Assessing Lidar Ratio Impact on CALIPSO Retrievals Utilized for the Estimation of Aerosol SW Radiative Effects across North Africa, the Middle East, and Europe

Abstract

North Africa, the Middle East, and Europe (NAMEE domain) host a variety of suspended particles characterized by different optical and microphysical properties. In the current study, we investigate the importance of the lidar ratio (LR) on Cloud-Aerosol Lidar with Orthogonal Polarization–Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIOP-CALIPSO) aerosol retrievals towards assessing aerosols’ impact on the Earth-atmosphere radiation budget. A holistic approach has been adopted involving collocated Aerosol Robotic Network (AERONET) observations, Radiative Transfer Model (RTM) simulations, as well as reference radiation measurements acquired using spaceborne (Clouds and the Earth’s Radiant Energy System-CERES) and ground-based (Baseline Surface Radiation Network-BSRN) instruments. We are assessing the clear-sky shortwave (SW) direct radiative effects (DREs) on 550 atmospheric scenes, identified within the 2007–2020 period, in which the primary tropospheric aerosol species (dust, marine, polluted continental/smoke, elevated smoke, and clean continental) are probed using CALIPSO. RTM runs have been performed relying on CALIOP retrievals in which the default and the DeLiAn (Depolarization ratio, Lidar ratio, and Ångström exponent)-based aerosol-speciated LRs are considered. The simulated fields from both configurations are compared against those produced when AERONET AODs are applied. Overall, the DeLiAn LRs leads to better results mainly when mineral particles are either solely recorded or coexist with other aerosol species (e.g., sea-salt). In quantitative terms, the errors in DREs are reduced by ~26–27% at the surface (from 5.3 to 3.9 W/m²) and within the atmosphere (from −3.3 to −2.4 W/m²). The improvements become more significant (reaching up to ~35%) for moderate-to-high aerosol loads (AOD ≥ 0.2).

1. Introduction

Atmospheric aerosols, via their interaction with the shortwave (SW) and the longwave (LW) radiation fluxes, perturb the radiation fluxes at the surface, within the atmosphere (ATM), and at the top of the atmosphere (TOA) [1,2]. The aerosol-induced imbalances of the Earth-atmosphere system radiation budget results from the absorption and the scattering, both of SW and LW radiation, as well as from the re-emission of LW radiation. Such disturbances are taking place either in a direct or in a semi-direct or in an indirect way. The direct radiative effects (DREs) correspond to those attributed to the direct aerosol–radiation interactions (REari) [2]. Furthermore, aerosols can indirectly modulate atmospheric dynamics, cloud formation, and precipitation due to rapid adjustments [3,4,5,6,7]. Aerosols can also serve as cloud condensation nuclei (CCN) and ice nuclei (IN), modifying clouds’ microphysical, macrophysical, and optical properties [8,9,10,11]. Through these complex processes, aerosols modulate the Earth’s radiation budget by modifying cloud albedo (referred to as Twomey effect, first indirect effect, or aerosol–cloud interactions; REaci) and cloud lifetime (referred to as second indirect effect or rapid adjustments to the REaci) [3,5]. An overall assessment of the induced radiative impacts is expressed by the aerosol effective radiative forcing (ERFari), a measure linking changes of the net TOA radiative fluxes with those of the near-surface temperature, thus quantifying aerosols’ crucial role on climate.

Numerical simulations have demonstrated that aerosols tend to cool the Earth-atmosphere system, at a global scale and over the long-term, partly counterbalancing the planetary warming attributed to greenhouse gases [12]. Nevertheless, global climate models continue to give diverging results of the ERFari and ERFaci magnitude, as stated in the latest report of the Intergovernmental Panel on Climate Change [13]. This poses limitations for a robust assessment of aerosols impact upon climate. Among several possible reasons, the obtained declinations arise from the pronounced spatiotemporal heterogeneity of aerosols’ properties, as well as from the misrepresentation of the aerosol burden vertical distribution [1,12,14,15,16,17]. The aforementioned factors, particularly those related to aerosol properties, govern at a large degree aerosol–radiation interactions, thus highlighting the importance of a reliable and accurate aerosol typing. Ground-based measurements, acquired using instrumentations deploying passive or active remote sensing techniques, have provided critical insight on aerosol-speciated intensive and extensive radiative properties [18,19,20,21,22,23]. Spaceborne observations, providing near-global coverage, have effectively overcome the spatial constraints of ground-based stations [24,25,26,27,28,29,30]. Passive sensors, such as the MODerate resolution Imaging Spectroradiometer (MODIS), onboard Terra, and Aqua satellites, have supplied the scientific community with multi-year records of the columnar aerosol optical depth (AOD) since 2000 [31,32,33,34,35]. Advanced aerosol observations have been enabled by sensors such as the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) onboard the CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) satellite, deploying active remote sensing techniques [36]. In contrast to passive sensors (e.g., MODIS), CALIOP provides vertically resolved aerosol retrievals, both in daytime and nighttime, at fine horizontal and vertical resolution, and at the same accuracy regardless the underlying surface type [37].

Numerous studies have been performed emphasizing on the representation of aerosols’ regime, both at regional and global scales, relying on CALIPSO spaceborne observations [38,39,40,41,42,43,44,45,46,47,48,49,50,51]. On the contrary, few research works have utilized CALIPSO retrievals for assessing the aerosol-induced radiative impacts within the Earth-atmosphere system. Ref. [52] exploited 1-year CALIOP and MODIS retrievals and numerical simulations (SPRINTARS; mSpectral Radiation-Transport Model for Aerosol Species) for the estimation of aerosol DREs at the surface and TOA, under clear sky, cloudy sky, and all-sky conditions. Ref. [53] computed aerosol DREs under all-sky conditions within the Earth-atmosphere system over a 3-year period (2007–2009) and at a global scale by combining vertically resolved CALIOP AODs and the FORTH radiation transfer model [54,55]. Recently, ref. [56] developed a CALIOP-MODIS synergistic algorithm under the assumption that the aerosol burden mainly consists of four components (water soluble, light absorbing, dust, and sea salt particles), and they estimated the surface and TOA SW DREs under clear sky conditions. All the aforementioned studies have pointed out the importance of a reliable representation of optical properties in the aerosol models.

Undoubtedly, CALIPSO observations have mutually contributed to the remote sensing community and in aerosol research. However, their utilization in studies related to aerosol–radiation interactions necessitates a better knowledge of their advantages and drawbacks prior to a robust assessment of aerosol DREs. According to [57], CALIPSO AOD underestimations of the order of 13% with respect to AERONET could be attributed to the mistyping of aerosol layers or to the incorrect modelling of aerosol microphysics for the different aerosol types. In daytime conditions, CALIPSO does not detect tenuous aerosol layers due to the strong background illumination reducing the signal-to-noise ratio [37]. Under the presence of opaque layers, the transmitted signal is partially or totally attenuated, thus making the detection of the layers below infeasible [27]. Aerosol classification errors have notified by [58], who demonstrated a very good detection when mineral particles are probed, in contrast to polluted dust and smoke (one of the most absorbing aerosol species), while a moderate level of agreement was found for marine and polluted continental particles. Over the polar regions, limitations in the CALIPSO classification scheme [44] forced dust from Asian deserts [59] and smoke from Canadian fires [60] to be classified either as clean or polluted continental. Ref. [61] compared the CALIPSO and CALIPSO-SODA LRs for each aerosol type over oceans and they notified the significance of redefining the dusty marine subtype, stating that a portion of these samples should be classified as clean marine.

