b 900fiPa-600hPc Thermal Wind (m)
Fig. 7b As Fig. 6a, but for the most intense cyclone in PROMES SCEN simulation
PRUDENCE simulations shows an average SST increase of 3.1°C (Somot et al., 2007) for the same emissions scenario, which is about half a degree less than the SST increase of HadCM3 model. The GCMs can also differ in the frequency and duration of blocking episodes, which would modify the environment for the development of tropical cyclones.
The observed SSTs (Smith et al., 1996) used in the present study show a limited interannual variability, due to the way the database was obtained (Somot et al., 2007). If large positive SST anomalies (like the one observed in the summer of 2003) are more frequent in the future, SSTs at the end of the summer for certain years might be clearly above the ones used in the present study, increasing the risk of tropical cyclone development.
The use of RCMs which are coupled to the Mediterranean Sea can offer important information. Somot et al. (2007) show that the use of an atmosphere-ocean RCM reduces slightly the average summer SST increase (to 2.9°C, compared with 3.1°C in the uncoupled simulation), but the impact on the intensity of tropical cyclones could be larger than implied by such a small average difference, as a coupled simulation would include the sea surface cooling typically observed under such cyclones. This limiting factor would depend on the depth of the warm sea mixed layer, as a deep mixed layer can even lead to a positive feedback on the intensity of tropical cyclones, as indicated e.g. by Shay et al. (2000).
Another important uncertainty is the emissions scenario. The A2 scenario used here is based on high emissions, but the global temperature increase in the specific HadAM3H A2 simulation is of 3.18°C between CTRL and SCEN runs (Frei et al., 2006), which is at the middle of the range projected by IPCC (Intergovernmental Panel on Climate Change, 2007). Lower emissions scenarios should be analysed to determine the effect of mitigation measures on tropical cyclone risk.
Regarding the synoptic environment under which the SCEN tropical cyclones develop, the tropical cyclones identified in the present study seem to follow a mechanism similar to the one proposed by Emanuel (2005) for the formation of Mediterranean tropical cyclones from upper-level cut-off lows. The cyclones with a clear tropical structure detected in the present study form from a cut-off low which triggers convection over the sea and undergoes a tropical transition afterwards (results not shown). The development of a strong tropical cyclone needs an adequate environment, and that seems to be provided by long-lasting blocking anticyclones over northern Europe. This synoptic setting explains also why these cyclones may move westward over long distances, as seen particularly in one of REMO-SCEN most intense cyclones, steered by the easterly winds on the southern part of the blocking anticyclone. The duration of these blocking anticyclones can be very long. For example, high pressures last during two weeks over northern Europe when the most intense REMO-SCEN cyclone develops.
Finally, some aspects of the RCM setup can affect the results. The domain of the models, together with the cyclone detection method used, has probably led to missing cyclone centres over the warmer waters of the southern and eastern Mediterranean Sea. The resolution (about 50 km) is certainly larger than that of many GCMs used up to now in analysing Mediterranean cyclones, but is still too coarse to fully resolve tropical cyclones. Resolution increases could lead to better simulation of tropical cyclone processes, and perhaps to more intense cyclones, if the tipically observed relationship between resolution and intensity is valid here.
The results presented here give strong support to the use of RCM ensembles. The use of only one RCM can lead either to underestimations of the risk or to over-estimations of the risk. The use of the cyclone phase space analysis seems to be a promising tool in analysing structural changes of cyclones, due to its objective and gradual character, as there are no arbitrary limits separating in a binary way extratropical from tropical cyclones. The large frequency of vertical level geopotential data needed by this method may limit its application, due to the possible absence of such data in the output from climate simulations.
Despite the need for more studies that take into account the different sources of uncertainty, the present results give plausibility to the future development of tropical cyclones over the Mediterranean Sea. At least some increase in tropical characteristics seems likely for the analysed emissions scenario, following the results presented here. A theoretical mechanism exists for explaining the possible development of tropical cyclones, and the fact that for present climate conditions some Mediterranean cyclones develop already partially tropical characteristics is a strong reason for taking this possibility into account.
Was this article helpful?