Direct and semi-direct aerosol radiative effect on the Mediterranean climate variability using a coupled regional climate system model

2.1 The CNRM regional climate system model: CNRM-RCSM4

The present work has been carried out with the fully coupled regional climate system model developed at CNRM. It includes the regional climate atmospheric model ALADIN-Climate (Radu et al. 2008; Déqué and Somot 2008; Farda et al. 2010; Colin et al. 2010; Herrmann et al. 2011), the land–surface model ISBA (Noilhan and Mahfouf 1996), the river routing scheme TRIP (Decharme et al. 2010) and the regional ocean model NEMOMED8 (Beuvier et al. 2010). We use here the version 5.2 of ALADIN–Climate, which has a similar physical package to the global climate model ARPEGE–Climate (Voldoire et al. 2012) used in the CMIP5 exercise. It is a bi-spectral semi-implicit semi-lagrangian model, with a 50 km horizontal resolution. ALADIN includes the Fouquart and Morcrette radiation scheme (FMR15, Morcrett 1989), based on the ECMWF model incorporating effects of greenhouse gases, direct effects of aerosols as well as the first indirect effect of sulfate aerosols (Twomey 1977). More information about the other parameterizations used is presented in Farda et al. (2010). Contrary to the globe, the domain is not periodic, so a bi-periodicization is achieved in gridpoint space by adding a so-called extension zone used only for Fourier transforms. The non-linear contributions to the equations are performed in gridpoint space. The domain chosen for this study is presented in Fig. 1, includes the official domain of the Med-CORDEX program, and represents 90x128 points, including a 11-point wide bi-periodization zone in addition to the more classical 8-point wide relaxation zone using the Davies technique.

The ocean model NEMOMED8 (Beuvier et al. 2010) is the regional eddy-permitting version of the NEMO-V2.3 ocean model (Madec 2008). NEMOMED8 covers the Mediterranean Sea (without the Black Sea) plus a buffer zone including a part of the near Atlantic Ocean, where a three-dimensional damping is performed towards temperature and salinity (monthly data, NEMOVAR-COMBINE reanalysis, Balmaseda et al. 2010), so that the circulation through the strait is simulated with realistic Atlantic waters. A sea level relaxation is also applied on this Atlantic part of the domain towards the same reanalysis. This version has a horizontal resolution of \(1/8^{\circ } \hbox {x}\, 1/8^{\circ }\hbox {cos(lat)}\), namely between 9 and 12 km, and 43 vertical levels, with layer thickness increasing from 6 m to 200 m. Its grid is tilted and stretched at the Gibraltar Strait to better follow the SW-NE axis of the strait and to increase the local resolution up to 6 km. The partial steps definition of the bottom layer is used, and the surface is parameterized with the free surface configuration, filtered formulation. Figure 1 shows the bathymetry of NEMOMED8 over the Mediterranean Sea.

The TRIP river routing model, developed by Oki and Sud (1998), is used here to convert the simulated runoff by the ISBA land surface scheme into river discharge using a river channel network at 0.5° resolution (Decharme et al. 2010; Szczypta et al. 2012). It is based on a single prognostic reservoir whose discharge is linearly related to the river mass, using a uniform and constant flow velocity. The 0.5° grid is cut on a domain that covers the whole drainage area of the Mediterranean and the Black Sea (represented respectively in red and black in Fig. 1), except for the Nile river basin for which climatological values are used. Coupling between all the different components is achieved by the OASIS3 coupler (Valcke 2013) at a one day frequency. The flux scheme is based on the work of Louis (1979) for the boundary layer turbulence physics, while the interpolation of the wind speed from the first layer of the model (about 30 m) to the 10m height follows Geleyn (1988). This model configuration called CNRM-RCSM4 (see the technical note for details, Sevault et al. 2013) is involved in the Med-CORDEX (www.medcordex.eu) and HyMeX (www.hymex.org) programmes. The main difference between the version of Sevault et al. (2013) and the one used in the present study is that spectral nudging is not applied in the latter.

