Methods and Data

Climate change simulations. The multi-model ensemble simulations which are the basis of the present study were performed within PRUDENCE project (Christensen et al., 2002). A description of the nine regional climate models and an analysis of their results can be found e.g. in Jacob et al. (2007) or Deque et al. (2007). The domains cover most of Europe, but differ particularly in the location of the southern and eastern boundaries. This is the reason of the differences in the extension of the Mediterranean basin covered by the models. The horizontal resolution is between 50 and 55 km. Two 30-year time slice simulations have been performed in this study: the control run for 1960-1990 period, with observed greenhouse gases levels, and a scenario run for 2070-2100 period, considering SRES-A2 greenhouse gases evolution (IPCC, 2000). Lateral boundary conditions were provided by simulations performed with the Hadley Centre atmospheric global climate model (HadAM3H) (Pope et al., 2000).

The analysis is performed for the month of September, in order to focus on the possible development of cyclones with tropical characteristics. In this month, the SSTs are near their maximum annual value and the summer subsidence over the Mediterranean Sea is also weakening. The summer subsidence is associated to the northward displacement of the subtropical high, and limits strongly summer precipitation over the region.

Observed SSTs (1960-1990) are used for present climate, and future climate SSTs are calculated by adding a constant increment and a positive trend to the monthly values used in the control simulation. This increment is the 30-year monthly mean difference between the two periods in corresponding runs performed with the ocean-atmosphere coupled model HadCM3 (Johns et al., 2003), and the trend has been taken from the same simulation. SSTs reach values of up to 30°C (averaged over the Mediterranean Sea) at the end of the scenario simulation. The SST values (averaged over the Mediterranean Sea) for both present and future climate simulations are shown in Fig. 1.

Cyclone detection method. An objective cyclone detection method (Picornell et al., 2001) based on sea level pressure (SLP) fields is used. The SLP output from the models has been smoothed with a Cressman filter with a radius of 200 km in order to filter out smaller scale noisy features that appear usually in SLP fields. SLP minima are detected in the filtered fields. Too weak cyclones are removed by applying an SLP gradient threshold within a radius of 400 km around the SLP minima. This radius is the reason why the so called cyclone detection area is smaller

22 —i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i 1 2 3 4 5 6 7 9 9 10 11 12 13 14 15 16 17 IB 13 20 21 22 21 24 25 26 27 29 29 30 31

Fig. 1 Evolution of September SST (averaged over the Mediterranean Sea) for control simulation (grey line, 1960-1990) and for scenario simulation (black line, 2070-2100)

22 —i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i—i 1 2 3 4 5 6 7 9 9 10 11 12 13 14 15 16 17 IB 13 20 21 22 21 24 25 26 27 29 29 30 31

YEARS (FROM SIMULATION START)

Fig. 1 Evolution of September SST (averaged over the Mediterranean Sea) for control simulation (grey line, 1960-1990) and for scenario simulation (black line, 2070-2100)

than the respective RCM domain. Figure 2 shows the smallest and the largest detection area among the nine models. The largest detection area covers almost the whole Mediterranean Sea, while the smallest one covers only the northernmost part. For the present study, the geostrophic vorticity at cyclone centre has been used for measuring cyclone intensity.

The cyclone detection method allows also to calculate cyclone tracks, but it uses the horizontal wind at 700 hPa to this effect. This field is not available for the analysed RCMs. Due to this, a subjective method has been applied in order to determine cyclone duration and track for the most intense cyclones of each simulation. To this effect, the daily SLP field has been plotted with a 2 hPa interval, and only those lows with at least one closed contour have been considered. Tracks have been obtained by linking the daily cyclone centres using criteria of proximity and similarity. The fact that only the most intense cyclones were analysed this way simplified this task.

Vertical structure of the cyclones. The objective analysis of the vertical structure of the cyclones for determining if the cyclones have tropical or extratropical characteristics has been applied following the published description of the cyclone phase space method (Hart, 2003), based on the geopotential fields between 900 and 300 hPa.

