The dynamics and predictability of Mediterranean cyclones leading to high impact weather

The dynamics and predictability of Mediterranean cyclones leading to high impact weather

Christoph Kottmeier, Ulrich Corsmeier, Claus-Jürgen Lenz

Institut für Meteorologie und Klimaforschung, Karlsruhe Institute of Technology

Summary

This project focuses on the dynamics of Mediterranean cyclones and the factors that determine their predictability. A special focal point will be the prognostic identification of cyclones leading to High Impact Weather (HIW) in the Mediterranean with storms, heavy precipitation and flash floods especially in the area south of the Alps. Additional criteria to the predicted intensity of the cyclones will be considered, because there are numerous cyclones which are not intense but cause severe weather. We will aim at quantifying the sensitivity of Mediterranean cyclogenesis to large-scale forcing due to an upper-level PV streamer at the dynamical tropopause relative to the impact of moist processes and surface fluxes. The role of embedded convection in the energy and water cycle of cyclones will be investigated in detail by numerical modelling and measurements (funded by other sources) in order to distinguish between cyclones with and without potential for HIW. It is foreseen to concentrate on four typical synoptic scenarios: (i) high amplitude trough approaching the Mediterranean from the west, (ii) remnant circulation in the lower troposphere reactivated by an upper tropospheric trough, (iii) streamer reaching the Mediterranean without cyclogenesis, and (iv) lee-cyclogenesis south of the Alps. With its model studies this project will contribute to future missions of the new German research aircraft HALO, to HyMeX the "Hydrological Cycle of the Mediterranean Experiment", and to THORPEX activities coordinated by the THORPEX European Regional Committee. The HALO-project NEPTUN on the “The western Mediterranean as a sensitive region for cyclone formation causing heavy-rain events” has been accepted as a HALO-DEMO-Mission, being coordinated by the applicants of this proposal.

 

Recent Results from the MED Project

1. High Impact Weather and Synoptic Scale Flow
 
The main threat in the western Mediterranean are local and regional thunderstorm clusters connected with extremely high precipitation and related flash floods (e.g. Atrani near Amalfi, 9 September 2010) as well as large scale rain due to synoptic lifting amplified by orography and embedded convective clusters (e.g. southeastern France, 6-8 September 2010; Meteo France, 2010). To estimate the frequency of high precipitation events in the northern part of the Mediterranean region the daily precipitation sums between 1971 and 1995 from the MAP project (Frei and Schär, 1998; Frei, 2004), have been used.
In the upper row of Fig. 1 the absolute frequency of days with precipitation sums exceeding 50 mm (left) and 100 mm (right) are shown. The highest occurrence of precipitation > 50 mm/day is found in the Ticino region and in the most southeasterly part of the Alps. The percentage of high precipitation days amounts up to 3 % in the 25 years period. This is an average of 11 days per year. Further regions with an enhanced occurrence of daily precipitation exceeding 50 mm are the Cevennes in southern France, the Ligurian coastal mountain range and the edge of the Alps in the Italian province Piedmont. The occurrence of daily precipitation sums > 100 mm is mainly restricted to the above mentioned areas with the highest frequency again in the Ticino region, the most southeasterly Alps and the Cevennes. These results were achieved with rain-gauge measurements converted to grid cells with a considerable size of about 400 square kilometres. For single rain gauges the precipitation amount and the frequency of high precipitation events may be considerably higher.
 
In the bottom panel of Fig. 1 the large scale synoptic weather situations according to Hess and Brezowski (1999) are shown contributing to precipitation > 50 mm. The most frequent synoptic weather situation (left) leading to high precipitation events in the Ticino region, at the Ligurian coastal mountain ranges is “trough over Western Europe”. This weather situation further contributes to high precipitation in the southeasternmost Alps, together with the situation “trough over Central Europe” and synoptic scale westerly flows. In the province Piedmont weather with southeasterly and easterly current lead mostly to high precipitation. The most frequent weather situation for high precipitation in the Cevennes is again “trough over Western Europe”. In addition in the southern part of this region the “high pressure ridge Central Europe”, often coupled with deep pressure in the Mediterranean and hence an easterly flow at the French coast, is frequent. The second most frequent synoptic weather situation (right) leading to high precipitation seems to be less systematic in most regions. Therefore the case selection can be focused on large scale synoptic situations with a trough over western Europe and situations with large scale flow from the south and southeast towards the Alps.
 