Open image in new window

2.2 Aerosols

The model ALADIN-Climate incorporates a radiative scheme to take into account the direct and semi-direct effects of five aerosol types (sea salt, desert dust, sulfates, black and organic carbon aerosols) through AOD climatologies. In the previous studies using this model, the climatology used was Tegen et al. (1997). In a recent paper (Nabat et al. 2013), this climatology has been compared to other data sets over the Mediterranean region, showing its limits, notably in underestimating desert dust. Nabat et al. (2013) have proposed a new AOD monthly interannual climatology over the period 2003–2009, based on a combination of satellite-derived and model-simulated products. The satellite instrument MODIS has indeed the best scores against ground-based measurements from AERONET, but does not enable to separate the aerosol content into different species. Consequently, models, which are able to reproduce the main aerosol patterns over the basin, have been included in this new data set: the regional climate model RegCM-4 (Giorgi et al. 2012) for dust and sea–salt aerosols, and the MACC reanalysis (Benedetti et al. 2011) for sulfates, black and organic carbons. The aim is to have the best possible estimation of the atmospheric aerosol content for these five most relevant species. More details about the construction of this new climatology, which is used in the present work, can be found in Nabat et al. (2013). Its main patterns will be presented in Sect. 4.1.

2.3 Simulations

Different configurations have been used for simulations in the present work, depending on the presence of aerosols, and on the use of the whole RCSM or the atmospheric model only (see Table 1). The simulation defined as the reference for this study, called C-AER, has used the fully coupled RCSM described previously with the mentioned aerosol fields. The lateral boundary conditions are provided by the ERA-INTERIM reanalysis (Dee et al. 2011). The simulation covers the period 2003–2009, after a spin-up period of 2 years (beginning in 2001) with the identical configuration, which is necessary to stabilize the land surface (notably soil moisture) and surface ocean. This spin-up period begins after an initial spin-up of 48 years to get a stabilized ocean. An ensemble of 6 simulations has been carried out using the same methodology, with different initializations of the atmosphere in 2001. The atmospheric conditions of the six members have been taken from January months from previous years. The objective is to address aerosol–climate issues without depending on the internal variability of the RCM (Christensen et al. 2001; Lucas-Picher et al. 2008). Indeed, because of non-linearities in the climate system, small differences in the initial state can have important consequences in the simulations (Laprise 2008).

2.4 Observational data

In order to evaluate the performance of CNRM-RCSM4, we use several datasets from ground-based measurements and satellite products. The European daily high-resolution gridded data set E-OBS (Haylock et al. 2008) provides surface temperature and is based on about 2,300 ground-based stations. The Climatic Research Unit (CRU) of the University of East Anglia (Harris et al. 2013) provides 2m-temperature and precipitation at a \(0.5^{\circ }\times 0.5^{\circ }\) resolution in the version TS3.1 used in the present work. It includes most of the land weather stations data around the world. Precipitation will be also compared to the GPCC product (Schneider et al. 2008), a homogenised gridded monthly precipitation dataset at 0.5° resolution. Surface radiation is evaluated against ground-based measurements from the GEBA (Gilgen and Ohmura 1999) and BSRN networks (Ohmura et al. 1998), as well as against satellite data, namely Surface Radiation Budget (SRB) and International Satellite Cloud Climatology Project (ISCCP). SRB data used in the present work comes from the NASA Global Energy and Water-cycle Experiment (GEWEX) SRB project (version 3.0 : GEWEX and QC products), and provides surface radiation at \(1^{\circ }\times 1^{\circ }\) resolution (Stackhouse et al. 2000). The ISCCP provides radiance measurements with a spatial resolution of \(2.5^{\circ }\times 2.5^{\circ }\) (Zhang et al. 2004). At the top of the atmosphere (TOA), radiation is compared to Clouds and the Earths Radiant Energy System (CERES)–Energy Balanced and Filled (EBAF) data (Loeb et al. 2009) at a \(1^{\circ }\times 1^{\circ }\) resolution.

Sea–surface temperature observations are given by Marullo et al. (2007), which is a product based on daily series of optimally interpolated SST maps at high resolution \((1/16^{\circ })\) computed from time series of satellite infrared AVHRR. Sensible and latent heat fluxes will be evaluated against reconstructions from the objectively analyzed air–sea fluxes (OAFlux). This product (Yu et al. 2008) provides a global analysis of the ocean latent and sensible heat fluxes at a \(1^{\circ }\times 1^{\circ }\) resolution, based on satellite observations, surface moorings, ship reports and atmospheric model reanalysed surface meteorology. Comparisons between these gridded data sets and CNRM-RCSM4 will be presented in the following section, using averages on different zones (EURS, EURN, ...) defined in Fig. 2.

