The variability of aerosol optical properties over the Mediterranean basin, in the ultraviolet (UV) and visible (VIS) spectral regions, is examined, using ground-based data from eight stations of the AErosol RObotic NETwork (AERONET), climatological values from the AeroCom (Aerosol Comparisons between Observations and Models) project and data from the MODerate resolution Imaging Spectroradiometer (MODIS) on-board the Terra and Aqua satellites. The aerosol optical properties from the three datasets are used as input data in a radiative transfer model, in order to estimate the differences in irradiance at the ground under cloud-free skies. The MODIS aerosol optical depth climatology shows better agreement with AERONET data. The highest difference in the monthly average values is equal to 0.09 at 550 nm, while the differences between the AERONET and the AeroCom climatologies reach 0.25 and 0.15 in the UV and VIS wavelengths, respectively. As a result, the AERONET modelled VIS and UV irradiances are closer to MODIS, with the absolute differences in average values reaching 15 (6%) and 2 W m–2 (6%), respectively, while absolute differences with AeroCom irradiances can reach up to 30 (12%) and 4.5 W m–2 (12%). The differences are higher in FORTH Crete, Nes Ziona, Sede Boker and Blida, which are influenced by desert dust aerosols. The separate effect on irradiances due to differences in aerosol optical depth and single scattering albedo from the three datasets has been also examined; according to results, the effect of aerosol optical depth is dominant. At this stage, the use of MODIS aerosol climatology would be recommended for calculations of UV and VIS irradiance in the Mediterranean basin. However, the AeroCom climatology, with some further improvements in the dust cycle, could be a valuable tool in future studies in this area, due to the provision of natural and manmade aerosol optical properties with higher spectral resolution.
Aerosols emitted in the atmosphere through natural processes (e.g. dust, sea salt, volcano eruptions) and anthropogenic activities (e.g. fossil fuel, industries, transport) have a direct effect, by scattering and absorbing the incoming solar radiation, and indirect effect, by altering cloud properties, in the Earth's energy balance. The radiative forcing efficiency of aerosols holds one of the highest uncertainties (IPCC 2007) and ground-based measuring networks, global aerosol models and satellite instruments have been developed to estimate their optical properties, formation, transportation and role in the climate system.
The AErosol RObotic NETwork (AERONET) is a worldwide ground network established by NASA for the study of atmospheric aerosols (Holben et al., 2001). It comprises over 100 stations which continuously provide data of aerosol optical properties, based on direct sun and sky radiance measurements by Cimel sun photometers (Holben et al., 1998, Dubovik and King, 2000, Dubovik et al., 2000, Dubovik et al., 2002, Gobbi et al., 2007). The AERONET data have been used extensively in studies related to the validation of satellite aerosol products retrievals (Chu et al., 2002, Torres et al., 2002, Kahn et al., 2005, Vidot et al., 2008, De Meij and Lelieveld, 2011), validation of other ground-based instruments aerosol properties (Gröbner and Meleti, 2004, Cheymol et al., 2009), the spatio-temporal variability of aerosols characteristics (Smirnov et al., 2002, Bäumer et al., 2008, Georgoulias and Kourtidis, 2011) and the calculation of aerosol direct radiative forcing (García et al., 2006, Bergamo et al., 2008, Clerici and Mélin, 2008).
The Aerosol Comparisons between Observations and Models (AeroCom) project is an international effort aiming to a better understanding of the effect of aerosols in climate. Estimations from more than 14 global models were compared with each other and with satellite and ground data, in order to improve the aerosol modules in the global models and minimize the uncertainty of their radiative effect (Kinne et al., 2006). These data, representing the AeroCom ensemble model, are provided with a spatial resolution of 100 × 100 km (1° × 1°). An initial assessment of AeroCom aerosol properties was done by Kinne et al. (2006). In general, a quite successful agreement was found in the annual global average of aerosol optical depth (AOD), at 550 nm, with AeroCom values being at 0.11–0.14 and AERONET at 0.135. However, differences in regional distribution and the contribution of dust and carbonaceous aerosols have been revealed (Textor et al., 2007), indicating that further improvement and understanding of the aerosol processes, chemistry, transport and removal mechanisms is needed. An estimation of the aerosol direct radiative forcing, based on nine AeroCom models, was provided by Schulz et al. (2006), giving a global annual value of −0.22 W m–2 at the top of the atmosphere. The dust transport and deposition was addressed in Huneeus et al. (2011), where 15 AeroCom models were compared with each other in the reproduction of the dust characteristics and its direct radiative effect. According to results, considerable differences were revealed among the models in the simulation of the dust cycle and, generally, the AOD and the particle size were overestimated.
