Climate Dynamics 15, p. 551-559
The NAO index, which is based on sea level pressure fluctuations over the North Atlantic in the 300-year control run of this model, is only moderately increasing within the 240-year scenario run, its long-term trend exceeding the variability of the control climate not before the end of the simulation. In contrast, the steadily growing storm track activity over northwestern Europe already surpasses the standard deviation defined from the control run after about 160 years. This effect is associated with a change of the NAO pattern. A determination of the centers of action for subsequent 10-year periods based on Empirical Orthogonal Functions shows a systematic northeastward shift of the NAO`s northern variability center from a position close to the east coast of Greenland, where it is also located in the control run, to the Norwegian Sea.
1. Introduction
2. Data
3. NAO and storm track in the control run
4. Changes of NAO and storm track in the scenario run
5. Discussion and Conclusions
Acknowledgments
References
Baroclinic wave activity plays a central role for local weather in the mid-latitudes. Following Blackmon (1976), this activity is often quantified by the variability of the bandpass (2.5-8 day) filtered geopotential height fields at 500 hPa. This quantity exhibits two main maxima in boreal winter, one over the North Atlantic (called the Atlantic storm track) and one over the North Pacific (the Pacific storm track). As the storm tracks are large-scale phenomena of the atmospheric circulation, they are generally well simulated by General Circulation Models of the Atmosphere (GCMs; see D`Andrea et al., 1998 for an intercomparison). Therefore storm tracks should constitute a suitable feature for assessing the greenhouse gas induced anthropogenic climate change in mid-latitudes.
After validating the storm track representation in a number of different GCM simulations against observational data, Cubasch et al. (1997) considered storm track intensity changes for triple CO2 concentration. A remarkable common signal in the scenario runs was an increase of mean storm track activity over the East Atlantic and Europe. This result was obtained for different GCMs (the LMD gridpoint model (Sadourny and Laval, 1984), the ECHAM spectral model (Roeckner et al., 1992)) at different resolutions (T21, T42) and employing different oceanic forcings (fully coupled; prescribed SST). Similar results were found by Carnell et al. (1996) with the United Kingdom Meteorological Office Coupled Atmosphere-Ocean Model. Thus, this tencency seems to be a stable feature of many GCM simulations for increasing greenhouse gas concentrations, in spite of, at least, one exception: Zhang and Wang (1997) find reduced synoptic activity (approx. 20% for the storm track over Europe) in a 2·CO2 equilibrium experiment using the NCAR CCM1 and a mixed layer ocean model. Simulated increases in gales and the intensity of winter rainfall events over Europe, which may be related to a local increase of baroclinic activity in a greenhouse gas scenario, were reported by Carnell et al. (1996), Lunkeit et al. (1996), Cubasch et al. (1995), Hall et al. (1994) and by Gregory and Mitchell (1995).
Another phenomenon responsible for climate variability in Europe is the North Atlantic Oscillation (NAO). It is usually quantified as the difference of normalized pressures over the Azores/Lisbon and Iceland (e.g., Hurrell, 1995a). The relationship between the NAO and local climate in Europe has already been mentioned by Defant (1924). Examples for more recent work on the importance of the NAO for local weather include studies with respect to climate in the Alps (Wanner et al., 1997) and winter rainfall over Europe (Hurrell and van Loon, 1997; Hurrell, 1995a), in particular over Portugal (e.g., Ulbrich et al., 1998; Rodo et al., 1997; Zorita et al., 1992). It was noted that the increase of the NAO index over the recent years may have contributed much to the observed wintertime hemispheric warming trend, and the NAO variability has become one of the key issues of WMO`s Research programme on Climate Variability and Prediction (CLIVAR, 1997).
