Large atmospheric shortwave radiative forcing by Mediterranean aerosols derived from simultaneous ground-based and spaceborne observations and dependence on the aerosol type and single scattering albedo
 Aerosol optical properties and shortwave irradiance measurements at the island of Lampedusa (central Mediterranean) during 2004–2007 are combined with Clouds and the Earth's Radiant Energy System observations of the outgoing shortwave flux at the top of the atmosphere (TOA). The measurements are used to estimate the surface (FES), the top of the atmosphere (FETOA), and the atmospheric (FEATM) shortwave aerosol forcing efficiencies for solar zenith angle (θ) between 15° and 55° for desert dust (DD), urban/industrial-biomass burning aerosols (UI-BB), and mixed aerosols (MA). The forcing efficiency at the different atmospheric levels is derived by applying the direct method, that is, as the derivative of the shortwave net flux versus the aerosol optical depth at fixed θ. The diurnal average forcing efficiency at the surface/TOA at the equinox is (−68.9 ± 4.0)/(−45.5 ± 5.4) W m−2 for DD, (−59.0 ± 4.3)/(−19.2 ± 3.3) W m−2 for UI-BB, and (−94.9 ± 5.1)/(−36.2 ± 1.7) W m−2 for MA. The diurnal average atmospheric radiative forcing at the equinox is (+7.3 ± 2.5) W m−2 for DD, (+8.4 ± 1.9) W m−2 for UI-BB, and (+8.2 ± 1.9) W m−2 for MA, suggesting that the mean atmospheric forcing is almost independent of the aerosol type. The largest values of the atmospheric forcing may reach +35 W m−2 for DD, +23 W m−2 for UI-BB, and +34 W m−2 for MA. FETOA is calculated for MA and 25° ≤ θ ≤ 35° for three classes of single scattering albedo (0.7 ≤ ω < 0.8, 0.8 ≤ ω < 0.9, and 0.9 ≤ ω ≤ 1) at 415.6 and 868.7 nm: FETOA increases, in absolute value, for increasing ω. A 0.1 increment in ω determines an increase in FETOA by 10–20 W m−2.
 Aerosols influence many atmospheric processes and may largely affect the radiative budget of the atmosphere and climate on local, regional, and global scales [Forster et al., 2007]. As a consequence of the perturbation to the radiative field, among other effects, aerosols may modify the atmospheric thermal structure, influence dynamical processes [e.g., Kishcha et al., 2003], cloud formation and occurrence [e.g., Ackerman et al., 2000], and the boundary layer structure [e.g., Heinold et al., 2008]. The radiative perturbation induced by aerosols is generally expressed as radiative forcing, which is the difference between the net radiative flux at a given atmospheric level with and without aerosols. The atmospheric forcing (AF) is the amount of energy per unit time stored in the atmosphere due to the presence of aerosols. AF depends primarily on the aerosol amount, size distribution, absorption properties, and vertical distribution. The single scattering albedo, that is, the ratio between the aerosol scattering and extinction coefficients, plays a central role. AF is zero for nonabsorbing particles and is positive for values of the single scattering albedo smaller than 1.
 The determination of AF requires measurements of vertical profiles of the net radiative fluxes and estimates of the aerosol-free fluxes. In few cases direct measurements of the atmospheric net fluxes have been used to derive the aerosol-induced atmospheric heating rate [Gao et al., 2008]. In most cases AF is derived from radiative model calculations based on observed aerosol properties [e.g., Haywood et al., 2001; Formenti et al., 2002; Meloni et al., 2004; Pace et al., 2005; and many others], often as the difference between the radiative forcing at the top of the atmosphere (TOA) and at the surface. Few studies use ground-based and satellite measurements of radiative fluxes to verify and constrain radiative transfer calculation against observed fluxes [Podgorny et al., 2000; Léon et al., 2002; Christopher et al., 2003]. In a limited number of cases AF is derived uniquely from observations, without support of modeling analyses, which in most cases require assumptions on some of the aerosol properties. For example, the direct method [Satheesh and Ramanathan, 2000] uses measurements of radiative fluxes and aerosol optical depth to derive the forcing efficiency (FE), which is the radiative forcing produced by a unit aerosol optical depth, from the slope of the net flux plotted versus the aerosol optical depth. The direct method has been applied to surface and satellite measurements of radiative fluxes obtained in the same region to determine estimates of surface, TOA, and atmospheric aerosol radiative effects [Satheesh and Ramanathan, 2000]. Recent studies have used accurate observations of the outgoing solar and IR fluxes from satellite instruments, such as Clouds and the Earth's Radiant Energy System (CERES) or the Geostationary Earth Radiation Budget [Slingo et al., 2006], in combination with satellite measurements of aerosol optical depths by the Visible Infrared Scanner [e.g., Loeb and Kato, 2002], the Moderate Resolution Imaging Spectroradiometer (MODIS) [Christopher and Zhang, 2002; Zhang et al., 2005], and/or the Multiangle Imaging Spectroradiometer [Patadia et al., 2008], to determine radiative effect at TOA on a regional and global scale.
 Aerosol radiative effects on a regional scale may be much larger than at the global scale due to the large influence of local/regional sources, aerosol properties and atmospheric residence time, hygroscopicity, and the different aerosol removal mechanisms.
