Assessment of an ensemble of ocean–atmosphere coupled and uncoupled regional climate models to reproduce the climatology of Mediterranean cyclones
- 962 Downloads
This study aims to assess the skill of regional climate models (RCMs) at reproducing the climatology of Mediterranean cyclones. Seven RCMs are considered, five of which were also coupled with an oceanic model. All simulations were forced at the lateral boundaries by the ERA-Interim reanalysis for a common 20-year period (1989–2008). Six different cyclone tracking methods have been applied to all twelve RCM simulations and to the ERA-Interim reanalysis in order to assess the RCMs from the perspective of different cyclone definitions. All RCMs reproduce the main areas of high cyclone occurrence in the region south of the Alps, in the Adriatic, Ionian and Aegean Seas, as well as in the areas close to Cyprus and to Atlas mountains. The RCMs tend to underestimate intense cyclone occurrences over the Mediterranean Sea and reproduce 24–40 % of these systems, as identified in the reanalysis. The use of grid nudging in one of the RCMs is shown to be beneficial, reproducing about 60 % of the intense cyclones and keeping a better track of the seasonal cycle of intense cyclogenesis. Finally, the most intense cyclones tend to be similarly reproduced in coupled and uncoupled model simulations, suggesting that modeling atmosphere–ocean coupled processes has only a weak impact on the climatology and intensity of Mediterranean cyclones.
Intense cyclones in the MR mainly form due to synoptic scale atmospheric systems, such as upper-tropospheric Rossby waves and stratospheric air intrusions (Fita et al. 2006; Claud et al. 2010; Flaounas et al. 2015a). They mainly occur in winter and autumn and they are favored at the leeward side of the Alps and Atlas Mountains, over the Gulf of Lions, the Ionian, Adriatic and Aegean Seas (e.g., Alpert et al. 1990; Trigo et al. 1999; Maheras et al. 2001; Campins et al. 2011; Lionello et al. 2016). These cyclones have a strong impact on the Mediterranean climate, its variability and extremes (e.g., Trigo et al. 2000; Lionello et al. 2006; Gaertner et al. 2007; Lionello and Giorgi 2007; Walsh et al. 2014; Zappa et al. 2015). Indeed, several studies showed the significant contribution of cyclones to the majority of rainfall and wind extremes (e.g., Jansà et al. 2001; Nissen et al. 2010; Reale and Lionello 2013; Raveh-Rubin and Wernli 2015), to the formation of the most prominent high impact weather events (Llasat et al. 2010), as well as to the modulation of the Mediterranean hydrological cycle (Romanski et al. 2012; Flaounas et al. 2015b).
Given their importance to the Mediterranean climate, cyclones proper representation by models is crucial for studying climate dynamics and impacts. To this end, reanalysis data is considered to provide realistic results due to the assimilation of a plethora of observations into atmospheric models. However, global reanalyses are typically available with relatively coarse resolutions that prohibit the detailed reproduction of cyclone meso-scale dynamics and weather extremes. To address this issue, RCMs have been long ago employed to analyze climate dynamics across different spatial scales (e.g., Giorgi 1990; Deque and Piedelievre 1995). In fact, several recent studies demonstrated the benefits from the use of RCMs in reproducing climate patterns at local scales within the MR and for improving the quality of input data for impact studies in areas of complex geography (Flaounas et al. 2013; Guyennon et al. 2013; D’Onofrio et al. 2014; Calmanti et al. 2015). Despite their ability to resolve fine-scale atmospheric features, RCM results are associated with significant uncertainties. For instance, RCMs are known to be particularly sensitive to the physical parameterization of atmospheric processes, to the boundary conditions, horizontal/vertical resolution (e.g., Roeckner et al. 2006; Sanchez-Gomez et al. 2011; Herrmann et al. 2011; Flaounas et al. 2011; Di Luca et al. 2014), and internal variability (e.g., Christensen et al. 2001; Sanchez-Gomez and Somot, 2016). Especially the latter is a primary source of uncertainty and leads to spread in RCM behavior under the same external conditions.