Observational studies confirmed that NAO and Atlantic storm track are closely related (Hurrell and van Loon, 1997; Rogers, 1990; Hurrell, 1995b). During positive NAO phases, the storm track is intensified and shifted northward. Another aspect of this relationship is described by Rogers (1997) who finds an association of increasing baroclinic wave activity over the Northeast Atlantic and Europe and a northeastward shift of the subpolar low and the subtropical high. He also notes that increased wave activity is accompanied by an increase of the mean pressure gradient between the centers. Mächel et al. (1998) showed significant correlations between the positions of Icelandic Low and Azores High, their intensity and the pressure gradient between them. These observational results corroborate studies employing simplified numerical models producing an area of low mean surface pressure northeast of a simulated storm track (Frisius et al., 1998; Hoskins and Valdes, 1990). This suggests that the storm track and the mean pressure pattern may be mutually dependent on each other.
The close relation between NAO, baroclinic wave activity, location of the main pressure centers and European climate is also reflected in findings about recent climatic trends over the East Atlantic and Europe: The NAO has come into a positive phase in the early 1990`s, rising from negative values in the 1970`s (Hurrell and van Loon, 1997; Hurrell, 1995a; Wallace et al., 1998; Schmutz and Wanner, 1998). During the same period an increase of baroclinic wave activity can be found over the Northeast Atlantic, which is detectable in parameters like cyclone core pressures (e.g. Haak and Ulbrich, 1996; Stein and Hense, 1994; Schinke, 1993) and SLP variability (Schmith et al., 1998; Rogers, 1997; Born and Flohn, 1997; Born, 1996). Schönwiese et al. (1993) reported decreasing/increasing mean winter SLP over northern/southern Europe for the time period 1961-1990. This corresponds to an increase in the frequency of westerly flow regimes during boreal winter (Schmutz and Wanner, 1998; Bardossy and Caspary, 1990). A relation between these recent changes in weather regime frequency and local rainfall over Europe seems to be well established (Buishand and Brandsma, 1997; Bardossy and Caspary, 1990). With respect to windspeed, there is a recent tendency towards increased winds in Northern Europe corresponding to the rise in pressure gradients and baroclinic activity (The WASA Group, 1998). This effect is, however regional in character. The synoptic patterns producing increased gales in Northern Europe are, for example, leading to decreased numbers of gales over Switzerland (Schiesser et al., 1997).
Apparently, the trends in different parameters observed during the past decades represent aspects of a mode of atmospheric variability. It is not really clear, however, if this mode is excited by increasing greenhouse gas concentrations. Recent trends in the NAO, the Atlantic/European storm track activity, and the mean sea level pressure (SLP) could also be part of a regular variability, i.\x11e. not anthropogenically influenced, and the simulated climate signal over Europe could be a sampling effect of low frequency variability reproduced by a GCM. In this paper we address the latter problem by investigating a control run for present day climate (section 3) and a greenhouse gas (GHG) scenario run of the same coupled Atmosphere-Ocean GCM (section 4). Non-linear trends of the NAO index, the storm track activity, and the mean SLP will be computed in order to demonstrate their relationship with transient GHG forcing. Evidence is given for a northeastward shift of the northern NAO variability center.
1. Introduction
The NAO index is obtained as the difference between two area averaged and normalized monthly mean sea level pressure (SLP) anomalies. The selected areas represent the teleconnectivity centers for winter SLP over the North Atlantic (for definition see Wallace and Gutzler, 1981) and are located northwest of Portugal [11-14°W, 40-43°N] and over Iceland [17-20°W, 65-68°N]. They are in excellent agreement with the observed centers of teleconnectivity according to Wallace and Gutzler (1981). As in the observational data, the teleconnectivity maxima are found in the vincinity of the variability maxima obtained from an EOF analysis of the SLP over the North Atlantic. The first EOF (explaining 46% of the total variance in winter) produces the typical NAO pattern whose subarctic center possesses a zonally elongated structure (Fig. 1, top panel), with a primary maximum over southeast Greenland and a secondary maximum over Iceland. This structure agrees well with the leading EOF of observational SLP data from the North Atlantic sector in January (Glowienka-Hense, 1990).
