Atmospheric Processes over the Broader Mediterranean Region 1980–2024: Effect of Volcanoes, Solar Activity, NAO, and ENSO

Abstract

The Mediterranean region is regarded as a hot spot on Earth because of its placement at the junction of many aerosols. Numerous studies have demonstrated that the North Atlantic Oscillation (NAO), which is closely related to the El Niño–Southern Oscillation (ENSO) phenomenon, influences the weather in the area. However, a recent study by the same author examined the ENSO effect on atmospheric processes in this area and discovered a slight but notable influence. This study builds on that earlier work, but it divides the Mediterranean region into four smaller regions during the same time span as the previous study, which is extended by two years, from 1980 to 2024. The division is based on geographical, climatological, and atmospheric process features. The findings demonstrate that volcanic eruptions significantly affect the total amount of aerosols. Additionally, the current study reveals that the Granger-causality test of the physical phenomena of solar activity, ENSO, and NAO indicates that all have a significant impact, either separately or in combination, on the atmospheric process over the four Mediterranean regions, and this effect can last up to six months. Moreover, a taxonomy of the different forms of aerosols across the four subregions is given.

1. Introduction

Electromagnetic energy from the Sun makes up the atmospheric radiation that traverses the Earth’s atmosphere. This radiation is affected by clouds, particles, and gases in the atmosphere [1]. Atmospheric aerosols are solid or liquid particles that flow in the air for periods of hours to days and are crucial to the Earth’s climate system [2,3], and its energy balance because they scatter and absorb solar radiation (direct effect) at both short- and long-wavelengths [4]. They also indirectly affect precipitation and cloud albedo [5,6,7]. Additionally, depending on their size, concentration, chemistry, and height in the atmosphere, atmospheric aerosols have a substantial impact on global climate change [8]. Dust, sea salt, black carbon, and volcanic debris are examples of primary aerosols, whereas sulphates, nitrates, ammonium, and secondary organic compounds are examples of secondary aerosols, which are produced by chemical processes [9]. As a result, atmospheric aerosols come from human activities, natural sources, or a mix of both [10]. Black carbon is primarily an absorbing aerosol, while sulphates, nitrates, ammonium, and sea-salt particles are examples of solely scattering aerosols. The direct role of atmospheric aerosols is related to their capacity to scatter and absorb solar radiation, while the indirect impact is related to their role as cloud condensation nuclei [11], which affects cloud albedo and modifies precipitation patterns in a particular location.

Even though their optical properties are currently well understood, there are still many unknowns regarding the climatic effects of atmospheric aerosols due to their wide variety of types, physicochemical and optical properties, impact on meteorology through weather modification, and atmospheric mixing processes [12]. Additionally, a rise or fall in aerosol concentration can proportionately diminish or increase solar radiation, which is particularly visible in clear-sky conditions because air particles interact with solar radiation through the mechanisms of scattering and absorption. Because of this, atmospheric aerosols are a major source of uncertainty in climate projections [13], and knowledge of them is necessary to make precise predictions about the amount of solar energy available in a given place [14,15].

On Earth, there are numerous large-scale atmospheric and oceanic circulation phenomena that influence weather both locally and globally. Among them, the El Niño–Southern Oscillation (ENSO) and the North Atlantic Oscillation (NAO) play important roles in global and regional climates, respectively. The significance of the Arctic Oscillation (AO), ENSO, and NAO circulations was examined by Yoo et al. [16], who concurred with [17] that ENSO activity is significant since future changes in NAO are anticipated to be more reliant on ENSO. Abish and Mohanakumar [18] demonstrated that during El Niño years, aerosol concentrations over the Indian subcontinent rise noticeably; high aerosol index values (i.e., 1.7) over the area make this clear. European weather visibly and strongly responds to the peak phases of NAO (both positive and negative), according to Laken and Stordal’s [19] analysis of European weather data from the Hess–Brezowsky Large Weather Scenario. On the other hand, these authors discovered no meaningful connection between ENSO and/or solar cycle and European weather. On the contrary, Kirov and Georgieva [20] found a close connection between solar activity and NAO and ENSO circulations. Mezzina et al. [21] underlined that the ENSO phenomenon helps make surface weather forecasts over the North Atlantic’s mid-latitudes more predictable. In coupled climate models, Casselman et al. [22,23,24] also verified the teleconnection between ENSO and NAO circulation patterns. The timing of the link between ENSO and the tropical North Atlantic was re-examined by García-Serrano et al. [25]. In the spring (March–May) of the following year, Liu et al. [26] discovered a strong positive long-term correlation between the tropospheric CO levels over the European sector of the North Atlantic and the winter ENSO event. The South Asian monsoon has a significant impact on the Eastern and Central Mediterranean climates, according to Alpert et al. [27]. Throughout the past 120 years (1901–2020), Urdiales-Flores et al. [28] showed that the Mediterranean basin has warmed more quickly that the rest of the world; they ascribed the greater warmth to the combined impacts of the region’s lower soil moisture and decreased aerosol pollution. Park [29] found a remote influence of ENSO on Mediterranean sky conditions during late summer and autumn. A recent study by Alsubhi and Ali [30] demonstrated the effect of ENSO on the dust outbreaks over the Arabian Peninsula.

Although the teleconnection between the ENSO and NAO circulations has been established as previously shown, there has been little attention paid to the question of whether ENSO may have an impact on atmospheric processes over the Mediterranean (i.e., atmospheric radiation and atmospheric aerosols). The only work on this topic is a recent study by the same author [31] that found a statistically significant relationship between ENSO and Ångström’s exponent, and between ENSO and direct aerosol radiative forcing over the larger Mediterranean region. On the other hand, little research has been conducted on how the ENSO phenomenon affects aerosol particles in the wider area surrounding the event. However, Westervelt et al. [32] stated that regional decreases in aerosol emissions are usually responsible for a shift to the positive ENSO phase (i.e., an El Niño event). Huang et al. [33] found that ENSO-related sea-surface temperature anomalies in the tropical Eastern Pacific have a substantial effect on suspended dust particles in the Arabian Peninsula to the Central Asia region. Banerjee and Kumar [34] have shown that El Niño can change the intensity of dust subsidence across the Indian subcontinent by remotely adjusting the strength of convection over the India–Pacific region. According to Brönnimann et al. [35] and Toniazzo and Scaife [36], both phases of ENSO can influence NAO. Leamon [37] discovered a highly significant association between the occurrence of the termination of the last five solar cycles and the Pacific’s Ocean shift from El Niño to La Niña. Huo et al. [38] studied the influence of solar activity in the North Atlantic and the tropical Pacific regions, but found no solar effect in the simulated sea-level pressure/sea-surface temperature coupled mode.

As far as the Mediterranean is concerned, many researchers think that the region is a hot spot for aerosol events from a number of sources, including air pollution from big cities or industrial areas [39,40], dust from the Sahara Desert [41,42,43,44,45], eruptions from volcanoes [46,47], and wildfires [48]. Figure 1 in [49] provides valuable information about the transport of Sahara dust and air pollutants throughout the Mediterranean region.

There has been some hazy evidence of a direct ENSO influence on the Mediterranean climate, despite the fact that the impact of the ENSO phenomenon on atmospheric processes has not been thoroughly investigated for the Mediterranean region up to now [50]. Conversely, there are studies that look at how ENSO affects weather patterns in many parts of the world. For example, the El Niňo/La Niňa phenomena have been demonstrated to have a significant influence on worldwide weather patterns, including monsoons [51,52,53], and the North Atlantic Ocean close to the east coast of the United States and Western Europe [21,54].

The present study is a continuation of a recent work by the same author [31]. That study examined the Mediterranean region as a unified area; its main conclusions were as follows: (i) positive correlation was found for TAOD₅₅₀, AE_470–870 with ONI or NAOI, and a negative one for DARF, netSWR_BOA,CS,A with either ONI or NAOI; (ii) a positive trend for netLWR_BOA and a negative trend for COD, TAOD₅₅₀, SAOD₅₅₀, and AE_470–870 in the period 1980–2024; (iii) the majority of the atmospheric aerosols over the entire Mediterranean region were classified as CMA and UIA types; (iv) there was a noticeable drop in DARF from cold to warm ONI events; and (v) DARF exposed a positive trend with time (1980–2022). (The explanation of the symbols used here are given in Table 2, Section 2). For further research, that article suggested that “a future study might perform analyses on parts of the Mediterranean region in order to achieve better spatial resolution”. This future research is therefore conducted here.

The questions to be investigated are, therefore, as follows: (i) Are the effects of solar activity, NAO, and ENSO on the atmospheric processes found in [31] valid in smaller areas of the Mediterranean too? (ii) How do volcanic eruptions affect the status of atmospheric aerosols over the Mediterranean? To do this, a preliminary condition is posed: the Mediterranean domain considered in [31] is split into four smaller sectors, which are configured in Section 2. Solar activity (in terms of the sunspot number, SSN) has also been added in the present study because some authors have shown a dependence of NAO events on SSN (e.g., [20,55]).

The work is divided into sections. Section 1 gives the necessary state of knowledge on the subject. Section 2 provides the data and the methodology used in this work. Section 3 deploys the results of the study in two parts: Part A. Physical approach; and Part B. Statistical approach. Section 4 provides the main findings of the work and gives thoughts for future research. Finally, Acknowledgements, Appendix A, and References follow.

2. Materials and Methods

As mentioned above, the present work is a logical extension to [31]. The difference between that work and the present study is the investigation of the effects from ENSO and volcanic eruptions on the atmospheric processes over four adjacent regions that make up the broader Mediterranean area used in [31]. These four subregions are shown in Figure 1 and have the following geographical coordinates: West Mediterranean (or WestMed in brevity, 6° W–10° E, 30° N–45° N); Central Mediterranean (CentMed, 10° E–0° E, 30° N–45° N); East Mediterranean (EastMed, 20° E–35° E, 30° N–41° N); and Balkans and Black Sea (BalBSea, 20° E–35° E, 41° N–45° N). The reasoning behind the division of the entire Mediterranean area into four subregions was dictated by the following considerations: (i) WestMed (Spain, South France, Northwest Italy, North Morocco, North Algeria) has distinct Atlantic influence, aerosol inflow from Iberia and Sahara, and high solar flux; (ii) CentMed (Italy, North Tunisia, Malta, West Greece, North Libya) lies in a transitional zone with complex aerosol mixing, volcanic activity, and high solar irradiance; (iii) EastMed (Greece, Turkey, Cyprus, Israel, North Egypt) is characterised by strong eastern aerosol transport, semi-arid climate, intense solar radiation; and (iv) BalBSea (Balkans, Bulgaria, Black Sea) has features of continental climate, unique aerosol pathways, and lower solar insolation.

Table 1. Justification of criteria for dividing the entire Mediterranean into four segments. The criteria refer to the climatology, atmospheric aerosols, geography, and solar irradiance levels in the four areas.

A. Climatology	B. Aerosols	C. Geography	D. Solar radiation	References
Western and eastern basins have different precipitation regimes. Northern areas (Atlantic, Black Sea) experience more continental weather patterns.	Western: Atlantic and Saharan dust. Central: mixed influence. Eastern: Middle East and Anatolian sources. Northern: European industrial and biomass-burning aerosols.	Each subregion clusters seas and coasts with shared topography and oceanographic features.	Highest in East and Central Mediterranean. Lower in the northern and western parts due to higher cloudiness and latitude.	[56,57] [58] Mediterranean Biogeographic Region—Wikipedia (https://en.wikipedia.org/wiki/Mediterranean_Biogeographic_Region (accessed on 31 July 2025)) Clouds and solar radiation | Copernicus (https://climate.copernicus.eu/esotc/2023/clouds-and-solar-radiation (accessed on 31 July 2025))

provides more details on the above criteria.