4.1 Aerosol variability

This section aims at evaluating the mean direct effect of aerosols on the Mediterranean climate. Simulations have been carried out with CNRM-RCSM4 with (C-AER) and without (C-NO) aerosols. Aerosols are included in ALADIN-Climate using monthly interannual climatologies (Nabat et al. 2013). Figure 8a presents the seasonal average AOD from this climatology over the period 2003–2009, while monthly averages per aerosol type are shown in Fig. 9 over Europe, the Mediterranean Sea and northern Africa. These 3 zones correspond to the merging of the zones defined in Fig. 2, namely respectively EURN+EURS+EURC, MEDW+MEDE+AEG, and AFRW+AFRE, and will also be used in the following sections. Aerosol content over the Mediterranean region is affected by a strong seasonal cycle, essentially controlled by desert dust and sulfates. Dust aerosols are indeed transported over the Mediterranean Sea during successing outbreaks, occuring first in spring in the eastern basin, and then moving westwards in summer. These outbreaks are due to intense dust uplift over the Libyan and Egyptian deserts in spring (Moulin et al. 1998) and favourable circulation patterns (Gkikas et al. 2012). In summer, frequent low pressure systems over northern Morocco favour the dust export over the western basin (Moulin et al. 1998). Anthropogenic sulfate aerosols are prevailing over Europe, especially in spring and summer, and can also extend up to the Mediterranean Sea. Local maxima can be noticed in the Po Valley and in central Europe. The other aerosols (organic, carbon and sea–salt aerosols) have weaker AOD. It should be noted that sea salt particles are maximal in winter over the Mediterranean Sea because of the associated strong winds.

4.2 Aerosol direct and semi-direct radiative forcing

Figure 8 presents the average aerosol net direct radiative forcing (DRF) at the surface and at the TOA for SW (b) and LW (c) radiation calculated in the C-AER simulations. Aerosol DRF is calculated in-line during the simulation by calling radiation code twice (with and without aerosols), in order to have only the direct aerosol effect. Average values of these parameters over Europe, the Mediterranean Sea and northern Africa are indicated in Table 2, while monthly averages per aerosol type are shown in Fig. 9. All these forcings show a strong seasonal cycle similar to the aerosol load (Fig. 8a).

Open image in new window

Open image in new window

The scattering and absorption of the incident radiation by the various aerosols cause a negative direct radiative forcing in SW radiation at the surface. Over Europe, the surface SW DRF is estimated at −19.4 Wm⁻² and reaches its maximum in summer (−25 Wm⁻² in July). It is dominated throughout the year by the sulfate forcing with a maximum in absolute terms in June, and accentuated in July by organic and dust aerosols (Fig. 9). Even if the dust AOD is stronger than the organic AOD, the organic forcing is higher than the dust one over Europe, because of different optical properties. Organic aerosols notably include black carbon, which is very absorbing. Figure 8 also reveals local maxima in Europe in spring (Po Valley, Benelux) and summer (central Europe) associated with industrial pollution and biomass burning. Over the Mediterranean Sea, sulfates, dust and organic aerosols are responsible for a high surface SW DRF (−20.9 Wm⁻² on average), especially in spring and summer (Fig. 9). Sulfates are prevailing in spring and autumn, and exert a strong impact on surface radiation because of their weaker asymmetry factor. Figure 9 shows that sea salts have a weaker contribution in the aerosol radiative effect, with a surface forcing maximal in winter \((-1\, \hbox {Wm}^{-2})\). Compared to local calculations, Benas et al. (2011) have estimated the total surface SW DRF at \(-26 \,\pm\, 16\,\hbox {Wm}^{-2}\) in Crete over a long-time period (2000–2010), using calculations performed with a radiative transfer model based on Terra and Aqua MODIS data, which is consistent with C-AER surface SW DRF (\(-20.3 \,\pm\, 0.1\,\hbox {Wm}^{-2}\) on average in Crete). The large difference in the error range comes from the calculation method: Benas et al. (2011) have provided the simple standard deviation from instantaneous values whereas we have calculated a confidence interval at the level 0.05 from the 6 members of the simulation ensemble. Over northern Africa dust aerosols lead to a high SW surface DRF in spring \((-23.3\, \hbox {Wm}^{-2})\) and in summer \((-30.2\, \hbox {Wm}^{-2})\). For the time, such estimations are rare in literature and two previous studies have been added in Table 2. The RegCM columns correspond to values established by Nabat et al. (2012) from a 10-year simulation (2000–2009) using the RCM RegCM-4 (Giorgi et al. 2012). On average, RegCM SW surface radiative forcing is weaker than in ALADIN, certainly due to sulfate aerosols in this RegCM simulation (Nabat et al. 2012) that have been shown to be underestimated over Europe, especially in summer, probably due to the emission datasets. However the seasonal cycle of dust particles notably is similar with a maximum in spring and summer. In addition to RCM estimations, Papadimas et al. (2012) have also calculated the surface SW radiative effect over the broader Mediterranean basin (29–46.5°N, 10.5°W–38.5°E) over the period 2000–2007 using MODIS observations. Their results are compared in the last two columns of Table 2 against the averages for-C-AER over the same box: results are close to the C-AER simulation at the surface with a slight overestimation by C-AER (\(-19.9\, \hbox {Wm}^{-2}\) on average for CNRM-RCSM4 against −16.5 Wm⁻²).