One of the specific objectives of the Coordinated Regional Downscaling Experiment (CORDEX) initiative is to provide a coordinated framework for the systematic investigation of RCM uncertainties in order to provide meaningful atmospheric data on fine scales (Giorgi et al. 2009). The MED-CORDEX initiative is the declination of CORDEX devoted to the MR (www.medcordex.eu; Ruti et al. 2015). In line with the objectives of CORDEX, in this study we assess the capacity of twelve RCM simulations to realistically reproduce Mediterranean cyclones, especially the most intense systems, which are also expected to provoke the majority of regional climate extremes and the strongest socio-economic impacts. These simulations are produced by seven different models, where two models have been used as stand-alone atmospheric RCMs, while the other five have been used both as atmospheric RCMs and as interactively coupled RCMs with oceanic models.
This study aims first at contributing to the better understanding of RCM deficiencies when used for climatological studies on cyclone-related impacts and dynamics. Second, this study assesses the capacity of RCMs to realistically reproduce intense Mediterranean cyclones, and finally it investigates the potential added value in using coupled models for reproducing intense Mediterranean cyclones. The RCMs assessment is achieved by comparing their cyclone climatologies to the one produced by the reanalysis, which also provided the RCMs initial and lateral boundary atmospheric conditions. The next section presents the models and the cyclone tracking methods. Section 3 provides a cyclone climatology comparison between ERA-Interim (ERAI; Dee et al. 2011) and the MED-CORDEX models and finally, in Sect. 4 we summarize our methods and results, and present the main conclusions of this study.
2 Models, cyclone tracking and methodology
2.1 Regional climate models and tracking methods
Short descriptions of the regional climate model simulations
Coupled model version
IPSL-WRF311 WRF version 3.1.1 (Skamarock and Klemp 2008) model at 20 km horizontal resolution and 28 vertical levels. Grid nudging is applied on wind, water vapor and temperature to relax WRF outputs each 6 h towards ERAI. Among the different physical parametrizations offered by the model, here we use the Kain-Fritsch convection scheme (Kain 2004), the WSM5 microphysics scheme (Hong et al. 2004) and the RRTMG longwave and shortwave radiation scheme (Clough et al. 2005)
ENEA-PROTHEUS Composed of RegCM and the MITgcm developed by Marshall et al. (1997a, b) and validated over the Mediterranean area by Sannino et al. (2009). RegCM3 receives the instantaneous values of SST every 6 h. The MITgcm receives from RegCM3 the 6-hourly cumulated rainfall and the surface forcing fields for the ocean, i.e. wind stress, sensible heat flux, latent heat flux, long and short wave incident radiation (Artale et al. 2010)
University of Frankfurt -COSMO-CLM CCLM version 4.21 (Rockel et al. 2008) is used for atmospheric-only simulations with horizontal grid resolution of 50 km and 32 vertical levels
COSMO-CLM/NEMO-MED12 The coupled model consisted of the atmospheric model CCLM and the ocean model NEMO-MED12. NEMO-MED12 is the regional part of the global ocean model NEMO v3.2, specially tuned for the Mediterranean Sea (for more details see, e.g., Lebeaupin Brossier et al. 2011; Akhtar et al. 2014)
University of Belgrade—EBU EBU (Eta Belgrade University) model (a version of NCEP’s ETA model) at 0.33° (~50 km) horizontal resolution and 26 vertical levels
University of Belgrade—EBU/POM EBU-POM is atmospheric-ocean two-way coupled regional model, with the atmospheric component EBU and the ocean component POM (Princeton ocean model). Models exchange atmospheric surface fluxes and SST every atmospheric model time step, which is of the order of minutes (Djurdjevic and Rajkovic 2008; Krzic et al. 2011)
CNRM-ALADIN52 ALADIN-Climate RCM version 5.2 (Colin et al. 2010; Herrmann et al. 2011). The Med-Cordex simulations, used in the current study, are described in details in Tramblay et al. (2013). They have 50 km horizontal resolution and 31 vertical levels
CMCC-CCLM COSMO-CLM version 4-8-19 (Rockel and Geyer 2008) model with 50 km horizontal resolution and 45 vertical levels. Type of convection parameterization: Tiedtke scheme
No coupled simulation