 
Fig. 1: Absolute frequency of days between 1971 and 1995 with precipitation exceeding 50 mm/day (top left) and 100 mm/day (top right); large scale weather situations generating most frequent (bottom left) and second most frequent (bottom right) daily precipitation > 50 mm (red: “trough over Western Europe“, green: “trough over Central Europe“, yellow-orange: easterly/southeasterly flow, violet: westerly flow, red-orange: “deep pressure area Britain”, turquoise: “high pressure ridge Central Europe”)
 
 
2. A Case Study of High Impact Weather
 
Within PANDOWAE-MED a typical case of Mediterranean cyclogenesis was selected for COSMO simulations and sensitivity studies which took place from 28 to 29 October 2008. Starting on 27 October, a strong trough protruded from the Northern Atlantic to Western Europe and the Iberian Peninsula on 29 October 2008 (Fig. 2, upper left). This large scale synoptic situation fits quite well in the above mentioned results of the highest probability of high precipitation events in case of the weather situation “trough over Western Europe”. On 27 October, 21 UTC the cold front reached the western Mediterranean Sea east of the Pyrenées indicated by a marked cyclonic change of wind direction and a postfrontal increase of wind speed (Fig. 2 lower left). Until 29 October the cold air proceeds from the southwest via the Strait of Gibraltar into the Mediterranean Sea leading to low tropospheric wind shear, convergence lines (Fig. 2, lower right) as well as local shallow deep pressure areas in the vicinity of the Spanish coast and the Balearian Islands (Fig. 2, upper right). Later a second deep pressure system developed over northern Italy on the leading edge of the PV streamer of the approaching trough (not shown). Wind shear and temperature differences near the sea surface gave rise to highly variable surface fluxes of up to 250 W m-2 sensible heat flux and 500 W m-2 latent heat flux (Fig. 3). At convergence lines with locally low wind speed (south of the Balearian Islands or west of Sardinia) significantly reduced fluxes occur. The high spatial and temporal variability of the flux pattern modifies the energy input into the atmospheric system via the lower boundary on small scale.
 

Fig. 2: 500 hPa topography in gpdm (top left) and mean sea level pressure in hPa (top right) both on 29 October 2008, 00 UTC. Temperature in 2 m height in °C and wind vector in 10 m above ground in ms-1 on 27 October 2008, 21 UTC (bottom left), and equivalent potential temperature in 850 hPa in °C with wind vector in m s-1 on 29 October 2008, 00 UTC (bottom right).
 
Fig. 3: COSMO simulation of sensible heat fluxes (left) and latent heat fluxes (right) of the cyclogenesis between 28 October and 30 October 2008. Surface fluxes on 29 October, 00 UTC after 48 hours simulation time (upper panel) and on 29 October, 12 UTC after 60 hours (lower panel).
 
  
3. Predictability of the HIW-Case
 
For this case we calculated a lagged-average forecast (LAF) ensemble (Hoffmann and Kalnay, 1983) using the COSMO model (version 4.6), driven the GME model of the DWD (Majewski et al., 2002). The model simulations were started with a time lag of 12 hours in the period between 25 October 2008, 00 UTC and 28 October 2008, 12 UTC. In Fig. 4 the temporal development of the area-averaged mean sea level pressure (top left) and the pressure minimum of the developing cyclone (bottom left) in a model sub-domain as given in Fig. 3 is shown. In the left column the temporal development in real time starting on 25 October 2008, 00 UTC and ending on 31 October 2008, 18 UTC can be found. In the top right figure the bias of the area-averaged pressure with respect to GME analysis data depending on the simulation time is shown. The lower left picture figures the bias of the pressure minimum depending on real time.
 
After 36 hours, the area-averaged pressure in the subdomain decreased from 1026 hPa to about 1008 hPa. Most of the LAF runs show a negative pressure bias, especially the first run started on 25 October 2008, 00 UTC. The second decrease in area-averaged pressure after hour 120 is due to a new cyclone moving from the Atlantic Ocean to the Bay of Biscay. The area-averaged pressure bias depending on forecast time shows systematically for all runs a bias of ± 1 hPa or less until a lead time of 42 hours. After this time the bias increases strongly to more than 3 hPa.
 
The pressure minimum in the subdomain started to decrease after 48 hours (27 October 2008, 00 UTC). After a strong decrease the deepening of the cyclones slowed down until 30 October 2008, 00 UTC (after 120 hours) corresponding to a strong deepening of the cyclone in the Lion’s Gulf reaching the leading edge of the PV streamer mentioned above. As in the area averaged pressure, the minimum pressure values are again mostly lower than those by the analysis data, even with a higher amount of up to 7 hPa. The start of the cyclogenesis or the start of the minimum pressure decrease between 48 hours and 60 hours seems to be delayed by all COSMO runs indicated by the positive pressure minimum bias. After 60 hours the cyclogenesis is mostly simulated too intensive in the LAF runs compared to the analysis data.
 