The time series of the resulting NAO-index, smoothed with a 4-year low-pass filter for clearer display, is shown in Fig. 1 (bottom panel). A spectrum analysis (not shown) reveals that the spectral densities are almost equal at all frequencies, with significant peaks around 7.5 and 30 years (see Christoph et al., 1998, for further discussion). In particular, there is no trend in the control run`s NAO time series.
The simulated storm track is computed with a recursive bandpass filter (Christoph et al., 1995). The average over all boreal winters of the entire control run is shown in Fig. 2 (top panel). Clearly, the Atlantic and the Pacific storm tracks are reproduced realistically with respect to their locations. Intensities are about 10% lower than those computed from ERA (ECMWF Re-Analyses) data.
The close relationship between the NAO index and the Atlantic storm track intensity is confirmed by the distribution of correlation coefficients. Fig. 3 shows a center of strong positive correlation southeast of Iceland (maximum correlation is 0.7). A center of negative correlation is located southwest of Portugal (maximum correlation is -0.7). This indicates an association of an increasing NAO index with a northward shift of the storm track, which is in agreement with observational results. The relevance of the NAO for storm track variability is further confirmed by performing a linear regression analysis (not shown) which produces maxima over the same areas.
In order to study a possible change of the NAO pattern as a function of the storm track activity over northwestern Europe (average over the area 6°W - 20°E and 40 - 70°N), we performed EOF analyses of monthly mean SLP fields in winter for those 10-year periods when storm track activity is above / below one standard deviation. The positions of the northern center as determined from the EOFs are all located close to the center for the whole run, i.e. over southeastern Greenland (no figure). The southern centers are found to be much more variable with respect to the longitude of their location, reaching from ˜ 20°W to 30°W (i.e., closer to the Azores region) during intense storm track decades, and from ˜ 0° to 20°W (i.e., towards gulf of Biscay) for weak storm track decades.
4. Changes of NAO and storm track in the scenario run
|
We start our investigation of the scenario run by asking if there is a systematic change of the NAO index with respect to the control run. In order to provide optimal comparability with the control run`s index, we used the same areas for averaging and performed the normalizations with the SLP standard deviations of the control run rather than those of the scenario run. The time series (Fig. 4) shows an increase of the NAO index with rising greenhouse gas forcing, but it is not before the end of the run that the quadratic curve fit (here this type of fit optimally represents the non-linear trend) to the data emerges from the band of the control run`s standard deviation. The number of winters, however, with index values exceeding the upper margin of the band is much larger towards the end than at the beginning of the run (Fig. 4).
|
The relative change of storm track intensities for two decades chosen (marked by horizontal bars in Fig. 5) from the scenario run is computed (Fig. 2, bottom panel). These periods represent an anthropogenic forcing of ˜ 1.2 W/m2 and ˜ 6 W/m2, with CO2 concentrations having approximately doubled in the second decade relative to the first one. While there is a 10-15% increase of storm track activity over the North Sea area (in agreement with several other greenhouse gas scenario simulations, see Cubasch et al., 1997), a two sided t-test applied to each individual grid point reveals that the signal is not statistically significant at the 95% confidence level. The lack of statistical significance is clearly due the large interannual variability of stormtrack activity visible in Fig. 5. Nevertheless, the two decades chosen reflect the temporal evolution of the long-term trend. Considering potential physical mechanisms for the increase in strom track activity, we found increasing baroclinicity over the entire northeast Atlantic as estimated by the maximum Eady growth rates (for definition see Hoskins and Valdes, 1990). Distinguishing upper and lower tropospheric growth rate values revealed a dominance of the upper tropospheric change (not shown). In addition, an increase in precipitation rates was detected over the storm track area, which suggests that diabatic heating effects are also contributing to the increased baroclinic wave activity.
The complete time series of Northwest-European storm track activity in the scenario run (Fig. 5) indicates falling values for the first decades of the run, followed by a steady increase of storm track activity. The initial decrease is also visible in the respective timeseries of the control run and may thus be assigned to long term variability (no figure). The quadratic curve fit to the data shown in Fig. 5 emerges from the control run`s band of variability at an anthropogenic forcing of ˜ 3 W/m2, so that the signal becomes significant much earlier than that of the NAO.