UCLM-PROMES PROMES version (Domínguez et al. 2010) with 50 km horizontal resolution and 37 vertical levels
No coupled simulation
To identify and track cyclones, six different cyclone tracking methods have been applied, based on either sea level pressure (SLP) or relative vorticity at 850 hPa (RV), i.e., the two atmospheric fields that are commonly used for tracking cyclones in gridded datasets (Hoskins and Hodges 2002; Neu et al. 2013). SLP is a low frequency field that reflects the atmospheric mass distribution, and is representative mainly of synoptic-scale atmospheric processes. On the other hand, relative vorticity is a high frequency field that is representative of the atmospheric circulation. Method 1 (Flaounas et al. 2014) and method 2 (Ayrault and Joly 2000) are both RV-based methods, identifying cyclones as local maxima of RV at 850 hPa. Methods 3–6 are all based on SLP. In particular, method 3 (Kelemen et al. 2015) and method 6 (Picornell et al. 2001) identify cyclones as local SLP minima, while method 4 (Lionello et al. 2002; Reale and Lionello 2013) and method 5 (Wernli and Schwierz 2006) perform a processing of SLP fields prior to determining cyclone centers as local SLP minima. All methods are described in detail in the Appendix and have been previously applied in the MR, among other methods that were used to study the MR climatology of cyclones (e.g., Murray and Simmonds 1991; Trigo et al. 1999; Pinto et al. 2005; Bartholy et al. 2009; Campins et al. 2011). Our motivation in using a variety of cyclone tracking methods is the fact that their results may present significant variability, even when considering same periods and input datasets (e.g., Bartholy et al. 2009; Campins et al. 2011; Lionello et al. 2016). Uncertainties exist due to differences in the mathematical and physical definitions of cyclones among the methods, as well as due to the tracking tools’ sensitivity to several factors such as the horizontal grid spacing of the input datasets (Kouroutzoglou et al. 2011; Hanley and Caballero 2012; Rudeva et al. 2014). The sensitivity of extratropical cyclone climatologies due to the use of different tracking methods has been thoroughly analyzed within the framework of the IMILAST project (Neu et al. 2013), while Lionello et al. (2016) specifically addressed this question for the MR.
Figure 1 provides an example of the uncertainty in cyclone tracking due to the use of different tracking methods. It illustrates the tracks of a medicane (merging of the words Mediterranean Hurricane) that occurred in December 2005 (Fita et al. 2007). The tracks are calculated by the six tracking methods, applied to ERAI, and are compared to the manually analyzed track, by connecting the ERAI local SLP minima at consecutive times. It is evident that methods present a common part of their tracks with the reference, i.e., they all succeeded to capture the cyclone’s eastwards displacement over the Mediterranean Sea. However, the lifetime (equivalent to the number of track points), the initial and final location, as well as the cyclone’s exact path differ from method to method. The use of multiple tracking methods constitutes an added value permitting the interpretation of RCM performances in reproducing cyclone tracks from different physical perspectives (Neu et al. 2013).
For the assessment of the models’ capacity to reproduce the Mediterranean cyclone climatology, we use the ERAI reanalysis as a reference. Therefore, we assess the realistic reproduction of the cyclone tracks by the RCMs, as well as the divergence of their results with respect to the driving dataset. In order to perform a meaningful and fair intercomparison between the models and ERAI, all 6-hourly RCM outputs of SLP and wind (used to calculate relative vorticity) have been upscaled to the ERAI grid (at 0.75° spatial resolution). Upscaling regional models is opposed to the motivation of performing climate downscaling, however it provides a straightforward and fair intercomparison of cyclone climatologies between RCMs and ERAI (Kouroutzoglou et al. 2011). Upscaling RCM fields (with resolutions of 20–50 km; Table 1) is expected to have a detrimental impact only on the detection of small and weak cyclonic features with characteristic lengths inferior to ~80 km (this is the average 0.75° × 0.75 grid spacing in the Mediterranean region). Such cyclonic systems unlikely have a strong impact on climate dynamics and extremes.