Fig. 2 Largest (continous box) and smallest (dashed box) detection areas in the regional climate model ensemble

Results

Frequency and Intensity of Cyclone Centres

As explained before, the detection area differs among the nine models. The southern extension of the detection area has been used for dividing the RCM ensemble in two groups for the following analysis. Those models reaching 36.5°C or further to the south were included in one group for which the most detailed analysis has been done, whereas models reaching only up to 39.5°C were included in the other.

As a measure for extreme intensity, the 95th percentile of geostrophic vorticity at cyclone centre has been used. Table 1 shows the values of this statistic for control (CTRL hereafter) and scenario (SCEN) simulations for the two groups of models, ordered from smaller to larger southern extension of the detection area. Among the models covering only the northern part of the Mediterranean Sea, two show almost no difference between CTRL and SCEN simulations, while the others show some intensity increase in the SCEN run. The increasing tendency of extreme intensity is more definite among the models covering also part of the southern Mediterranean Sea. Most of them show a moderate to strong increase in the intensity of extreme cyclones, with only one exception (RegCM) showing almost no increase. The spread in intensity among the models is much larger for SCEN than for CTRL runs, particularly for the models with larger detection area.

The following discussion concentrates on the group of five models with larger Mediterranean domain. The clearer intensity change in this group compared to the remaining models might suggest that it's related to cyclones over southern and eastern areas of the Mediterranean Sea. In order to analyse this possible

Table 1 Intensity of cyclone centres

Model

CTRL

A2 SCEN

CHRM (41.5°C)

186

182 (-2%)

HIRHAM (40.50C)

162

212 (+31%)

RACMO (40.50C)

198

231 (+17%)

RCAO (39.50C)

185

187 (+1%)

HadRM3H (36.50C)

169

361 (+114%)

CLM (35.50C)

216

274 (+27%)

PROMES (35.50C)

257

300 (+17%)

REMO (33.50C)

221

491 (+122%)

RegCM (32.50C)

205

214 (+4%)

95th percentile of geostrophic vorticity at cyclone centre (x 10-6 s-1). The percentage variation from CTRL to A2 SCEN simulations is indicated in brackets in the third column. The southern limit of the cyclone detection area is indicated in brackets after the model name, in the first column. The two groups of models mentioned in the text are separated by a blank line.

95th percentile of geostrophic vorticity at cyclone centre (x 10-6 s-1). The percentage variation from CTRL to A2 SCEN simulations is indicated in brackets in the third column. The southern limit of the cyclone detection area is indicated in brackets after the model name, in the first column. The two groups of models mentioned in the text are separated by a blank line.

Fig. 3 Mediterranean subdomains for intensity change analysis: western (W), northern (N), southern (S) and eastern (E) subdomains relationship, the Mediterranean Sea has been divided in four subdomains, shown in Fig. 3. The northern subdomain corresponds roughly to the area of Genoa cyclones. The warmer waters occur in the southern and eastern subdomains, which are not fully included in all the five models. If a direct relationship exists between higher SSTs and the cyclone intensity increase, this should show up in the spatial distribution of cyclone centres.

Figure 4 shows, for all five models and for every subdomain, the frequency of cyclone centres (all low pressure centres during the 30-year simulation period) and the average intensity of the four most intense cyclone centres, as a measure of the change in extreme intensity from CTRL to SCEN run. The model with the largest intensity increase (REMO) shows a clear increase in the number of cyclone centres and in their extreme intensity particularly over the southern and eastern subdomains, as expected if the SST increase is a major reason for these changes. But the picture isn't that clear when we look at the other models. HadRM3H shows also the largest intensity increase over the southern subdomain, but the number of detected cyclone centres is very low there. This is in part due to the more limited detection domain for this model. A visual inspection of the SLP field at the original HadRM3H domain reveals that the cyclone detection method misses several intense cyclone centres over the southern and eastern Mediterranean areas. If these missed cyclones would have been taken into account, a clearer relationship between intensity increase and SSTs for this model would have been apparent. But the largest frequency and extreme intensity increases for CLM occur over the western