In order to gain better spatial coverage, measuring systems in satellites have been used to observe the Earth's atmosphere, land and ocean. MODerate resolution Imaging Spectroradiometer (MODIS) is the instrument onboard the NASA's Terra and Aqua satellites, which has been providing data regarding the aerosol optical characteristics since February 2000 (Terra) and July 2002 (Aqua). The two satellites are in a sun-synchronous, near-polar, circular orbit and view the entire Earth's surface every 1–2 days (Remer et al., 2005). MODIS data have been used to study the seasonal variation of aerosols and discrimination of their different types, biomass burning aerosols and dust events (Arola et al., 2007, Santese et al., 2007), as well as calculate the aerosol direct radiative effect (Zhou et al., 2005). The new algorithm used for processing the acquired data, labelled Collection 5, (Levy et al., 2007a, 2007b) provides more accurate results when compared to AERONET. Remer et al. (2008) showed that the new algorithm improves the retrievals over land; over non-bright surfaces, the expected error in AOD over land is within ± (0.05 + 0.15 AOD). However, the reliability of data over bright surfaces (e.g. deserts) is still limited. Papadimas et al. (2009) compared the MODIS Collections 4 and 5 and AERONET data over the Mediterranean basin and stated that their correlation coefficient was increased from 0.66 to 0.76 and the overall differences were within the expected range of uncertainties of the two datasets.
The area of this study, the Mediterranean basin, exhibits a large variety of aerosols, such as desert dust from North Africa and the Middle East, maritime aerosols from the Mediterranean Sea, as well as anthropogenic aerosols from the European coastal cities, resulting in significantly higher concentrations (from 2 to 10 times) than over the least polluted environments at northern latitudes (Lelieveld et al., 2002). Several studies have been conducted for this area, using both ground-based and satellite retrieved datasets. Based on MODIS data, Papadimas et al. (2008) showed that there has been a 20% decrease in AOD in the region, during the period 2000–2006, observed mainly in the western parts of the Iberian, Italian and Balkan peninsulas and the Southern Anatolian peninsula. Koukouli et al. (2010) revealed a decreasing trend in aerosol load between −2.5 and −4.1% per year over South Balkan/Eastern Mediterranean region. Based on the same dataset, Gkikas et al. (2009) studied the frequency and intensity of aerosol episodes, showing that the western and central Mediterranean areas are subject to strong aerosol events, associated with sea salt and biomass burning cases. Eastern Mediterranean regions are influenced, besides the local anthropogenic sources, by the desert dust aerosols and their transport, reaching high values in spring and summer (Hatzianastassiou et al., 2009, Querol et al., 2009, Gerasopoulos et al., 2011, Kalivitis et al., 2011).