CALIOP is an elastic lidar, and for the derivation of the extinction coefficient from the retrieved backscatter coefficient it is required to have a predefined extinction-to-backscatter ratio (referred as lidar ratio, LR). LR is highly variable either among aerosol species or even for the same aerosol type due to its dependency on origin-related optical/microphysical properties [62]. In the CALIOP extinction retrieval algorithm, this crucial parameter [40] depends on the assigned type retrieved from the aerosol algorithm, and its value is derived from a combination of measurements, modelling, and cluster analysis of Aerosol Robotic Network (AERONET) multi-year observations [37,44,63,64]. Previous studies [42,45,65,66] have shown that an increase in the dust LR reduces the underestimation of AOD (resulting from the vertical integration of the extinction coefficient) against MODIS and AERONET. Their findings highlight the importance of an accurate definition of LR leading to a more reliable lidar-derived AOD, a critical and quite sensitive parameter when studying aerosol radiative effects [53,55]. Ground-based lidars have provided unprecedented LR measurements [60,62,67,68,69,70,71,72,73] thanks to their advanced observational capabilities. A comprehensive overview of these measurements is presented by [74], who developed the DeLiAn database containing state-of-the-art observations acquired using ground-based lidars operating in different regions of the planet (in terms of aerosols’ regime).

In the current study, the region of interest (ROI) comprises North Africa, the Middle East, and Europe, the so-called NAMEE domain. NAMEE is an ideal region for aerosol studies, hosting a variety of natural and anthropogenic aerosols either originating within its geographical limits or advected from outer regions. Dust particles emitted from the Sahara Desert and Middle East, encompassing the most active dust sources worldwide [75], are advected towards the Mediterranean and northwards under the prevailing atmospheric circulation [76,77,78]. Under wind stress over the sea surface, sea salt particles are ejected into the atmosphere [79], contributing to the natural component of the aerosol budget of the ROI. The Atlantic Ocean, the Mediterranean Sea, and the Red Sea are the main sources of marine aerosols in NAMEE. Fine particles produced by anthropogenic activities in urban and industrialized areas, which can be transported at a regional scale [80], constitute a major component of the aerosol burden resulting in air quality degradation with concomitant health impacts [81,82]. During summer, smoke particles at large concentrations are emitted from wildfires taking place across Central Europe, the surrounding area of the Black Sea, and the Mediterranean basin [76,83]. Furthermore, smoke particles from Canadian and US wildfires undergo long-range transport as they traverse the Atlantic Ocean before reaching Europe [84,85,86].

In the present study, we employ aerosol data from ground-based AERONET observations and retrievals from the CALIOP lidar onboard the CALIPSO satellite, along with a radiative transfer model (libRadtran [87]) to estimate the aerosol DREs within NAMEE. Our focus is to investigate the role of LR on the estimated DREs at the surface, within the atmosphere, and at TOA. To accomplish this goal, we are using as inputs to the RTΜ AODs derived using the default (CALIOP) and the updated (DeLiAn) LRs, as well as AERONET AODs which serve as reference. In Section 2, the utilized datasets are presented. In Section 3, we describe the UVSPEC RT model configuration. Section 4 presents the selection of the study cases, the applied methodology for the discrimination of the individual aerosol species in aerosol mixtures on the CALIOP retrievals, as well as the treatment of the RTM inputs. In Section 5, we discuss the results of our study which are summarized in the Conclusions Section, along with the future aspects.

2. Datasets

2.1. CALIOP-CALIPSO Spaceborne Retrievals

From April 2006 to August 2023, the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) on board the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite acquired vertically resolved information on aerosols and cloud properties [37]. CALIOP is a dual-wavelength backscatter (532, 1064 nm) and single-wavelength polarization (532 nm) lidar obtaining aerosol and cloud properties at fine horizontal and vertical resolution [37]. CALIOP distinguishes aerosols and clouds and their respective subtypes whereas, thanks to the deployment of the polarization channel, can detect spherical (e.g., marine particles, liquid cloud droplets) and non-spherical suspended particles (e.g., dust, ice cloud particles) [44,58,64,88,89,90].

In our analysis, we process the CALIOP version 4.20 (V4.20) L2 aerosol profiles (released in November 2016) in which aerosols and clouds are categorized based on the L1 layer-integrated values of depolarization and attenuated backscatter along with external information (i.e., geographical location, surface type, and layer altitude). In the troposphere, the CALIOP aerosol classification scheme categorizes aerosols in seven subtypes, namely “marine”, “dust”, “polluted continental/smoke”, “clean continental”, “polluted dust”, “elevated smoke”, and “dusty marine” [44]. A series of quality filters is applied to reduce errors in the layer detection/classification errors and in the extinction, as well as biases caused by the negative signal anomaly. The implementation of the cloud aerosol discrimination (CAD) algorithm and the screening of misclassified cirrus fringes reduce, as much as possible, cloud contamination (e.g., [42,45,91]).

In the present study, we exploit quality-assured vertical profiles of the backscatter coefficient and the linear particle depolarization ratio at 532 nm over a 14-year period (2007–2020). The CALIPSO data have been extracted from the LIVAS (Lidar climatology of Vertical Aerosol Structure) database [92]. The ESA-CALIPSO dataset [93], EARLINET (The EARLINET publishing group 2000–2010), and AERONET products [94,95], in combination with aerosol models from the literature [96,97,98], were implemented for LIVAS development. From its establishment, LIVAS has contributed to various studies related to dust climatology and retrievals optimization [42,45,66,99], new dataset development [34], the assessment of aerosols’ impact on radiation [66,100,101,102], and model evaluation [103].

2.2. AERONET

AERONET (AErosol RObotic NETwork) is a worldwide network with more than 1000 automatic sun-photometers [94,95] providing columnar aerosol optical and microphysical properties. In the current study, AERONET sun-direct spectral measurements and almucantar retrievals [104,105] are utilized for the assessment of the CALIOP AODs and the representation of the aerosol-speciated spectral variation of the single scattering albedo (SSA) and the asymmetry parameter (ASYM). More specifically, we process the latest version (V3) of the quality-assured (Level 2.0) AERONET data available at the finest temporal resolution (i.e., all points) [106,107]. For the spectral match between ground-based (AERONET) and spaceborne (CALIOP) observations, we apply the Ångström formula (Equation (1)) in order to derive the AERONET AOD at 532 nm (AOD_λ). This is achieved relying on the AERONET AOD at 440 nm (AOD_λ0) and the Ångström exponent (å) in the spectrum range spanning from 440 nm to 675 nm.

$${\mathrm{A}\mathrm{O}\mathrm{D}}_{\mathsf{\lambda }}={\mathrm{A}\mathrm{O}\mathrm{D}}_{\mathsf{\lambda }0}·{\left(\frac{\mathsf{\lambda }0}{\mathsf{\lambda }}\right)}^{\dot{\mathrm{a}}\left(\frac{\mathsf{\lambda }0}{\mathsf{\lambda }}\right)}$$

(1)

Finally, the spectral signatures of the extensive (AOD) and intensive (SSA, ASYM) optical properties at 440, 675, 870, and 1020 nm are reproduced for each aerosol type according to the CALIPSO retrievals (see Section 4.3.1) and are used as inputs in the RTM.

2.3. DeLiAn

DeLiAn (Depolarization ratio, Lidar ratio, and Ångström exponent) is a newly developed dataset providing aerosol-speciated properties acquired using ground-based lidars operated in experimental campaigns (e.g., SAMUM 1/2, SALTRACE) or contribute to networks (e.g., EARLINET) [67,70,71,74,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127]. Ship-borne vertically resolved aerosol measurements in the Atlantic Ocean have been also obtained using Lidars mounted on the Meteor and the Polastern research vessels [72,128,129,130,131].

In the DeLiAn database, 13 aerosol categories are reported, incorporating pure aerosol species as well as dust mixtures. In this study, we exploit the measured DeLiAn LRs and the depolarization ratios (used for the discrimination of aerosol components in dust mixtures [119], as will be described in Section 4.2). The LRs for the CALIPSO aerosol types [44] and the corresponding values from DeLiAn are listed in

Table 1. Aerosol-speciated lidar ratios at 532 nm in the CALIPSO V4.2 algorithm and in the DeLiAn dataset.