Open image in new window

At the TOA, the SW DRF is also negative in CNRM-RCSM4 over the sea (−10.5 Wm⁻²) and Europe (−6.5 Wm⁻²), because of the scattering of the incident radiation by sulfate aerosols in the atmosphere. Over Crete, C-AER estimates it at \(-8.4\,\hbox {Wm}^{-2}\) against \(-6 \,\pm\, 5\,\hbox {Wm}^{-2}\) for Benas et al. (2011). The situation is different over northern Africa where dust aerosols prevail and where the surface albedo is higher. These two elements lead to an enhanced absorption of reflected radiation, hence a positive TOA SW DRF in summer (+0.8 Wm⁻²). Compared to values from Papadimas et al. (2012) and RegCM in table 2 over the broader Mediterranean basin, C-AER TOA forcing is higher, especially over Europe and the Mediterranean Sea, which is probably due to differences in surface albedo.

With regards to LW radiation, only dust aerosols have a significant impact because of their microphysical properties, and notably the presence of coarse dust mode. Sulfate and organic aerosols have weaker effects on LW. Surface and TOA LW RF are consequently higher in northern Africa in summer, reaching +4.8 and +3.1 Wm⁻², respectively. RegCM averages are similar over Europe and the Mediterranean Sea, but higher at the surface over northern Africa (+6.5 Wm⁻² in summer for RegCM). Uncertainties in the dust LW absorption coefficient are probably responsible for this difference. It should be recalled here that the LW scattering effect of mineral dust is not taken into account in CNRM-RCSM4 (Dufresne et al. 2002).

Figure 10 presents the semi-direct aerosol effect, resulting from calculations achieved thanks to the C-NO simulations. The SW (resp. LW) semi-direct effect has been calculated as the difference between the effect of aerosols on SW (LW) radiation (i.e. C-AER–C-NO) and the SW (LW) DRF. The SW surface semi-direct effect (Fig. 10a) is globally weaker than the direct effect, reaching +5.7 Wm⁻² on average over the Mediterranean Sea, +5.0 Wm⁻² over Europe and +4.7 Wm⁻² over northern Africa at the surface for SW radiation, but contributes to reduce the negative aerosol radiative forcing. This is due to changes in cloud cover and circulation as explained in more details in the following section. In terms of surface downward LW radiation (Fig. 10b), semi-direct effect is limited (on average −0.1 Wm⁻² over the Mediterranean Sea, +0.6  Wm⁻² over Europe and +2.0 Wm⁻² over northern Africa). This LW effect is also probably related to changes in cloud cover by aerosols (see Sect. 4.3).

4.3 Impact of aerosol forcing on the Mediterranean climate variability

After these direct and semi-direct radiative forcing calculations, the comparison with the C-NO simulation now enables to estimate the consequences for different meteorological parameters as well as ocean–atmosphere fluxes. Figure 11 presents the seasonal average difference between C-AER and C-NO in terms of 2m-temperature (a), SST (b), sensible (c) and latent (d) heat loss, surface specific humidity (e), precipitation (f), cloud cover (g) and sea surface salinity (h). First, the negative surface radiative forcing described in the previous paragraph has repercussions on surface temperature. Over Europe, Fig. 11a shows that 2m-temperature is reduced by 0.4 °C on average because of the aerosol dimming (sulfates essentially), and up to 1 °C in summer in eastern and central Europe (see Table 3 for seasonal averages). A similar decrease is observed over northern Africa (−0.5 °C on average) because of dust aerosols, but contrary to Europe, this effect is also important in winter. These values are important compared to the biases mentioned in the previous section, and their robustness has been proven using the simulation ensembles. Figure 11 indeed only shows significant values at the level 0.05. With regards to SST (Fig. 11b), the decrease is estimated at 0.5 °C on average over the Mediterranean Sea, and can reach 1 °C in summer in the Adriatic Sea and near the African coast, which is consistent with the study of Yue et al. (2011) at the global scale. A contrast can also be noticed in winter, spring and autumn between the western and the eastern basins. Moreover, the impact remains high in autumn (−0.5 °C) despite a decrease in the aerosol load. Like 2m-temperature, the aerosol impact on SST is of the same rough size as the bias in RCMs (e.g. Artale et al. 2010; L’Hévéder et al. 2012).

