This study considered two aerosol data sources. AERONET provides ground-based AOD measurements, while MODIS estimates AOD from satellite radiometers.

AERONET is a global network of sun photometers established by the National Aeronautics and Space Administration (NASA) and extensively developed by other partner agencies, institutes, and universities worldwide [20]. This program aims to assess the optical properties of aerosols and validate satellite retrievals of aerosol optical properties. AERONET observations can be used as reference values to verify and evaluate the accuracy of the satellite products since the measurements employed CIMEL CE-318 sun photometers [21].

Solar extinction measurements are taken with a high spatial resolution (15 min) in the wavelength range of 340 nm to 1640 nm. AERONET contains three levels of AOD quality: Level 1.0 (unfiltered raw data), Level 1.5 (clouds filtered), and Level 2 (quality assured) [22]. High-accuracy aerosol data at 500 nm wavelength from Level 2 of version 3 were used for MODIS validation.

AERONET lacks AOD products at 550 nm. Since AERONET provides the Angström exponent (AE) values, the estimation of AOD at 550 nm employed AERONET AOD at 500 nm and 440 nm wavelengths using the following equation that relates AOD at the desired wavelength (λ) and AOD at the reference wavelength (λ₀) [23].

$\text{Angström exponent }AE=\frac{\text{In}\left(AO{D}_{\lambda }/AO{D}_{{\lambda }_0}\right)}{\text{In}\left(\lambda /{\lambda }_0\right)}$ (1)

MODIS is one of the first satellite-based passive radiometers, part of NASA's Earth-observing system (EOS) program designed to retrieve aerosols over land and ocean. MODIS provides large-scale geophysical data sets such as atmospheric, oceanic, and terrestrial information based on 36 spectral channels with a temporal resolution of 1–2 days [24]. MODIS level 2 (MOD04_L2) provides aerosol parameter data with a spatial resolution of 10 ×10 km [25]–[27]. This analysis used aerosol products from MOD04_L2.

Two sites with distinct climate types in Morocco, where high-performance ground-based radiation measurements are available, enable us to investigate aerosols' impact on solar irradiance components. The irradiance data come from high-precision ground-based meteorological stations installed at Green Energy Park (GEP, Benguerir) and Tantan.

The weather stations are equipped with high-performance sensors to measure GHI, DHI, and DNI. GHI measurement uses a Kipp&Zonen CMP21 pyranometer classified in the highest possible ISO pyranometer performance category [28]. The sensors are characterized by a broad spectral bandwidth from 310 nm to 2800 nm, with a response time of < 1 s [29]. A similar pyranometer measures DHI by blocking the direct irradiance with an opaque ball placed above the instrument. The DNI is measured using a Kipp&Zonen CHP1 (first-class) pyrheliometer characterized by a spectral band from 200 nm to 4000 nm and a response time of 1 s [30]. The sensors are mounted on an automatic sun tracker Kipp&Zonen SOLYS 2 to maintain the instrument's alignment with the solar disk [31]. Figure 1 illustrates the sensors used to measure the three components of solar radiation installed at the site of Benguerir as an example. Table 1 presents the irradiance component measured and the sensors used.

Figure 1.

Example instrumentation at the Benguerir site: Kipp&Zonen pyranometer CMP21 (a); a shading ball for DHI measurement (b); Kipp&Zonen pyrheliometer CHP1 (c)

As already mentioned, the first aim of this study is to validate aerosol data from MODIS products using ground AOD measurements from the AERONET network over Morocco. Daily AOD values from MODIS and AERONET at 550 nm were compared for validation. The validation relied on the best-known statistical indicators: the Root Mean Square Error (RMSE), the Mean Bias Error (MBE), and Correlation Coefficient (CC). According to the studies [32], [33] and after several validations of MODIS against ground measurements worldwide, the authors confirmed that eq. (5) defines the expected error (EE) of MODIS relative to AERONET. For valid AOD data, the requirement is that 66% of the data fall within the EE envelope [34].

$RMSE=\sqrt{\frac{1}{N}{\displaystyle \sum _{i=1}^N{\left(AO{D}_{\left(MODIS\right)i}-AO{D}_{\left(AERONET\right)i}\right)}^2}}$ (2)

$MBE=\frac{1}{N}{\displaystyle \sum }_{i=1}^N\left(AO{D}_{\left(MODIS\right)i}-AO{D}_{\left(AERONET\right)i}\right)$ (3)

$CC=\frac{\mathrm{cov}\left(AO{D}_{\left(MODIS\right)},AO{D}_{\left(AERONET\right)}\right.}{\sigma AO{D}_{\left(MODIS\right)}\sigma AO{D}_{\left(AERONET\right)}}$ (4)

$EE=\pm \left(0.15\chi +0.05\right)$ (5)

where $\mathrm{cov}$ and $\sigma$ are the covariance and the variance, respectively. In our case, x represents the AOD measured from AERONET.