The fact that the change of the NAO index with increasing greenhouse gas forcing is much less pronounced than that of the European storm track activity may be assigned to a change in the spatial characteristics of the NAO. Therefore we performed 24 EOF analyses for subsequent 10-year periods, each one comprising 120 monthly means taken from the scenario run, and determined the centers of variability. These centers are depicted in Fig. 6. Those from the first decades (with small to moderate forcing) are marked by circles while those from the later ones (moderate to strong forcing) are marked by dots. As long as the greenhouse gas forcing is small the subpolar centers of NAO variability are located closely to the respective center determined for the entire control run (cf. Fig. 1, top) which is marked by a grey square over Southeastern Greenland. When the anthropogenic forcing exceeds 3 W/m2 around year 2020, the northern center moves towards more easterly positions. We would like to mention that, although the seasonal cycle has been removed from the data sets, a possible influence of summer months on the shift of the NAO centers seen in Fig. 6 cannot be excluded a priori. Performance of an EOF analysis based on winter months only for the period 2020 to 2099 demonstrates, however, that the displacement of centers is an effect clearly visible during winter (Fig. 7). Also note that this shift is not simply a switch between two minor maxima being part of the same elongated structure found in all EOFs, but a real displacement of the pattern (compare Figs. 1 top and 7). Thus, consideration of a spatially fixed NAO index is inadequate for an appraisal of greenhouse gas induced changes.
In order to further assess the greenhouse gas induced storm track change over Europe, we performed a global EOF analysis of the bandpass filtered geopotential heights over all 240 boreal winters. The resulting first EOF (Fig. 8, top panel) explains 34% of the total variance. The principal component (PC) belonging to this EOF is lowpass filtered and then compared to the anthropogenic greenhouse gas forcing function of the scenario run in Fig. 8 (bottom panel). The almost parallel development of the two curves strongly suggests that the leading EOF captures the greenhouse gas signal with respect to storm tracks. While there are also large contributions of the EOF over the Southern Hemisphere (indicating a poleward shift and intensification of the local storm track), one particular change suggested by the EOF is the increase of storm track intensities over the eastern North Atlantic and Europe. The resemblance between the time series of storm track activity averaged over Europe (Fig. 5) and the first PC of the global storm track (Fig. 8, bottom) suggests that storm track intensification over Europe is indeed a stable feature of anthropogenic climate change.
The same kind of analysis was performed for the winter means of sea level pressure, as both observations and model studies suggest a relationship between this parameter and storm track activity. The first PC (Fig. 9, bottom panel), explaining 79% of the total variance of the greenhouse gas run, develops in parallel to the radiative forcing (note that the spin-up phase of the SLP-PC is an artifact of the low-pass filter applied). For the Northern Hemisphere high latitudes, the EOF shows decreasing surface pressure (Fig. 9, top). Concerning the Northeast Atlantic in particular, the decrease becomes larger towards the northeast, i.e. the Iceland and Norwegian sea region. Bearing in mind the relationship between surface pressure and baroclinic activity as produced by a number of simplified models (see introduction), the local increase of the Atlantic storm track intensity shown in Fig. 8 is in agreement with the reduction in SLP described above.
In accordance with a number of other GCM studies, the coupled atmosphere-ocean model ECHAM4/OPYC3 produces an intensification of the 500 hPa storm track over Northern Europe with increasing greenhouse gas concentrations. We have identified this change to be part of the first global EOF of storm track intensity, and conclude that it is an anthropogenic effect rather than long term model variability. As storm track activity is correlated with the NAO (using a spatially fixed index based on the monthly mean sea level pressure in the 300-year control run), a tendency towards higher NAO index values detected in the scenario run is not unexpected. This long term trend, however, is not emerging from the present day climate variability as rapidly as the storm track intensity. This is likely due to a northeastward shift of the NAO centers with enhanced greenhouse gas concentrations which a spatially fixed index cannot capture, thus leading to an underestimation of the change.