Cute Pixel Art Dog

Fig. 3 Mediterranean subdomains for intensity change analysis: western (W), northern (N), southern (S) and eastern (E) subdomains

Fig. 4 Number of cyclone centres during 30 years (upper graph) and geostrophic vorticity at cyclone centre (x10~6s_1) (average of the 4 most intense cyclone centres, lower graph) for the different models, for every subdomain (horizontal axis) and for CTRL (grey bar) and A2 SCEN (white bar) simulations. The scales are different for each model

Fig. 4 Number of cyclone centres during 30 years (upper graph) and geostrophic vorticity at cyclone centre (x10~6s_1) (average of the 4 most intense cyclone centres, lower graph) for the different models, for every subdomain (horizontal axis) and for CTRL (grey bar) and A2 SCEN (white bar) simulations. The scales are different for each model subdomain, whereas in PROMES these two quantities increase similarly over three subdomains, including the northern one. Finally, RegCM (which gives no overall intensity increase) shows even some intensity decrease over the southern subdomain. For this last model, the clearest change is a frequency increase over the western subdomain, which seems to be linked mainly to more cyclone centres forming north of the Atlas mountain range.

In summary, the model ensemble shows no systematic relationship between higher SSTs and more intense cyclone centres, though a clearer relationship may be masked by the limited extension to the south and east of the detection areas.

Fig. 5 SCEN simulation: evolution of SST (°C) (black line, left vertical axis) during the period 2071-2100 and evolution of the number of intense cyclones per year (grey line, right vertical axis) for all the five models analysed. The 10 most intense cyclones have been taken for every model

An interesting related question is if the most intense SCEN cyclones develop at the end of the 30-year period, when the trend leads to the the highest SSTs (Fig. 1). For the same group of five models previously considered, we have taken the ten most intense cyclones of every model, and summed for all models the number of these cyclones occurring every year. This aggregate frequency is shown in Fig. 5, compared to the SST evolution. There seems to be no trend in the number of most intense cyclones, though the highest value occur in the last two years. A visual inspection reveals no correlation with SST either, with some peaks in the number of intense cyclones coinciding even with the lowest SSTs.

This is an aggregate result for the five models. If we look only at the model with the strongest intensity increase, we see that the three most intense cyclones occur during the last decade (figure not shown). This model showed also a relationship between higher SST areas and the intensity increase. SSTs seem therefore to play a role in the intensity increase, but other factors contribute to it.

Analysis of Tropical Characteristics of Cyclones

We will analyse now if the cyclone intensity increase, found in most models of the RCM ensemble, is associated at least in part to tropical characteristics. To this end, the cyclone phase space method described by Hart (2003) has been applied to two of the models. The analysis of other models with this method has not been possible, due to the absence of the necessary daily upper-level data.

The cyclone phase space method uses 3 parameters: one indicates if the cyclone is thermally asymmetric (i.e. has a frontal structure) or symmetric in the horizontal, the other two describe if the cyclone has a cold core or a warm core in the lower troposphere (900-600 hPa) and in the upper troposphere (600-300 hPa). A tropical cyclone is a thermally symmetric, full-tropospheric warm-core cyclone.

The two models analysed are REMO (which shows the strongest intensity increase) and PROMES (which shows the weakest intensity increase among the five models, when we exclude the model showing virtually no intensity change). This way we can get a first approximation of the relationship between intensity changes and vertical structure changes of cyclones in the RCM ensemble.

Specifically, the four most intense cyclones of CTRL and SCEN runs have been analysed for these two models. Figure 6a shows the cyclone phase space evolution

T3 200.

T3 200.

ASYMMETRIC COLD-CORE

1 1

ASÏMM. WARM-CORE

_________45d8L

SYMMETRIC COLO-CORE !

.J.--J--.J

1 |

DEEP WARM-CORE

i

SHALLOW WARM-CORE

DEEP COLD-CORE 1 1 1 1

600 h Po Thermol Wind

200.