Figure 3 displays the density scatter plots between the daily average of MODIS and AERONET AOD at 550 nm. The dotted line shows the linear regression of the scatterplots represented by y-equation, and the two red lines represent the envelope of the expected error for MODIS. Table 3 represents the comparison results. For Oujda, Saada, and Rass El Ain sites, the outcome is that more than 66% of data points fall within the EE envelope with values of 71%, 69%, and 68%, with high correlation coefficients of 0.77, 0.75, and 0.82 and low MBE values of 0.005, 0.009, and −0.02, respectively. For Ouarzazate, although the correlation is 0.78, the percentage of values within EE is only 53%. For this site, MODIS overestimates AERONET with 44% of points outside the EE with an MBE of 0.06. Despite these errors, the performance of MODIS can still meet the requirements. For the Oukaimden site, only a few points were available, which makes MODIS validation challenging. Here, a large percentage of data is below the EE with a considerable error value of 0.11 and a low correlation coefficient of 0.65, implying MODIS does not perform well enough. In the case of the Dakhla site, a large percentage of data fell below the EE (65%), with a high MBE (0.17) and a low correlation coefficient (0.57). Moreover, MODIS cannot capture the variability of ground-based AOD. According to previous studies, the errors between MODIS and AERONET can be mainly due to the following reasons: aerosol model assumptions (0−20%), instrument calibration (2−5%), and pixel selection (0−10%) [38]. Except for the Oukaimden and Dakhla sites, the statistical performances show that MODIS AOD retrieval is relatively satisfactory over the AERONET study sites.

Figure 3.

Density scatters plot between daily AOD from AERONET and MODIS at 550 nm for all six AERONET sites over Morocco; the two red lines represent the envelope of the expected error (EE) for MODIS, and the dotted grey line represents the regression result

Classification of aerosols types

The classification of aerosol types is of great importance for climate modelling [39] as it improves the accuracy of aerosol radiative impact assessment. In this section, each site's aerosol classification was performed to identify the region's prevalent aerosol type. The classification uses the relationship between the AOD and the Angström exponent to indicate the size of AOD particles by setting threshold values where each group represents an AOD type. This classification builds on previous studies [40], [41] to define the appropriate thresholds for each type of AOD.

The selected threshold values are as follows:

• clean continental (CC): AOD < 0.2 with AE > 1.0

• clean marine (CM): AOD < 0.2 with AE < 0.9

• anthropogenic/burning (AB): AOD > 0.3 with AE > 1.0

• coarse/dust (CD): AOD > 0.6 with AE < 0.7.

The remaining cases not belonging to any of the above groups are characterized as mixed types (Mix) or indeterminate aerosols.

Figure 4.

Scatter plots between AOD at 550 nm and Angström exponent of 440−870 nm and various aerosol types

Figure 4 shows scatter plots between AOD at 550 nm and Angström exponent of 440−870 nm, indicating various aerosol types' contribution for each AERONET site. It was found that every type of aerosol makes contributions with different values for each site.

The graphs show that the dominant aerosol types at most sites are CM (35%−73%) and MIX (32%−50%). On the other hand, CD (1%−12%) and AB (0%−4%) are the least frequent types. The most significant percentage of AOD of CM type was recorded for the Ouarzazate site with a percentage of 73%, Oukaimden (51%), Saada (51%), and Dakhla (36%). In comparison, the lowest contribution was recorded at the Rass El Ain site (23%). MIX aerosols are present in large percentages at Rassel Ain (50%) and Saada (32%). For the site of Oujda, CC (32%), CM (35%), and MIX (33%) aerosol types are present with nearly equal contributions: The largest percentage of CD (12%) was recorded at the Dakhla site. In contrast, only small percentages of CD (1%−5%) are present for the other sites. The contribution of aerosols type AB is absent in most sites, except for Rass El Ain, where a contribution of 4% was observed.

The purpose of validating MODIS with AERONET ground-based measurements was to exploit its data to investigate the impact of AOD on solar irradiance. The validation was done using solar irradiance measurements available in Tantan and Benguerir sites. As depicted in Figure 5, the study sites are located in a region with high AOD at 550 nm averages, especially during May, June, July, and August.

Figure 5.

Monthly AOD from MODIS [42]

Figure 6 represents the monthly variations of AOD for the two sites from 01/01/2017 to 31/12/2020, collected from MODIS [42], while Table 4 represents the seasonal patterns of AOD for each station. Figure 6 and Table 4 show a high AOD in summer and a low AOD in winter and fall for both stations. The AOD values are very high in the Tantan site in summer, with a value of 0.50, which can be explained by the phenomena of transport of aerosols from the desert to the west (Figure 7). While for the same summer, the AOD is 0.22 for Benguerir. The minimum AOD value appears in autumn with a value of 0.17 for the Tantan site and 0.11 for the Benguerir site. Figure 7 shows a large plume of African desert dust blowing westward over the Atlantic Ocean of the Moroccan coast and arcing northeastward to the Canary Islands in the eastern Atlantic [43]. Figure 8 displays pictures taken at the Green Energy Park (Benguerir) during a dust storm. In summer, cloudy conditions are rare in Morocco, and aerosols may have a crucial role in attenuating the solar radiation that reaches the earth's surface. The effect of aerosols resides not only in the attenuation of solar radiation. They are also responsible for the performance degradation of PV panels and the decrease of CSP mirrors reflectance that directly affects the energy production.