 In the Mediterranean, Markowicz et al.  applied the so-called hybrid method to derive surface, TOA, and atmospheric forcing using ground-based and satellite measurements by CERES during 14 days in 2001. The hybrid method uses radiative transfer calculations to estimate the aerosol-free fluxes needed to derive the forcing.
 In a recent study [Di Biagio et al., 2009] we estimated the surface shortwave radiative forcing for different aerosol types observed at Lampedusa (35.5°N, 12.6°E, central Mediterranean) during 2004–2007. In that study we applied the direct method to derive the aerosol radiative forcing at the surface as a function of the solar zenith angle, aerosol type, and aerosol single scattering albedo. The different aerosol types were identified on the basis of the value of the aerosol optical depth and its wavelength dependence, relating aerosol optical properties and backward air mass trajectories, following Pace et al. . The aerosol single scattering albedo was determined as by Meloni et al.  from measurements of optical depth and diffuse-to-global radiation ratio, using radiative transfer model calculations.
 In this study the analysis by Di Biagio et al.  is extended to investigate the role of single scattering albedo and aerosol type on TOA and atmospheric forcing. Accurate determinations of shortwave fluxes and aerosol optical depth and identification of cloud-free periods at the surface are combined with satellite observations from CERES. Colocated, simultaneous ground-based and satellite observations are selected to obtain a consistent data set. The direct method is independently applied to the surface and the TOA observations for different aerosol types and values of the single scattering albedo. The atmospheric forcing is then derived as the difference between TOA and surface forcing for the identified aerosol types. This study, to our knowledge, is the first one allowing the determination of the atmospheric forcing for different aerosol types and single scattering albedo based only on observations. This result is particularly relevant in the Mediterranean region, where the large variability in aerosol amount and properties may lead to significant radiative effects.
 This study is based on simultaneous colocated measurements of the instantaneous shortwave fluxes at TOA, Iup, and at the surface, Idown, and of the aerosol optical properties from the ground obtained during the period June 2004 to August 2007. The outgoing upward shortwave flux at TOA is derived from CERES observations [Wielicki et al., 1996] on board Terra and Aqua satellites. CERES Single Scanner Footprint (SSF) TOA/Surface Fluxes and Clouds data are considered in this study: we use edition 2B (from June 2004 to August 2006) and edition 2F (from September 2006 to August 2007) data from Terra and edition 2B (from June 2004 to April 2006) and edition 2C (from May 2006 to August 2007) data from Aqua. The CERES SSF product contains instantaneous filtered radiances in the shortwave (0.2–5 μm), total (0.2–100 μm), and window (8–12 μm) channels at a resolution of 20 km2 at nadir, combined with MODIS information on aerosol and cloud properties. For each channel, the filtered radiance is first converted to unfiltered radiance through a spectral correction algorithm to compensate for the spectral response of the radiometer and then converted to TOA fluxes by applying an angular dependence model (ADM); in the shortwave case the ADM depends on the observation geometry, surface type, and cloud conditions [Loeb et al., 2005]. In this study, only CERES observations with the whole footprint over the ocean and the field of view (FOV) center at the ocean surface within an area of 1.5° latitude by 1.5° longitude around Lampedusa are considered. Iup data are also corrected for a scaling factor that accounts for spectral darkening of the CERES shortwave optics [Matthews et al., 2005]. The uncertainty associated with the determination of Iup is calculated as the quadratic composition of the uncertainty in the shortwave filtered radiance (∼1% over cloud-free ocean [Wielicki et al., 1996]), the uncertainty in converting filtered to unfiltered radiance (∼1% over cloud-free ocean [Loeb et al., 2001]), and the uncertainty on the dependence of the ADM on solar zenith angle, θ, and aerosol fine mode fraction (∼5% and ∼3% respectively [Loeb et al., 2006]). The calculated total uncertainty in Iup is less than 6%. Iup data from Terra and Aqua are used together without any correction. As discussed by Loeb et al. , at the latitude of Lampedusa the difference between Terra and Aqua instantaneous TOA fluxes is negligible. Moreover, Iup measurements corresponding to different versions of the CERES-SSF data (editions 2B and 2F for Terra and editions 2B and 2C for Aqua) are used together: this has no impact on the results since no changes are applied to the CERES measurements, while only MODIS data are modified in the different versions.
 The aerosol optical properties and surface downward fluxes are measured at Lampedusa with a multifilter rotating shadowband radiometer (MFRSR) and three Eppley precision spectral pyranometers (PSP). The instruments are installed on the roof of the ENEA Station for Climate Observations. The instruments have the horizon free of significant obstacles.
 The MFRSR [Harrison et al., 1994] measures global and diffuse irradiances in six narrowband channels centered at 415.6, 495.7, 614.6, 672.6, 868.7, and 939.6 nm and in one broadband channel (300–1100 nm). The data are averaged over 1 min. Direct irradiances are calculated, for the different channels, as the difference between global and diffuse irradiances and are used to derive the aerosol optical depth, τ. The MFRSR calibration procedure and the aerosol optical depth derivation algorithm are described in detail by Pace et al. . The estimated uncertainty in τ is ∼0.02 [Pace et al., 2006]. The Ångström exponent, α, is defined as the negative slope of τ versus λ on a logarithmic scale and is calculated from the values of τ at 415.6 and 868.7 nm.