Cyclone intensity for the RV-based methods 1 and 2 is measured by the maximum RV in the cyclone center. However, methods 1 and 2 apply data filtering before identifying cyclones and hence their cyclone intensity measures are not directly comparable to the other methods. To overcome this inconsistency, the cyclone track points from methods 1 and 2 have been re-attributed to the raw relative vorticity values of the RCM simulations (after regridding to the ERAI grid). For SLP-based methods 3–6, cyclone intensity is determined by the minimum central SLP value along the cyclone track. All six tracking methods are applied to all RCM outputs, as well as to ERAI for the 20-year period of 1989–2008 and within the domain shown in Fig. 1. For all methods, we retain only the cyclones that exist for at least one day (i.e., tracks with at least five 6-hourly track points), as also done in Neu et al. (2013).
2.2 Using a multi-method approach for tracking cyclones
The number of cyclones tracked by each method for the uncoupled (first row in each cell) and coupled models (second row in each cell)
In this study we focus on the most intense Mediterranean cyclones, however the cyclones’ intensity might be sensitive to the ERAI grid spacing, for both SLP and RV-based cyclone tracking methods (Murakami and Masato 2010; Kouroutzoglou et al. 2011). To avoid defining arbitrary thresholds on cyclone intensity and in order to analyze a statistically robust ensemble of intense Mediterranean cyclones, in this study we choose to analyze the 500 most intense cyclones from each method. We thus form a set of intense cyclones for ERAI and one for each RCM. These datasets each contain a total of 3000 intense cyclones (i.e., 500 cyclones from each cyclone tracking method). To simplify the analysis, we retain only the mature stage of the cyclones (the time and location of the track points with minimum SLP or maximum RV). Note that in each dataset the same cyclone may be detected by more than one method. We remove these multiples by considering that two cyclones are the same if their centers are located within no more than 3° in longitude and latitude, and if both cyclones occur within a time difference of no more than 12 h. This methodology to diagnose “identical” cyclone tracks is simpler than the one applied by Blender and Schubert (2000) or Froude et al. (2007), who used a criterion based on the whole cyclone tracks. However, here we treat a complex dataset with tracks reproduced from different methods and hence of different track characteristics (e.g., track lengths of the same cyclones might be considerably different; Fig. 3a). Removing multiples leads to 1600–1800 intense cyclone tracks per dataset (hereinafter referred to as the 500 most intense cyclones). Figure 3c shows the seasonal cycle of the 500 most intense cyclones in ERAI. All methods show a lower frequency of intense cyclones in summer and a higher one in spring and winter, in agreement with previous studies of intense Mediterranean cyclones (e.g., Campins et al. 2011; Cavicchia et al. 2014; Flaounas et al. 2015b).
3 Cyclone tracks in the MED-CORDEX models
3.1 Assessment of RCMs to reproduce cyclone climatology in the Mediterranean region
The results when using the RV-based methods 1 and 2 show that the RCMs present a small statistical spread (i.e., the models have similar values of the RMSE and standard deviation). The results of method 1 show that almost all RCMs have CCD RMSE values smaller than 15 %, standard deviations that are close to the reanalysis (about 15 %) and correlation scores between 0.6 and 0.95. On the other hand, according to method 2, the RCMs present correlation scores between 0.4 and 0.8, as well as RMSE values of more than 20 %. Methods 1 and 2 reproduce more cyclones than the other methods, reflecting their ability to also capture weak cyclonic circulations. The similar spread of RCMs when using these methods suggests that all models perform similarly in capturing these features. However, the poorer statistical scores for method 2, than for method 1, are most likely due to very weak cyclones that are filtered out in method 1 (RV less than 3 × 10−5 s−1) and which are expected to be less consistent between the RCMs and the reanalysis. Indeed, the reproduction of weak and small cyclones are expected to be particularly sensitive to model resolution and thus are expected to be more frequent in the cyclone climatology of RCMs than of ERAI.