Aerosol Subtype		CALIPSO V4 LR (sr)	DeLiAn LR (sr)/δ532
Dust	Saharan dust	44	53.1/0.28
	Middle Eastern dust	44	37.4/0.28
Marine	Clean marine	23	21.9/0.01
	Dried marine	23	26.9/0.15
Polluted continental/smoke		70	71.8/0.03
Elevated smoke		70	71.8/0.03
Clean continental		53	56.2/0.03

. Note that we assign different dust LRs over the Sahara Desert and Middle East. Ultimately, we differentiate between the LR and the depolarization ratio of marine particles in wet and dry conditions [132] by exploiting the relative humidity (RH) from the MERRA-2 reanalysis provided in the CALIPSO auxiliary files.

2.4. CERES

The Clouds and the Earth’s Radiant Energy System (CERES) project aims to improve our knowledge of the Earth’s radiation budget based on measurements acquired with broadband scanning radiometers mounted on several satellites [133]. The observed radiances within the field-of-view of the CERES instrument (footprint nominal diameter of 20 km at nadir) are combined with fine-resolution retrievals (e.g., aerosols and cloud properties) derived using satellite imagers (e.g., MODIS). The purpose of this combination is to accurately determine the scene type and convert the measured radiances into fluxes. This conversion is achieved by utilizing empirical angular distribution models (ADMs) that are specific to each scene type [134,135]. The main products of CERES consist of radiances, radiative fluxes, and cloud properties distributed at various processing levels [134]. Here, we use the instantaneous TOA-reflected SW fluxes from the level 2 CERES Single Scanner Footprint (SSF) Edition4A SSF product derived by retrievals from the Aqua satellite (2002–present).

2.5. BSRN

The Baseline Surface Radiation Network (BSRN) operates under the World Climate Research Programme (WCRP). BSRN provides high-quality radiation measurements at 73 sites worldwide spread across different climate zones, which generally span the wavelength range of 250 to 3000 nm for the SW and 4000 to 40,000 nm for the LW radiation [136,137]. Apart from the global horizontal irradiance (GHI or surface solar radiation, measured with a pyranometer) and its standard deviation (GHI_std) on a minute basis, the BSRN stations can measure the diffuse solar flux (DIF, with a shaded pyranometer), the direct normal irradiance (DNI, measured with a pyrheliometer), and the long-wave downward radiation (LWD, measured with a pyrgeometer). The data are archived at the Alfred Wegener Institute (AWI) in Bremerhaven, Germany, and are freely available (https://bsrn.awi.de/ (accessed on 13 December 2023)). For the purposes of the current research, we analyze the GHI for the validation of the simulated downwelling SW radiation reaching at the surface.

3. LibRadtran

The RTM simulations are performed with the UVSPEC model of the libRadtran radiative transfer package [87,138], employing the DIScrete Ordinates Radiative Transfer (DISORT) solver [139,140]. The synthetic extraterrestrial spectrum of [141] was used for the incoming solar spectral irradiance at TOA. The RTM calculations are carried out for 2085 wavelengths in the SW spectral range (250–5000 nm) using a four-stream plane parallel approximation. The model accounts for the scattering and absorption of solar radiation from atmospheric gases and aerosols and the reflection from the Earth’s surface. We employ the representative wavelength approach (REPTRAN [142]) for the molecular absorption band parametrization. The aerosol data required as inputs in the RTM consists of the spectrally and vertically resolved aerosol extinction coefficient and the single scattering albedo. The phase function moments are computed from the aerosol asymmetry parameter, assuming that the phase function follows the Henyey–Greenstein approximation [143]. We provide a detailed description of the RTM input data and their preprocessing in Section 4.3.

The model simulates the downwelling and the upwelling shortwave (solar) fluxes at the Earth’s surface, at 121 layers within the atmosphere, and at TOA. The vertical resolution is 100 m up to 8 km, becoming coarser at higher altitudes. For the calculation of the clear sky (i.e., cloud-free) DREs, we contrast the simulated net (downwelling–upwelling) solar fluxes at the Earth’s surface (NETSRFC), within the atmosphere (ATM), and at the Top of the Atmosphere (TOA), simulated using the RTM in an aerosol-free atmosphere and when the suspended particles are treated as a radiatively active substance. More specifically, the DREs (in W/m²) are computed based on Equation (2), in which the terms F_i,aer and F_i,noaer correspond to the net solar fluxes with and without aerosols, respectively, whereas the subscript i represents either NETSRFC or ATM or TOA. In addition, the calculated DREs must satisfy Equation (3).

$${\mathrm{D}\mathrm{R}\mathrm{E}}_{\mathrm{i}}={\mathrm{F}}_{\mathrm{i},\mathrm{a}\mathrm{e}\mathrm{r}}-{\mathrm{F}}_{\mathrm{i},\mathrm{n}\mathrm{o}\mathrm{a}\mathrm{e}\mathrm{r},}$$

(2)

$$\mathrm{i}=\mathrm{N}\mathrm{E}\mathrm{T}\mathrm{S}\mathrm{R}\mathrm{F}\mathrm{C},\mathrm{A}\mathrm{T}\mathrm{M},\mathrm{T}\mathrm{O}\mathrm{A}$$

$${\mathrm{D}\mathrm{R}\mathrm{E}}_{\mathrm{T}\mathrm{O}\mathrm{A}}={\mathrm{D}\mathrm{R}\mathrm{E}}_{\mathrm{N}\mathrm{E}\mathrm{T}\mathrm{S}\mathrm{R}\mathrm{F}\mathrm{C}}+{\mathrm{D}\mathrm{R}\mathrm{E}}_{\mathrm{A}\mathrm{T}\mathrm{M}}$$

(3)

DREs quantify the aerosol-induced radiative perturbations, and are mainly driven by the aerosol optical properties, the solar zenith, the radiation fluxes, and the surface albedo in a cloud-free atmosphere. Nevertheless, these dependencies are critical when an intercomparison between DREs is attempted. In order to eliminate the contribution of the non-aerosol factors (i.e., solar zenith angle), we have also calculated the Aerosol Radiative Budget Efficiency (ARBE [144]) by normalizing DREs with the aerosol-free fluxes (Equation (4)). ARBE is a unitless parameter, expressing the perturbation of the radiation budget and highlights the importance of aerosol optical properties. In the present study, we calculate ARBE as the absolute percentage change in solar radiative fluxes (%) [144] according to the following equation:

$${\mathrm{A}\mathrm{R}\mathrm{B}\mathrm{E}}_{\mathrm{i}}=\left|{\mathrm{D}\mathrm{R}\mathrm{E}}_{\mathrm{i}}\right|/{\mathrm{F}}_{\mathrm{I},\mathrm{n}\mathrm{o}\mathrm{a}\mathrm{e}\mathrm{r}},$$

(4)

$$\mathrm{i}=\mathrm{N}\mathrm{E}\mathrm{T}\mathrm{S}\mathrm{R}\mathrm{F}\mathrm{C},\mathrm{A}\mathrm{T}\mathrm{M},\mathrm{T}\mathrm{O}\mathrm{A}$$

4. Methodology

4.1. Definition of the Study Cases

A first criterion for the definition of the study cases concerns the coincident availability of ground-based (AERONET) and spaceborne (CALIPSO) observations. For the AERONET-CALIPSO collocation, we have identified CALIPSO overpasses residing within a circle of 100 km radius centered at each AERONET site (Figure 1a). In addition, ground-based retrievals must be available within a time window of ±30 min centered at the satellite overpass time. From the selected CALIPSO overpasses, we consider only cloud-free scenes based on the CALIOP’s classification scheme. Overall, 2355 cases (scenes) satisfy the defined spatiotemporal criteria. At the next step, we ensure that across the CALIPSO scenes: (i) the laser beam penetrates throughout the atmosphere reaching at the ground, (ii) the number of the filtered backscatter coefficient records is less than 5%, and (iii) the ground elevation is relatively flat without abrupt changes (i.e., steep topography) [145]. By applying the aforementioned criteria, the raw matchups are reduced from 2355 down to 550 cases, constituting our final sample. In Figure 1, we present an example of a scene nearby the El Farafra station (Sahara Desert) where only dust aerosols are probed according to aerosol subtype product (left panel—Figure 1b). For the part of the CALIPSO orbit falling with the circle area around the ground site (red dashed line in Figure 1a), we are reproducing the curtain plot of the backscatter coefficient at 532 nm (central panel—Figure 1b). These retrievals are spatially averaged and are multiplied with the defined CALIPSO and the measured DeLiAn LRs (see Table 1) to derive the CALIPSO-based (red) and the DeLiAn-based (green) vertical profiles of the extinction coefficient and the columnar AODs at 532 nm (right panel—Figure 1b). For the ground-based measurements, apart from the “default” time window (±30 min), the sunphotometric AODs have been temporally averaged for three other time windows (±15, ±45, and ±60), noting possible temporal variations of the aerosol burden over the station. For this specific case study, the utilization of the revised dust LR (53 sr), instead of the CALIPSO default (44 sr), reduces the bias against AERONET AOD by ~14% (from −34.5% to −20.6%).