Fig. 6a Cyclone phase space for the most intense cyclone in REMO CTRL simulation. Upper frame: 900 hPa-600 hPa storm-relative thickness symmetry (m) versus 900 hPa-600 hPa thermal wind parameter (m). Lower frame: 600-300 hPa thermal wind parameter (m) versus 900 hPa-600 hPa thermal wind parameter (m). The most intense cyclone centre is highlighted with a circle for the most intense REMO-CTRL cyclone. This cyclone has a lifetime of 4 days. In the second day, it develops a full-tropospheric warm core, but it has a certain degree of thermal assymetry. Another REMO-CTRL cyclone (figure not shown) is able to develop a symmetrical warm core structure during two days, but the warm core is less intense and doesn't reach 300 hPa. Similar short-lived warm cores develop also in PROMES-CTRL simulation, where three of the four most intense cyclones show a warm core during 1 day. An example can be found in Fig. 7a), which shows the phase space evolution for the most intense PROMES-CTRL cyclone. The appearance of such short-lived features is consistent with the so called "medicanes" (Fita et al., 2007), which are ocasionally observed over the Mediterranean and show a partially tropical structure. It's noteworthy that the days with a most intense cyclone centre correspond to a warm core structure in the cyclones shown (the most intense cyclone centre is highlighted with a circle in all these phase space graphs).

REMO-SCEN most intense cyclones show a strong structure change. As can be seen in Fig. 6b, the most intense cyclone is able to develop a symmetric, full-tropospheric warm core structure during 6 days. The warm cores are also much more intense than for CTRL cyclones. The parameters reach values typical of a strong hurricane (Hart, 2003). The other three most intense cyclones (not shown) have a warm core structure during 5 to 8 days and, at the same time, low values of the thermal asymmetry parameter. These simulated cyclones are therefore tropical cyclones during part of their lifetime. They maintain a warm core structure for several consecutive days (up to 8), which indicates clearly that the feedback between latent heat fluxes and winds at the sea surface is operating as in fully developed tropical cyclones.

In contrast, the differences between PROMES SCEN and CTRL cyclones are smaller, in good correspondence with the smaller intensity change. The changes are in the direction of an increasingly tropical structure, as two SCEN cyclones show warm cores during more days (2 and 3) in their lifetime, and the intensity of the upper tropospheric warm cores is larger. This is associated to deeper warm cores. The most intense centres for every cyclone correspond to lows with warm core structure, though not necessarily with thermal symmetry, as seen in Fig. 7b).

The clear relationship between maximum intensity and warm cores could suggest that tropical characteristics are also behind at least part of the intensity increase observed in two other models (CLM and HadRM3H). The data needed for a cyclone phase space analysis are not available for these models, but we can look for surface features associated likely to a tropical structure. The most intense CLM-SCEN cyclone is particularly noteworthy in this respect: it deepens strongly while diminishing in size and the spatial distribution of the modelled precipitation is compact and rather symmetrical around the cyclone centre, with high precipitation values near the centre and no evidence of fronts. The radius of maximum wind is small, taking into account the horizontal resolution of 50 km. Several other cyclones simulated by these two models show similar characteristics of the precipitation and wind field, which makes likely that they have tropical characteristics during part of their lifetime.

100.

0 10

asymmetric cold-core

ASYMM. WARM-CORE

SYMMETRIC COLD-CORE 1 SYMM. WARM-CORE -1-1-1-1-1-

200.

SYMMETRIC COLD-CORE 1 SYMM. WARM-CORE -1-1-1-1-1-

0EEP WARM-CORE

1 x*t l/ M.

1 SHALLOW WARM-CORE .

DEEP COLD-CORE 1

1 1 1

-

200.

900hPa-600hPa Thermal Wind (m)

200.

900hPa-600hPa Thermal Wind (m)

Fig. 6b As figure 6a, but for the most intense cyclone in REMO SCEN simulation

Was this article helpful?

0 0
Survival Treasure

Survival Treasure

This is a collection of 3 guides all about survival. Within this collection you find the following titles: Outdoor Survival Skills, Survival Basics and The Wilderness Survival Guide.

Get My Free Ebook


Post a comment