Methods 3 and 6 detect the fewest cyclones (Table 2) and therefore the model comparison with ERAI in Fig. 4 reveals small RMSE and standard deviation values, compared to the other methods. A low RMSE does not imply that RCMs present “better” results when using these methods. In fact, the RMSE tends to provide relatively small values for weak and smooth fields (as the ones produced by methods 3 and 6) and tends to “punish” models that produce more variable fields with peaks at the wrong locations (the so-called double-penalty problem discussed in the verification literature, e.g., Wernli et al. 2008). In fact, for these methods, the CCD is higher over the Adriatic, Ionian and western Mediterranean seas (Fig. 2, Online resource 1), where the most intense cyclones are expected to occur (Flaounas et al. 2015b). Indeed, methods 3 and 6 apply a rather strong criterion on cyclone intensity, detecting only cyclones with SLP lower than 1013 hPa and with a SLP gradient greater than 0.5 hPa per 100 km−1, respectively. It is plausible thus to suggest that most of the RCMs capture correctly the occurrence of the most intense cyclones in the central Mediterranean. For methods 3 and 6 only UCLM-PROMES (as well as ENEA-REGCM3 for method 6) does not present similar performance than the other RCMs, mainly due to their higher CCD close to the Alps.
While strong cyclones seem to be adequately captured by all RCMs (Fig. 2 and Online resource 1), method 5 presents a wide statistical spread in the RCMs performance. In fact, method 5 tends to capture the intense winter cyclones with closed contours of SLP and therefore similar statistical scores should be expected by all RCMs in Fig. 4. In contrast, method 5 presents the widest spread of the RCMs performance among the methods, with the COSMO-CLM, the CNRM and CMCC-CCLM having a similar CCD standard deviation as in ERAI (about 20 %) and the highest spatial correlation with the reanalysis. Further analysis suggests that the better statistical scores of these RCMs mostly reflect their capacity to capture the cyclones occurring close to the Alps, the Adriatic Sea and Cyprus. In these areas intense cyclones are most frequent and thus have a strong impact on statistical scores of method 5. Method 4 tends to smooth the CCD spatial variability over these areas (Fig. 2 and Online Resources 1) and hence it comes as no surprise that the COSMO-CLM, the CNRM and the CMCC-CCLM RCMs also present, as in method 5, the best performance compared to the other models.
Figure 4 shows that the coupled versions of the model simulations yield only small deviations of their statistical scores when compared with the atmosphere-only models. Online Resources 2 shows the CCD differences between the atmosphere–ocean coupled and uncoupled model versions. Indeed, the CCD differences exceed 5 % for methods 1, 2 and 5, while differences for methods 3, 4 and 6 are less than 2 %. Such small CCD differences are supported by the fairly equal number of cyclone tracks between the coupled and uncoupled models, regardless of the method used (Table 2). Larger differences are shown in Online Resources 2 for method 5 close to the Turkish Aegean and southern coasts. This suggests that SST differences near these mountainous coasts may affect significantly the pattern of enclosing SLP contours and thus may modulate CCDs between coupled and uncoupled RCMs. This behavior of Method 5 in the eastern Mediterranean is observed for all RCMs.
3.2 Intense cyclone climatologies in the RCMs
Percentage of the 500 most intense cyclones, equally present in both ERAI and the MED-CORDEX models
Percentage of the 500 most intense cyclones equally present in both atmospheric models and their coupled atmosphere–ocean versions
Again, the impact of air–sea coupling on cyclone intensity is rather insignificant for most models (Fig. 8). Significant differences between the structures of intensity distributions might be observed for specific models and sub-regions as for instance between the RV median of COSMO-CLM and ENEA-REGCM coupled and uncoupled versions (Fig. 8f). Thus, it seems that our results converge to the fact that the air–sea coupling effect is rather weak for the cyclones’ spatiotemporal variability and intensities. In fact, the impact of SST to cyclone dynamics has been thoroughly investigated in past papers, where positive and negative SST anomalies have been applied (Homar et al. 2003; Katsafados et al. 2011; Miglietta et al. 2011; Tous et al. 2012). Results showed that when applying significant SST anomalies (e.g., +4 K) there were indeed considerable changes in cyclone dynamics and tracks. However, the coupled RCMs used here were shown to present an average SST bias of less than 1 K (Ruti et al. 2015) and therefore coupling could not be expected to have a strong impact on cyclones intensity. In addition, the weak impact of coupling air–sea interactions on cyclone intensity and tracks maybe due to the RCMs resolution. For instance, Akhtar et al. (2014) showed that the coupling effect in medicanes becomes significant for model grid spacings of the order of 10 km, while it is rather weak for coarser grids. Similar results have been reached by Gaertner et al. (this issue) who showed that improved horizontal resolution is important for RCMs to reproduce the most intense Mediterranean cyclones. It seems that the original resolution of the models (grid spacing of 20–50 km) is rather coarse for resolving realistically the feedback of air–sea interaction to cyclone dynamics.