 The PSP measures global downward irradiance, Idown, with a uniform responsivity in the spectral range 0.285–2.8 μm. The data are acquired every 30 s. Three different PSPs were alternately installed at the ENEA station during the period June 2004 to August 2007. The measurements of the PSPs are corrected for the thermal offset signal following the method of Dutton et al.  and for the cosine response of the instruments as discussed by Di Biagio et al. . The corrections, calibrations, and estimate of the measurement errors for the three pyranometers are described by Di Biagio et al. . The total uncertainty in Idown is less than 5%.
 The moderate resolution transmittance (MODTRAN 4) radiative transfer model [Anderson et al., 1995] is used to assess the influence of the differences in spectral regions measured at the surface (between 280 and 2800 nm) and from satellite (between 200 and 5000 nm). The calculations are made for different aerosol types and solar zenith angles. The difference between surface net fluxes in the region 280–2800 nm and in the region 280–5000 nm is always smaller than 1%; it is somewhat larger (up to 3.5%) at TOA. The fraction of terrestrial radiation below 5 μm is considered in the calculations. The surface irradiance between 200 and 285 nm is negligible, and we consider the spectral region 0.2–5 μm throughout the following analysis, neglecting the net fluxes between 200 and 285 nm, and between 2800 and 5000 nm at the surface.
 To permit the utilization of data corresponding to different periods of the year, Iup and Idown are scaled to the mean Sun-Earth distance. Only data for cloud-free intervals over both the entire CERES FOV and Lampedusa must be considered in this analysis. CERES data from a limited area around Lampedusa (35.5° ± 0.75°N, 12.6° ± 0.75°E) are used, and cloud-free data are selected using the cloud screen algorithm described by Meloni et al. . This algorithm is an adaptation of the method by Long and Ackerman  and uses measurements of surface global and diffuse irradiances. The original method by Long and Ackermann has been modified to allow a better discrimination between desert dust, frequently observed at Lampedusa, and clouds [Meloni et al., 2007].
3. Determination of the Direct Shortwave Aerosol Radiative Forcing
3.1. Radiative Forcing Calculations
 The shortwave aerosol radiative forcing (RF) at the surface (RFS) and TOA (RFTOA) can be written as follows:
where AS is the surface shortwave albedo, Fnet,S and Fnet,TOA are the shortwave net fluxes at the surface and TOA, respectively, and the subscript a-f identifies aerosol-free conditions. By convention, the flux is positive downward at the surface and upward at TOA. At fixed solar zenith angle, θ, the forcing efficiency at the surface (FES) and TOA (FETOA) can be defined as follows:
Following Satheesh and Ramanathan , the value of FE is derived calculating the slope of the linear fit between experimental values of Fnet and τ. The atmospheric forcing efficiency, FEATM(θ), is calculated as the difference between FE(θ) at TOA and at the surface. RF(θ) is obtained multiplying FE(θ) by τ.
 The dependence of FE on solar zenith angle is due to the combination of different factors. As θ increases, the path of the solar radiation in the atmosphere, the attenuation, and the fraction of diffuse radiation increase, especially at shorter wavelengths. Consequently, the forcing efficiency displays a dependence on the solar zenith angle [e.g., Nemesure et al., 1995; Formenti et al., 2002; Meloni et al., 2005], and estimates of the forcing efficiency must be derived at fixed θ.
 Radiative transfer model calculations show that FES is relatively constant for solar zenith angles smaller than 45° and increases (in absolute value) for increasing values of θ; the value of solar zenith angle where FES starts increasing depends on the aerosol properties [Formenti et al., 2002; Meloni et al., 2005]. The forcing at TOA decreases with θ (sign included) for solar zenith angles smaller than 60°–75°, depending on the aerosol properties, and increases for larger angles [Nemesure et al., 1995; Formenti et al., 2002; Meloni et al., 2005].
3.2. Aerosol Types
 The aerosol radiative effect depends on a variety of parameters related to the size, shape, amount, vertical distribution, and composition of the particles. The size distribution influences the aerosol radiative effect through two mechanisms: by regulating the intensity of the aerosol-radiation interaction in the different portions of the spectrum (i.e., how the aerosol optical depth varies with wavelength), and by determining the angular distribution of the scattering, through the phase function. Both mechanisms should produce a dependence of the forcing efficiency on the Ångström exponent. The composition of the particles also plays a large role since it determines (together with size) the relative weight of scattering and absorption, as described by the single scattering albedo. The wavelength dependence of the single scattering albedo also plays a role in determining the dependence of FE on aerosol type [Meloni et al., 2006] and θ. Thus, the main parameters that drive the aerosol forcing efficiency, which are independent of the absolute value of the optical depth, are Ångström exponent and single scattering albedo.