Open image in new window

Table 3

Annual (left column of each cell) difference between the C-AER and the C-NO simulation ensemble means in terms of 2m-temperature (°C) and precipitation (mm/day) over northern Africa, the Mediterranean Sea and Europe, as well as SST (°C) over the Mediterranean Sea

Parameter	Northern Africa		Mediterranean Sea		Europe
2m-temperature	−0.5	−0.5	−0.4	−0.3	−0.4	−0.2
		−0.5		−0.4		−0.4
		−0.4		−0.4		−0.4
		−0.5		−0.4		−0.4
SST			−0.5	−0.4
				−0.5
				−0.6
				−0.5
Precipitation	0.0	0.0	−0.2	−0.2	−0.1	−0.1
		0.0		−0.1		−0.2
		−0.1		−0.1		−0.2
		0.0		−0.2		−0.1

Seasonal averages (DJF / MAM/JJA/SON) are indicated on the right column of each cell for the same regions and parameters. The confidence interval at the level 0.05 calculated from the ensemble spread is for all the regions and seasons ±0.0 °C for 2m and SST, and ±0.0 mm/day for precipitation

As a consequence, the total heat budget at the surface between the two simulations has been changed by aerosols (see Table 4). Over northern Africa, the decrease in SW downward radiation is counterbalanced by a decrease in LW up radiation, as well as in sensible heat loss. The total budget difference caused by aerosols, i.e. the residual of the energy balance, is sligthly negative in spring \((-0.2 \,\hbox {Wm}^{-2})\), and slightly positive in autumn \((+0.2 \,\hbox {Wm}^{-2})\), resulting in a zero annual effect as expected. The soil’s thermal capacity is indeed not high enough to maintain these differences. Over Europe, the decrease in SW downward radiation is counterbalanced by the same fluxes, plus the latent heat loss. This decrease in evaporation causes a positive impact on the heat budget in spring \((+0.4\, \hbox {Wm}^{-2})\), for a zero annual effect. Over the Mediterranean Sea, the annual impact is negative \((-1.5 \,\hbox {Wm}^{-2})\), because the latent heat loss and LW upward flux do not counterbalance totally the SW effect.

As a result of the decrease of the latent heat loss, surface specific humidity also decreases with the presence of aerosols. Figure 11e presents the average seasonal difference of surface specific humidity between C-AER and C-NO. This diminution ranges from −0.1 g/kg in winter to −0.5 g/kg in summer near the western coasts and in the Ionian basin. Coastal land regions are also affected in summer, notably eastern Spain, Tunisia and Italy, indicating probably a reduction of convection because of the increase of atmospheric stability due to the presence of aerosols. Over central Europe, a slight increase in summer can be noted probably due to a modification in atmospheric circulation. In line with this decrease in humidity and increase in stability, the comparison of simulations reveals a decrease of cloud cover (Fig. 11g) with the adding of aerosols, ranging from −1 to −3 % over the Mediterranean Sea and Europe, in spring and summer, but also in autumn over the sea. Consequently, aerosols also tend to reduce precipitation (Fig. 11f): −0.2 mm/day on annual average over the Mediterranean Sea and −0.1 mm/day over Europe, and up to −0.5 mm/day in Italy and Croatia in spring (seasonal values are recalled in Table 3). These modifications of precipitation and evaporation (also reduced by aerosols, figure not shown), lead to a decrease of sea surface salinity (SSS, Fig. 11h). Indeed, evaporation is strongly reduced by aerosols over the Mediterranean Sea (−0.37 mm/day), which prevails over the reduction in precipitation (−0.15 mm/day). The total water budget of the Mediterranean Sea could also be affected by changes in river discharges caused by changes in precipitation over land, but runoffs only decrease by −0.01 mm/day in C-AER. Consequently, SSS is found to decrease by −0.10 psu on average under the effects of aerosols, with a maximum in autumn (−0.12 psu, Fig. 11h). The contrast between the aerosol impact on SSS in north-western and south-eastern Mediterranean is similar to the impact on latent heat fluxes.