Similar results have been reported from the analysis of aerosol data from single ground-based sites, selected for this study. The Sede Boker station (31°N, 35°E), in the Negev desert of Israel, exhibits a high AOD in spring due to desert dust and in late summer-early autumn attributed to anthropogenic aerosols. However, the radiative effect of mineral dust is considered as dominant (Derimian et al., 2006). The FORTH station (35°N, 25°E) in the island of Crete, Greece, presents a spring maximum due to dust transport and a winter minimum. During summertime, the transport of industrial and biomass burning aerosols is dominant (Fotiadi et al., 2006). This site is ideal for the study of transport and mixing of different aerosol types. Thessaloniki (41°N, 23°E), in northern Greece, is also a crossroad for a variety of aerosols, being additionally influenced by aerosols of industrial origin from the northeastern European countries (Balis et al., 2004). Kelektsoglou and Rapsomanikis (2011) revealed a summer maximum and a winter minimum, but reported high values in spring and autumn as well. More absorbing particles were found during autumn and winter, mainly due to emissions of aerosols from local pollution sources. A 9-year (1997–2005) period study of the AOD in the ultraviolet (UV) region and the PM10 concentrations showed a decreasing tendency, but the levels remained high (Kazadzis et al., 2007). The station at the IMS-METU site (37°N, 34°E) in Erdemli, Turkey is ideal for the study of mineral dust characteristics, because it is affected by dust transport from both the Sahara and Middle East deserts. Sahara dust episodes occur more frequently in spring, whereas dust from the Middle East deserts affects aerosol concentration at higher altitudes during summer and autumn (Kubilay et al., 2003). The station at Lecce University (40°N, 18°E), south Italy, is in a remote location, not influenced by industrial sources. It is generally characterized by clean continental aerosols and a mixing of water-soluble and more absorbing particles is evident (Barnaba et al., 2007).
In this study, the differences in aerosol optical properties (optical depth, single scattering albedo and asymmetry factor) in the UV and visible (VIS) spectral regions over eight sites in the Mediterranean basin are examined, using data from the AERONET, the AeroCom climatology and the MODIS satellite instrument. A radiative transfer model is used to estimate the uncertainties in UV and VIS irradiance reaching the ground that arise from the use of the satellite and model-derived climatologies.
2. Data and methology
The monthly climatological values of AOD, single scattering albedo and asymmetry factor retrievals from eight AERONET stations across the Mediterranean Sea were selected for this study (Figure 1), covering the time period 1996–2008. More recent data were not used, since the AeroCom climatology derived from model runs averaged for 3–10 years (Kinne et al., 2006) and is considered to be representative for the 1998–2007 time period (Pappas et al., 2011). AERONET level 2.0 data are used, which are pre- and post-field calibrated, automatically cloud cleared and manually inspected. More specifically, the instruments are calibrated before they are positioned in the field and after a measurement period. A cloud-screening algorithm is used by applying criteria for the short-time period variability (1 min) as well as the hourly and diurnal time period variations of AOD. Finally, data are visually inspected for abnormal data points. The use of level 2.0 data and the different starting dates of operation at each station reduced the total amount of years from 1.5 (at Thessaloniki) to 9.2 (at Sede-Boker) (Table 1). For the theoretical calculations, the monthly mean values of the α and β Ångström coefficients, the single scattering albedo and the asymmetry factor were used as input data. For calculations of UV irradiance, the α and β Ångström coefficients were derived from the optical depths at 340 and 380 nm. AERONET does not provide the single scattering albedo and asymmetry factor values in the UV region; so, measurements at the lower available wavelength (440 nm) were also used for UV calculations. Similarly, in the VIS spectral region, the Ångström coefficients were calculated from the optical depths at 440 and 675 nm. The average values of the single scattering albedo and the asymmetry factor at those two wavelengths were used in model calculations.
Table 1. Time periods of available level 2.0 AERONET AOD data, used for each station
Oct. 2003–Jun. 2008 [2.890 years]
Feb. 2003–May 2008 [4.340 years]
Jan. 2003–Mar. 2008 [3.668 years]
Nov. 1999–Jan. 2009 [4.707 years]
Mar. 2003–Dec. 2008 [3.734 years]
Feb. 2000–Mar. 2008 [5.386 years]
Jan. 1996–Jan. 2008 [9.170 years]
Sep. 2005–Jan. 2008 [1.482 years]
For UV and VIS irradiance calculations based on AeroCom climatology, the Ångström coefficients, derived from the optical depths at two UV (345 and 380 nm) and VIS (440 and 665 nm) wavelengths, were used. The average values of single scattering albedo and asymmetry factor at these wavelengths were used for the model calculations. The Ångström coefficients, single scattering albedo and asymmetry factor, that were used as input in the model calculations, refer to the monthly mean values for the time considered.