4.2. Separation of Dust and Non-Dust Components in Dusty Mixtures

Based on the CALIOP classification scheme [44,146], aerosols are categorized in ten types, in which seven of them correspond to tropospheric particles and the remaining ones represent stratospheric aerosols (see Section 2.1). Ref. [66] demonstrated that within the NAMEE domain and over the study period (2007–2020), the frequency of occurrence of the stratospheric aerosols in the CALIOP records is negligible ($~$0.8%), and for this reason they are omitted in our analysis. Among the tropospheric aerosol species, dust clearly dominates ($~$51.8%) over other types such as marine ($~$10.5%), polluted continental/smoke ($~$4.6%), and elevated smoke ($~$2.8%). The dominant dust presence is further reinforced, as indicated by the relatively high percentages of the dusty marine and polluted dust categories, which together make up 27% of the overall CALIOP aerosol sample. In dust conditions as well as in dust mixtures, we separate the dust and non-dust components by implementing the methodology presented in [119], which has been widely applied in numerous studies [65,147,148,149,150,151,152,153,154,155]. The discrimination method takes advantage of the depolarization ratio (δ), which strongly depends on the particles’ sphericity. For spherical particles, δ takes low values, whereas the opposite is evident when suspended particles of irregular shape (i.e., dust) are probed.

Under dust conditions or in polluted dust mixtures, we are assuming that the non-dust component consists of polluted continental/smoke or elevated smoke, adopting the CALIOP’s aerosols typing nomenclature. The separation of the aerosol load on its counterparts within a layer is made according to Equation (5) by setting depolarization ratios equal to 0.28 and 0.03 for the dust (δ_d) and non-dust components (δ_nd), respectively. The depolarization values have been extracted from the DeLiAn dataset [74]. Based on Equation (5), the dust backscatter coefficient (β_d) is obtained, contributing to the total measured amount (β_p), whereas the portion of the backscattered signal attributed to non-dust particles is calculated via the subtraction of these two terms (Equation (6)). It is notified that when the measured depolarization is lower than δ_nd, only non-dust particles exist whereas, and when δ_p is higher than δ_d, then pure dust conditions are valid. In contrast to the depolarization, the intensive aerosol optical properties (SSA, ASYM) of the non-dust species are not identical. Since they are used as inputs to the RTM (see Section 4.3.1), it is necessary to make a discrimination between the two non-dust types. Following the principles of the CALIOP classification scheme [44], we assign a layer as a polluted continental/smoke when its top is below 2.5 km (a.g.l.), otherwise it is assumed to be elevated smoke. In dusty marine mixtures, the same approach is followed, but instead of using a constant δ_nd threshold, the corresponding level is defined dynamically (yielding values from 0.01 to 0.15) taking into account the ambient relative humidity that can affect the marine particles’ shape (cubic-like), as has been shown in [132]. Based on [149], β_d and β_nd are determined according to the following equations:

$${\mathsf{\beta }}_{\mathrm{d}}={\mathsf{\beta }}_{\mathrm{p}}\frac{\left({\mathsf{\delta }}_{\mathrm{p}}-{\mathsf{\delta }}_{\mathrm{n}\mathrm{d}}\right)\left(1+{\mathsf{\delta }}_{\mathrm{d}}\right)}{\left({\mathsf{\delta }}_{\mathrm{d}}-{\mathsf{\delta }}_{\mathrm{n}\mathrm{d}}\right)\left(1+{\mathsf{\delta }}_{\mathrm{p}}\right)},{\mathsf{\delta }}_{\mathrm{n}\mathrm{d}}\le {\mathsf{\delta }}_{\mathrm{p}}\le {\mathsf{\delta }}_{\mathrm{d}}$$

(5)

$${\mathsf{\beta }}_{\mathrm{n}\mathrm{d}}={\mathsf{\beta }}_{\mathrm{p}}-{\mathsf{\beta }}_{\mathrm{d}}$$

(6)

Finally, after eliminating the stratospheric aerosols and disentangling dust and non-dust components in dusty mixtures, the resulting CALIPSO aerosol subtypes are five, namely dust, marine, polluted continental/smoke, elevated smoke, and clean continental.

4.3. RTM Inputs

In this Section, we provide a detailed description of the UVSPEC RT model configuration and the treatment (i.e., preprocessing) of the required inputs comprising aerosol, surface, and atmospheric data. A schematic overview of the model setup is illustrated in Figure 2.

4.3.1. Aerosol Optical Properties

For the calculation of DREs, the spectral signatures of AOD, asymmetry parameter (ASYM) and single scattering albedo (SSA) are required. For this reason, we combine the CALIPSO vertical speciated AOD profiles at 532 nm with lookup tables, providing the spectral optical properties of the five aerosol types considered in this study. These lookup tables are based on AERONET spectral observations (at 440, 675, 870 and 1020 nm) when the aerosol load consists only in single “pure” aerosol types, according to the collocated CALIPSO data [44]. Figure 3 depicts the spectral: (i) SSA, (ii) ASYM, and (iii) AOD for dust (yellow), marine (cyan), clean continental (green), polluted continental/smoke (orange), and elevated smoke (black) aerosols. Among aerosol types, the highest AODs are found for elevated smoke, while the lower values are recorded for clean continental particles. A common feature is the decrease in AOD for increasing wavelengths, which, however, is steep for fine aerosols (smoke) and smooth for coarse aerosols (e.g., dust, sea-salt) [156,157,158,159]. Due to the high scattering efficiency of marine particles, SSA values are close to unity and rather stable throughout the shortwave spectrum [98]. A different SSA spectral signature is revealed for mineral dust, showing a decreasing absorption efficiency at longer wavelengths [160]. For the remaining three aerosol types (elevated smoke, polluted continental/smoke, and clean continental), SSA values are lower, mainly attributed to the presence of fine absorbing particles, gradually decreasing from 440 nm to 1020 nm [161,162]. In the asymmetry parameter, there is a clear discrimination between fine and coarse aerosols with lower (less forward scattering) and higher (more forward scattering) values, respectively [163,164].

We adjusted the spectral AERONET AOD from the lookup tables (AERONET) to match the CALIPSO observations at 532 nm (derived either with the default or the Delian LR) for each aerosol type and layer:

$$\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}=\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }=532 \mathrm{n}\mathrm{m}\right)}_{\mathrm{cal}\_\mathrm{obs},\mathrm{type}}\frac{\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }\right)}_{\mathrm{type}}}{\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }=532 \mathrm{n}\mathrm{m}\right)}_{\mathrm{type}}}$$

(7)

where AOD_{cal_obj,type} is the CALIPSO AOD observations for an aerosol type at 532 nm. AOD(λ)_type and AOD(λ)_layer,type represent the spectral AOD from the lookup tables and the respective adjusted values, respectively. It should be noted that the AOD (λ = 532 nm)_type is derived using interpolation between 440 and 675 nm using the AERONET Ångström exponent. Then, the total AOD, the SSA, and the ASYM at each layer are calculated according to the following formulas:

$$\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}={\sum }_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}},$$

(8)

$${\mathrm{SSA}\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}={\sum }_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}\frac{{\mathrm{AOD}\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}{\mathrm{SSA}\left(\mathsf{\lambda }\right)}_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}}{\mathrm{A}\mathrm{O}\mathrm{D}{\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}},$$

(9)

$$\mathrm{A}\mathrm{S}\mathrm{Y}\mathrm{M}{\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}={\sum }_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}\frac{{\mathrm{AOD}\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}{\mathrm{SSA}\left(\mathsf{\lambda }\right)}_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}{\mathrm{ASYM}\left(\mathsf{\lambda }\right)}_{\mathrm{t}\mathrm{y}\mathrm{p}\mathrm{e}}}{{\mathrm{AOD}\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}{\mathrm{SSA}\left(\mathsf{\lambda }\right)}_{\mathrm{l}\mathrm{a}\mathrm{y}\mathrm{e}\mathrm{r},\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}},$$

(10)

where SSA_type and ASYM_type are the aerosol-speciated SSA and ASYM derived from the lookup tables, AOD_type is the adjusted speciated AOD, and AOD_layer,total, SSA_layer,total, and ASYM_layer,total correspond to the aerosol optical properties at each layer.