To answer the research questions posed in Section 1, the Giovanni platform, which can be accessed without cost at https://giovanni.gsfc.nasa.gov/giovanni (accessed on 8–25 October 2023) [59], provided month-to-month data for the years 1980–2024 (45 years or 4.5 decades) at a spatial resolution of 0.5° × 0.625°. More specifically, for each subregion, the following data were downloaded: netSWR at the bottom-of-the-atmosphere (BOA) with and without aerosols in the atmosphere and under clear-sky conditions, netSWR_BOA,CS,A and netSWR_BOA,CS,NA, respectively, in Wm⁻² (MERRA-2 model; Giovanni files M2TMNXRAD v5.12.4); netSWR at the top-of-the-atmosphere (TOA) with and without aerosols in the atmosphere and under clear-sky conditions, netSWR_TOA,CS,A and netSWR_TOA,CS,NA, respectively, in Wm⁻² (MERRA-2 model; Giovanni files M2TMNXRAD v5.12.4); netLWR at BOA with and without aerosols in the atmosphere and under clear-sky conditions, netLWR_BOA,CS,A and netLWR_BOA,CS,NA, respectively, in Wm⁻² (MERRA-2 model; Giovanni files M2TMNXRAD v5.12.4); netLWR radiation at TOA with and without aerosols in the atmosphere and under clear-sky conditions, netLWR_TOA,CS,A and netLWR_TOA,CS,NA, respectively, in Wm⁻² (MERRA-2 model; Giovanni files M2TMNXRAD v5.12.4); scattering aerosol optical depth at 550 nm, SAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4); desert-dust aerosol optical depth at 550 nm, DDAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4); total aerosol optical depth at 550 nm, TAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4); black-carbon aerosol optical depth at 550 nm, BCAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4); stratospheric aerosol optical depth at 550 nm, StAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXRAD v5.12.4); sea-salt aerosol optical depth at 550 nm, SSAOD₅₅₀ (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4); cloud optical depth, COD (MERRA-2 model; Giovanni file M2TMNXRAD v5.12.4); and Ångström’s exponent in the spectral band 470–870 nm, AE_470–870 (MERRA-2 model; Giovanni file M2TMNXAER v5.12.4). Although the MERRA-2 scientific team in charge evaluates the data for quality before publishing it on the Giovanni website, further quality checks were performed on the downloaded data to look for outliers, missing, and out-of-range values; however, none were discovered.

As far as the accuracy in the estimation of AOD by the MERRA-2 reanalysis is concerned, Shaheen et al. [60], in a study of AOD trends over the East Mediterranean in the period 2000–2018, concluded the following: (i) MERRA-2 AOD estimates show a systematic low bias compared to AERONET measurements, particularly under high-AOD conditions associated with desert dust; and (ii) MODIS observations perform better during episodic events but have larger spread in monthly mean values. On the other hand, Su et al. [61], analysing AOD data from MERRA-2, MODIS, and AERONET in the period 1980–2023, found that MERRA-2 AOD estimates had the greatest accuracy with an expected error ±0.05 ± 20% compared to AERONET measurements. Mukkavilli et al. [62] used AOD data over Australia in the period 2002–2012 taken from NASA’s MERRA-2 and the Monitoring Atmospheric Composition and Climate (MACC) programme of the European Centre for Medium-Range Weather Forecasts (ECMWF), evaluating them against concurrent ground-based measurements from the Aerosol characterisation via Sun photometry: Australian Network (AeroSpan), which is part of the global Aerosol Robotic Network (AERONET) and found that MERRA-2 reanalysis overestimates monthly AOD twice as much compared to AeroSpan/AERONET ground observations but corelates better with AeroSpan/AERONET than ECMWF/MACC. Khatibi and Krauter [63] evaluated MERRA-2 air temperature, wind speed, and solar radiation datasets against measured data from Meteonorm and Deutscher Wetterdienst (DWD, Germany) over eight stations around the world and found high correlation coefficients from 0.95 to 0.99 for solar radiation, 0.99 for air temperature, and from 0.81 to 0.99 for wind speed. They concluded that MERRA-2 datasets are valuable, considering their global coverage and availability.

In addition to the above datasets, monthly mean data was downloaded during the same time period for the sunspot number, SSN, the North Atlantic Oscillation Index, NAOI, and the Oceanic El Niño Index 3.4, ONI_3.4. For more information about the indices, the reader is referred to Section 2 in [31] and

Table 2. List of the parameters used in the present work with a short description.

Parameter	Remarks
ONI	Index of the El Niño–Southern Oscillation: a standard for tracking the oceanic part of the ENSO climate pattern. The current standard is ONI3.4. In the rest of the text, Figures and Tables ONI implies ONI3.4. The NOAA’s scientists have determined the (−0.5, +0.5) zone as neutral ENSO occurrences; further, they have defined warm El Niňo events for ONI ≥ +0.5 and cold La Niňa ones for ONI ≤ −0.5 (see https://cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php (accessed on 8–25 October 2023)).
NAOI	Index of the North Atlantic Oscillation: a measure of the surface sea-level pressure difference between the subtropical High at Azores and the subpolar Low over Iceland. A neutral zone similar to that for ONI is established here. Therefore, it makes sense to use the same delimitation as ONI; neutral NAO events in the band (−0.5, +0.5), positive NAO cases for NAOI ≥ +0.5, and negative NAO events for NAOI ≤ −0.5.
SSN	Sunspot number: the number of sunspots on the surface of the Sun associated with the 11-year solar activity.
TAOD550	Total aerosol optical depth at 550 nm: a measure of the total solar radiation extinction by aerosols at 550 nm.
SAOD550	Scattering aerosol optical depth at 550 nm: a measure of the solar radiation extinction by scattering aerosols at 550 nm.
AAOD550	Absorbing aerosol optical depth at 550 nm: a measure of the solar radiation extinction by absorbing aerosols at 550 nm.
DDAOD550	Desert-dust aerosol optical depth at 550 nm: a measure of the solar radiation extinction by scattering desert-dust aerosols at 550 nm.
StAOD550	Stratospheric aerosol optical depth at 550 nm: a measure of the solar radiation extinction by scattering stratospheric aerosols at 550 nm (mainly sulphates erupted by volcanoes).
SSAOD550	Sea-salt aerosol optical depth at 550 nm: a measure of the solar radiation extinction by sea-salt particles at 550 nm.
BCAOD550	Black-carbon aerosol optical depth at 550 nm: a measure of the solar radiation extinction by absorbing black-carbon aerosols at 550 nm.
AE470–870	Ångström’s exponent: a measure of the aerosols size in the visible band of 470–870 nm. A typical value of 1.3 is often used.
SSA	Single-scattering albedo: a measure of the scattering property of aerosols; if SSA = 1, the particles are purely scatterers; if SSA ≈ 0, the particles only absorb.
DARF	Direct aerosol radiative forcing: it refers to the scattering and/or absorption of solar radiation by atmospheric aerosols. Positive DARF values indicate local/regional warming, while negative values indicate local/regional cooling.
COD	Cloud optical depth: a measure of the solar radiation extinction by clouds. It plays significant role in the Earth’s energy budget.
netSWRBOA,CS,A	Net SW radiation at BOA, CS, and A: difference between the incoming and reflected radiation at the surface of the Earth under clear skies and presence of aerosols in the atmosphere.
netSWRBOA,CS,NA	Net SW radiation at BOA, CS, and NA: difference between the incoming and reflected radiation at the surface of the Earth under clear skies and absence of aerosols in the atmosphere.
netSWRTOA,CS,A	Net SW radiation at TOA, CS, and A: difference between the incoming and outgoing radiation at 100 km from the surface of the Earth under clear skies and presence of aerosols in the atmosphere.
netSWRTOA,CS,NA	Net SW radiation at TOA, CS, and A: difference between the incoming and outgoing radiation at 100 km from the surface of the Earth under clear skies and absence of aerosols in the atmosphere.
netLWRBOA,CS,A	Net LW radiation at BOA, CS, and A: difference between the incoming and emitted infra-red (IR) radiation at the surface of the Earth under clear skies and presence of aerosols in the atmosphere.
netLWRBOA,CS,NA	Net LW radiation at BOA, CS, and A: difference between the incoming and emitted IR radiation at the surface of the Earth under clear skies and absence of aerosols in the atmosphere.
netLWRTOA,CS,A	Net LW radiation at TOA, CS, and A: difference between the incoming and outgoing IR radiation at 100 km from the surface of the Earth under clear skies and presence of aerosols in the atmosphere.
netLWRTOA,CS,NA	Net LW radiation at TOA, CS, and A: difference between the incoming and outgoing IR radiation at 100 km from the surface of the Earth under clear skies and absence of aerosols in the atmosphere.

in the present work. Moreover, it is important to mention that the SSN, NAOI, and ONI data is not site-specific.

The netSWR and netLWR data was used to determine DARF in the atmosphere. This study has taken into account clear-sky (CS) conditions because, as said in the Introduction, “…a rise or fall in aerosol concentration can proportionally reduce or increase solar radiation, especially noticeable under clear-sky conditions”. The following are the net-radiation expressions [1]:

DARF = (Q_A − Q_NA)_TOA − (Q_A − Q_NA)_BOA,

(1)

Q_A = netSWR_A − netLWR_A,

(2)

Q_NA = netSWR_NA − netLWR_NA,

(3)

netSWR or netLWR = F↓ − F↑.

(4)

In the equations above, Q is the difference between netSWR and netLWR, whereas netSWR or netLWR is the difference between the downward (F↓) minus the upward (F↑) radiation fluxes (by convention, downward fluxes are positive and upward fluxes are negative). Aerosol-laden and aerosol-free atmospheres are denoted by the subscripts A and NA, respectively. In all computations, the four Mediterranean subregions shown in Figure 1 have been regarded as distinct entities; that is, all examined parameter values are time series of monthly data averaged over the selected Mediterranean domains between 1980 and 2024.

The absorption, scattering, and total aerosol optical depths at the wavelength of 550 nm are interrelated through the equation [1]:

TAOD₅₅₀ = SAOD₅₅₀ + AAOD₅₅₀.

(5)

Since AAOD₅₅₀ values are not available on the Giovanni platform, they were computed from Equation (5) by subtracting SAOD₅₅₀ from TAOD₅₅₀. Additionally, the single-scattering albedo, SSA, was estimated by the following equation:

SSA = SAOD₅₅₀/TAOD₅₅₀.

(6)

In Equation (5), light-scattering entities in the atmosphere are considered the desert-dust particles, volcano-related stratospheric aerosols, sea-salt, and, partially, clouds as they may have either a scattering or an absorbing role (e.g., [64,65,66]). Therefore, SAOD₅₅₀ in Equation (5) can be replaced with the sum of the optical depths of desert-dust aerosols, stratospheric aerosols, sea-salt, and clouds. As far as the inclusion of COD in Equation (5) is concerned, this was performed by dividing COD by 1000 in order to have a compatible magnitude with those of other optical depths. Moreover, because of the clouds’ double role as scatterers or absorbers, one can consider that COD = COD_sca + COD_abs, where COD_sca = 0.85∙COD refers to their scattering property and COD_abs = 0.15∙COD to their absorbing role (e.g., [64]). These coefficients represent globally averaged fractions derived from radiative transfer studies; while they may vary depending on cloud microphysics and wavelength, they offer a reasonable first-order estimation for climatological analysis [67]. Thus, the scattering and absorbing CODs to be included in Equation (5) can be expressed as follows:

COD_sca = 0.85∙(COD/1000),

(7a)

COD_abs = 0.15∙(COD/1000).

(7b)

Therefore, SAOD₅₅₀ and AAOD₅₅₀ can be rewritten as follows:

SAOD₅₅₀ = DDAOD₅₅₀ + StAOD₅₅₀ + SSAOD₅₅₀ + CODsca + OSAOD₅₅₀,

(8)

AAOD₅₅₀ = BCAOD₅₅₀ + COD_abs + OAAOD₅₅₀.

(9)

In Equation (8), the latter parameter implies the optical depth of other scattering particles not explicitly designated here (cf. nitrates, ammonium salts, secondary organic aerosols, secondary aerosols). In Equation (9), the absorbing aerosols are black carbon and clouds; the latter parameter means the optical depth of other absorbing particles not explicitly referred to here (cf. brown carbon, soot aggregates). An approximation of TAOD₅₅₀ in Equation (5) can thus be rewritten as TAOD’₅₅₀ by taking into consideration Equations (8) and (9):

TAOD’₅₅₀ ≈ (DDAOD₅₅₀ + StAOD₅₅₀ + SSAOD₅₅₀ + COD_sca) + (BCAOD₅₅₀ + COD_abs).