 The aerosol observations obtained at Lampedusa in the period June 2004 to August 2007 simultaneously with satellite and radiative flux ground-based measurements are classified in three categories on the basis of the particle optical properties, following the method of Pace et al. : desert dust (DD) for τ ≥ 0.15 and α ≤ 0.5, urban/industrial-biomass burning aerosols (UI-BB) for τ ≥ 0.1 and α ≥ 1.5, and mixed aerosols (MA) for τ > 0 and 0.5 < α < 1.5, τ < 0.15 and α ≤ 0.5, and τ < 0.1 and α ≥ 1.5. This method is based on the analysis by Pace et al. , who relate aerosol optical properties measured at Lampedusa and air mass trajectories. Their analysis shows that particles of different origin and type correspond to different regions of the τ-α graph. Figure 1 shows the behavior of the Ångström exponent versus the optical depth at 495.7 nm for the three identified aerosol classes observed at Lampedusa during June 2004 to August 2007. DD are particles transported from Africa and are characterized by large dimensions and moderate to high columnar load. UI-BB are small particles generally coming from central-eastern Europe, in some cases produced in forest fires [Pace et al., 2005, 2006]. The MA class includes cases of pure marine aerosols and mixtures of different particle types (dust, pollution, biomass burning, and marine particles). As discussed by Pace et al. , clean marine conditions are rare at Lampedusa and are generally associated with subsiding air masses which have spent a small fraction of time over continental areas. Due to the small number of cases, pure marine aerosols are not considered as a separate class. In addition, their low aerosol optical depth (between 0.07 and 0.11 [Pace et al., 2006]) makes unpractical the use of the direct method for the determination of the forcing. Because of the complexity of the Mediterranean environment, the mixed aerosol is the most frequent class at Lampedusa. Strong events of DD and UI-BB are generally observed in spring and summer, while MA is more frequent in winter. DD, UI-BB, and MA represent, respectively, 27%, 6%, and 67% of all the observed data (AD). The average optical depth and Ångström exponent for the different aerosol types are reported in Table 1.
Table 1. Averages of the Aerosol Optical Depth at 495.7 nm, Ångström Exponent, and Single Scattering Albedo for the Three Different Aerosol Types Over June 2004 to August 2007a
τ at 495.7 nm
ω at 415.6 nm
ω at 868.7 nm
Standard deviations of the averages are also shown. α, Ångström exponent; τ, aerosol optical depth; and ω, single scattering albedo. Abbreviations are as follows: AD, all the observed data; DD, desert dust; MA, mixed aerosols; UI-BB, urban/industrial-biomass burning aerosols.
0.31 ± 0.13
0.25 ± 0.15
0.76 ± 0.03
0.89 ± 0.05
0.21 ± 0.09
1.60 ± 0.09
0.91 ± 0.06
0.81 ± 0.04
0.14 ± 0.08
0.85 ± 0.37
0.81 ± 0.06
0.82 ± 0.07
0.19 ± 0.12
0.73 ± 0.46
 Following Meloni et al. , estimates of the single scattering albedo, ω, are retrieved at 415.6 and 868.7 nm at θ = 60° for the observed cases of DD, UI-BB, and MA. The method of Meloni et al.  is based on the measurements of the diffuse-to-global irradiance ratio and τ at 415.6 and 868.7 nm, both obtained from MFRSR measurements; at a fixed solar zenith angle and value of τ, the diffuse-to-global irradiance ratio depends on the aerosol single scattering albedo and the value of ω is derived using radiative transfer model calculations. The estimated uncertainty in ω is <0.05 at 415.6 nm and <0.07 at 868.7 nm [Meloni et al., 2006]. The single scattering albedo varies between 0.7 and 1.0 for all the aerosol types. Average values of ω for the different aerosol types are reported in Table 1. The single scattering albedo increases (decreases) for increasing wavelength for DD (UI-BB), while it presents an almost neutral behavior for MA. A more detailed description of the optical properties of DD, UI-BB, and MA is reported by Di Biagio et al. .
 For overhead Sun the median of the solar spectrum is around 680 nm, and, depending on the aerosol properties, a dependence of FE on the aerosol size necessarily arises when the forcing efficiency is derived using the optical depth at 500 nm. At low values of α, as is the case for DD, the aerosol optical depth at 500 nm is representative of the optical depth throughout the spectrum. Conversely, for large values of α the optical depth decreases rapidly with wavelength, and τ at 500 nm may not be representative of the optical depth effective for the whole solar spectrum. As shown by di Sarra et al. , due to this effect, FES for UI-BB changes greatly when calculated with respect to the optical depth at 671 nm instead of 500 nm; conversely, small changes appear for DD and MA.
3.3. Surface Shortwave Albedo
 A parameterization of the surface shortwave albedo, AS, is needed to evaluate Fnet,S and to determine FES(θ) (equation (3)).
 The Station for Climate Observation is located on a 50 m high plateau along the northeastern coast of Lampedusa, about 15 m from the coastline. Thus, contributions of both land and ocean have to be taken into account in the determination of AS. The surface albedo is calculated as the weighted average of land and ocean albedo over a circle of 5 km radius around the measurement station (28% contribution from land and 72% from ocean). The monthly mean MODIS measurements at Lampedusa in 2003 and 2004 are used for the land albedo. The ocean albedo is calculated as a function of θ, τ, and wind speed following Jin et al. . A detailed description of the albedo determination is reported by Di Biagio et al. . The overall surface shortwave weighted albedo varies from 0.06 to 0.21. The uncertainty in AS is smaller than 6%. The overall error in Fnet,S increases by less than 0.3% due to the uncertainty in AS, and this contribution has been neglected.