Climate trends observed over the North Atlantic and Europe have been suggested to be unprecedented in recent decades by some researchers (see Wallace et al., 1998, with respect to a positive trend in SLP variability; Schmutz and Wanner, 1998 with respect to the increase of the NAO index and of the frequency of westerly flow regimes), while others have stressed the fact that the recent trend does not emerge from long term variability (e.g., The WASA group, 1998; Rogers, 1997). Apparently, our model results corroborate with these trends. This agreement can, however, not be interpreted as a proof of greenhouse gases already changing European climate. We have shown that there is large interannual variability with respect to storm track activity, and the changes on this time scale are to be distinguished from the long term trend. The combination of large interannual variability and long term trends means that a consideration of short observation or simulation periods (e.g. 6 years, as by the WASA group, 1998) as a proxy for the trend may be misleading, resulting in too large or too small estimates of greenhouse gas induced changes. This does not only apply to the intensity of the storm track over Europe, but also to the NAO, whose amplitude is subject to considerable long term fluctuations in the control run for present day climate.
With respect to the physical mechanisms producing the European storm track signal we found a consistent increase of upper tropospheric baroclinicity over the Northeast Atlantic. This is in close agreement with the work of Lunkeit et al. (1996) who found that it is upper level baroclinicity rather than lower level baroclinicity which is responsible for increasing storm track activity in a greenhouse gas scenario. This reasoning is consistent with the mechanisms brought forward to explain present day variability in baroclinic wave activity: As the NAO comes into a positive phase, there is increased advection of cold air from the Labrador Sea, strengthening the meridional temperature gradient and thus increasing baroclinicity over the eastern North Atlantic (Born and Flohn, 1997) . Increased zonal winds due to the rising pressure gradients are increasing air-sea fluxes, thus enforcing baroclinicity (Hoskins and Valdes, 1990). This leads to an increased growth of baroclinic waves and thus to larger storm track activity downstream (i.e., over the Northeastern Atlantic). Barotropic decay of the disturbances produces reduced mean surface pressure poleward and increased pressure equatorward of the disturbance path (e.g., Frisius et al., 1998; Hoskins and Valdes, 1990; Grotjahn and Lai, 1991). Such a decrease of mean pressure at the downstream tail of the storm track is in fact part of the Greenhouse Gas signal in our simulation. It should be noted that is also produced by other scenario runs (e.g., Cubasch et al., 1997; Carnell et al., 1996; Hall et al., 1994).
This mechanism is, however, not sufficient to explain the shift of NAO variability centers in the scenario run. Our study gave no evidence for a displacement of the northern NAO center with increasing storm track activity over Europe in the control run. Thus, this effect in the greenhouse gas run must be distinguished from simulated present day variability. It is an open question what causes this shift. One can assume that it is connected with the anthropogenic change in the Atlantic storm track. The activity increase of the storm track is produced by the well known upper tropospheric temperature rise at low latitudes resulting in enhanced meridional temperature gradients and thus in strengthened upper-air baroclinicity. Another contributing factor could be the enhancement of baroclinic wave activity through increased precipitation rates over the North Atlantic, the latter being a consequence of increased latent heat content in a warmer troposphere.
One would certainly like to know if there is any evidence for such a shift of the NAO in observational data.To our knowledge, there are no studies yet discussing a long term variability of NAO centers. It is interesting to note in this context that Schmith et al. (1998) and Rogers (1997) find an association between increasing baroclinic variability and a pressure decrease over northwestern Scandinavia and an increase in the Biscay region. The pattern of this pressure decrease is not identical with the typical NAO variability, but it is not clear if it is an indicator for a displacement of the NAO poles or of a change in mean pressure over arctic and subarctic regions of the North Atlantic sector as expected from a regular increase of baroclinic activity.
5. Discussion and Conclusions
|
Last Modified: Sun Jun 13 12:36:43 2004 |
|