(10)

In Equation (10), the first parenthesis refers to SAOD₅₅₀ and the second to AAOD₅₅₀; both OSAOD₅₅₀ and OAAOD₅₅₀ have been omitted. According to the above, the downloaded TAOD₅₅₀ can be considered as the observed values, while TAOD’₅₅₀ may refer to the estimated TAOD₅₅₀ time series. This is said because TAOD₅₅₀ comes directly from MERRA-2 reanalysis calculations (cf. observations) and TAOD’₅₅₀ is the summation (cf. estimation) of the optical depths of atmospheric aerosols.

NOAA uses the ONI_3.4 standard to classify El Niño (warm) and La Niña (cold) episodes in the East Pacific Ocean (https://cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php (accessed on 8–25 October 2023)). This index [68] is used to monitor the oceanic part of the ENSO climate pattern. ONI_3.4 is computed as a running 3-month mean sea-surface temperature (SST) anomaly for the El Niño region (5° S–5° N, 120° W–170° W). If the SST anomaly for five consecutive overlapping 3-month periods is >+0.5 °C, an event is considered warm; if it is <−0.5 °C, it is considered cold.

The sea-level pressure difference between the subpolar Low over Iceland and the subtropical High at the Azores serves as the basis for NAOI. When NAOI is positive, the central region of the North Atlantic Ocean, the East United States, and West Europe are experiencing above-normal pressures, whereas the high latitudes of the North Atlantic are experiencing below-normal pressures [69]. The Gulf Stream, which starts in the Gulf of Mexico and moves northward across the Atlantic, is solely linked to the NAO phenomenon. Temperatures above normal in the East United States and North Europe and below normal in Greenland and frequently in South Europe and the Middle East are associated with high positive NAOI phases. They provide below-normal precipitation patterns throughout South and Central Europe and above-normal precipitation over North Europe and Scandinavia [70]. Temperature and precipitation patterns during severe adverse (negative) NAOI phases are opposite of those during significant positive NAO occurrences. The NAOI calculation procedure is explained in [71].

The Earth’s energy budget is determined by COD, a fundamental cloud feature that comes from ground-based or satellite observations (e.g., [64,65,66]).

In the Ångström’s wavelength-dependent formula for characterising how atmospheric aerosols attenuate solar light, AE is the exponent, which can be calculated through the following equation [1]:

TAOD_λ = β·λ^AE

(11)

where λ is the wavelength (in nm). The present work has taken AE in the visible wavelength band, i.e., 470–870 nm. The value of AE = 1.3 is commonly considered in the international literature; above this threshold, aerosols are characterised as fine, and below this value, aerosol particles are characterised as coarse.

The SSN is associated with the 11-year solar activity. Number 25 is the current solar cycle, which began in December 2019 and is still running. The examined period from 1980 to 2024 includes solar cycles 22–25.

Table A1. List of volcanoes with volcanic explosivity index, VEI, in the period 1980–2024. Note that various volcanoes have been active through several years.

Year	Volcano, Country	VEI
1980	St. Helens, USA Etna, Italy Stromboli, Italy	5 2 2
1981	Pagan, USA Alaid, Russia St. Helens, USA Etna, Italy Stromboli, Italy	4 4 5 2 2
1982	Galunggung, Indonesia El Chichón, Mexico Pagan, Mexico St. Helens, USA Etna, Italy Stromboli, Italy	4 5 4 5 2 2
1983	Colo, Indonesia Galunggung, Indonesia Pagan, USA St. Helens, USA Etna, Italy Stromboli, Italy	4 4 4 5 2 2
1984	Pagan, USA St. Helens, USA Etna, Italy Stromboli, Italy	4 5 2 2
1985	Pagan, USA St. Helens, USA Etna, Italy Stromboli, Italy	4 5 2 2
1986	Chikurachki, Russia Klyuchevskoy, Russia St. Helens, USA Augustine, USA Etna, Italy Stromboli, Italy	4 4 5 4 2 2
1987	Klyuchevskoy, Russia Etna, Italy Stromboli, Italy	4 2 2
1988	Klyuchevskoy, Russia Etna, Italy Stromboli, Italy	4 2 2
1989	Klyuchevskoy, Russia Etna, Italy Stromboli, Italy	4 2 2
1990	Kelud, Indonesia Klyuchevskoy, Russia Etna, Italy Stromboli, Italy	4 4 2 2
1991	Cerro Hudson, Chile Pinatubo, Philippines Etna, Italy Stromboli, Italy	5 6 2 2
1992	Spurr, USA Etna, Italy Stromboli, Italy	4 2 2
1993	Lascar, Chile Etna, Italy Stromboli, Italy	4 1 2
1994	Rabaul, Papua N. Guinea Etna, Italy Stromboli, Italy	4 3 2
1995	Rabaul, Papua N. Guinea Etna, Italy Stromboli, Italy	4 3 2
1996	Etna, Italy Stromboli, Italy	3 2
1997	Etna, Italy Stromboli, Italy	3 2
1998	Etna, Italy Stromboli, Italy	3 2
1999	Sheveluch, Russia Etna, Italy Stromboli, Italy	4 3 2
2000	Ulawun, Papua N. Guinea Sheveluch, Russia Etna, Italy Stromboli, Italy	4 4 3 2
2001	Sheveluch, Russia Etna, Italy Stromboli, Italy	4 3 2
2002	Reventador, Ecuador Ruang, Indonesia Sheveluch, Russia	4 4 4
2003	Reventador, Ecuador Sheveluch, Russia	4 4
2004	Manam, Papua N. Guinea Sheveluch, Russia	4 4
2005	Manam, Papua N. Guinea Sheveluch, Russia	4 4
2006	Rabaul, Papua N. Guinea Manam, Papua N. Guinea Sheveluch, Russia	4 4 4
2007	Rabaul, Papua N. Guinea Manam, Papua N. Guinea Sheveluch, Russia	4 4 4
2008	Kasatochi, USA Okmok, USA Chaiten, Chile Rabaul, Papua N. Guinea Manam, Papua N. Guinea Sheveluch, Russia	4 4 4 4 4 4
2009	Sarychev Peak, Russia Chaiten, Chile Rabaul, Papua N. Guinea Manam, Papua N. Guinea Sheveluch, Russia	4 4 4 4 4
2010	Merapi, Indonesia Eyjafjallajokull, Iceland Chaite, Chile Rabaul, Papua N. Guinea Sheveluch, Russia	4 4 4 4 4
2011	Nabro, Eritrea Puyehue-Cordon Caulle, Chile Grimsvotn, Iceland Chaite, Chile Sheveluch, Russia	4 5 4 4 4
2012	Nabro, Eritrea Puyehue-Cordon Caulle, Chile Sheveluch, Russia	4 5 4
2013	Sinabung, Indonesia Etna, Italy Sheveluch, Russia Stromboli, Italy	4 3 4 2
2014	Manam, Papua N. Guinea Semeru, Indonesia Kelud, Indonesia Sinabung, Indonesia Etna, Italy Sheveluch, Russia	4 4 4 4 3 4
2015	Wolf, Ecuador Calbuco, Chile Manam, Chile Semeru, Indonesia Sinabung, Indonesia Sheveluch, Russia Etna, Italy	4 4 4 4 4 4 3
2016	Manam, Papua N. Guinea Semeru, Indonesia Sinabung, Indonesia Etna, Italy Sheveluch, Russia	4 4 4 3 4
2017	Semeru, Indonesia Manam, Papua N. Guinea Sinabung, Indonesia Etna, Italy Sheveluch, Russia	4 4 4 3 4
2018	Manam, Papua N. Guinea Semeru, Indonesia Sinabung, Indonesia Etna, Italy Sheveluch, Russia Stromboli, Italy	4 4 4 3 4 2
2019	Ulawun, Papua N. Guinea Sinabung, Indonesia Manam, Papua N. Guinea Etna, Italy Semeru, Indonesia Sheveluch, Russia Stromboli, Italy	4 4 4 3 4 4 2
2020	Soufriere, Saint Vincent and the Grenadines Taal, Philippines Manam, Papua N. Guinea Semeru, Indonesia Etna, Italy Sheveluch, Russia Stromboli, Italy	4 4 4 4 3 4 2
2021	Hunga Tonga-Hunga Ha’apai, Tonga Fukutu-Oka-no-Ba, Japan Soufriere, Saint Vincent and the Grenadines Manam, Papua N. Guinea Semeru, Indonesia Sheveluch, Russia Etna, Italy Stromboli, Italy	5 4 4 4 4 4 3 2
2022	Hung Tonga-Hunga Ha’apai, Tonga Manam, Papua N. Guinea Semeru, Indonesia Sheveluch, Russia Etna, Italy Stromboli, Italy	5 4 4 4 3 2
2023	Manam, Papua N. Guinea Semeru, Indonesia Sheveluch, Russia Etna, Italy Stromboli, Italy	4 4 4 2 2
2024	Manam, Papua N. Guinea Semeru, Indonesia Sheveluch, Russia Etna, Italy Stromboli, Italy	4 4 4 2 2

(see Appendix A) gives a list of the energetic volcanoes in the period of the study with VEI (volcanic explosivity index) magnitudes 4–6 taken from the Smithsonian Institution, USA (https://volcano.si.edu). VEI is a measure of the size of explosive volcanic eruptions and was defined by [72]. The VEI scale spans between 0 and 8; VEIs up to 6 have the following definitions: VEI = 0, category: effusive, column height: <100 m, ejecta volume: <10⁴ m³, periodicity: constant, tropospheric/stratospheric circulation: negligible/none; VEI = 1, category: gentle, column height: 100 m⁻¹ km, ejecta volume: >10⁴ m³, periodicity: daily, tropospheric/stratospheric circulation: minor/none; VEI = 2, category: explosive, column height: 1–5 km, ejecta volume: >10⁶ m³, periodicity: 2 weeks, tropospheric/stratospheric circulation: moderate/none; VEI = 3, category: severe, column height: 3–15 km, ejecta volume: >10⁷ m³, periodicity: 3 months, tropospheric/stratospheric circulation: substantial/possible; VEI = 4, category: catastrophic, column height: >10 km, ejecta volume: >0.1 km³, periodicity: 18 months, tropospheric/stratospheric circulation: substantial/definite; VEI = 5, category: cataclysmic, column height: >10 km, ejecta volume: >1 km³, periodicity: 12 years, tropospheric/stratospheric circulation: substantial/significant; VEI = 6, category: colossal, column height: >20 km, ejecta volume: >10 km³, periodicity: 50–100 years, tropospheric/stratospheric circulation: substantial/substantial. The selection of volcanoes with VEI ≥ 4 was based on their severity and the potential circulation of the debris in the troposphere/stratosphere. An exception was made for the volcanoes Etna and Stromboli in Italy [73], which showed VEIs between 2 and 3 in the period of the study. This exception is justified by the fact that both volcanoes are within the studied region; no matter how low their activity is, they may affect atmospheric loading and composition in the CentMed area [74].

Table 2 provides a summary of the parameters examined in this study, along with relevant commentary to provide an overview of the variables. It is important to note that this list includes factors like solar fluxes and clouds that do not represent aerosol properties and large-scale circulation patterns; nevertheless, they are directly related to aerosol properties, and it is therefore desirable to examine how they depend on SSN, NAO, and ENSO.

3. Results

This section is divided into two parts: the first part examines the physical properties of the aerosols and atmospheric parameters mentioned in Section 2, while the second part investigates the relationships among the parameters using statistical procedures.

3.1. Part A—Physical Approach

3.1.1. Observed and Estimated Aerosol Optical Depth

The definitions of the observed and estimated TAOD are provided and elucidated by Equations (5) and (10), respectively. To demonstrate the accuracy, Figure 2 illustrates the annual mean values of the estimated TAOD’₅₅₀ and the observed TAOD₅₅₀, along with their percentage difference, defined as follows:

ΔTAOD₅₅₀ (%) = 100·(TAOD₅₅₀ − TAOD’₅₅₀)/TAOD₅₅₀.