4 Summary and discussion
In this study we analyzed the climatology of Mediterranean cyclones for the 20-year period of 1989–2008, as represented by ERAI and twelve RCM simulations. The simulations have been performed in the framework of the MED-CORDEX project by seven atmospheric RCMs. Five models were also coupled with an oceanic model. Our results are based on the application of six cyclone tracking methods, which emphasize different aspects of cyclone characteristics. Consequently, our study permits a wide assessment of the model’s performance, covering multiple aspects of cyclone-related physics. This avoids false conclusions based upon a single, subjectively chosen analysis method.
Considering the RCMs capacity to realistically reproduce the climatology of Mediterranean cyclones, the interpretation of the results may vary according to the tracking method. With RV-based method 1, all RCMs have higher scores than with the RV-based method 2, which suggests that the RCMs tend to reproduce more accurately the stronger cyclones since the weaker ones are filtered out in method 1.
Comparing the performance of RCMs using different SLP-based methods, the results become less homogeneous (Fig. 4). Methods 3 and 6, which tend to identify only the most intense cyclones, show a rather robust performance of all RCMs (except of PROMES and ENEA-REGCM, which have a relatively high RMSE). For methods 4 and 5, the RCMs COSMO-CLM, CNRM and CMCC-CCLM perform best with the lowest RMSE, high correlation scores and standard deviation equal to ERAI. The realistic reproduction of cyclones near the Alps is the key feature that determines the good performance of these 3 RCMs.
Focusing on the most intense cyclones, all RCMs reproduce the spatial distribution of their tracks. However we show that the RCMs tend to produce more intense cyclones over land areas. It is noteworthy that RCMs have higher horizontal resolutions than ERAI and thus offer the potential of resolving finer scale atmospheric dynamics (e.g., related to interactions with orography). The feedback of fine scale atmospheric processes to meso-scale systems such as Mediterranean cyclones might have a non-detrimental effect on cyclone characteristics such as intensity and tracks. For instance, the RCMs tendency to produce more intense cyclones over land could be also attributed to the models’ capacity to resolve more realistically, e.g., the impact of orography on cyclone dynamics. Consequently, our results concerning the reproduction of intense cyclones in the RCMs should not be regarded as a metric of the RCMs deficiency when compared to ERAI, but rather as the RCMs behavior when reproducing dynamics related to intense cyclones.
Nudging is beneficial for reproducing the ERAI cyclones. Indeed, the WRF model, which used grid nudging reproduced about two thirds of ERAI intense cyclones, compared to the other RCMs that reproduced about one third. Grid nudging was also beneficial for WRF to perform closest to ERAI in terms of reproducing the cyclones’ seasonal cycle. This performance of WRF, differing from the other RCMs, is due to the relaxation of its fields towards ERAI (for wind, temperature and water vapor), which in turn also served as our reference dataset. Therefore, for reasons of consistency the WRF nudged model could not be taken into account for a fair RCM intercomparison.
Air–sea coupling was shown to have a rather weak impact on the cyclone climatology and their intensities. Concerning the cyclone climatology, the weak effect of air–sea interactions is explained by the fact that cyclogenesis of intense cyclones is mainly due to upper tropospheric forcing. On the other hand, the weak effect of air–sea interactions on cyclone intensities could be due to the resolution of RCMs that is anyway coarser than what is typically needed to resolve such fine scale processes, as well as their feedback to the dynamics of the cyclonic systems.