 At the surface, Fnet,S is measured over a mixture of land and ocean, and FES(θ) is derived accordingly. As stated above, only CERES measurements with the whole footprint occupied by the ocean surface are selected, and thus the calculated FETOA corresponds to the aerosol forcing over the ocean. The MODTRAN 4 radiative transfer model [Anderson et al., 1995] is used to determine the value of C, which is a correction factor defined as the ratio between the surface net flux over the ocean, Fnet,O, and Fnet,S. The surface net flux is simulated as a function of θ, τ, and aerosol type, considering pure ocean surface and the land/ocean mixed shortwave albedo AS. Three different aerosol types are considered: desert dust, urban-industrial, and maritime aerosols. The desert dust and the urban-industrial types are used to derive C for cases of DD and UI-BB, respectively. For MA the correction factor is calculated as the average of the value of C obtained for the three different aerosol types.
 The dependence of C on the solar zenith angle is smaller than 0.07% for 15° ≤ θ ≤ 45° for all the aerosol types and has been neglected. C is thus calculated for θ = 25° as a function of the aerosol type for values of τ between 0 and 0.5: C decreases for increasing τ for all aerosol types and varies between 1.089 and 1.083 for DD, between 1.089 and 1.088 for UI-BB, and between 1.089 and 1.084 for MA. The retrieved values of C are fitted with a second-degree polynomial as a function of τ for each aerosol type, and the fitting polynomial is applied to convert Fnet,S to Fnet,O. The uncertainty in C is about 0.5%. The overall error in Fnet,S increases by less than 0.03% due to the uncertainty in C, and this contribution has been neglected.
 The derivation of the forcing efficiencies proceeds as follows: (1) simultaneous and colocated ground-based and CERES observations at Lampedusa of Idown and Iup are selected, (2) AS is calculated taking into account solar zenith angle, wind velocity, and aerosol optical depth and is used to derive Fnet,S, (3) the factor C and Fnet,S are used to calculate Fnet,O, and (4) the forcing efficiencies at the surface and at TOA are derived as explained in section 3.1 at fixed solar zenith angles and for different aerosol types.
4. Results and Discussion
 FETOA(θ) and FES(θ) are calculated for the different aerosol types and AD, calculating the slope of the least squares linear fit between net radiative fluxes versus τ(θ) for four intervals of solar zenith angle, 15° ≤ θ ≤ 25°, 25° ≤ θ ≤ 35°, 35° ≤ θ ≤ 45°, and 45° ≤ θ ≤ 55°. Data points more than two standard deviations from the fitting line are removed to eliminate possible residual contamination by clouds. The fitting line is recalculated after removal of the outliers. The uncertainty in the retrieved values of FETOA(θ) and FES(θ) is estimated from the uncertainty in the slope of the least squares linear fit, taking into account the estimated uncertainties on the radiative fluxes. The uncertainties in θ, AS, and C are small and have been neglected. FEATM(θ) is calculated as the difference between FETOA(θ) and FES(θ), and the uncertainty in the retrieved atmospheric forcing efficiency is calculated with the error propagation formula. We stress that the subtraction of the forcing efficiencies at the surface and TOA is possible here because only simultaneous and colocated observations at the ground and from satellite are selected.
Figure 2 displays the behavior of the upward TOA shortwave flux and surface shortwave net flux versus the aerosol optical depth at 495.7 nm for 25° ≤ θ ≤ 35° for DD, UI-BB, and MA. Least squares linear fits are also shown for the different aerosol types and for AD. Table 2 shows the retrieved values of FETOA, FES, and FEATM for the different aerosol classes and the different solar zenith angle intervals. A small number of UI-BB aerosol cases are present in the data set, and FETOA may be retrieved only for 15° ≤ θ ≤ 25° and 25° ≤ θ ≤ 35°. Estimates of FES for MA at 45° ≤ θ ≤ 55° and for UI-BB at all solar zenith angles intervals are not possible due to the limited data.
Table 2. Instantaneous Shortwave Top of Atmosphere, Surface, and Atmospheric Forcing Efficiencies Retrieved at Four Intervals of Solar Zenith Angles Between 15° and 55° for the Three Different Aerosol Types and for All the Observed Dataa
Solar Zenith Angle Interval (°)
FETOA ± σFE (W m−2)
FES ± σFE (W m−2)
FEATM ± σFE (W m−2)
Number of Points TOA
Number of Points Surface
The number of data points considered at top of atmosphere and at the surface are also reported. FETOA, FES, and FEATM are top of atmosphere, surface, and atmospheric forcing efficiencies, respectively.
−36.4 ± 6.1
−247 ± 22
230 ± 26
−65.0 ± 5.1
−151 ± 18
86 ± 19
−64 ± 21
−136 ± 12
72 ± 24
−36 ± 12
−49.2 ± 8.8
−38.0 ± 4.1
−309 ± 16
271 ± 17
−58.2 ± 3.8
−236 ± 16
178 ± 16
−72.3 ± 6.8
−212 ± 33
148 ± 35
−164 ± 13
−45.7 ± 2.6
−293 ± 10
247 ± 10
−69.2 ± 2.4
−233.1 ± 9.7
164 ± 10
−72.7 ± 5.9
−244 ± 33
171 ± 34
4.1. Surface Forcing Efficiencies
 This study includes only data measured simultaneously to satellite overpasses, and a more limited number of surface observations are included in the data set with respect to the analysis by Di Biagio et al. . In that analysis all cloud-free data in the period 2004–2007 are used to derive the surface forcing; their results are believed to be representative for the average properties of the identified aerosol types in the central Mediterranean. The consistency of the retrieved values of FES with respect to those derived by Di Biagio et al.  is used here as a method to assess the representativeness of the retrieved results. Thus, we assume that the data set used in this study describes the average properties of the considered aerosol type when the values of FES are in agreement with those of Di Biagio et al. ; if this is not the case, we assume that the data set is biased toward particular aerosol conditions which, due to the data selection process, play a large role.