(12)

The plots in Figure 2 indicate that in all four Mediterranean subregions, the mean errors ΔTAOD₅₅₀ are mainly within the band ±6% of the TAOD₅₅₀ values or lower; more specifically, the mean ± std of ΔTAOD₅₅₀ are as follows: −0.247% ± 2.218 (WestMed), −2.019% ± 4.121 (CentMed), −2.858% ± 2.303 (EastMed), and +0.488% ± 2.796 (BalBSea), which indicate that TAOD’₅₅₀ overestimates TAOD₅₅₀ in WestMed, CentMed, and EastMed, and underestimates it over BalBSea. Nevertheless, the error introduced is very low, as it lies in the band (−2.86%, +0.49%), and the approximation of TAOD’₅₅₀ to TAOD₅₅₀ can therefore be considered accurate. This means that the estimation of the total aerosol extinction factor can be derived from the summation of the solar radiation attenuations outlined in Equation (10), in case of TAOD₅₅₀ unavailability (see also Figure 2). It is noteworthy that the ΔTAOD₅₅₀ values are negative (i.e., TAOD₅₅₀ < TAOD’₅₅₀) in WestMed, CentMed, and EastMed, and positive (i.e., TAOD₅₅₀ > TAOD’₅₅₀) only over BalBSea. This observation can be attributed to the fact that TAOD₅₅₀ intrinsically contains the effect of clouds (i.e., COD), which is found to have a lesser real impact (lower cloudiness) over the WestMed, CentMed, and EastMed subregions than the simulated TAOD’₅₅₀. In fact, the central and eastern areas of the Mediterranean region present fewer clouds, according to a study on cloudiness over the Mediterranean [75]. Despite the procedure for estimating TAOD’₅₅₀, the present work uses the observed TAOD₅₅₀ values in all calculations.

The observed TAOD₅₅₀ values from Figure 2 during the research period over all four Mediterranean subregions are reproduced in Figure 3. In this graph, linear regression fits to the TAOD₅₅₀ annual time series are presented, having the following expressions, significant at the 99.9% confidence interval (CI): TAOD₅₅₀ = −1.6 × 10⁻³·t + 3.439, R² = 0.401 (WestMed), TAOD₅₅₀ = −2.2 × 10⁻³·t + 4.268, R² = 0.510 (CentMed), TAOD₅₅₀ = −1.2 × 10⁻³·t + 2.715, R² = 0.272 (EastMed), and TAOD₅₅₀ = −3.6 × 10⁻³·t + 7.503, R² = 0.738 (BalBSea). The EastMed’s lower R² score indicates that the linear regression model poorly captures the relationship between t and TAOD₅₅₀, at least in relation to the expressions for the other three subregions. However, it is noteworthy that the EastMed’s R² is greater than those for the other three areas when considering the TAOD_550,WVE values (see Figure 5, Section 3.1.2).

3.1.2. Effect of Volcanoes

Strong volcanic eruptions that release sulphur into the stratosphere are the main cause of stratospheric aerosols [76]. Because of the massive eruptions from Mt. El Chichón and Mt. Pinatubo volcanoes in 1982/1983 and 1992/1993, respectively, the impact of volcanic debris cycling in the atmosphere is clearly visible in Figure 3. Sulphating particles make up the majority of the debris left over during volcanic eruptions. The two large StAOD₅₅₀ peaks in Figure 4 in 1982/1983 and 1992/1993, as a result of the eruptions from the aforementioned volcanoes, demonstrate the importance of these particles in the global atmospheric aerosol loading.

The dashed lines in Figure 4 are linear fits to the StAOD₅₅₀ data series with the following expressions: StAOD₅₅₀ = −0.19 × 10⁻²·t + 3.786, R² = 0.434 (WestMed), StAOD₅₅₀ = −0.26 × 10⁻²·t + 542042, R² = 0.588 (CentMed), StAOD₅₅₀ = −0.17 × 10⁻²·t + 3.565, R² = 0.399 (EastMed), and StAOD₅₅₀ = −0.38 × 10⁻²·t + 7.823, R² = 0.750 (BalBSea). All regressions are significant at the 99.9% CI for the period of the study, where t is a year in the period 1980–2024. The lower R² score for EastMed is again noticeable and may be attributed to two factors: (i) EastMed is a transitional zone between continental, maritime, and desert influences, it receives episodic inputs from Saharan dust and volcanic plumes, which vary irregularly over time [77]; (ii) stratospheric aerosol transport is modulated by Quasi-Biennial Oscillation (QBO), Brewer–Dobson circulation, and tropopause height variability, which may affect EastMed more erratically than other regions [78]. On the other hand, the subregion is subject to complex synoptic systems and shifting wind regimes (e.g., Etesians, Sirocco), making aerosol transport highly variable and less predictable by a simple time-based model [79].

As seen in Figure 4, StAOD₅₅₀ falls between 0.03 and 0.3 in each of the four Mediterranean subregions. In a study on stratospheric aerosols from 1850 to 1990, Sato et al. [80] discovered global StAOD₅₅₀ values ranging from 0.001 to 0.125 (see their Table 2), quite comparable with those of the present study. According to the authors, the significant eruption of the Indonesian Krakatau volcano in August 1883 (VEI = 6) caused the high StAOD₅₅₀ value of 0.125 in 1885, which affected the globe for over two years.

The volcanic debris that circles the Earth at stratospheric heights and drastically scatters solar radiation obscures all other aerosol impacts. As a result, the sharp decline in TAOD₅₅₀ in all four Mediterranean subregions is plasmatic (see Figure 2 and Figure 3). This issue becomes clear by removing StAOD₅₅₀ from TAOD₅₅₀:

TAOD_550,WVE = TAOD₅₅₀ − StAOD₅₅₀,

(13)

where WVE = without volcanic eruptions. The result clearly demonstrates the impact of both natural and man-made aerosols; TAOD_550,WVE then represents the aerosol loading in the atmosphere free of volcanic debris. To graphically show this result, Figure 3 is converted to Figure 5, which shows TAOD_550,WVE in the four Mediterranean areas.

The annual average TAOD₅₅₀ values in the entire 45-year period are as follows: 0.203 ± 0.034 (WestMed), 0.228 ± 0.037 (CentMed), 0.205 ± 0.031 (EastMed), and 0.234 ± 0.056 (BalBSea); the corresponding TAOD_550,WVE values are as follows: 0.135 ± 0.013 (WestMed), 0.135 ± 0.017 (CentMed), 0.117 ± 0.014 (EastMed), and 0.084 ± 0.010 (BalBSea). From these values, one can easily conclude that the average TAOD_550,WVE levels are 33.5%, 40.8%, 42.9%, and 64.1% lower than the corresponding TAO_D550 in the four Mediterranean subregions, respectively. This result demonstrates unequivocally how volcanic particles outweigh the total atmospheric aerosol loading in the larger Mediterranean region. This phenomenon is thought to be global, but further relevant research is required to prove it. The comparable average values of TAOD₅₅₀ (as well as of TAOD_550,WVE) across the three Mediterranean subregions (WestMed, CentMed, and EastMed), and their almost half values over the BalBSea region, are another finding of the analysis, which indicates that global volcanic eruptions do not necessarily affect Mediterranean aerosols in a uniform manner.

The aforementioned observations immediately lead to the conclusion that there is no discernible trend in aerosols for a volcanic-debris-free atmosphere over the research period. In fact, TAOD_550,WVE = 2 × 10⁻⁴·t − 0.346, R² = 0.061 (WestMed), TAOD_550,WVE = 5 × 10⁻⁴·t − 0.938, R² = 0.179 (CentMed), TAOD_550,WVE = 5 × 10⁻⁴·t − 0.924, R² = 0.238 (EastMed), and TAOD_550,WVE = 2 × 10⁻⁴·t − 0.321, R² = 0.074 (BalBSea) are the linear regression fits to the annual TAOD_550,WVE values (see Figure 5). All regressions are insignificant at the 95% CI with the exception for EastMed, which is significant at the 99.9% CI and has a higher R² score; t denotes any year between 1980 and 2024. The slope is almost negligible (of the order of 10⁻⁴) in all linear regression expressions, resulting in a trend that is not significant throughout the 45-year period. Therefore, in regard to the climate change issue, it would be very interesting if climatologists could research how a (hypothetical) atmosphere devoid of volcanic debris might affect the Earth’s climate.

3.1.3. Effect of NAO and ENSO

In a recent study, Serykh [81] reaffirmed the teleconnection between the El Niño–Southern Oscillation and the tropics of the Indian and Atlantic Oceans in temperate and high geographical latitudes; this was also performed in other research works (e.g., [21,22,23]). Therefore, to show the effect of both the NAO and ENSO phenomena on TAOD₅₅₀, AE_470–870, DARF, SSA, netSWR_BOA,CS,A, and netLWR_BOA,CS,A, 3D contour plots have been prepared.

Figure 6 displays TAOD₅₅₀ as a function of ONI and NAOI in the four Mediterranean areas. It is clear that WestMed, CentMed, and EastMed favour higher TAOD₅₅₀ values in neutral/positive ONI and NAOI events (see Figure 6a, Figure 6b and Figure 6c, respectively). In contrast, the Balkans and Black Sea subregion may accommodate greater TAOD₅₅₀ values for negative ONI and positive NAOI phases (see Figure 6d).

In South Europe, below-normal precipitation patterns are often linked to positive NAO episodes, as shown in Section 2 [24]. The decreased washout impact of the atmospheric aerosols over the Mediterranean region (South Europe) in positive NAO phases explains the TAOD₅₅₀ patterns in Figure 6. The sole difference is that, unlike the other three areas, the BalBSea subregion has high TAOD₅₅₀ values for ONI ≤ 0 rather than ONI ≥ 0. Thus, the question that arises is why the ENSO effect reverses over the BalBSea area.

Quite opposite patterns to those of TAOD₅₅₀ were derived for DARF, almost symmetrical to the (NAOI, ONI) = (0, 0) point. Higher DARF values (higher warming effect) are shown over WestMed and CentMed under both negative ONI and NAOI phases (see Figure 7a and Figure 7b, respectively). The patterns over EastMed and BalBSea are more complex; EastMed has a secondary peak in the neutral zones of both ONI and NAOI (see Figure 7c), whereas BalBSea presents a spread pattern with high DARF values within the neutral zones of both ONI and NAOI, as well as in the neutral ONI/positive NAOI bands, and in the neutral NAOI/positive ONI zone (see Figure 7d). The positive DARF values (warming effect) vary between 1 Wm⁻² and 7 Wm⁻², being higher over WestMed and CentMed and lower in EastMed and BalBSea.

For all four Mediterranean subregions, including BalBSea, the AE_470–870 versus ONI and NAOI patterns (not shown here) were found to be similar to those of TAOD₅₅₀. The difference from TAOD₅₅₀ is that greater AE_470–870 values occur for ONI ≥ 0 and NAOI ≤ 0. This new result indicates that over all four Mediterranean subregions, positive ONI and positive/negative NAOI phases favour fine-mode aerosols (AE_470–870 > 1.3).

In all four Mediterranean domains, the 3D plots for SSA versus ONI and NAOI (not shown here) display broad peaked values in the upper left corner, i.e., positive/neutral ONI and negative NAOI values. In general, an aerosol scattering profile over the larger Mediterranean region with SSA values greater than 0.90 emerges.

The netSWR_BOA,CS,A parameter yielded results that were strikingly comparable to the SSA patterns (not displayed here). This suggests that the scattering aerosol particles are the main source of solar radiation attenuation.

Higher values for the netLWR_BOA,CS,A parameter appear symmetrically to the NAOI = 0 axis compared with those for the netSWR_BOA,CS,A parameter (not displayed here) for WestMed, CentMed, and EastMed. These high values are found in the contour plot’s lower left corner, specifically for both negative ONI/NAOI phases.

Figure 8 shows the annual TAOD₅₅₀ values as a function of ONI in the four selected subregions of the Mediterranean. The graph shows that the majority of the TAOD₅₅₀ data points in all four areas lies within the neutral ONI band, fewer cases in the cold ONI zone, and less in the warm band, i.e., 11, 27, and 7 years (or 155, 324, and 125 months) for cold, neutral, and warm ENSO phases, respectively. This result is true for all the parameters considered in this study. In other words, ENSO seems to regulate the atmospheric aerosol circulation over the broader Mediterranean region. Similar hints have been given in a study by Westervelt et al. [32], which showed that positive ENSO phases (i.e., El Niño events) are associated by regional reductions in aerosol emissions. Also, Huang et al. [33] showed that suspended dust particles over the region from the Arabian Peninsula to Central Asia are significantly impacted by ENSO-related SST anomalies in the tropical Eastern Pacific. Moreover, Banerjee and Kumar [34] demonstrated that, by remotely altering the strength of convection across the India–Pacific region, El Niño can modify the intensity of dust subsidence over the Indian subcontinent. Nevertheless, an immediate conclusion from Figure 8 is that the aerosol loading over the broader Mediterranean area is more frequent during neutral ENSO phases in the majority of the cases (27 years out of the 45-year period in each Mediterranean area).