Assessing the quality of RCMs reproduce prominent atmospheric flow systems such as cyclones is of paramount importance for understanding the model uncertainties in climate processes and extremes and this study shows the potential of tracking methods for a diagnostic description of RCM results. Though cyclone tracks are diagnostics on the spatiotemporal occurrence of cyclones and do not provide direct information on the complex multi-scale physical interactions that have a strong impact on cyclogenesis and cyclone intensification, they may be used to identify the factors responsible for the quality of RCMs results. The present work opens perspectives for the use of tracking methods to interpret RCMs outputs even beyond MED-CORDEX, not only for impact-related studies but also for Mediterranean cyclone dynamics.
This work is part of the Med-CORDEX initiative (www.medcordex.eu). Med-CORDEX is a coordinated contribution to CORDEX that is supported by HyMeX (www.hymex.org) and MedCLIVAR (www.medclivar.eu) international programs. M.A. Gaertner and R. Romera were funded by the Spanish Government and the European Regional Development Fund, through Grants CGL2007-66440-C04-02, CGL2010-18013 and CGL2013-47261-R. S. Calmanti would like to acknowledge financial support from the EU funded SPECS project (Grant # 3038378). N. Akhtar would like to thank the Center for Scientific Computing (CSC) of the Goethe University Frankfurt am Main and the Deutsches Klimarechenzentrum (DKRZ) for providing computational facilities. M.Reale in the last stage of this work was supported by OGS and CINECA under HPC-TRES program award number 2015-07. Finally, H. Wernli would like to thank M. Sprenger for his help in adapting the cyclone tracking method 5 to data from regional models.
- Ayrault F, Joly A (2000) Une nouvelle typologie des dépresssions météorologiques: classification des phases de maturation. Compte-Rendus à l’Académie des Sciences, Sciences de la Terre et des planètes 330:167–172Google Scholar
- Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Beljaars ACM, van de Berg L, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer AJ, Haimberger L, Healy SB, Hersbach H, Hólm EV, Isaksen L, Kållberg P, Köhler M, Matricardi M, McNally AP, Monge-Sanz BM, Morcrette JJ, Park BK, Peubey C, de Rosnay P, Tavolato C, Thépaut JN, Vitart F (2011) The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q J R Meteorol Soc 137:553–597. doi:10.1002/qj.828 CrossRefGoogle Scholar
- Drobinski P, Anav A, Lebeaupin Brossier C, Samson G, Stéfanon M, Bastin S, Baklouti M, Béranger K, Beuvier J, Bourdallé-Badie R, Coquart L, D’Andrea F, De Noblet-Ducoudré N, Diaz F, Dutay JC, Ethe C, Foujols AM, Khvorostyanov D, Madec G, Mancip M, Masson S, Menut L, Palmieri J, Polcher J, Turquety S, Valcke S, Viovy N (2012) Model of the regional coupled earth system (morce): application to process and climate studies in vulnerable regions. Environ Model Soft 35:1–18CrossRefGoogle Scholar
- Gaertner MA, González-Alemán JJ, Romera R, Domínguez M, Gil V, Sánchez E, Gallardo C, Miglietta MM, Walsh K, Sein D, Somot S, Dell’Aquila A, Teichmann C, Ahrens B, Buonomo E, Colette A, Bastin S, van Meijgaard E, Nikulin G (this issue) Simulation of medicanes over the Mediterranean Sea in a regional climate model ensemble: impact of ocean–atmosphere coupling and increased resolutionGoogle Scholar
- Giorgi F, Jones C, Asrar G (2009) Addressing climate information needs at the regional level: the CORDEX framework. WMO Bull 58:175–183Google Scholar