Based on the described methodology, we created two datasets relying on the default CALIPSO (“CALIPSO default”) and the DeLian-based (“DeLian”) lidar ratios for the derivation of the vertical AOD profile at 532 nm. Furthermore, a third dataset was created by adjusting the DeLian AOD profile to match the columnar AOD, as measured using AERONET. The DREs were calculated for the three datasets after converting the AOD to layer extinction coefficient and the asymmetry parameter to phase function moments.

4.3.2. Surface and Atmospheric Data

Apart from aerosol optical properties, ancillary surface and atmospheric datasets are processed and are used as inputs to the radiative transfer model. For the land–ocean mask, we are relying on the International Geosphere-Biosphere Programme (IGBP) data available in the MODIS MCD12C1 V061 products, whereas for the snow-cover mask, we are utilizing the daily MODIS MCD43C1 V006 data. Both MODIS products are provided at 0.05° × 0.05° spatial resolution, and they are regridded at 0.25° × 0.25° cells, encapsulating an AERONET station. We are considering the surface type as “ocean” or “snow” when the majority (>50%) of the aggregated fine-resolution cells are similarly classified in the IGBP database. For these cases, the spectral surface albedo is calculated from the libRadtran built-in IGBP library. On the contrary, in snow-free land, we used as input the Ross–Li Bidirectional Reflectance Distribution Function (BRDF) model parameters (volumetric, isotropic, and geometric [165]) for two spectral bands (visible and near-infrared), taken from the MODIS daily snow-free BRDF/albedo dataset (MCD43c2 v061), which are averaged over a 0.25° × 0.25° cell. For the pressure, temperature, air density, ozone, oxygen, water vapor, carbon dioxide, and nitrogen dioxide vertical profiles, we select one of the five standard atmospheres provided by libRadtran (i.e., midlatitude summer, midlatitude winter, subarctic summer, subarctic winter, and tropical). The ozone and water vapor profiles are scaled to match with the respective hourly averaged data from the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2 reanalysis [166]), provided at a 0.5° × 0.625° latitude-longitude grid (collection M2T1NXSLV [167]).

5. Results

5.1. Assessment of Lidar Ratio Impact on CALIOP AODs

An introductory analysis emphasizing on the assessment of the mid-visible (532 nm) CALIOP AODs versus the corresponding AERONET measurements is presented in this Section. Our aim is to highlight the impact of the defined LR (CALIOP-based, DeLiAn-based) on CALIOP’s performance in terms of reproducing the columnar AODs, one of the main drivers of the aerosol-induced DREs [17,144]. In Figure 4, we present the scatterplots between spaceborne (CALIPSO; y axis) and ground-based (AERONET; x-axis) AODs for the entire CALIPSO-AERONET collocated samples (550 cases; Figure 4a), as well as considering only the matchups where moderate-to-high AODs (≥0.2) are measured at the AERONET stations (197 cases; Figure 4b). This screening is crucial for assessing LR impact on the calculated AODs since the aerosol-induced DREs are intensified at high load conditions. The assessment analysis is jointly performed for the default CALIOP (red) and the DeLiAn-based LRs (green). Based on our results, for the entire sample, slight improvements are revealed when the DeLiAn lidar ratios are adopted. In quantitative terms, the slope of the linear regression and the Mean Bias Error (MBE) are slightly improved, whereas the correlation coefficient and the Root Mean Square Error (RMSE) values are similar/identical (Figure 4a). Similar findings are drawn when AODs exceed 0.2 (Figure 4b). In both categories, it is found that CALIOP underestimates AODs in agreement with previous studies [17,27,66,92,144,168]. This drawback is mainly attributed to the inability of the CALIOP sensor to detect tenuous aerosol layers, particularly during daytime due to the strong illumination background [27,37], whereas aerosol layers’ misclassification and possible cloud contamination, mainly by water clouds [146], can also affect the quality of the retrieved AODs.

An alternative way to present the assessment of the CALIOP AODs is illustrated in the box-whisker plots of Figure 5, in which the AERONET, the default CALIOP, and the DeLiAn-based AODs are compared when considerable aerosol loads (AOD ≥ 0.2) are recorded. Moreover, the boxplots are clustered separately for mineral particles, dust–marine, or dust–pollution–smoke mixtures recorded in the vicinity of AERONET stations. The remaining possible combinations are grouped under the “other” category whereas the overall results from the 197 cases are also provided. Among the aerosol categories, it is revealed that under pure dust conditions, the median AERONET and CALIOP AODs are exact (0.38) when the DeLiAn LR is applied on the CALIPSO dust backscatter coefficient profiles. On the contrary, when the default dust LR is used, CALIPSO underestimates the columnar AOD by ~11% (0.04). In environments with a coincident presence of dust and marine particles, the employment of the DeLiAn LRs leads to a reduction in the median CALIOP-AERONET AOD departures from −26% to −16%. Similar findings, albeit less pronounced, are found in scenes where dust particles along with pollution or smoke aerosols are probed. More specifically, the negative CALIOP-AERONET AOD biases are reduced from −10% to −3% thanks to the consideration of the DeLiAn LRs. For the rest of the cases (i.e., “other” group), the deviations are negligible. Overall, the consideration of the default and the DeLiAn LR yield median CALIOP AODs equal to 0.25 and 0.27, respectively, whereas those from AERONET are equal to 0.30. This indicates that the existing negative CALIOP-AERONET declinations are reduced from 17% down to 10% thanks to the employment of the DeLiAn LR, particularly in pure dust and dust-dominated atmospheric scenes.

5.2. Assessment of Lidar Ratio Impact on the Direct Radiative Effects (DREs)

Following the discussion in the previous Section regarding the lidar ratio impact on the CALIOP AODs, a subsequent analysis focusing on the aerosol-induced clear sky SW DREs is presented here. In the upper row of Figure 6 are the results for the total AERONET-CALIOP matchups, and in the bottom row are the scatterplots when AERONET AODs are higher/equal than/to 0.2. The comparison between the calculated DREs based on AERONET (x-axis) and CALIOP (y-axis) is performed for the DREs at the surface (NETSRFC; left column), within the atmosphere (ATM; middle column), and at the Top of the Atmosphere (TOA; right column). In the RTM simulations, all the inputs are identical, except AODs, thus facilitating the quantification of the LR impact on DREs computed based on the default (red circle points) and the DeLiAn (green circle points) values, under the assumption that the AERONET-based DREs serve as reference. In the Supplementary Materials, the AODs for each case within the aerosol categories (see Section 5.1) along with the solar zenith angle (SZA) and the surface albedo are provided.