Figure 6.

Monthly variability of AOD value at 550 nm at Benguerir (left) and Tantan (right)

Figure 7.

Dust/aerosol transport to the Canary Islands from the Sahara (image from MODIS [43])

Figure 8.

Real pictures taken at the Green Energy Park (Benguerir, Morocco) during a dust storm on 21 and 22 April 2018

Solar irradiance is sensitive to various meteorological parameters. In addition to solar geometry parameters such as solar zenith angle, clouds represent the major factor attenuating solar radiation. However, atmospheric aerosols mainly affect solar irradiance in regions with low cloud cover and predominantly clear skies, such as Morocco [5].

Under clear skies, atmospheric aerosols are considered the most important factor affecting solar radiation. Depending on their chemical and physical properties, they scatter and/or absorb solar radiation. Aerosols tend to decrease the DNI and increase the DHI. Since the GHI is the sum of the two components, it is less impacted when compared to DNI. Therefore, technologies using GHI such as PV are more advantageous than those using DNI like solar thermal or concentrated PV.

Photovoltaic and thermal power production strongly correlate to GHI and DNI, respectively. This study aims to evaluate the impact of AOD on these two components of solar irradiance using one-year data for each site (Benguerir and Tantan). The study period for Benguerir is from 01/01/2020 to 31/12/2020, while for Tantan, the data cover the period between 01/01/2017 and 31/12/2017. Irradiance data were collected from ground-based weather stations, while AOD data for the same periods came from MODIS.

The assessment of the impact of AOD considers only clear sky days. Such days are selected using the clear sky index K_t defined by the ratio between the GHI and the GHI_cls under clear sky conditions, see eq. (6). GHI_cls values are obtained from the clear sky model McClear [44]. Clear days correspond to a K_t > 0.7 [45], [46].

$K_t=\frac{GHI}{GH{I}_{cls}}$ (6)

Loss coefficients are defined to calculate the loss percentage of the irradiance caused by aerosols. For GHI, the loss coefficient LGHI is calculated according to global irradiance at the top of the atmosphere TOA. For DNI, the loss coefficient L_DNI is calculated according to the extraterrestrial radiation on a normal plane E₀. TOA and E₀ are obtained from the SODA portal [47]. The loss percentages for GHI and DNI are represented by the equations below:

$L_{GHI}=\frac{GHI}{TOA}\times 100$ (7)

$L_{DNI}=\frac{DNI}{E_{.0}}\times 100$ (8)

Figure 9 shows the variation of daily averages of GHI, DNI, and DHI according to different AOD values. An important feature is that the magnitude of the irradiance varies considerably with the AOD value. DNI is the most impacted component by AOD; its value decreases with the increase of aerosol content in the atmosphere, while DHI increases with the increase of AOD values. The GHI also decreases with increasing AOD but is less affected due to the compensation effect than the DNI. For the Benguerir site, the DNI decreases from 440 W/m² to 137 W/m², the GHI from 345 W/m² to 232 W/m² for AOD values of 0.02 to 0.6, respectively. Similar trends appear in the Tantan site: the DNI increases from 427 W/m² to 104 W/m², the GHI from 365 W/m² to 204 W/m², and the DHI increases from 48 to 117 for an AOD of 0.02 and 0.6, respectively.

Figure 9.

Variations of solar irradiation components according to the AOD value under clear days for Benguerir and Tantan

For a deep characterization of AOD's impact on solar radiation, Figure 10 illustrates the losses rate values calculated from eq. (7) and eq. (8) for both DNI and GHI. Indeed, the results described in Figure 10 and Table 5 show that the loss rate of DNI is higher than that of GHI. The minimum loss values for DNI are 67.89% and 66.95% at AOD = 0.02 for Tantan and Benguerir sites, respectively. In contrast, the maximum values reach 92.15% and 89.97% for the AOD value of 0.6 for Tantan and Benguerir, respectively. In the case of GHI, at a low AOD value of 0.02, the losses are 22.14% for Tantan and 24.85% for Benguerir. For a higher AOD value of 0.6, the losses of GHI are 53.61% for Tantan and 39.18% for the site of Benguerir. The main conclusion is that DNI and GHI decrease with increasing aerosol content in the atmosphere while DHI increases with increasing AOD values.

Figure 10.

The percentage of GHI and DNI loss due to AOD for Benguerir and Tantan