To summarize, the direct effect of aerosols is responsible for a significant decrease in surface temperature over Europe and northern Africa, as well as a decrease in SST over the Mediterranean Sea. It can also be seen as a reduction of the activity of the hydrological cycle with a decrease in precipitation, cloud cover and humidity, which will be discussed in details in the next section. The robustness of these results has been confirmed by the simulation ensembles.

Open image in new window

4.4 Impact of ocean–atmosphere coupling on the aerosol climatic response

Until now the present study has used simulations from a fully coupled modeling system, including atmosphere, ocean, land surface and rivers. In this section and in order to understand the role of ocean–atmosphere coupling, simulations have been carried out with the atmosphere model ALADIN only (including the land surface scheme). Figure 12 presents the average summer difference beween the fully coupled simulation ensemble mean differences (C-AER—C-NO) and the same differences in forced simulations (F-AER—F-NO), for different parameters : semi-direct SW (a) and LW (b) aerosol radiative forcings, 2m-temperature (c), surface specific humidity (d), latent (e) and sensible (f) heat losses, precipitation (g) and cloud cover (h). The summer season has been chosen because it is the period of the year when the aerosol forcing is maximal (see Fig. 11).

As in coupled simulations, aerosols scatter and absorb the incident radiation, resulting in a decrease in surface SW radiation. The aerosol DRF is consequently similar in coupled and forced simulations (figure not shown). On the contrary the semi-direct aerosol effect shows significant differences (Fig. 12a and b): the SW semi-direct radiative forcing is higher up to +4 Wm⁻² in coupled simulations, whereas the semi-direct LW effect is higher in coupled simulations in coastal regions of the Mediterranean Sea Fig. 12b. SW and LW radiation has actually been modified in the response in forced simulations because of changes of cloud cover (h) and humidity (d) over the sea essentially, and regions under maritime influence. This could imply aerosols have less impact on clouds in forced simulations over the Mediterranean Sea and its surroundings.

As SSTs are prescribed, no impact of aerosols has been noticed in 2m-temperature over the Mediterranean Sea in forced simulations, as Fig. 12c shows a decrease in temperature similar to the coupled simulations. Over land surfaces, the decrease is generally of the same intensity as in coupled simulations, except over coastal regions (e.g. northern Egypt, Italy, eastern Spain, ...) where the temperature decrease is less important without the ocean–atmosphere coupling. This is due to maritime advection that can bring cooler air in coupled simulations, where aerosols have been able to decrease SST. As SSTs are prescribed, the decrease of the latent heat loss with aerosols (e) is weaker in forced simulations, and almost equal to zero over the sea. As a consequence, surface specific humidity aerosol effect difference between coupled and forced simulations (Fig. 12d) is negative over the Mediterranean Sea, indicating that there is almost no humidity decrease without ocean–atmosphere coupling. Coastal regions such as the North-African coast, eastern Spain and Italy are also affected by this decrease in humidity in coupled simulations only. With regards to precipitation (g), the drying over these coastal regions (Spain, Tunisia, Sicilia) is consequently reinforced in coupled simulations. In forced simulations, the reduction of the impact on the hydrological caused by the prescription of SSTs alleviates this drying effect. The same phenomenon can be noted for cloud cover (h).

To conclude, this comparison between forced and coupled simulations highlights the importance of using a fully coupled RCSM to investigate the response of Mediterranean climate to aerosol direct and semi-direct SW-LW radiative forcing. Otherwise the aerosol effect could be underestimated as shown in previous studies (e.g. Yue et al. 2011).