In the shortwave spectrum, DRE_NETSRFC is negative, since aerosols induce a surface cooling due to scattering and absorption of the incoming solar radiation by the overlying aerosol layer(s). On the contrary, aerosol particles heat the atmosphere by absorbing a portion of the incoming shortwave radiation, as indicated by the positive DRE_ATM values. The combined effect of the perturbations at the surface (DRE_NETSRFC) and within the atmosphere (DRE_ATM) determines the signal at TOA, representing the disturbances to the Earth-atmosphere system’s radiation budget. Positive DREs at TOA indicate a planetary warming, whereas negative values correspond to a planetary cooling. According to our model calculations, positive DRE_TOA is found over most arid regions, especially in North Africa, under the presence of intense dust loads. The warming effect at TOA results from the multiple scattering between the relatively absorbing dust particles and the underlying highly reflective surface [2,17,53,55,144,169,170,171,172]. On the other hand, large negative DRE_TOA values are evident above “dark” maritime surfaces when dust or dust-dominated layers are advected. Over downwind oceanic regions receiving dust particles from the nearby deserts, DRE_TOA yields maximum negative values (in absolute terms) at a global scale [53,171]. In Europe, pollution or smoke-absorbing particles are met more frequently than other aerosol species, but they induce a rather weak imbalance in the radiation budget attributed to low AODs. Sea salt particles are mainly accumulated at low concentrations and, in combination with their strong scattering efficiency (SSA values are close to unity), cause a weak TOA cooling, which is mainly driven by the surface cooling, since the atmospheric warming effect is negligible.

In the scatterplots of Figure 6 are depicted the CALIOP DREs, with respect to those computed based on AERONET AODs. For all cases (upper row in Figure 6), it is noted that a reduction in the MBE (from CALIOP to DeLiAn LRs) for the NETSRFC (from 5.3 to 3.9 W/m²) and the ATM (from −3.3 to −2.4 W/m²). At the surface and within the atmosphere, the MBE reductions reveal a strengthening of the cooling and warming effects, respectively, attributed to the employment of the DeLiAn lidar ratios. Even though the underestimation of the CALIOP-based DREs still exists, the MBE reductions indicate a noticeable shift towards the desired direction (i.e., minimization of the CALIOP-AERONET deviations). At TOA, the 93% of the matchups corresponds to negative DREs (points residing in the bluish sector of the 2D space), and only the 7% indicate a planetary warming (i.e., positive DREs; points residing in the reddish sector of the 2D space). Overall, the suppression of the MBE levels from 2.0 W/m² (CALIOP LR) to 1.5 W/m² is mainly driven by the “improvements” seen on the negative DRE_TOA values (from 2.2 to 1.6 W/m²). An additional positive impact of the DeLiAn LRs appears in the increase in the slope values (the regression line comes closer to the 1:1 line) at all levels within the Earth-atmosphere system. By contrasting the CALIOP and the DeLiAn DREs, it is found that the correlation coefficients do not change, whereas the RMSE levels are slightly reduced for the latter ones. It must be noted that the obtained evaluation metrics are driven by the cluster of the CALIOP-AERONET collocated records associated with low-to-moderate AODs and DREs. For this reason, we repeated the analysis, focusing on cases with moderate-to-intense loads (AERONET AOD ≥ 0.2; 197 cases) inducing stronger perturbations on the radiation fields. Again, the adoption of the DeLiAn LRs improves the level of agreement between spaceborne-based and AERONET-based DREs, particularly at the surface and in the atmosphere, and at a lesser degree at TOA, as indicated by the offset of the evaluation scores towards the ideal values (i.e., MBE→0; slope→1; RMSE→0; r→1).

Following the same presentation format as for AOD, in Figure 7 are depicted the box-plots for the aerosol-speciated DREs at the ground (upper panel), within the atmosphere (middle panel) and at TOA (bottom panel), for the entire AERONET-CALIPSO collocated sample (550 cases). According to our results, the dust-induced DREs show the highest sensitivity on the LR changes, with the most noticeable improvements found for the DRE_NETSRFC. In quantitative terms, the CALIOP-based median values (−20.7 W/m²) are lower by 29% than those of the AERONET-based DREs (−29.1 W/m²). Thanks to the implementation of the DeLiAn LRs, the underestimation of the surface cooling is reduced by $~$50% (from 8.4 to 4 W/m²), leading to the final deviation of 14%. Within the atmosphere, the CALIOP-AERONET and the DeLiAn-AERONET percentage deviations are equal to 30% and 24%, respectively, showing an improvement under pure dust conditions. On the contrary, the CALIOP and the DeLiAn DRE_TOA levels are similar and lower than those based on AERONET. In atmospheric scenes where dust and marine particles coexist, the AERONET-satellite DRE declinations are equal to −20% (NETSRFC), −23% (ATM), and –23% (TOA) when the default CALIPSO LRs are used. The corresponding levels for the DeLiAn LRs are equal to −16%, −19%, and −18%, respectively. Improvements are also found at all levels of the Earth-atmosphere system when mineral particles are probed along with pollution/smoke aerosols. More specifically, the DRE_NETSRFC values are reduced from −46% down to −39%, the DRE_ATM from −51% down to 44%, and the DRE_TOA from −48% down to −42%. In the marine (M) and the pollution–smoke (P/S) categories, we do not obtain any significant variation on DREs by updating the CALIOP LR based on the observed values (DeLiAn), since its modification is minor (see Table 1). This finding suggests that the underestimated CALIOP-based DREs, compared to AERONET, can be attributed to first-order factors such as the underestimation of AOD due to undetected tenuous aerosol layers, rather than the definition of the lidar ratio. In the “other” category, the improvements do not exceed 4%, regardless of which DREs are investigated. Overall, the consideration of the DeLiAn LRs suppresses the CALIPSO-AERONET negative biases on DREs from −37% down to −34% at the surface, from −39% down to −31% within the atmosphere, and from −40% down to −37% at TOA.

In Figure 8 are illustrated the scatterplots between AERONET (blue points), CALIOP (red points), and DeLiAn (green points) DREs (y-axis) versus AERONET AODs (x-axis). The slope corresponds to the DRE change rate with respect to AOD and expresses the aerosol-induced radiation perturbations per AOD unit, a measure known as aerosol radiative efficiency [144,173,174,175]. It is clarified that for its calculation, we are including all cases characterized either by different aerosol regimes or solar zenith angles or underlying surface types. Therefore, it must be used with caution concerning its magnitude. Nevertheless, it is emphasized that our priority in the current study is to assess the impact of the lidar ratio. Aligned to this goal, we are comparing the obtained findings between the two CALIOP-based RTM simulations against those derived when the AERONET AODs are exploited. At the surface and within the atmosphere, there appeared a quasi-linear dependency between AODs and DREs, indicating a strengthening of the cooling and warming effects, respectively, as more aerosol particles are accumulated. This is expressed in quantitative terms by the sign and the absolute values of the correlation coefficients. Among the three datasets, the maximum slope (~−107 W/m² per AOD unit) is recorded for CALIOP and the minimum for AERONET (~−129 W/m² per AOD unit), whereas the DeLian slope resides in between (−117 W/m² per AOD unit). A similar ranking for the slopes’ absolute magnitudes it is recorded within the atmosphere, yielding values equal to ~87 W/m² (CALIOP), ~94 W/m² (DeLiAn), and ~103 W/m² (AERONET) per AOD unit. Therefore, the consideration of the DeLiAn LRs leads to stronger surface cooling and atmospheric warming and reduces the negative CALIOP-based departures approximately by ~8% at both levels. At TOA, the contribution of external factors (e.g., surface albedo) distorts the quasi-linear DREs-AODs dependency, and for this reason the collocated points are not “aligned” across the slope line, but spread in the two-dimensional space. Based on AERONET, the tropospheric aerosols induce a planetary cooling equal to ~26 W/m² per AOD unit, whereas according to CALIOP and DeLiAn, the corresponding values are lower by ~24% and ~11%, respectively.