The influence of evaporation on the water cycle and the cyclone development has been examined by varying the transition coefficient over water in the COSMO model. The coefficient has been varied that the area-averaged evaporation was about 50 % (rat_sea = 100) and about 150 % (rat_sea = 2) of the evaporation by the reference model run (rat_sea = 20). Both model runs were started on 27 October 2008, 00 UTC and compared to the corresponding reference run from the LAF ensemble prediction run. As can be seen in Fig. 5 (top left) the area averaged evaporation grows with increasing lead time except the peaks around noon, caused by the enhanced evaporation of the land surfaces during daytime. The hourly evaporation rates at the end of the simulation are 1.8 mm/day in case of rat_sea = 100 and 5.0 mm/day in case of rat_sea = 2. Maximum values of evaporation (not shown) are between 10 mm/day (rat_sea = 100) and 30 mm/day (rat_sea = 2). As expected the area-averaged precipitation is increasing with increasing evaporation over the sea and with time, due to the initialization of the model with identical data. At the end of the forecast the precipitation rate differs by a factor of 3 (Fig. 5, top right)
 
The effect of the different evaporation rates can also be seen in the area-average of the vertical water column (not shown). Enhancing the evaporation does not result exclusively in higher precipitation, but in addition in higher water column values. This means not only the water cycle is enforced by higher evaporation, but additional water remains in the atmosphere. In the lower panel of Fig. 5 the bias of area-averaged pressure (left) and the bias of the pressure minimum (right) are shown. The bias of the area-averaged pressure is positive in the case of reduced evaporation and strongly negative by enhanced evaporation. The development of the small-scale shallow cyclones on the leeside of the Iberian Peninsula at first and the development of the large scale Mediterranean cyclone over northern Italy secondly are dependent on the evaporation and the intensity of the water cycle. This can be seen when looking at the bias of the pressure minimum. In all cases COSMO calculates a too strong development of the northern Italy cyclone, but in the case of enhanced evaporation the core pressure deviates by more than 10 hPa from the corresponding value deduced from the analysis.
 
Fig. 4: Temporal development of area averaged pressure reduced to mean sea level starting on 25 October 2008, 00 UTC (top left); bias of area-averaged mean sea level pressure depending on simulation time (top right); temporal development of minimum value of mean sea level pressure starting on 25 October 2008, 00 UTC (bottom left) and bias of minimum value of mean sea level pressure starting on 25 October 2008, 00 UTC (bottom right).
 
Fig. 5: Temporal development of area-averaged evaporation (top left); area-averaged precipitation (top right); bias of area-averaged mean sea level pressure (bottom left), and bias of minimum value of mean sea level pressure (bottom right). All abscissas start on 27 October 2008, 00 UTC and end on 30 October 2008, 06 UTC.
 
 
4. Large Scale Forcing as given by EOF Analyses
 
For the October 2008 case the analysis of the large scale upstream conditions forming the upper level trough using the TIGGE archive is still ongoing. Taking 165 different ensemble runs from 8 global forecasting systems, EOF analyses have been made and the most meaningful meteorological variables with respect to Mediterranean cyclone formation (geopotential 500 hPa and temperature 850 hPa) have been identified to be used for clustering in a reasonable number of cluster classes. In the top row of Fig. 6 the ensemble mean is shown for the temperature on 850 hPa pressure level (shaded) and the standard deviation components of EOF1 and EOF2. It can be seen that the highest differences in the 850 hPa temperature prediction are found on the front- and on the backside of the trough-like tongue of cold air ranging from the polar region over the British Isles to the west coast of the Iberian Peninsula. For the 124 gpdm and 136 gpdm cluster mean isohypses, shown in the lower left figure, 4 clusters are roughly similar in northern and western Europe, however the intensity and extension of the trough over northwestern Africa is differing. In the cluster represented by the red graphs the trough axis is located much further to the northwest, but with a higher intensity in its northern part over the British Isles.
 
 
Fig. 6: Mean value of temperature (shaded) and EOF1 and EOF2 on 850 hPa pressure level (top row), mean 850 hPa level isohypses (bottom left), mean 850 hPa level isotherms (bottom right). All runs started on 12 UTC 22 October 2008 (7 day forecast).
 
The 0° C isotherms on the 850 hPa level show 3 similar cases with a cold tongue over Western Europe (blue, green, and pink graphs), whereas 2 clusters (the turquoise and red lines) show weaker cold air penetration on the eastern Atlantic Ocean and a considerably upstream delay. The delay in the “red cluster” is due to the delay in trough propagation, whereas the cold air penetration to the south is linked in the “turquoise cluster” to a secondary trough at 20° West and 40° North.
 
The results of the model runs performed in PANDOWAE-MED show a couple of factors like
  • storing capacity of water vapour column,
  • propagation of upper level trough,
  • intensity of upper level PV streamer,
  • low level wind direction,
  • lowest level wind speed enhancing turbulence and surface fluxes,
  • dominating influence of latent heat flux,
  • orographic influence in upstream flow,
  • and orograhic blocking
influencing the development of the scale and intensity of the Mediterranean cyclones and the associated HIW. Further investigation of these factors as well as their predictability should be made by additional model sensitivity studies of the considered cases as well as by extending this research to further cyclogenesis events and statistics.