4.5 Aerosol effects on the Mediterranean Sea

The ocean–atmosphere coupling also enables to study the effect of aerosols on the Mediterranean Sea. As shown previously, aerosols lead to a decrease of SST and an increase of SSS, resulting in changes in water density and water circulation. The prevailing effect has been found to be an increase of sea surface density, potentially favouring ocean deep convection and reinforcing the regional thermohaline circulation. Figure 13 presents the volume of water with density higher than \(29.1\, \hbox {kg/m}^3\) in the Gulf of Lions (a) and \(29.3\, \hbox {kg/m}^3\) in the Adriatic Sea (b). In these two basins, an increase of the volume of dense water is noticed in C-AER in winter, when deep water is formed. In particular, during the winter 2004–2005, known to be a period of strong ocean convection in the Gulf of Lions (Schroeder et al. 2008), the formation of deep water has increased from 1.4 Sv in C-NO to 2.2 Sv in C-AER. The confidence interval calculated from the ensemble spread shows the robustness of these results. Besides, this amplification of oceanic convection with aerosols can also be seen in the meridional (Fig. 13c, d) and zonal (Fig. 13 e−f) overturning streamfunctions for the C-AER simulation and the C-AER—C-NO difference. These functions are calculated from the vertical integration of the meridional (resp. zonal) current velocities averaged along the longitudes (resp. latitudes), as in Somot et al. (2006), thus showing the main circulations in a x-z (resp. Y-z) plan. The increase in the deep water formation in the Adriatic Sea in C-AER leads to an increase in the outflow at the Otranto Strait, with dense water reaching higher depths in the Northern Ionian Sea in C-AER (Fig. 13c, d). With regards to the zonal overturning streamfunction (Fig. 13e, f), the circulation of the surface Atlantic water eastward as well as the westward path of the Levantine intermediate water in the under layer are simulated both in C-AER and C-NO. However, an amplification of the formation of deep water either coming from the Adriatic Sea (16–20°E below 2,000 m) or coming from the Aegean Sea (20–25°E) is found in C-AER (Fig. 13f), thus reinforcing the thermohaline circulation. To summarize, aerosols tend to densify surface water, thus amplifying ocean convection and deep water formation in winter.

Open image in new window

Due to the complexity of aerosol-climate processes involved over the Mediterranean, an elaborated study is presented in the following section using two case studies that illustrate some important processes and possible positive feedbacks in the aerosol–atmosphere–ocean–land system.

5.1 Dust effect in the eastern basin

As mentioned previously, dust aerosols are generally transported from the Libyan desert to the Mediterranean from spring to autumn. These dust outbreaks have significant effects on the vertical profiles of different parameters such as temperature and humidity (e.g. Miller and Tegen 1998). Figure 14 presents the aerosol effects over the eastern Mediterranean during June 2007. This month was indeed submitted to larger than usual dust loads over the Mediterranean Sea (0.14 for AOD at 550 nm over MEDE) and northeastern Africa (0.32 over AFRE), as shown in Fig. 14 (b, top left). Figure 14a presents the vertical profile differences for coupled and forced simulations in temperature, humidity and geopotential, as well as the aerosol vertical distribution, over the eastern Mediterranean Sea (MEDE) and northeastern Africa (AFRE).

Over AFRE, near-surface temperature profiles confirm the cooling due to the absorption and scattering of the incoming solar radiation thus prevented from reaching surface. The decrease in temperature due to aerosols reaches −0.3 °C in forced simulations and −0.4 °C in coupled simulations. These profiles also enable us to highlight the warming between 600 and 950 hPa (between +0.3 and +0.4 °C), due to the SW absorption by dust particles present at the same heights. The simulation ensembles show a weaker uncertainty on the cooling in surface temperature than in the intensity of the warming in the mid-troposphere. This warming can also be noticed over the Mediterranean Sea (between +0.5 and +0.7 °C). However, the decrease in surface air temperature is only present in coupled simulations (−0.3 °C), as SST can be modified by dust aerosols.

As a result, the reduction in near-surface temperature has lead to a decrease in latent heat loss over the Mediterranean Sea (figure not shown), and thus to specific humidity in the lower troposphere (blue curves in Fig. 14a, up to −0.8 g/kg at 1,000 hPa). Over the continent (AFRE), this decrease in humidity is also more important in coupled simulations, reaching −0.4 g/kg between 925 hPa and the surface, than in forced simulations (−0.06 g/kg). Figure 14 (b, top right) shows that the eastern Mediterranean Sea is affected by northerly winds in summer, advecting maritime air to the African continent. As a result, this decrease in humidity has been advected over northeastern Africa. Cooler air is also advected at the surface, reinforcing the decrease in near-surface temperature due to the direct aerosol effect, which explains the differences mentioned below between coupled and forced simulations in the aerosol impact.

These changes in temperature, humidity and air–sea fluxes also affect the atmospheric circulation. Figure 14 (b, bottom) presents the aerosol impact in coupled simulations on wind and geopotential at the surface (left) and at 850 hPa (right). Aerosols tend to stabilize the atmosphere above 900 hPa over the eastern Mediterranean Sea and 850 hPa over notheastern Africa, generating an anticyclonic wind anomaly. At the surface, the difference between C-AER and C-NO shows a decrease in sea level pressure, and a northly wind anomaly over the eastern basin, reinforcing the average northly wind over this region, and its consequences for the aerosol impact. As a consequence, this first case study clearly illustrates a positive feedback between dust aerosols, atmosphere, and ocean, which is stronger in fully coupled simulations than in atmosphere-only ones.