5.3. Assessment of Lidar Ratio Impact on ARBE

A direct case-by-case comparison of DREs is always questionable, since several factors, regulating the sign and the magnitude of the induced radiative perturbations, vary among the atmospheric scenes. A more robust approach is feasible when ARBE is considered. ARBE represents the aerosol-induced perturbation (expressed in percentage) of the radiation fields with respect to those in an atmosphere without the presence of suspended particles (isolation of aerosol effects from the solar geometry effects), and it is calculated by normalizing DREs with the aerosol-free radiation fluxes (Equation (4)). The ARBE values for NETSRFC, ATM, and TOA have been summarized in the boxplots of Figure 9, and they are presented for the six aerosol categories, as well as for the whole AERONET-CALIPSO collocated sample (550 cases). Among the three levels of the Earth-atmosphere system, the largest perturbations are primarily found within the atmosphere (warming) and secondarily at the surface (cooling), whereas their net effect results in weak radiative disturbances at TOA. Focusing on the aerosol types, it is revealed that under pure dust conditions, the largest ARBEs are recorded, whereas moderate radiative imbalances are evident in cases where mineral and marine particles coexist. In dust scenes, the AERONET-based NETSRFC ARBEs are higher than those of CALIOP and DeLiAn by ~20% and ~6%, respectively. Similarly, within the atmosphere, the revised dust LR improves the level of agreement between AERONET and CALIPSO by reducing the negative biases from ~33% (default LR) down to ~19% (DeLiAn) with respect to AERONET. This positive tendency is also evident at TOA, but the median ARBE values in the three datasets are low (up to 0.6%). For the dusty marine (D+M) cases, the consideration of the DeLiAn LR leads to a suppression of the CALIPSO-AERONET negative biases, as for the dust scenes, but to a lesser extent. In the remaining aerosol categories, the observed improvements are either weak or negligible throughout the Earth-atmosphere system. For the entire sample (‘ALL’ in x-axis), we found that when we apply a more realistic LR, the CALIPSO-AERONET biases at the surface (~−37%), within the atmosphere (~−36%), and at TOA (~−34%) are slightly reduced by ~3%, ~6%, and 4%, respectively.

5.4. Evaluation of Simulated Radiation Fields at the Surface and at TOA

In the last Section, we evaluated the RTM-simulated SW fluxes of the downwelling radiation reaching at the ground versus BSRN measurements and the upwelling radiation at TOA against those observed using CERES. For the intercomparison between modelled and observed TOA SW radiation fluxes, we have detected all CERES overpasses (footprints) falling within a circle of 100 km radius centered at each AERONET site. The maximum time offset between the model and CERES observations was set to ±5 min to avoid variations (depending on the solar zenith angle) of the TOA radiation. Based on the cloud CERES-SSF data, we exclude all the cloud-contaminated footprints. Considering the strong dependency of the TOA-reflected radiation on the surface albedo, we apply an additional criterion related to the surface type. Within the CERES footprint, we extract the percentage coverage of the surface types, as these are classified in the IGBP database (included in the SSF data). For the model runs in which the predominant surface type is “ocean”, we exclude the CERES footprints containing “land” and vice versa (Section 4.3.2). From the remaining CERES footprints, we choose the nearest one to the AERONET site, avoiding possible inhomogeneous spatial patterns of the upwelling SW radiation at TOA. Our defined criteria are fulfilled in 369 out of the 550 cases.

Figure 10 presents the scatterplots between the CALIPSO-based (red), the DeLiAn-based (green), and the AERONET-based (blue) simulated TOA fluxes versus the observed ones using CERES, for all cases (369; upper panel) and for those when AERONET AOD exceeds 0.2 (144; bottom panel). The computed correlation coefficients indicate that the model is well configured, being able to capture the case-by-case variability, as well as indicating that the spaceborne observations have been treated appropriately. Nevertheless, in our simulations, the model underestimates the upwelling SW fluxes at TOA. These negative biases can be attributed to: (i) the AOD underestimation using CALIOP (e.g., undetected tenuous layers) (relevant for the CALIPSO-based RTM runs), (ii) the misrepresentation of the intensive aerosol optical properties (e.g., SSA) (relevant for all RTM runs), (iii) uncertainties in the aerosols’ macrophysical properties (vertical profile of aerosol layers) (relevant for all RTM runs), and (iv) the impact of non-aerosol factors such as the surface reflectance (relevant for all RTM runs).

Among the three RTM input datasets, the minimum MBEs (−4.0 W/m²) are achieved for AERONET, whereas the corresponding levels for CALIPSO and DeLiAn are equal to −6.0 W/m² and −5.4 W/m², respectively, when all cases are studied. Therefore, we obtain a slight reduction in the underestimation of the order of 15% (or 0.6 W/m² in absolute terms) when the revised LRs are adopted. Under moderate-to-high AOD conditions, the improvements become more noticeable, since the DeLiAn MBEs (−5.0 W/m²), instead of the CALIPSO (−6.1 W/m²) ones, are closer to those of AERONET (−3.4 W/m²). This translates into reductions by 1.1 W/m² and 47% expressed in absolute and percentage terms, respectively.

An assessment analysis has also been conducted for the surface solar radiation (SSR) at BSRN stations within the study region. In the RTM runs, we utilize as inputs the CALIPSO aerosol extinction coefficient based on the default and the revised lidar ratio. The collocation procedure, the quality screening, as well as the derivation/definition of aerosol optical properties are the same as for the model runs discussed in the previous Section. We exclude from the analysis elevated BSRN stations. The comparison between modelled and measured SSR is performed only when the BSRN measurements have been acquired under cloud-free conditions during the CALIPSO overpass time. To exclude cases where clouds, undetected using CALIPSO, affected the BSRN SSR measurements, we investigated the diurnal cycles of the measured SSR and DNI, along with the SSR standard deviation (GHI_std) within fine resolution time steps (i.e., 1-min). Moreover, we have calculated the clear sky theoretical diurnal cycle of SSR and DNI (pvlib python library [176]). We consider that the time periods close to CALIPSO overpasses are cloud-free when (i) no abrupt DNI changes occur, (ii) the GHI_std is near-zero, and (iii) the measured SSR and DNI are close to their theoretical clear sky values. In the Appendix A (Figure A1), we present two examples of the cloud screening process. This methodology yielded 35 suitable cases for intercomparison between the modelled and the measured SSRs. Figure 11 shows the scatterplots between the BSRN SSR and the respective modelled data based on the CALIPSO (red) and DeLian (green) LRs, along with the computed statistical metrics, the linear fit, and the y = x line. The high R and slope values ($~$1), along with the low RMSE and MBE levels, indicate a very good model performance. However, there are not clear differences on RMSEs and MBEs between the CALIOP-based (15.92 W/m² and 10.61 W/m²) and the DeLiAn-based (15.48 W/m² and 9.82 W/m²) RTM runs. We note that this result is expected, since, in the selected atmospheric scenes, weak aerosol loads are recorded (AODs < 0.25). Even though our findings are restricted by the limited number of cases, we can conclude at some point that the overestimation of the simulated SSR is attributed mainly to the underestimation of AOD using CALIOP and likely by an underestimation of the SSA (becoming more critical when strong/moderate light absorbing particles are probed), since the surface albedo uncertainty plays a trivial role on the SSR amount.

6. Conclusions

Aerosols’ burden across North Africa, Europe, and the Middle East (NAMEE domain) is defined by the coexistence of suspended particles of natural and anthropogenic origin. Under this complex regime, aerosol load is characterized by a pronounced heterogeneity, posing challenges in the discrimination among their respective types and subsequently on the assessment of their climatic role. The overarching goal of the present study is to investigate the aerosol-induced shortwave (SW) direct radiative effects (DREs) in cloud-free atmospheres by combining spaceborne retrievals (Cloud-Aerosol Lidar with Orthogonal Polarization–Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIOP-CALIPSO)), radiative transfer simulations (libRadtran), sunphotometric observations/retrievals (AERONET), and a collection of ancillary datasets (DeLiAn, MODIS, MERRA-2, CERES, and BSRN). More specifically, we emphasize the importance of a realistic definition of the aerosol-speciated lidar ratio (LR) on the quality-assured CALIPSO retrievals, which are used as inputs in radiative transfer model (RTM) simulations for the estimation of DREs. As a reference for LR, we utilize measurements derived using state-of-the-art ground-based lidars operating in places affected by different aerosol species. The aforementioned data have been assembled in the DeLiAn database. In our analysis, we exclude stratospheric aerosols (as these are defined in the CALIPSO aerosol classification scheme), since their occurrence over the study period (2007–2020) is very rare in the region of interest (NAMEE domain). Based on the CALIPSO classification, tropospheric aerosols are categorized in seven types, including mixtures such as “dusty marine” and “dust-pollution”. Through the implementation of well-established decoupling techniques (relying on the linear particle depolarization ratio), we conclude that there are five aerosol types, namely dust, marine, polluted continental/smoke, elevated smoke, and clean continental. Taking advantage of the highly accurate AERONET aerosol optical depths (AODs) and the availability of intensive aerosol optical properties (single scattering albedo, asymmetry parameter), the RTM simulations have been performed for 550 atmospheric scenes probed using CALIOP when CALIPSO flies nearby the ground stations (≤100 km).