 FES depends on the surface albedo, and values of FES larger than those obtained by Di Biagio et al.  are expected over ocean. By comparing the values of FES in Table 2 with those obtained by Di Biagio et al. we obtain a good agreement for six of nine determinations. Taking into account the estimated uncertainties, FES is larger (in absolute value) at 15° ≤ θ ≤ 25° for both DD and MA with respect to the values obtained for the same aerosol classes at θ = 20° by Di Biagio et al. , while it is comparable or lower at 25° ≤ θ ≤ 35° and 35° ≤ θ ≤ 45° with respect to the values at θ = 30° and 40° reported by Di Biagio et al. .
 Experimental data in Table 2 show that FES decreases in absolute value for increasing θ between 15° and 45° for all the aerosol types. The same behavior was observed in previous work recently conducted at Lampedusa (di Sarra et al., 2008; Di Biagio et al., 2009).
 As shown in Table 2, the largest values of the FES are obtained for MA. Mixed aerosols display a low value of the single scattering albedo throughout the solar spectrum (ω ≈ 0.80 both at 415.6 and 868.7 nm): thus, while DD and UI-BB show high absorption at short and long wavelengths, respectively, MA absorbs over all the solar spectrum. This effect, in addition to the relatively high values of the optical depth throughout the spectrum with respect to the value at 500 nm, contributes to produce the observed largest value of FES for MA.
4.2. Top of Atmosphere and Atmospheric Forcing Efficiencies
 The number of data points obtained at the surface is in general lower than at TOA because several satellite observations falling in the selection region are associated with a single surface measurement. The values of FETOA must be seen with some caution when the derivation of the forcing efficiency at the surface is not available; in these cases the values of FETOA, as discussed above, may be biased toward few specific cases.
 The largest value of FETOA, −164 W m−2, is obtained for MA at 45° ≤ θ ≤ 55°. Due to the limited number of data points, no estimate of the surface forcing efficiency is possible, and this result is probably not representative for the average MA class.
 FETOA increased in absolute value with θ (see Table 2) for all aerosol classes. As discussed by Nemesure et al. , the dependence of FETOA on solar zenith angle is primarily determined by the increase of the upscatter fraction with θ.
 FEATM decreases for increasing θ both for DD and MA. Largest values are obtained for MA. Also in this case, this effect is believed to be due to the combination of a relatively small Ångström exponent and a low single scattering albedo, which implies relatively elevated values of the absorption optical depth throughout the solar spectrum.
4.3. Dependence on the Single Scattering Albedo
 Measurements corresponding to DD and MA in the solar zenith angle interval 25° ≤ θ ≤ 35° are grouped in three classes of single scattering albedo (0.7 ≤ ω < 0.8, 0.8 ≤ ω < 0.9, and 0.9 ≤ ω ≤ 1) for ω at both 415.6 and 868.7 nm. FETOA is then calculated separately for each class. The determinations of ω are made at θ = 60° [Meloni et al., 2006], and we assume that the estimated single scattering albedo remains constant between 25° ≤ θ ≤ 35° and θ = 60°. Figure 3 shows the behavior of the upward shortwave TOA flux versus τ at 495.7 nm for 25° ≤ θ ≤ 35° for DD and MA and for the various classes of ω at 415.6 and 868 nm. Least squares linear fits are shown for the different classes of ω. Table 3 shows the retrieved values of the instantaneous FETOA. FETOA is always negative and increases in absolute value for increasing ω for MA: a 0.1 increment in ω at 415.6 and 868.7 nm determines an increase in FETOA by 10–20 W m−2, corresponding to about 25%. Due to the large uncertainties on the retrieved forcing efficiencies for DD, no clear conclusion can be drawn for this type of particle.
Table 3. Instantaneous Shortwave Top of Atmosphere Forcing Efficiency at 25° ≤ θ ≤ 35° for Desert Dust and Mixed Aerosols for Different Values of the Single Scattering Albedo
Single Scattering Albedo ω
Top of Atmosphere Forcing Efficiency (W m−2)
−70.0 ± 5.0
−42.5 ± 7.8
−71 ± 25
−56.7 ± 5.5
−64 ± 21
−42.5 ± 8.7
−62 ± 10
−63.7 ± 6.6
−62.3 ± 7.7
−72 ± 16
Di Biagio et al.  analyzed the dependence of the surface forcing efficiency on the aerosol single scattering albedo: FES at 60° decreases with increasing ω at 868.7 nm for DD and MA, while it decreases with increasing ω at 415.6 nm for MA and increases for DD. A 0.1 increment in the single scattering albedo at 868.7 nm produces a reduction in FES by 25–30 W m−2 for both DD and MA. These values correspond to a reduction by about 10%–15% for MA and 15%–20% for DD. Combining the results of this work with those obtained by Di Biagio et al., we derive an estimate of the sensitivity of the atmospheric forcing on the aerosol single scattering albedo: an increase of 0.1 in ω at 868.7 nm determines a reduction of FEATM by 30%–45% for MA.