From the MODIS dataset, the monthly climatological averages of Ångström exponent and AOD at 550 nm over each site for the period 2000–2008 were taken into account. The daily averaged values, derived from the Terra and Aqua satellite estimations when they were available, were used to calculate the monthly climatological averages for the selected time period. Due to the limitation of MODIS operation in VIS and infrared wavelengths, the AOD was extrapolated in the UV, using the Ångström formula, τ(λ1) = τ(λ2) × (λ1/λ2)–α. As MODIS dataset does not provide estimations about the single scattering albedo and the asymmetry factor, spectrally independent values (0.95 and 0.67, respectively) were used in model calculations of irradiance. These values were taken as representative of the Mediterranean basin and similar values are found by Bergamo et al. (2008).
The estimation of the irradiance at the ground was done by using the UVspec code from the LibRadtran package (Mayer and Kylling, 2005). Typical, for the mid-latitudes, vertical profiles of the basic atmospheric gases, aerosols, pressure and temperature were used. Aerosols were further described with the assigned values for α and β Ångström coefficients, single scattering albedo and asymmetry factor. The total ozone column was set to 320 DU. The surface reflectance was set to 0.03 in the UV and 0.10 in the VIS region. The value for the reflectance in the UV agrees with the ones given in Herman and Celarier (1997) for the Mediterranean area, while the value in the VIS was chosen as a combination of the ones given in Bowker et al. (1985), for sand, water and concrete surfaces. Because the study focuses on differences between the modelled irradiances for the three datasets, the accuracy in the input reflectances does not affect the results. The purpose of this study was limited to the estimation of differences in cloud-free irradiance due to the use of different aerosol datasets. At each station and for the 15th day of each month, the monthly climatological values of aerosol optical properties from AERONET, AeroCom and MODIS were taken into account to calculate the UV (280–400 nm) and VIS (400–700 nm) integrated irradiance from sunrise to sunset in 30-min steps including the time of local noon.
3. Results and discussion
The monthly climatological averages of AOD from AERONET, AeroCom and MODIS, all extrapolated at 550 nm, using the Ångström formula, over each station are presented in Figure 2. Due to the limited availability of MODIS data over Sede Boker, only results for AERONET and AeroCom are presented. Generally, MODIS data seem to be in better agreement with AERONET values over all other sites. At Blida, a very good agreement between them is revealed, while AeroCom values are significantly lower (down to 0.24) from April to November. At Lecce University and FORTH Crete, the AOD is overestimated by AeroCom during most months of the year and the peak value of May is not evident in the other two datasets. Lower differences among the three datasets are revealed at Carpentras, Thessaloniki and IMS-METU Erdemli during some months. In contrast with the results from the North African site of Blida, the ground-based data are significantly lower than satellite and model estimations at the Middle East sites of Nes Ziona and Sede Boker. Figure 3 presents the differences in AOD at 380 nm. MODIS AOD has been extrapolated in the UV using the Angström formula, τ(λ) = β × λ–α, where α and β are the Angström coefficients. The average difference at each site is within ±0.03, relatively to the difference at 550 nm. However, in certain months, the differences are significantly higher.
In general, the highest differences from the AERONET data are revealed over the Middle East and North African sites and the island of Crete and, most probably, are relevant to the limited accuracy of MODIS aerosol products over bright surfaces and the imprecise simulation of the dust cycle in the AeroCom models. It is out of the scope of this study to examine further the possible reasons of difference in AOD among the three datasets. It should be also noted that the AERONET data at the selected sites are not available during the exact time period of MODIS data or the period that AeroComm climatology is considered as representative. However, the presentation of these results highlights the uncertainties of the satellite and modelled climatologies over the Mediterranean basin and indicates the importance to examine their effect in calculations of UV and VIS irradiance reaching the ground.
Solar irradiance reaching the ground is also affected by the single scattering albedo and, less significantly, by the asymmetry factor of aerosols. The average annual values of the AERONET and AeroCom single scattering albedo over each station at similar wavelengths in the VIS spectral region and at 380 nm for AeroCom are presented in Table 2. In all cases, the average overestimation of single scattering albedo by the AeroCom climatology is around 0.05. Similar results are revealed when comparing the AeroCom values in the UV region with the AERONET ones at 440 nm.
