In contrast, the model simulated a decrease in the number of extreme wind events and in baroclinicity for the regions where Lothar and Martin occurred. It should also be mentioned that low resolution climate models are unable to produce small scale cyclones and their associated wind fields.
Extreme winds were produced by the three cyclones that hit Europe (Denmark, Northern France to southern Germany and Switzerland, Southern France) in December 1999. For some of the recording stations, the observed gust wind speeds were of record-breaking intensity. For each event, there was a band of extremely high baroclinicity near the cyclone track over the Eastern North Atlantic and extending partly into Europe. High equivalent potential temperatures were observed south of the track. These large scale characteristics are common to European storm events, though the details of the synoptic development are rather different.
A number of station data were taken from the Berliner Wetterkarte issued by the Berlin Meteorological Institute. The ECMWF analysis was obtained with the permission of the German Weather Service. NCEP reanalysis data were provided by the German Climate Computing Centre (DKRZ). Information on damages and casualties was taken from a bulletin on the ,December gales Lothar and Martin in Europe" by the Munich Re. We wish to thank M. Kurz from the German Weather Service, DWD, for providing the manual analysis map in Fig. 1d. This work was supported by the German Science Foundation (SFB419), the German Ministry of Research (BMBF grant No. 01 LA9861/5), and the European Union (project STOEC, grant ENV4CT97-0499). J. Pinto is supported by the Portuguese Office for Science and Technology (Program PRAXIS XXI, grant BD/15775/98).
Part of the data mentioned is taken from the following sources:
|
|
Last Modified: Sun Jun 13 12:37:32 2004 |
|