4.4. Daily Mean Forcing Efficiencies
 Model estimates of the ratio, r, between the instantaneous and the daily mean forcing efficiency, FEd, at TOA for the equinox and the summer solstice are retrieved from MODTRAN calculations. Three different particle types (desert dust, urban-industrial, and maritime aerosols) and four intervals of solar zenith angle (15°–25°, 25°–35°, 35°–45°, and 45°–55°) are considered. The desert dust and the urban-industrial types are used to derive r for cases of DD and UI-BB, respectively, while for MA r is calculated as the average of the values obtained for the three different aerosol types. FEd is calculated at TOA and at the surface for both the equinox and the summer solstice and for each aerosol type, multiplying the instantaneous values of FETOA (values in Table 2) by the corresponding r and then averaging the obtained values. The uncertainty in the retrieved FEd is calculated with the error propagation formula.
 For the surface forcings r is calculated from the instantaneous and the daily mean forcing efficiencies reported by Di Biagio et al. . The atmospheric FEd is calculated as the difference between the daily mean forcing efficiency obtained at TOA and at the surface. For the UI-BB class, we use the values of FEd at the surface by Di Biagio et al. .
Table 4 shows the retrieved FEd at TOA, surface, and in the atmosphere at the equinox and summer solstice for DD, UI-BB, and MA. FEd at TOA is largest for DD, while FEd at the surface is largest for MA; lowest values of FEd are obtained for UI-BB at both TOA and the surface. The atmospheric FEd is largest for MA, intermediate for UI-BB, and lowest for DD both at the equinox and at the summer solstice. These results indicate that the atmospheric forcing is ∼30%–50% of the surface forcing for DD, ∼70% for UI-BB, and ∼60% for MA. This large radiative effect may induce a significant perturbation to the atmospheric thermal structure, stability, and cloud properties.
Table 4. Top of Atmosphere, Surface, and Atmospheric Daily Forcing Efficiencies at Equinox and Summer Solstice for the Three Different Aerosol Typesa
FEd at the Equinox (W m−2)
FEd at the Summer Solstice (W m−2)
The ratio f between the surface and TOA daily efficiency at the equinox and Δτa,700 (see text) are also reported. Abbreviations are as follows: FEd, daily forcing efficiency; TOA, top of atmosphere.
Table 5. Daily Average Forcing Efficiencies Derived Over the Sea at Top of Atmosphere, Surface, and Atmosphere for Different Aerosol Types in the Mediterranean Basina
(FEd)TOA (W m−2)
(FEd)S (W m−2)
(FEd)ATM (W m−2)
The ratio f between the surface and the top of atmosphere daily efficiency is also calculated. (FEd)TOA, (FEd)S, and (FEd)ATM are the daily average forcing efficiencies derived over the sea at top of atmosphere, surface, and atmosphere, respectively. Abbreviations are as follows: aut, autumn; spr, spring; sum, summer; win, winter.
 As mentioned above, size distribution and single scattering albedo play a central role in determining the aerosol direct forcing.
 With the aim of disentangling the complex dependencies of FEd we have calculated the change in absorption aerosol optical depth at 700 nm, Δτa,700 (Table 4), that is, approximately at the median of the solar spectrum, corresponding to a unit change in aerosol optical depth at 500 nm. The calculation is made using the mean optical properties of DD, UI-BB, and MA and the single scattering albedo at 868.7 nm, shown in Table 1. The value of Δτa,700 is largest for MA and smallest for DD (it is 0.101, 0.111, and 0.135 for DD, UI-BB, and MA, respectively), as is the case for FEATM. This suggests that, once size distribution differences are taken into account, the change in absorption optical depth is the single main factor driving the redistribution of the radiation in the atmosphere.
4.5. Daily Mean Radiative Forcings
 The daily mean radiative forcing, RFd, is calculated by multiplying the FEd values in Table 4 by the average optical depth for DD, UI-BB, and MA. The obtained values of RFd at the equinox and summer solstice are shown in Figure 4. RFd values at TOA and at the surface are largest for DD due to the high value of both forcing efficiency and average optical depth. Similar values of the atmospheric RFd are obtained for all the aerosol types. Largest atmospheric RFd values are observed for UI-BB at the equinox and for DD at the summer solstice.
 The obtained values suggest that the mean atmospheric RFd is reasonably independent of the aerosol type since variations in the forcing efficiency appear to be balanced, on the average, by different mean values of the aerosol optical depth.
 By multiplying the atmospheric FEd by the highest value of the optical depth observed for each aerosol class (0.88 for DD, 0.44 for UI-BB, and 0.45 for MA) we derive an atmospheric RFd at the summer solstice of +35 W m−2 for DD, +23 W m−2 for UI-BB, and +34 W m−2 for MA. These values represent an estimate of the maximum obtained atmospheric RF.
4.6. Comparison With Other Studies
 In Table 5, the results of our analysis are compared with estimates of the aerosol radiative forcing obtained over the sea in a selection of other studies conducted in the Mediterranean region in recent years. The reported results span a relatively large interval of values, indicating the variability in aerosol optical properties and distribution in the Mediterranean basin.