Table 2. Annual average values of SSA from AERONET and AeroCom over the selected sites
The monthly mean percentage (%) differences of calculated average UV irradiance between the AeroCom climatology and the AERONET measurements are presented in Figure 4, for each station. The combined differences are those resulting from the use of both AOD and SSA from AeroCom dataset. The separate effect of each factor is investigated and the differences due to the use of only SSA or AOD are also presented in the figure. In general, the greatest monthly differences are within ±15% and correspond to almost ±4.5 W m–2 in UV irradiance reaching the ground. During the April–September period, the UV irradiance derived from the AeroCom climatology is underestimated at most sites. The underestimation is higher (14%) for sites of the Eastern Mediterranean that lie close to desert regions (Sede Boker, Nes Ziona, FORTH Crete). For the same time period, less UV values (down to −7%) are calculated at Lecce University and IMS-METU Erdemli. At Thessaloniki and Carpentras, a quite good agreement (within 5 and 8%) is revealed for almost all months. On the contrary, the use of AeroCom climatology leads to the overestimation of UV irradiance up to 15% at another site that is affected dominantly from desert dust, the North African station of Blida. Significant overestimations (above 10%) can be observed also at specific sites and months, like February at IMS-METU Erdemli and Nes Ziona and November–December at Thessaloniki. Similar results were calculated for UV irradiance at local noon (not shown here), but, due to the weaker effect of aerosols on irradiance at lower solar zenith angles, the overall agreement is within ±12%. The maximum percentage difference corresponds to ∼7 W m–2 in UV irradiance reaching the ground.
In general, the differences in calculated irradiances are due to AOD values, while the effect of SSA is of secondary importance. However, some exceptions are revealed at some stations during specific months. For example, the high difference of 12% at Nes Ziona in February is almost entirely attributed to the difference in SSA values between the AeroCom and AERONET datasets. For the majority of months, the differences due to the use of SSA from AeroCom are around 2%, except in Blida, where the differences vary between 0 and 6%. In general, increased irradiance values are calculated by using the AeroCom SSA, because it is higher than the corresponding values from the AERONET database, as already seen in Table 2.
The mean percentage differences between the modelled VIS irradiance, as derived from AeroCom and AERONET datasets are presented in Figure 5. Percentage differences up to ±12% are revealed which correspond to discrepancies in irradiance up to 30 W m–2. Similar calculations for local noon, reveal differences up to 10% (35 W m–2). AOD is again the major factor responsible for the irradiance differences. Differences due to SSA range between 0 and 4% and are positive in their majority.
Figures 6 and 7 present the monthly mean percentage differences between the modelled UV and VIS irradiances, respectively, as derived from MODIS and AERONET datasets. In each figure, the upper panel presents the differences that arise with the use of AOD from MODIS and the yearly constant value of 0.95 for SSA, while in the lower panel SSA from the AERONET dataset is used in the calculation of the modelled MODIS irradiances, so the differences are attributed directly to the MODIS AOD. The agreement between UV irradiance calculations with MODIS and AERONET climatologies, when compared with the AeroCom-AERONET differences, is significantly better. During winter and spring, MODIS-derived UV irradiance is lower by ∼3%, while the average difference is reduced to 0% for the rest of the year. Again, the highest differences during summertime are calculated over Blida (6%). The corresponding difference in average and local noon UV irradiance reaching the ground is 2 and 3.3 W m–2, respectively. If SSA from AERONET is used in the calculations of the modelled MODIS UV irradiances, then increased differences are revealed, because AERONET SSA, as given in Table 2, is usually lower than 0.95. The use of the same SSA value as model input reveals the real effect of AOD differences between the two datasets in model calculated UV irradiance. In this case also, the effect of SSA is of secondary importance.
At all sites, the MODIS-derived VIS irradiance is lower than the AERONET during most months (Figure 7). However, the maximum underestimation in average values is 7% (15 W m–2). At local noon, differences up to 20 W m–2 (6%) are calculated. The highest differences are revealed over FORTH Crete and Nes Ziona stations during the winter and spring, respectively. During summertime, when the effect of aerosols on solar irradiance is more pronounced due to decreased cloudiness, the overall difference ranges from −4 to +2%. Again, the use of AERONET SSA in the calculation of MODIS irradiances, increases the differences between the two datasets, but the increase is lower than in the UV, because aerosols affect less the irradiance at VIS wavelengths than the UV ones.