- Herrmann M, Somot S, Calmanti S, Dubois C, Sevault F (2011) Representation of daily wind speed spatial and temporal variability and intense wind events over the Mediterranean Sea using dynamical downscaling: impact of the regional climate model configuration. Nat Hazards Earth Syst Sci 11:1983–2001. doi:10.5194/nhess-11-1983-2011 CrossRefGoogle Scholar
- Kelemen FD, Bartholy J, Pongracz R (2015) Multivariable cyclone analysis in the Mediterranean region. Idojaras Q J Hung Meteorol Serv 119(2):159–184Google Scholar
- Lionello P, Bhend J, Buzzi A, Della-Marta PM, Krichak SO, Jansa A, Maheras P, Sanna A, Trigo IF, Trigo R (2006) Cyclones in the mediterranean region: climatology and effects on the environment. In: Lionello P, Malanotte-Rizzoli P, Boscolo R (eds) Mediterranean climate variability, developments in earth and environmental sciences, vol 4. Elsevier, Amsterdam, pp 325–372Google Scholar
- Lionello P, Trigo IF, Gil V, Liberato MLR, Nissen KM, Pinto JG, Raible CC, Reale M, Tanzarella A, Trigo RM, Ulbrich S, Ulbrich U (2016) Objective climatology of cyclones in the Mediterranean region: a consensus view among methods with different system identification and tracking criteria. Tellus A. doi:10.3402/tellusa.v68.29391 Google Scholar
- Llasat MC, Llasat-Botija M, Prat MA, Porću F, Price C, Mugnai A, Lagouvardos K, Kotroni V, Katsanos D, Michaelides S, Yair Y, Savvidou K, Nicolaides K (2010) High-impact floods and flash floods in Mediterranean countries: the FLASH preliminary database. Adv Geosci 23:47–55. doi:10.5194/adgeo-23-47-2010 CrossRefGoogle Scholar
- Neu U, Akperov MG, Bellenbaum N, Benestad R, Blender R, Caballero R, Cocozza A, Dacre HF, Feng Y, Fraedrich K, Grieger J, Gulev S, Hanley J, Hewson T, Inatsu M, Keay K, Kew SF, Kindem I, Leckebusch GC, Liberato MLR, Lionello P, Mokhov II, Pinto JG, Raible CC, Reale M, Rudeva I, Schuster M, Simmonds I, Sinclair M, Sprenger M, Tilinina ND, Trigo IF, Ulbrich S, Ulbrich U, Wang XL, Wernli H (2013) IMILAST: a community effort to intercompare extratropical cyclone detection and tracking algorithms. Bull Am Meteorol Soc 94:529–547. doi:10.1175/BAMS-D-11-00154.1 CrossRefGoogle Scholar
- Ruti P, Somot S, Giorgi F, Dubois C, Flaounas E, Obermann A, Dell’Aquila A, Pisacane G, Harzallah A, Lombardi E, Ahrens B, Akhtar N, Alias A, Arsouze T, Aznar R, Bastin S, Bartholy J, Beranger K, Beuvier J, Bouffies-Cloche S, Brauch J, Cabos W, Calmanti S, Calvet JC, Carillo A, Conte D, Coppola E, Djurdjevic V, Drobinski P, Elizalde A, Gaertner M, Galan P, Gallardo C, Gualdi S, Goncalves M, Jorba O, Jorda G, Lheveder B, Lebeaupin-Brossier C, Li L, Liguori G, Lionello P, Macias-Moy D, Onol B, Rajkovic B, Ramage K, Sevault F, Sannino G, Struglia MV, Sanna A, Torma C, Vervatis V (2015) MED-CORDEX initiative for Mediterranean Climate studies. Bull Am Meteorol Soc. doi:10.1175/BAMS-D-14-00176.1 Google Scholar
- Sanchez-Gomez E, Somot S (2016) Impact of the internal variability on the cyclone tracks simulated by a regional climate model over the Med-CORDEX domain. Clim Dyn 1–17Google Scholar
- Sevault F, Somot S, Alias A, Dubois C, Lebeaupin-Brossier C, Nabat P, Adloff F, Déqué M, Decharme B (2014) A fully coupled Mediterranean regional climate system model: design and evaluation of the ocean component for the 1980–2012 period. Tellus A 66:23967. doi:10.3402/tellusa.v66.23967 CrossRefGoogle Scholar
- Tramblay Y, Ruelland D, Somot S, Bouaicha R, Servat E (2013) High-resolution Med-Cordex regional climate model simulations for hydrological impact studies: a first evaluation of the ALADIN-Climate model in Morocco. Hydrol Earth Syst Sci 17:3721–3739. doi:10.5194/hess-17-3721-2013 CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.