 Few studies, considering those dedicated to the estimation of the radiative forcing of Mediterranean aerosols, have analyzed the dependence of the forcing at TOA on the aerosol absorption. Vrekoussis et al.  studied, for a mixture of dust, nonsea sulfate, and biomass burning aerosols observed at Crete (eastern Mediterranean) during 2001–2002, the dependence of the forcing at TOA on the values of the single scattering albedo at 550 nm and retrieved efficiencies between −99 and −26 W m−2 for single scattering albedo between 0.98 and 0.74.
Saha et al.  derived, for continental pollution aerosols in southern France during 4 days in 2006, a TOA forcing efficiency of −10.6, −13.7, −16.6, and −17.5. The aerosol events were characterized by values of the single scattering albedo at 525 nm of 0.76, 0.85, 0.80, and 0.89. The TOA forcing efficiency does not show a clear dependence on ω.
Santos et al.  derived, for aerosols observed in a coastal site in Portugal during 2004 and 2005, values of FETOA between −40 and −55 W m−2 for desert dust particles with ω at 550 nm between 0.93 and 0.96. For forest fire/urban aerosols with ω at 550 nm between 0.90 and 0.95 they derived forcing efficiency at TOA between −10 and −30 W m−2.
 Colocated simultaneous surface and satellite measurements of the aerosol optical properties, surface shortwave irradiance, and outgoing shortwave flux at the top of atmosphere obtained in the period June 2004 to August 2007 are used to derive the direct shortwave aerosol radiative forcing at the surface, TOA, and in the atmosphere. Three different aerosol types are distinguished: desert dust, urban/industrial-biomass burning aerosols, and mixed aerosols. The aerosol radiative forcing at surface and TOA are derived for each aerosol class by applying the direct method [Satheesh and Ramanathan, 2000]. FEATM is derived as the difference between TOA and surface forcing.
 The results of our analysis may be summarized as follows:
 1. The average optical depth is 0.31 for DD, 0.21 for UI-BB, and 0.14 for MA. The average Ångström exponent is 0.25 for DD, 1.60 for UI-BB, and 0.85 for MA. The average single scattering albedo at 415.6/868.7 nm is 0.76/0.89 for DD, 0.91/0.81 for UI-BB, and 0.80/0.82 for MA.
 2. Estimates of FES, FETOA, and FEATM are derived for solar zenith angles between 15° and 55° and for the different aerosol classes. FES decreases in absolute value with θ for both DD and MA; estimates of FES are not available for the UI-BB class. FETOA increases in absolute value for increasing θ for both UI-BB and MA; FETOA for DD increases for θ between 15° and 35°, while is almost constant for higher values of the solar zenith angle. FEATM decreases with θ for all the aerosol types.
 3. The daily mean aerosol forcing efficiency is derived at the equinox and at the summer solstice. FEd is largest for DD at TOA and for MA at the surface; lowest values are obtained for UI-BB both at TOA and at the surface. The atmospheric FEd is largest for MA and lowest for DD. FEd at TOA at the equinox is (−45.5 ± 5.4), (−19.2 ± 3.3), and (−36.2 ± 1.7) W m−2 for DD, UI-BB, and MA, respectively. FEd at the surface is (−68.9 ± 4.0) W m−2 for DD, (−59.0 ± 4.3) W m−2 for UI-BB, and (−94.9 ± 5.1) W m−2 for MA. The atmospheric FEd is (+23.4 ± 6.7) W m−2 for DD, (+39.8 ± 5.4) W m−2 for UI-BB, and (+58.7 ± 5.4) W m−2 for MA. These results indicate that the atmospheric forcing is ∼30%–50% of the surface forcing for DD, ∼70% for UI-BB, and ∼60% for MA.
 4. Values of the daily mean aerosol radiative forcing, RFd, at TOA and at the surface are largest for DD due to the high value of both forcing efficiency and average optical depth. The atmospheric RFd at the equinox and at the summer solstice is (+7.3 ± 2.5) and (+12.5 ± 3.1) W m−2 for DD, (+8.4 ± 1.9) and (+11.0 ± 2.9) W m−2 for UI-BB, (+8.2 ± 1.9) and (+10.7 ± 2.5) W m−2 for MA, respectively, suggesting that the mean atmospheric forcing is approximately independent of the aerosol type.
 5. Cases of DD and MA corresponding to the solar zenith angle interval 25° ≤ θ ≤ 35° are grouped in three classes of single scattering albedo (0.7 ≤ ω < 0.8, 0.8 ≤ ω < 0.9, and 0.9 ≤ ω ≤ 1) at 415.6 and 868.7 nm, and FETOA is calculated for each class. FETOA increases in absolute value for increasing ω for MA: a 0.1 increment in ω at 415.6 and 868.7 nm produces an increase in FETOA by 10–20 W m−2.
 6. The main parameters determining the forcing efficiency are the single scattering albedo and the Ångström exponent (i.e., the aerosol absorption and the size distribution). The atmospheric forcing efficiency seems to respond primarily to the aerosol absorption, while the TOA forcing efficiency shows a larger sensitivity to the Ångström exponent.
 This work was partly supported by the Aeroclouds and NOMAC projects, funded by the Italian Ministry for Universities and Research. CERES data were obtained from the NASA Langley Research Center Atmospheric Science Data Center. Contributions by F. Monteleone, S. Piacentino, and D. Sferlazzo are gratefully acknowledged. Helpful comments and suggestions by two anonymous reviewers are also acknowledged.