3.1.4. Effect of Desert Dust and Black Carbon

One may wonder how anthropogenic black-carbon aerosols and desert-dust particles impact the overall aerosol loading over the wider Mediterranean region. This section answers this question. To give an initial overview of both parameters, Figure 9 shows the TAOD₅₅₀ and TAOD_550,WDDBC curves (WDDBC = without desert-dust and black-carbon aerosols). These graphs show how the temporal history of the TAOD₅₅₀ across the research period is exactly mirrored by the TAOD₅₅₀ and TAOD_550,WDDBC values across the four Mediterranean subregions. The maxima in TAOD₅₅₀ and StAOD₅₅₀ in the years 1982/1983 and 1992/1993 in Figure 3 and Figure 4, respectively, provide confirmation that StAOD₅₅₀, SSAOD₅₅₀, and COD are the primary contributing elements to the total TAOD₅₅₀ after desert-dust and black-carbon aerosols are excluded.

To work out the contribution of TAOD_550,WDDBC to the total TAOD₅₅₀, one could find the ratio of the areas under the TAOD_550,WDDBC curves to those under TAOD₅₅₀, expressed as a percentage. To make things easier, instead of computing the areas under both curves of TAOD₅₅₀ and TAOD_550,WDDBC, one can convert the areas under the curves into equivalent orthogonal parallelograms by calculating their temporal averages in the whole period of the study. Then, the ratios of the areas of these orthogonal parallelograms are equal to those of the areas under the real curves. For these calculations, the annual averages of TAOD₅₅₀ and TAOD_550,WDDBC are estimated at 0.203 ± 0.034 and 0.094 ± 0.035 (WestMed), 0.228 ± 0.037 and 0.128 ± 0.039 (CentMed), 0.205 ± 0.031 and 0.121 ± 0.034 (EastMed), and 0.234 ± 0.056 and 0.183 ± 0.057 (BalBSea), respectively. The expression 100∙(TAOD₅₅₀ − TAOD_550,WDDBC)/TAOD₅₅₀) (in %) represents the contribution of all other aerosols (i.e., StAOD₅₅₀, SSAOD₅₅₀, and COD) to the total TAOD550, which is 53.7% (WestMed), 43.9% (CentMed), 41.0% (EastMed), and 21.8% (BalBSea). By subtracting these percentages from 100%, one can derive the contribution of DDAOD₅₅₀/BCAOD₅₅₀ to the total TAOD₅₅₀, which is 46.3% (WestMed), 56.1% (CentMed), 59.0% (EastMed), and 78.2% (BalBSea). These figures clearly show that there is an almost 50%/50% balance between the desert-dust/black-carbon aerosols and the stratospheric/sea-salt/cloud effects in the WestMed, CentMed, and EastMed areas. At first glance, the BalBSea subregion seems to be affected more by desert-dust and black-carbon aerosols than by volcanic eruptions. Nevertheless, desert dust is not generally present over BalBSea; this justifies the much lower TAOD_550,WVE average over BalBSea than those over the other three areas (see Section 3.2). This imbalance over BalBSea may assign its different patterns shown in Figure 6 and Figure 7 in comparison to those in the other three Mediterranean subregions.

3.1.5. Examination of Some Individual Aerosol Parameters

Figure 10 shows how TAOD₅₅₀, AAOD₅₅₀, SAOD₅₅₀, and DDAOD₅₅₀ relied on ONI during the study period across the four Mediterranean subregions. DDAOD₅₅₀ is linked to the movement of desert dust, primarily from the Sahara Desert, throughout the Mediterranean region [82]. A lower concentration of absorbing aerosols over the Mediterranean during the studied period is inferred from the graph’s very low AAOD₅₅₀ values, which indicate low attenuation of solar radiation by absorbing aerosols. This is in contrast to the higher concentration of scattering aerosols (higher SAOD₅₅₀ values). Therefore, the main factor reducing solar radiation as it passes through the atmosphere over the larger Mediterranean region is the scattering process. This result suggests that the only factors lowering solar irradiance throughout the Mediterranean are scatterers such as nitrates, sea-salt, and sulphate particles [9]. In a similar investigation for the Mediterranean as a whole, conducted during nearly the same experimental period as the current work, Kambezidis [31] made the same observation. Moreover, in a study on Greece’s climate and solar radiation, the same author [83] came to the opposite conclusion for the country between 2005 and 2016. This discrepancy was highlighted in [31], indicating that the Mediterranean needs to be examined in smaller sections, as is the case in the current study, in order to completely comprehend the aerosol influence throughout the region. Additionally, because dust aerosols appear to both scatter and absorb solar radiation (depending on the size of their particles), DDAOD₅₅₀, which is located between SAOD₅₅₀ and AAOD₅₅₀ (see Figure 10), displays a Janus face, that is, one scattering and one absorbing face [84]. The levels of DDAOD₅₅₀ approach those of AAOD₅₅₀ when one transitions from WestMed to BalBSea, suggesting that the absorbing “face” of DDAOD₅₅₀ may be taking the central stage. The methodology described in Kambezidis [31] can easily yield some interesting results when considering the averages of TAOD₅₅₀, AAOD₅₅₀, SAOD₅₅₀, and DDAOD₅₅₀ over each of the four subregions in the 45-year period (WestMed: 0.203 ± 0.034, 0.012 ± 0.001, 0.191 ± 0.034, 0.103 ± 0.013; CentMed: 0.228 ± 0.037, 0.012 ± 0.001, 0.216 ± 0.037, 0.093 ± 0.013; EastMed: 0.205 ± 0.031, 0.011 ± 0.001, 0.195 ± 0.031, 0.080 ± 0.012; BalBSea: 0.234 ± 0.056, 0.010 ± 0.001, 0.224 ± 0.055, 0.042 ± 0.007, respectively).

• WestMed. The expression TAOD₅₅₀ = SAOD₅₅₀ + AAOD₅₅₀ could alternatively be written as TAOD₅₅₀ = (DAOD₅₅₀ + OSAOD₅₅₀) + AAOD₅₅₀ (see Equations (8)–(10)). In the event of a Sahara dust outbreak, the scattering particles could include dust particles as well as other particles (such as sea-salt, sulphates, and nitrates, denoted as OSAOD₅₅₀ in the equation). By replacing the average values of TAOD₅₅₀ = 0.203, AAOD₅₅₀ = 0.012, SAOD₅₅₀ = 0.191, and DDAOD₅₅₀ = 0.103 in the last expression, it is easily found that the average OSAOD₅₅₀ is 0.088, which is equal to ≈7·AAOD₅₅₀. In the same way, DAOD₅₅₀ ≈ 9·AAOD₅₅₀. Therefore, TAOD₅₅₀ = (9·AAOD₅₅₀ + 7·AAOD₅₅₀) + AAOD₅₅₀ = 17·AAOD₅₅₀. This outcome indicates that the total solar radiation attenuation in the WestMed subregion is 17 times more than the attenuation resulting from aerosol absorption alone.
• CentMed. Following the same procedure, it is estimated that TAOD₅₅₀ = 19·AAOD₅₅₀. This implies that the total solar radiation is attenuated in the CentMed subregion 19 times by an atmosphere containing absorbing aerosols only.
• EastMed. Here, TAOD₅₅₀ = 19·AAOD₅₅₀, as in CentMed. This result shows the resemblance of the aerosol profiles in these two Mediterranean subregions (cf. Figure 10b,c).
• BalBSea. Now, TAOD₅₅₀ = 23·AAOD₅₅₀ having a higher coefficient than the one in the other three subregions (see the aerosol profiles in Figure 10d).
• Kambezidis [31] found the AAOD₅₅₀ coefficient to be 18 for the whole Mediterranean area in the same period, a value rather close to the average of the four coefficients in the present study (19.5).
• The immediate conclusion from the above results is that the contribution of absorbing aerosols to the total solar radiation attenuation becomes less and less important from WestMed to BalBSea as its role is progressively taken up by the scattering effect. This is also confirmed by the progressively increasing average SSA values over the four subregions: 0.939 (WestMed), 0.947 (CentMed), 0.948 (EastMed), and 0.958 (BalBSea).

Figure 10 shows that the annual values in the TAOD₅₅₀ and AAOD₅₅₀ time series occur within the neutral ONI zone most of the time. A remarkable observation, though, is that the DDAOD₅₅₀ and SAOD₅₅₀ time series are always occurring during cold La Niňa events.

EastMed’s average TAOD₅₅₀ value throughout the time period under study in this work is 0.206, which is quite consistent with the East Mediterranean’s 0.207 value discovered by Ozdemir et al. [85] between 1999 and 2018. The average TAOD₅₅₀ value for the Mediterranean region from 2005 to 2013 was found to be 0.150, according to Chiapello et al. [86]. This value is much lower than the average value found in the current investigation for the same experimental period (0.219). According to a research by Kaskaoutis et al. [87] on the aerosol loading over the Athens region in Greece, the city’s pollution influence on solar radiation resulted in a high yearly average of TAOD₅₅₀ = 0.35 between 2000 and 2005. Mallet et al. [88] found TAOD₄₄₀ values over the Mediterranean region in the period 1996–2012 in the range 0.15–0.32, these values include those of the present work (0.204–0.236).

Ångström’s exponent, AE_470–870, is a number that indicates aerosol size. A plot of AE_470–870 versus ONI similar to that of TAOD₅₅₀ in Figure 8 is shown in Figure 11a. For the four Mediterranean subregions, the distribution of the data pairs (AE_470–870, ONI) is shown here. Every year, it is noted that coarse-mode particles are present throughout the Mediterranean, irrespective of the ENSO phase. Only two years, 1983 and 1992, are associated with fine-mode aerosols, and these are related to the volcanic eruptions of Mt. El Chichón and Mt. Pinatubo (Mt. El Chichón, Mexico, 28 March–4 April 1982; Mt. Pinatubo, Philippines, 12–15 June 1991). Mt. El Chichón injected 7 × 10⁶ tons of SO₂ and 20 × 10⁶ tons of sulphuric acid aerosols into the atmosphere within a month after the eruption, causing a surface cooling in the northern hemisphere of ≈−0.2 °C [89,90]. The eruption occurred just as the 1982/1983 El Niňo event was in its starting phase; because of this, some scientists have supposed that the El Chichón eruption triggered a positive El Niňo phase [91]. Nevertheless, climate models and detailed studies of past volcano eruptions and El Niňo events have not shown a connecting line between the two events; therefore, the timing must be just a coincidence [91]. On the other hand, the Pinatubo eruption released a huge amount (17 × 10⁶ tons) of SO₂ into the stratosphere causing a global temperature decrease of ≈−0.5 °C and a reduction in solar light by 10% globally over 1992–1993 [92].

Figure 11b shows the temporal evolution of AE_470–870 over the 45-year period, with peaks in 1983 and 1992 corresponding to the Mt. El Chichón and Mt. Pinatubo eruptions, respectively. Higher AE_470–870 values are observed over the BalBSea subregion than the other three areas. The dashed lines are linear regression fits to the AE_470–870 data series with the following expressions: AE_470–870 = −8.1 × 10⁻³·t + 17.072, R² = 0.331 (WestMed); AE_470–870 = −7.4 × 10⁻³·t + 15.723, R² = 0.336 (CentMed); AE_470–870 = −8.2 × 10⁻³·t + 17.434, R² = 0.366 (EastMed); and AE_470–870 = −7.91 × 10⁻³·t + 17.002, R² = 0.481 (BalBSea). All regression equations are significant at the 99.9% CI. In all four subregions the trends are negative with negligible slopes, as shown by the very low coefficients in the regression lines.