Open image in new window

5.2 Aerosol effect during a European heat wave

The second aerosol event studied in this work occurs in July 2006, when western Europe was affected by a heat wave (Rebetez et al. 2008) with temperature anomalies ranging from 4 to 6 °C in a large zone (France, Benelux, Poland, Switzerland, Germany). The synoptic situation was characterized by high geopotentials over Maghreb, generating a powerful ridge over western Europe, and advecting dry air masses over these regions. The monthly average geopotential and wind at 700 hPa simulated in C-AER are presented in Fig. 16b. High geopotentials are reproduced by the model, showing this southwesterly flow from Morocco to Germany. Compared to the average 2003–2009, surface temperature in July 2006 is higher by 2.9 \(\,\pm\, 0.1\, ^{\circ }\hbox {C}\) in C-AER over EURS, showing the ability of CNRM-RCSM4 to reproduce this heat wave.

This month was also characterized by high dust loads, as shown in Fig. 15, notably over Spain and northwestern Africa. In parallel, sulfate aerosols were prevailing over France and western Europe. Total AOD at 550 nm reaches 0.47 in northwestern Africa (AFRW), 0.23 over the western Mediterranean Sea (MEDW) and 0.22 in southwestern Europe (EURS). These three zones used in this section have been defined in Fig. 2.

Figure 16 presents the vertical profiles (a) of the aerosol impact on temperature, geopotential and specific humidity over these three zones for coupled and forced simulations, as well as the aerosol impact on the atmospheric circulation at 700 hPa (b) and temperature (c). As in the first case over the eastern Mediterranean, the direct effect of dust aerosols causes a cooling at the surface over northwestern Africa (−0.2 °C), and a decrease in SST and near-surface air temperature over the western Mediterranean Sea in the fully coupled simulations. Between 600 and 1000 hPa, the absorption of solar radiation by dust aerosols tends to warm (on average +0.8 °C, see the increase in temperature at 700 hPa, Fig. 16c) and stabilize the atmosphere (see the geopotential profile, Fig. 16 a). This phenomenon increases the geopotential at 700 hPa (b), thus reinforcing the ridge over western Europe. As a consequence, the southwesterly flow responsible for dry and warm air masses advection is reinforced in C-AER compared to C-NO, and can also contribute to the increase in temperature in the mid-troposphere, but also at the surface over southwestern Europe. Figure 16 (c, right) shows an increase in 2m-temperature of 0.5 °C on average over EURS, and up to 1.2 °C locally in southeastern France and in Italy. The warming can also be due to the presence of dust aerosols themselves over this region, and dynamics is probably not the only explanation. At the surface, this study underlines that, in this specific case, the dust dynamic effects (here, increase of the advection of dry air masses) should certainly exceed the cooling effect over land due to the dust surface forcing.

Over the Mediterranean Sea, dust aerosols also prevent the radiation from reaching the surface, thus decreasing SST and latent heat loss (figure not shown). As a consequence, Fig. 16a shows a decrease in specific humidity in the lower troposphere (up to −1.1 g/kg), making the air even drier. This effect is attenuated in the atmosphere-only simulations because of the prescription of SST. This drying caused by aerosols is another factor tending to reinforce the ridge over western Europe. This aerosol effect on ocean–atmosphere fluxes explains why we have observed a more important warming due to aerosols in the coupled simulations than in the forced simulations (see the profiles over EURS and MEDW in Fig. 16a).

One important finding of this case is that, in some situations like this heat wave over Europe, the presence of dust particles associated with the ridge has lead to a positive feedback consisting in a reinforcement of the ridge and an extra warming in the whole troposphere, that exceeds the cooling effect due to the dust dimming. As a result, July 2006 has been noticed to be \(2.9 \,\pm\, 0.1^{\circ }\hbox {C}\) higher than average July in the C-AER simulations, and only \(2.4 \,\pm\, 0.1^{\circ }\hbox {C}\) higher in the C–NO simulation, indicating aerosols are responsible for \(0.5^{\circ }\hbox {C}\) of this heat wave, namely about 15 %. The atmosphere–ocean coupling also enables to take into account the drying of the lower atmosphere due to the decrease in SST, underlying the importance of taking into account the effect of particles on ocean–atmosphere fluxes and SST. It should also be noted that these results cannot be attributed to the RCM internal variability as they have been confirmed by the 6-member ensembles used here. Finally, this study also highlights that the radiative effect of dust needs to be treated in forecast models for improving the simulations of meteorological fields in such specific cases.

Open image in new window

Open image in new window