The aerosol optical properties from the AeroCom climatology and the MODIS satellite instrument were used to examine their differences from ground-based measurements and estimate the effect on cloud-free UV and VIS irradiance calculations over eight AERONET sites across the Mediterranean basin. The absolute differences in AOD climatologies from AeroCom and AERONET can reach up to 0.25 and 0.15, in the UV and VIS part of the spectrum, respectively. These maximum differences occur in spring, where the aerosol loads are higher. The maximum absolute difference in AOD from AERONET and MODIS at 550 nm, also occurs in spring, but is quite lower than the one from AeroCom, reaching the value of 0.09. During the rest of the year, the AERONET climatology is in better agreement with the MODIS than the AeroCom data.
The monthly mean values of aerosol optical properties from AERONET, AeroCom and MODIS were used to estimate the UV (280–400 nm) and VIS (400–700 nm) irradiance at the ground. Differences between AERONET and AeroCom in the UV can reach up to 7 (12%) and 4.5 W m–2 (15%) for local noon and average values, respectively. In the VIS spectral region, the corresponding values are 35 (10%) and 30 W m–2 (12%). The largest irradiance differences are observed at the North African and Middle East stations of the Mediterranean coast (Nes Ziona, Sede Boker, Blida) and the island of Crete.
The differences between the AERONET and the MODIS-derived VIS irradiances are lower, reaching 20 (6%) and 15 W/m2 (7%) for the local noon and the average values, respectively. The highest values correspond to the stations of FORTH Crete and Nes Ziona. On average, the difference in UV irradiance is between 0 and 3%. During winter and spring, MODIS-derived UV irradiance is lower by ∼3%, while the average difference is reduced to 0% for the rest of the year. Again, the highest differences during summertime are calculated over Blida (6%). The corresponding difference in average and local noon UV irradiance reaching the ground is 2 and 3.3 W m–2, respectively.
In general, the pre-mentioned differences in calculated irradiances are mostly attributed to the AOD values, while the effect of SSA is of secondary importance. The MODIS and the AeroCom climatologies of aerosol optical properties present the highest differences from the AERONET ground-based measurements, over sites where the presence of desert dust aerosols is dominant. MODIS estimations are closer to the AERONET values, which indicate the need for further improvement of the dust cycle in the AeroCom models.
A thorough comparison has been attempted among climatologies of aerosol optical properties in the Mediterranean basin, taken from AeroCom, MODIS and AERONET, and their effect on calculations of UV and VIS irradiance. According to the results, the satellite and modelled datasets still need improvements and the corresponding climatologies have to be further assessed. The MODIS climatology is in better agreement with ground-based measurements and could be preferred for calculations of UV and VIS irradiance. However, with some further improvements, the AeroCom climatology could be a significantly valuable tool, because it provides estimations of aerosol optical properties with higher spectral resolution about natural and anthropogenic aerosols.
The authors would like to thank the principal investigators of the AERONET stations, Brent Holben (Blida, IMS-METU Erdemli stations), Jean Philippe Morel, Didier Barraillé (METEO-FRANCE Carpentras station), Andrew Clive Banks (FORTH Crete station), Maria Rita Perrone (Lecce University station), Arnon Karnieli (Nes Ziona, Sede Boker stations), Spyros Rapsomanikis, Alkiviadis Bais (Thessaloniki station) for providing the data. We thank the AeroCom project (AeroCom: http://aerocom.met.no/aerocomhome.html) for use of their data. This study was mainly conducted and funded by project ‘Hellenic Network of Solar Energy’ (HNSE), funded by the General Secretariat for Research and Technology, Greek Ministry of Education, Lifelong Learning and Religious Affairs. The MODIS teams are acknowledged for providing the Collection 5 aerosol products from the Terra and Aqua satellites. The LibRadtran team (www.libradtran.org) is acknowledged for providing the model algorithm.