First, an assessment analysis for AOD (determining at a large degree DREs) has been conducted, showing that CALIPSO underestimates AOD with respect to AERONET. In quantitative terms, the negative CALIOP-AERONET biases reach down to −0.04 (all cases) and −0.06 (AERONET AOD ≥ 0.2) when the default CALIPSO LRs are considered. Thanks to the employment of the DeLiAn LRs, the negative declinations are suppressed by 25% and 32%, respectively. Among aerosol types, the most pronounced improvements are encountered mainly when dust particles are solely recorded, and secondarily in atmospheric scenes where dust coexists with other aerosols (i.e., marine, pollution, and smoke). Nevertheless, it is important to acknowledge that the enhancements attributed to the revised LRs are partially impeded by errors/shortcomings in CALIOP retrievals/observations (e.g., undetected tenuous aerosol layers).

The lidar ratio is a critical parameter for the computation of DREs involving elastic lidar-derived AOD. To assess its impact on the aerosol-induced radiative imbalances at the surface (NETSRFC), within the atmosphere (ATM), and at the Top of the Atmosphere (TOA), we performed four sets of RTM experiments for each study case. The control run corresponds to an aerosol-free atmosphere, whereas the other three simulations have been performed by utilizing AERONET AODs, the default CALIOP AODs (without modifying the raw aerosol-speciated LRs), and the CALIOP AODs relying on the DeLiAn aerosol-speciated LRs. As expected, aerosols tend to cool the surface (due to the reduction of the downwelling SW radiation) and warm the atmosphere (due to the absorption of the incoming SW radiation). At TOA, negative DREs (planetary cooling) dominate over “dark” surfaces, whereas their sign turns to positive (planetary warming) when absorbing particles (i.e., dust), accumulated at moderate-to-high concentrations, are suspended over bright surfaces (i.e., deserts). Regardless of which LR is assumed, CALIPSO underestimates the aerosol-induced radiative effects throughout the Earth-atmosphere system with respect to those derived using AERONET due to inherent deficiencies on the CALIOP lidar (i.e., undetected tenuous aerosol layers). However, the consideration of the DeLiAn LRs leads to obtain better results mainly when mineral particles are either solely recorded or coexist with other aerosol species (e.g., sea salt). In quantitative terms, the MBE values are reduced by ~26–27% for the NETSRFC (from 5.3 to 3.9 W/m²) and ATM (from −3.3 to −2.4 W/m²) DREs when the DeLiAn LRs are employed. The improvement becomes more significant at moderate-to-high load conditions (AOD ≥ 0.2), reaching up to ~35%. At TOA, positive tendencies of similar magnitude are obtained, becoming evident mostly at cases where dust or dust-dominated layers are advected over maritime areas, resulting in negative DREs (i.e., planetary cooling).

A fair case-by-case comparison of the aerosol-induced radiative effects has been also conducted, relying on the aerosol radiative budget efficiency (ARBE). Within the Earth-atmosphere system, the largest disturbances are primarily recorded within the atmosphere and secondarily at the surface, whereas their combined effect result in low ARBEs at TOA. Among the defined aerosol categories, the best agreement between AERONET-based and satellite-based simulations is revealed for dust when the DeLiAn LR has been adopted. Under pure dust conditions, the negative satellite-AERONET declinations reach down to ~−20% and ~−32% for NETSRFC and ATM, respectively, when the default CALIPSO LR is applied. Thanks to the consideration of the DeLiAn LR, there are improvements in the order of ~14% at both layers and of ~10% at TOA. When all cases are studied, the desired improvements are suppressed significantly, since in non-dust atmospheric scenes, the satellite-AERONET ARBE departures are lower. Based on the default CALIPSO LRs, the ARBE levels within the Earth-atmosphere system are underestimated by 34–37% with respect to those based on AERONET, whereas the consideration of the DeLiAn LRs introduces improvements ranging from 3% to 5%.

Finally, we evaluated the simulated SW radiation fluxes at the top of the atmosphere and at the surface against the clear sky satellite CERES-SSF observations and BSRN ground-based measurements, respectively. For the ΤOA upwelling irradiances, we found that the MBE values are reduced from −6.1 W/m² (CALIPSO LRs) down to ~5 W/m² (DeLiAn LRs) at moderate-to-high aerosols loads (AOD ≥ 0.2). For the downwelling SW radiation at the ground, the corresponding MBE values are equal to 10.6 W/m² and 9.8 W/m². The negative TOA MBEs (indicating that less radiation is reflected back to space from aerosols) along with the positive surface MBEs (indicating that more radiation reaches the ground) underscore the underestimation of aerosol load using CALIPSO, as well as demonstrating that first-order errors dominate over the lidar ratio adjustments. Overall, the implementation of the DeLiAn LR led to prominent improvements under pure dust and dust-dominated conditions. The observed differences can be considered lower and within the aerosol-measured uncertainties for the other aerosol types, but still in the right direction.

Our analysis emphasizes the importance of a realistic definition of the lidar ratio on CALIPSO retrievals towards assessing the aerosol SW direct radiative effects under the critical assumption of a reliable aerosols classification scheme on the spaceborne observations. Better aerosol typing and the construction of representative aerosol models will boost such efforts on satellite-based studies via the provision of reliable microphysical/optical properties governing aerosol–radiation interactions. Under the framework of MIRA (Models, In situ, and Remote Sensing of Aerosols; https://science.larc.nasa.gov/mira-wg/ (accessed on 15 January 2024), the aerosol community aims to integrate advanced measurements and optical calculations targeting to build a Table of Aerosol Optics (TAO), expanding existing historical databases [155]. Obtaining more information on aerosol properties would be extremely advantageous for elastic backscatter lidars (such as CALIOP), enhancing their retrieval quality and their applicability in aerosol-radiation studies (among others). The advent of the EarthCARE satellite (expected launch on May 2024), a joint mission of the European Space Agency (ESA), and the Japan Aerospace Exploration Agency (JAXA) will substantially upgrade the monitoring capacity from space [177]. Moreover, synergies with CALIPSO will facilitate a more robust aerosol typing via the harmonization of aerosol optical properties at 355 nm (EarthCARE) and 532 nm (CALIPSO) [178]. The HETEAC (Hybrid End-To-End Aerosol Classification) retrieval algorithm (developed for the EarthCARE (Cloud, Aerosol, and Radiation Explorer) satellite mission) classifies tropospheric aerosol species in two fine modes (weakly/strongly absorbing) and in two coarse modes (spherical, non-spherical), resembling typical conditions met in the real world [178]. For the first time, joint aerosol, radiation, and cloud observations will be acquired by four instruments (ATLID (atmospheric LIDAR), MSI (Multispectral Instrument), CPR (Cloud Profiling Radar), BBR (Broadband Radiometer)) mounted on the EarthCARE platform [179]. Thanks to this unique opportunity, radiative closure experiments [180], similarly done in the current study, will substantially advance our knowledge on aerosol–cloud-radiation interactions and on global radiative forcing [177,178]. Finally, the exploitation of synergies between CALIPSO with passive satellite sensors can enable the representation of dust lidar ratio spatial patterns, relying on the precise identification of depolarizing mineral particles using CALIOP and the reliable quantification of the columnar aerosol extinction from the passive spaceborne instruments.