A last investigation concerns the types of aerosols present in the atmosphere of the four Mediterranean subregions in the period of the study. To categorise aerosols into groups, several researchers have developed varying ranges for the TAOD₅₅₀ and AE_470–870 values (e.g., [93,94]). The current investigation adopts the aerosol-type discrimination criteria from [94]. These are as follows: (i) clean-maritime aerosols (CMA) for TAOD₅₅₀ < 0.15 and AE_470–870 < 1.30; (ii) urban/industrial/biomass-burning aerosols (UIA) for TAOD₅₅₀ > 0.20 and AE_470–870 > 1.00; (iii) coarse-mode, including desert-dust aerosols (DDA) for TAOD₅₅₀ > 0.25 and AE_470–870 < 0.70; and (iv) mixed-type aerosols (MTA) elsewhere. The different colouring of the data points in Figure 12 indicates the various types of aerosols present over the Mediterranean subregions on a monthly basis between 1980 and 2024. Most of them fall into the MTA and UIA categories. CMA is due to the large mass of water of the Mediterranean Sea, whereas the UIA are related to air pollution over the large urban agglomerations along the Mediterranean Sea coast, industrial activities in the surrounding countries, and occasional summertime wildfires. Although Figure 10d shows low annual DDAOD₅₅₀ values, it is interesting to note that there are no desert-dust aerosols over the BalBSea subregion (see Figure 12d). This discrepancy may be due to the categorisation criteria’s inability to detect low concentrations of desert-dust particles. For the purpose of comparison, the annual scales on both axes of the plots in Figure 12 have been purposefully left in place. However, Figure 12’s scatter plots demonstrate that MTA-type aerosols contribute significantly to the overall aerosol loading across the four Mediterranean subregions; in fact, the percentages of contributions from CMA, UIA, DDA, and MTA to the overall aerosol loading are 30.4%, 8.1%, 11.1%, and 50.4% (WestMed); 16.5%, 22.2%, 6.5%, and 54.8% (CentMed); 18.0%, 16.9%, 7.8%, and 57.4% (EastMed); and 12.4%, 52.2%, 0.00%, and 35.4% (BalBSea), respectively. The low percentages of desert-dust aerosols, in particular, are caused by the DDA months being solely spring (April, May, and occasionally March or June, not stated specifically here). Similar conclusions were reached by Kambezidis [31] for the entire Mediterranean region, Kaskaoutis et al. [42] for Solar Village in Saudi Arabia, and Kalapureddy et al. [94] for the Arabian Sea. This finding implies that the aerosols in question over the larger Mediterranean region are related to suspended dust aerosols from the Sahara. The prevalence of UIA over BalBSea (52.2%) in contrast to the other three regions (8.1%, 22.2%, and 16.9% for WestMed, CentMed, and EastMed, respectively) is another noteworthy finding. The reason for this is that while some Balkan counties (i.e., Bulgaria, Croatia, Romania, and Slovenia) have recently joined the European Union (EU), their adherence to relevant EU air pollution standards has not yet produced notable outcomes; on the other hand, other Balkan nations remain outside the EU. These facts may be ascribed to more intense air pollution and activities based on petrol and coal [95].

The DARF and AE_470–870 time series’ annual average values (represented by red curves) are displayed in Figure 13 for the whole Mediterranean region (allMed). Interestingly, the CentMed curves resemble the allMed curves more than the WestMed, EastMed, and BalBSea curves. The amount of the aerosols and the warming/cooling effect over CentMed may, therefore, play a major role and influence the adjacent WestMed and EastMed regions (see AE_470–870 in Figure 13a), as well as every other subregion (see DARF in Figure 13b).

The Pearson’s correlation coefficient, r, for the monthly mean values of DARF and AE_470–870 is displayed in

Table A2. Pearson’s correlation coefficient, r, between the monthly mean AE_470–870 and the monthly mean DARF values over the four Mediterranean domains in the period 1980–2024. All correlations are significant at the 99.9% CI. DARF has units of Wm⁻².

Parameter	Pair of Mediterranean Areas	r
AE470–870	WestMed, CentMed	0.777
	WestMed, EastMed	0.706
	WestMed, BalBSea	0.678
	CentMed, EastMed	0.935
	CentMed, BalBSea	0.895
	EastMed, BalBSea	0.932
	WestMed, allMed	0.887
	CentMed, allMed	0.968
	EastMed, allMed	0.944
	BalBSea, allMed	0.895
DARF	WestMed, CentMed	0.899
	WestMed, EastMed	0.797
	WestMed, BalBSea	0.871
	CentMed, EastMed	0.932
	CentMed, BalBSea	0.920
	EastMed, BalBSea	0.913
	WestMed, allMed	0.950
	CentMed, allMed	0.983
	EastMed, allMed	0.938
	BalBSea, allMed	0.939

(see Appendix A). As exhibited in Figure 13, the highest r values between DARF (CentMed) and DARF (allMed) and between AE_470–870 (CentMed) and AE_470–870 (allMed) are verified. In the Mediterranean region, it is especially fascinating to note how the r values of both parameters progressively increase as one moves from a left to a right domain. Regarding AE_470–870, for instance, the correlation coefficients between WestMed and CentMed are 0.777, and between CentMed and EastMed 0.935. Regarding DARF, the correlation between WestMed and EastMed is 0.797, while the correlation between WestMed and BalBSea is 0.913.

3.1.6. General Conclusions from Section 3.1.1 to Section 3.1.5

• Volcanic eruptions, depending on the amount of debris outflown in the atmosphere, can mask and modulate the optical properties of the natural and anthropogenic aerosols. In the present study, volcanoes from the (continuously updating) catalogue of the Global Volcanism Programme, Smithsonian Institute (https://volcano.si.edu), were selected in the period 1980–2024. They are characterised by VEIs in the range 4–6, except for Etna and Stromboli volcanoes in CentMed, that were “allowed” to have VEIs of 2 and 3.
• The ΔTAOD₅₅₀ values were found to be almost negative (i.e., TAOD₅₅₀ < TAOD’₅₅₀) in the inner part of the Mediterranean region (CentMed and EastMed), while ΔTAOD₅₅₀ varied between negative and positive values over WestMed and BalBSea in the whole period of the study. This indicates the smaller effect of clouds on the CentMed and EastMed areas than the other two subregions.
• In an old study, Dutton et al. [96] examined TAOD₅₀₀ at the Mauna Loa Observatory (MLO, 19.54° N) and found that the maximum values of the parameter were 0.28 and 0.23 from the El Chichón (1982) and Pinatubo (1991) eruptions, respectively. These values are comparable with the peaked values in Figure 2a–d of the present study. Indeed, the maximum annual TAOD₅₅₀ values for the El Chichón and Pinatubo eruptions are 0.289 and 0.319 (WestMed), 0.319 and 0.335 (CentMed), 0.274 and 0.303 (EastMed), and 0.354 and 0.354 (BalBSea), respectively. These higher maximum values are due to the fact that the MLO is located on the remote island of Hawaii, thus recording the background aerosol loading in the atmosphere. Similar observations to those at the MLO were made at some Antarctic stations by Herber et al. [97]. Sato et al. [80], in a study to investigate the stratospheric aerosol loading from volcanic eruptions in the period 1850–1990, found that El Chichón mostly affected regions in the northern hemisphere at latitudes 0° N–30° N just after its eruption in March/April 1982, and regions at latitudes 60° N–90° N at the end of the year; one can remark that these low-latitude areas in the northern hemisphere include most of the broader region of the Mediterranean in this study, which were affected by those eruptions (see Figure 3, Figure 4, Figure 11b and Figure 13a). A similar study by Long and Stowe [98] about the Pinatubo eruption found that the stratospheric aerosol loading was particularly high over areas with latitudes 20° S–20° N just after its eruption in June 1991. This confirms the lower TAOD₅₅₀ values recorded at MLO due to the Pinatubo eruption in comparison to those over the four Mediterranean subregions (see Figure 3, Figure 4, Figure 11b and Figure 13a). A recent work by Toohey and Sigl [99] reconstructed StAODs in the period from 500 BC to 1900 AD; they found peak StAOD₅₅₀ values ranging between 0.2 and 0.4, quite comparable to the two peaks shown in Figure 2 in the four Mediterranean subregions due to the El Chichón and Pinatubo eruptions.
• For the first time, it was demonstrated that the TAOD_550,WVE time series, i.e., atmospheric aerosols devoid of volcanic materials but including anthropocentric-origin ones, show rather constant levels (no significant trend) in any of the four Mediterranean subregions. This outcome indicates that a volcanic-debris-free atmosphere may have a lesser impact on the climate change phenomenon. Such a simulation has never been attempted to the knowledge of the author, and it would be interesting to examine the characteristics and evolution of a volcanic-debris-free climate in the past and the future.
• The desert-dust and black-carbon aerosols themselves account for half of the aerosol loading in the broader area of the Mediterranean region, especially over the Mediterranean Sea and the surrounding countries; an exception is the Balkans and Black Sea areas where the contribution from desert dust has a lesser or negligible effect.
• The free-of-volcanic-debris aerosols show lower (almost half) AOD values than those with volcanic debris over the Mediterranean (cf. TAOD₅₅₀ and TAOD_550,WVE in Figure 2 and Figure 4).
• The global volcanism seems to have had a greater influence over the Balkans and Black Sea (see Figure 4), and the least aerosol loading without volcanoes (see Figure 5) in the period 1980–2024.
• The above differentiation among the four Mediterranean subregions may have a major contribution to the patterns of TAOD₅₅₀ from the ENSO and NAO phases (see Figure 6a–d), especially over the BalBSea subregion (see Figure 6d).
• The majority of the annual TAOD₅₅₀ values are favoured by negative and neutral ENSO conditions (see Figure 8).
• The linear regression fits to TAOD₅₅₀ as a function of time (see Figure 3) give the following trends: −0.017/decade (WestMed); −0.024/decade (CentMed); −0.013/decade (EastMed); and −0.037/decade (BalBSea). All trends are negative with a greater slope over BalBSea and a smaller slope over EastMed. Kambezidis [31] found a trend over the entire Mediterranean in the period 1980–2022 equal to −0.018/decade, close to the average of the above four TAOD₅₅₀ slopes (−0.023/decade). Chiapello et al. [86] found a TAOD₅₅₀ trend of −0.06/decade over the Western Mediterranean in the period 2005–2013, which is almost three times higher than the corresponding average value in the present work. Shaheen et al. [60] found a TAOD₅₅₀ trend over EastMed of ≈−0.002/decade from MERRA-2 in the period 1980–2018, ten times lower than the average value in the present work. These varying slopes among different studies are due to the different periods examined, the source of the data used, and the areas covered; they, therefore, reflect the need to standardise the whole procedure for a more comparable way among various works.
• The linear regression fits to TAOD_550,WVE as a function of time (see Figure 5) derived the following trends: +0.003/decade (WestMed); +0.003/decade (CentMed); +0.006/decade (EastMed); and +0.002/decade (BalBSea). These trends are all positive, greater in EastMed, smaller over BalBSea, equal in the other two areas, and an order of magnitude lower than those for TAOD₅₅₀.
• The above result shows that the atmospheric aerosol loading over the entire Mediterranean in the period 1980–2024 has a negligible trend or is rather constant in the absence of global volcanic eruptions.
• There is an almost 50%/50% balance between the desert-dust/black-carbon aerosols and the stratospheric/sea-salt/cloud effect on the total aerosol loading over the WestMed, CentMed, and EastMed areas (see Figure 9). The BalBSea subregion is affected more by black-carbon aerosols than volcanic eruptions, probably because of its remote location.
• To include the cloud effect as a component of SAOD₅₅₀, the COD values can be divided by 1000 in order to derive comparable values with SAOD₅₅₀ (see Equation (9)).
• TAOD₅₅₀ was found as a function of AAOD₅₅₀ over the four areas: TAOD₅₅₀ ≈ 17·AAOD₅₅₀ (WestMed), TAOD₅₅₀ ≈ 19·AAOD₅₅₀ (CentMed, EastMed), and TAOD₅₅₀ ≈ 23·AAOD₅₅₀ (BalBSea). The higher weighting factor for BalBSea (i.e., 23) refers to the higher contribution of absorbing aerosols to the total aerosol loading over the region in comparison to that over the other three areas. This was justified in Figure 10 and Figure 12.
• Regardless of the ENSO phase, all AE_470–870 values indicate coarse-mode particles across the Mediterranean on a yearly basis. Fine-mode aerosols are linked to only two years: 1983 and 1992, related to the volcanic eruptions of Mt. El Chichón and Mt. Pinatubo, respectively.
• Most of the aerosol types fall into the MTA and UIA categories over the entire Mediterranean region. There is an absence of or at least a contribution to desert-dust aerosols over BalBSea.

3.2. Part B—Statistical Approach

All statistical tests were implemented by the freeware JASP 0.19.3 statistical software (see https://jasp-stat.org).

3.2.1. Nexus Between the Aerosol Parameters and SSN/NAOI/ONI

This section investigates the nexus among all the parameters considered in the present study with the SSN, NAO, and ENSO phenomena. This is performed by calculating Pearson’s correlation coefficient, r, and the coefficient of determination, R², to the monthly averaged values of the parameters. The statistical results are shown in

Table A3. Pearson’s correlation coefficient, r, and coefficient of determination, R², for any parameter considered in the present study with either SSN, NAOI, or ONI over the four Mediterranean domains in the period 1980–2024. Both r and R² statistical indices are applied to the monthly averaged values of the parameters. R² refers to the linear regression fits between the independent variable (any aerosol or irradiance parameter) as a function of the dependent variable (SSN, NAOI, or ONI). The annual averages of DARF, netSWR, and netLWR are expressed in Wm⁻². Note that the r or R² values with asterisk(s) indicate the significance level (* for CI = 95%, or *** for CI = 99.9%). All values are rounded to the third decimal digit.

Parameters in Pair	r				R2
	WestMed	CentMed	EastMed	BalBSea	WestMed	CentMed	EastMed	BalBSea
NAOI-ONI	+0.011				0.000
NAOI-SSN	+0.061				0.004
ONI-SSN	+0.018				0.000
TAOD550-NAOI	−0.059	−0.017	+0.205	+0.231	0.003	0.000	0.042	0.053
TAOD550-ONI	+0.084	+0.090 *	+0.201	+0.193	0.007	0.008 *	0.040	0.037
TAOD550-SSN	+0.173 ***	+0.206 ***	+0.293	+0.368 *	0.030 ***	0.043 ***	0.086	0.135 *
AAOD550-NAOI	−0.152 ***	−0.146 ***	−0.285	−0.031	0.023 ***	0.021 ***	0.082	0.001
AAOD550-ONI	−0.053	−0.072	−0.306 *	−0.122	0.003	0.005	0.093 *	0.015
AAOD550-SSN	+0.019	+0.013	−0.224	+0.106	0.000	0.000	0.050	0.011
SAOD550-NAOI	−0.052	−0.009	−0.215	+0.232	0.003	0.000	0.046	0.054
SAOD550-ONI	+0.092 *	+0.098 *	+0.211	+0.195	0.008 *	0.010 *	0.045	0.038
SAOD550-SSN	+0.181 ***	+0.215 ***	+0.299 *	+0.367 *	0.033 ***	0.046 ***	0.089 *	0.135 *
DDAOD550-NAOI	−0.144 ***	−0.108 *	−0.284	−0.191	0.021 ***	0.012 *	0.081	0.037
DDAOD550-ONI	−0.059	−0.098 *	−0.284 *	−0.333 *	0.004	0.010 *	0.110 *	0.111 *
DDAOD550-SSN	+0.052	+0.014	−0.011	−0.043	0.003	0.000	0.000	0.002
StAOD550-NAOI	+0.101 *	+0.121 *	+0.295 *	+0.231	0.10 *	0.015 *	0.087 *	0.053
StAOD550-ONI	+0.225 ***	+0.183 ***	+0.295 *	+0.213	0.051 ***	0.034 ***	0.087 *	0.046
StAOD550-SSN	+0.260 ***	+0.236 ***	+0.323 *	+0.353 *	0.068 ***	0.056 ***	0.104 *	0.124 *
BCAOD550-NAOI	−0.108 *	−0.117 *	−0.188	+0.026	0.012 *	0.014 *	0.035	0.001
BCAOD550-ONI	−0.034	−0.025	−0.144	+0.021	0.001	0.001	0.021	0.000
BCAOD550-SSN	−0.110 *	−0.057	−0.271	+0.115	0.012 *	0.003	0.073	0.013
SSAOD550-NAOI	−0.107 *	−0.061	−0.212	−0.278	0.012 *	0.004	0.045	0.077
SSAOD550-ONI	−0.074	+0.081	−0.082	−0.133	0.005	0.007	0.007	0.018
SSAOD550-SSN	−0.165 ***	+0.126 *	−0.306 *	−0.401 *	0.027 ***	0.016 *	0.093 *	0.160 *
AE470–870-NAOI	+0.197 ***	+0.160 ***	+0.324 *	+0.314 *	0.039 ***	0.026 ***	0.105 *	0.098 *
AE470–870-ONI	+0.219 ***	+0.233 ***	+0.331 *	+0.279	0.048 ***	0.054 ***	0.109 *	0.078
AE470–870-SSN	+0.131 *	+0.167 ***	+0.268	+0.292	0.017 *	0.028 ***	0.072	0.085
SSA-NAOI	+0.152 ***	+0.213 ***	+0.301 *	+0.202	0.023 ***	0.045 ***	0.090 *	0.041
SSA-ONI	+0.182 ***	+0.186 ***	+0.287	+0.210	0.033 ***	0.035 ***	0.082	0.044
SSA-SSN	+0.246 ***	+0.239 ***	+0.306 *	+0.293	0.060 ***	0.057 ***	0.094 *	0.086
DARF-NAOI	−0.161 ***	−0.150 ***	−0.294 *	−0.025	0.026 ***	0.022 ***	0.086 *	0.001
DARF-ONI	−0.023	−0.037	−0.334 *	−0.183	0.001	0.001	0.111 *	0.034
DARF-SSN	+0.016	+0.012	−0.228	+0.081	0.000	0.000	0.052	0.007
netSWRBOA,CS,A-NAOI	−0.194 ***	−0.195 ***	−0.027	−0.046	0.138 ***	0.038 ***	0.001	0.002
netSWRBOA,CS,A-ONI	−0.013	+0.013	−0.098	−0.143	0.000	0.000	0.010	0.020
netSWRBOA,CS,A-SSN	+0.019	+0.018	−0.088	−0.272	0.000	0.000	0.008	0.074
netLWRBOA,CS,A-NAOI	+0.145 ***	−0.183 ***	−0.154	−0.266	0.021 ***	0.033 ***	0.024	0.071
netLWRBOA,CS,A-ONI	+0.031	+0.069	+0.106	+0.074	0.001	0.005	0.011	0.005
netLWRBOA,CS,A-SSN	−0.064	+0.013	−0.237	−0.156	0.002	0.000	0.056	0.004
COD-NAOI	+0.076	+0.235 ***	−0.042	−0.316 *	0.006	0.055 ***	0.002	0.100 *
COD-ONI	−0.023	−0.030	+0.200	+0.005	0.001	0.001	0.040	0.000
COD-SSN	−0.080	−0.072	−0.113	−0.083	0.006	0.005	0.013	0.007

(see Appendix A) for the four Mediterranean subregions. It is interesting to note that Kambezidis [31] found significant nexus in the AE_470–870/NAOI and DARF/ONI time series pairs for the broader Mediterranean area in almost the same experimental period with that of the present work. From Table A3, the following outcomes can be drawn: (i) there are significant relations at 95%, or 99.9% CI (indicated by *, ***, respectively); (ii) high correlations with ONI exist for AAOD₅₅₀, DDAOD₅₅₀, StAOD₅₅₀, AE_470–870, and DARF over various parts of the Mediterranean; (iii) high correlations with NAOI occur for StAOD₅₅₀, AE_470–870, SSA, DARF, and COD over certain areas of the Mediterranean. In summary, there is a significant effect of (i) the ENSO phenomenon on some aerosol types (absorbing, desert-dust, and stratospheric aerosols), and some atmospheric parameters (AE and DARF), and (ii) the NAO phenomenon on stratospheric aerosols and some atmospheric entities (AE, DARF, SSA, and COD), mainly over WestMed and CentMed.

Table A8. Summary of the Pearson’s correlation coefficient, r, between parameter q and one of the three physical phenomena (SSN, NAOI, ONI) over the four Mediterranean domains in the period 1980–2024. Annual mean values have been used. The asterisks denote the significant level of the relationship (* for 95% CI, *** for 99.9% CI). The absence of an r value in a parameter pair implies insignificant level, i.e., a p-value > 0.05 (< 95% CI). DARF has units of Wm⁻².

Parameters in Pair	WestMed	CentMed	EastMed	BalBSea
TAOD550–SSN		+0.206 ***		+0.368 *
SAOD550–SSN		+0.215 ***	+0.299 *	+0.367 *
StAOD550–SSN	+0.260 ***	+0.236 ***	+0.323 *	+0.353 *
SSAOD550–SSN			−0.306 *	−0.401 *
SSA–SSN	+0.246 ***	+0.239 ***	+0.306 *
StAOD550–NAOI			+0.295 *
AE470–870–NAOI			+0.324 *	+0.314 *
SSA–NAOI	+0.246 ***	+0.239 ***	+0.306 *
DARF–NAOI			−0.294 *
COD–NAOI		+0.235 ***		−0.316 *
AAOD550–ONI			−0.306 *
DDAOD550–ONI			−0.284 *	−0.333 *
StAOD550–ONI	+0.225 ***		+0.295 *
AE470–870–ONI	+0.219 ***	+0.233 ***	+0.331 *
DARF–ONI			−0.334 *

(see Appendix A) gives a summary of the significant correlations of the parameters with SSN, NAOI, and ONI for |r| > 0.2.

Linear regression fits to TAOD₅₅₀ over the four Mediterranean subregions in the period of the study as a function of ONI (not imprinted in Figure 8), show increasing trends of TAOD₅₅₀ with the following expressions: 4.8 × 10⁻³·ONI + 0.203, R² = 0.001 (WestMed), 8.0 × 10⁻³·ONI + 0.228, R² = 0.022 (CentMed), 8.0 × 10⁻³·ONI + 0.205, R² = 0.032 (EastMed), and 13.2 × 10⁻³·ONI + 0.233, R² = 0.027, (BalBSea), all insignificant at the 95% CI. Nevertheless, these trends are considered quite low or negligible; they are therefore compatible with the small positive r values for TAOD₅₅₀/ONI and SAOD₅₅₀/ONI, and small negative r values for DDAOD₅₅₀/ONI and AAOD₅₅₀/ONI (see Table A3). In conjunction with this, Urdiales-Flores et al. [28] failed to associate a declining trend of TAOD₅₅₀ over the Mediterranean region with a significant effect of ENSO.

A more careful observation of the statistical results in Table A3 (see Appendix A) shows that all three natural phenomena (SSN, NAOI, and ONI) are linked with AE_470–870, SAOD₅₅₀, SSA, and StAOD₅₅₀ at a CI of 95% to 99.9% in all four areas or parts of the Mediterranean region, no matter of whether the correlation coefficient is positive or negative. In other words, there is a combined effect of the three mechanisms on the aerosol-particle size, their scattering capability, their scattering albedo, and the stratospheric aerosol concentration. Actually, the latter three parameters are related directly or indirectly to AE_470–870, because the scattering or absorption effects of aerosols on solar radiation depend on their size.

The relationship between the various aerosol properties could also be shown in Table A3. However, in order to save space and because the primary objective of the current study is to identify any (in)direct association between the ENSO phenomenon and the aerosol parameters, this was left out. It was preferred that a description of some statistical findings for the aerosol parameters be provided in the sections that follow.

3.2.2. Nexus Between the Aerosol Parameters and the Four Mediterranean Areas

This section investigates the nexus of TAOD₅₅₀ and AE_470–870 with the four selected areas that make up the broader Mediterranean region. The deeper meaning of this investigation is to examine statistically the possible dominance of the various aerosol types over the four subregions, or, in other words, to support the hypothesis that each subregion may exhibit distinct aerosol regimes; this was performed in Section 3.1.5 in a graphical way (see Figure 12). To implement this, the statistical process of Cluster Analysis (CA) was invoked; the analysis was performed on the monthly values of the above aerosol parameters over the four areas in the whole period of the study. The hierarchical analysis gave the following silhouette scores (SS) for four clusters: Cluster 1 SS = 0.194, Cluster 2 SS = 0.391, Cluster 3 SS = 0.352, and Cluster 4 SS = 0.562. It should be noted that SS is delimited in the range [−1, +1]; when SS → +1 means perfectly clustered, 0.5 ≤ SS ≤ 0.7 means reasonably well-clustered, 0.3 ≤ SS < 0.5 implies moderate structure, and SS < 0.3 refers to weak or overlapping clusters. Therefore, the adoption of SS = 0.562 represents well-separated and cohesive clusters. The next step was to associate clusters with subregions. For this reason, a Decision-Tree Classification (DTC) was applied between the subregions and TAOD₅₅₀ and AE_470–870. The results in

Table 3. Confusion matrix from a DTC analysis of the frequencies of the monthly mean modelled and observed TAOD₅₅₀ and AE_470–870 values.

Modelled Values	Observed Values
	WestMed	CentMed	EastMed	BalBSea
WestMed	12	6	6	0
CentMed	4	7	1	4
EastMed	3	8	9	1
BalBSea	4	6	7	21

give the frequency (in %) of the predicted (modelled) TAOD₅₅₀ and AE_470–870. Their observed diagonal values sum up to 49%, which means that the model correctly predicts the subregion about half the time, which is modest but meaningful given the complexity of aerosol dynamics.

The interpretation of the results in Table 3 are as follows.

• BalBSea. Correctly classified: 21%; often misclassified as CentMed (6%) and EastMed (7%); suggests overlap in aerosol signatures with neighbouring subregions.
• CentMed. Correctly classified: 7%; frequently confused with EastMed (1%) and WestMed (4%); indicates transitional aerosol behaviour.
• EastMed. Correctly classified: 9%; misclassified as BalBSea (1%) and WestMed (3%); suggests shared fine-particle regimes.
• WestMed. Highest correct rate: 12%; still confused with EastMed (6%) and CentMed (6%); likely reflects seasonal dust variability.

The scientific takeaways of the above results are that (i) TAOD₅₅₀ and AE_470–870 alone offer partial discrimination between subregions, and (ii) an overlap in regimes (e.g., dust vs. air pollution) reduces classification clarity. By considering the average TAOD₅₅₀ and AE_470–870 values over the study period, there is a connection between the four clusters and the subregions shown in

Table 4. Inter-period 1980–2024 mean TAOD₅₅₀ and AE_470–870 values for the four clusters and characteristic aerosol regimes over the four subregions. The symbols in parentheses denote the four dominant aerosol types set up in Section 3.1.5 (see Figure 12).

Cluster	Aerosol Regime	TAOD550	AE470–870	Dominant Subregion
1	Air pollution (UIA)	0.197	1.151	BalBSea, EastMed
2	Maritime/dust (DDA)	0.256	0.760	WestMed, CentMed
3	Secondary aerosols (MTA)	0.348	1.444	BalBSea
4	Clean-maritime (CMA)	0.137	0.844	WestMed, EastMed

.

The inter-relation of the aerosol types with the four clusters and the subregions can justify the geographical determination of the four Mediterranean areas given in Section 2 (see Table 1 and Figure 1).

3.2.3. Descriptive Statistics

Table 5. Annual averages, μ, standard deviations, σ, medians, m, and trends of the parameters considered, q, in the present work over the four Mediterranean domains as a function of time (years in the examined period 1980–2024). All AODs are considered at 550 nm; AE in the band 470–870 nm; netSWR and netLWR under the BOA, CS, and A conditions; DARF, netSWR, and netLWR statistics have units in Wm⁻². In the trend rows (fourth row in each Mediterranean area), the parameters downwards indicate the slope α (in units/decade), R², and CI (* for 95% CI, and *** for 99.9% CI), respectively; absence of * in the trend indicates significance at CI < 95%. All numbers have been rounded to the third decimal digit. Σ = statistic parameter.

Area	Σ	Aerosol/Atmospheric Parameter, q
		TAOD	AAOD	SAOD	DDAOD	StAOD	SSAOD	BCAOD	AE	SSA	DARF	netSWR	netLWR	COD
WestMed	μ	0.203	0.012	0.191	0.103	0.068	0.011	0.006	0.874	0.939	4.587	194.532	−99.375	15.144
	σ	0.034	0.001	0.034	0.013	0.037	0.001	0.001	0.185	0.010	0.457	0.907	1.547	1.354
	m	0.198	0.012	0.184	0.102	0.053	0.011	0.006	0.813	0.935	4.557	194.721	−99.505	15.108
	α R2 sign.	−0.020 0.402 ***	+4 × 10−4 0.263 ***	−0.020 0.415 ***	−4 × 10−4 0.006	−0.020 0.434 ***	+8 × 10−4 0.629 ***	+6 × 10−4 0.711 ***	−0.078 0.331 ***	−6 × 10−3 0.702 ***	+0.205 0.376 ***	+0.137 0.042 ***	+0.401 0.119 *	+0.049 0.003
CentMed	μ	0.228	0.012	0.216	0.093	0.093	0.020	0.006	0.928	0.946	3.980	209.044	−95.654	20.408
	σ	0.037	0.001	0.037	0.013	0.044	0.010	0.001	0.167	0.009	0.404	1.084	1.238	1.780
	m	0.220	0.012	0.207	0.094	0.075	0.018	0.007	0.882	0.942	3.995	209.298	−95.692	20.241
	α R2 sign.	−0.020 0.510 ***	0.000 0.208 *	−0.020 0.520 ***	−0.002 0.000	−0.029 0.588 ***	+0.003 0.196 *	0.000 0.336 ***	−0.078 0.348 ***	−0.006 0.694 ***	+0.156 0.265 ***	+0.303 0.137 *	+0.342 0.139 *	+0.020 0.000
EastMed	μ	0.205	0.011	0.195	0.080	0.089	0.018	0.006	0.935	0.948	3.258	214.748	−99.220	18.337
	σ	0.031	0.001	0.031	0.012	0.036	0.002	0.001	0.179	0.009	0.571	1.045	1.484	1.776
	m	0.198	0.011	0.188	0.081	0.073	0.018	0.006	0.869	0.943	3.385	214.751	−99.483	18.214
	α R2 sign.	−0.010 0.271 ***	+0.001 0.319 ***	−0.010 0.292 ***	+0.002 0.073	−0.020 0.399 ***	+0.001 0.719 ***	+0.000 0.592 ***	−0.078 0.366 ***	−0.005 0.601 ***	+0.274 0.426 ***	−0.098 0.017	+0.743 0.447 ***	+0.176 0.017
Bal BSea	μ	0.234	0.010	0.224	0.042	0.150	0.011	0.008	1.217	0.958	3.286	196.987	−91.253	21.024
	σ	0.056	0.001	0.055	0.007	0.058	0.001	0.001	0.149	0.009	0.418	1.608	1.818	2.161
	m	0.205	0.010	0.195	0.042	0.118	0.011	0.008	1.165	0.953	3.386	197.569	−91.087	20.602
	α R2 sign.	−0.039 0.738 ***	0.000 0.001	−0.039 0.741 ***	+0.001 0.015	−0.039 0.750 ***	+0.001 0.627 ***	0.000 0.020	−0.078 0.481 ***	−0.006 0.745 ***	+0.059 0.036	+0.841 0.494 ***	+0.694 0.262 ***	+0.147 0.008

gives the averages, standard deviations, medians, and trend values of the parameters considered in this study. The trend, α (in units/decade), was estimated from the linear regression fit to the data points of the parameter q in the form q (t) = a·t + b; q = any parameter in the headings of Table 5 (see columns 3–15); a, b = regression constants; t = any year within the period 1980–2024; α = 10·[q (2024) − q (1980)]/45 (in units/decade).

The statistical results in the above Table can give some valuable information.

• The magnitudes of the μ, σ, and m values for the parameters of TAOD₅₅₀, AAOD₅₅₀, SAOD₅₅₀, SSAOD₅₅₀, BCAOD₅₅₀, and SSA are similar throughout the four Mediterranean subregions. This suggests that the subregions have comparable aerosol loading patterns. The significance of large water bodies (Mediterranean Sea, Black Sea) must be acknowledged here, particularly for the SSAOD₅₅₀ and BCAOD₅₅₀ parameters, as these aquatic surfaces contribute to TAOD₅₅₀ with relatively high sea-salt aerosol loads. On the other hand, numerous large cities, industrial areas, and wildfires around the Mediterranean contribute to TAOD₅₅₀ with black-carbon aerosols, particularly over BalBSea (see µ in Table 5).
• In contrast, the DDAOD₅₅₀, StAOD₅₅₀, and AE_470–870 parameters exhibit a declining, rising, and increasing trend, respectively, from WestMed to BalBSea. This suggests that the overall particle sizes of the aerosols of volcanic debris and desert dust vary across the four Mediterranean subregions. While stratospheric aerosols appear to finally influence the Balkan peninsula and Black Sea area more (see Figure 4 and µ in Table 5), desert-dust outbreaks from the Sahara Desert dominate the West Mediterranean [100]. According to AE_470–870, fine-mode aerosols gradually take over from WestMed to BalBSea (see Figure 10a and µ’s in Table 5).
• As DARF decreases from WestMed to BalBSea, a comparatively greater warming effect was observed over the West Mediterranean than over the Balkans and Black Sea regions (see µ’s in Table 5). In contrast, COD is marginally rising in each of the subregions listed, which is consistent with [31], who noted it for the Mediterranean region as a whole. However, between 1979 and 2012, Kambezidis et al. [101] discovered rising trends in COD for mid- and high- level clouds and falling trends for low-level clouds. A modest increasing trend in COD was observed from WestMed to BalBSea (see COD μ’s in Table 5), taking into account all clouds in the current study. This discovery is consistent with the increasing COD in [101] and could be explained by the fact that MERRA-2 reanalysis data “see better” upper clouds.
• Table 5’s average yearly COD values are in line with those from other studies. For instance, in the Atacama Desert in Northern Chile, Luccini et al. [102] discovered COD levels of about 15 (at Arica) and about 11 (near Poconchile). It is important to keep in mind that the COD values indicate how frequently cloudiness occurs in a given area; they are lower in deserts and greater in the Mediterranean regions that include vegetation and water surfaces.
• For the majority of parameters (except from DDAOD₅₅₀ and COD in all four areas and for AAOD₅₅₀ and DARF over BalBSea), linear regression fits to the data points of the parameters as a function of time, t (years in the period 1980–2024), demonstrated significance at the 99.9%, or 95% CI. One can, therefore, forecast the parameter level shortly after 2024 with a reasonable degree of accuracy thanks to the high CIs (the accuracy assumes no additional high volcanic activity at the global level). For instance, according to the Giovanni platform (accessed on 27 January 2025), the μ_TAOD550 values for January 2025 for the four locations are 0.087, 0.118, 0.105, and 0.098, respectively. The January 2025 values, also calculated using the linear regression formulas (not displayed here for spacing reasons), were found to be 0.070, 0.100, 0.115, and 0.098, respectively, quite close to the observed (Giovanni) ones.
• Except for the DARF and COD parameters, which displayed positive trends (see α’s in Table 5), all of the aerosol and atmospheric metrics have trends that are marginally negative or even close to zero.
• The four Mediterranean subregions’ average yearly AE_470–870 values for the study period were determined. For the entire Mediterranean, Kambezidis [31] obtained an average AE_470–870 value of 0.939, which is very similar to the average of the four subregions (0.989). However, Ozdemir et al. [85] observed higher AE_470–870 values for the East Mediterranean between 1999 and 2018, in the range 1.15–1.66. From 2000 to 2015, Sharafa et al. [100] discovered that the average annual AE_470–870 values from AERONET stations in three sub-Saharan African nations ranged from 0.555 to 1.722. AE_440–870 fluctuated between 0.92 and 1.57 over the Mediterranean between 1996 and 2012, according to Mallet et al. [88]. These values are in line with the 0.874–1.217 values obtained in the current investigation.

3.2.4. Deseasonalisation; Stationarity; Cross-Correlation Analysis; Durbin–Watson Test; Granger-Causality Test

The statistics given in Table A2 and Table A3 provide information of a rather strong (i.e., CI = 99.9%) relationship between SSN/NAOI/ONI and some of the parameters examined. Nevertheless, these outcomes do not show if any of the NAOI, ONI, or SSN time series causes (influences) the existence of any of the aerosol parameters. In other words, one would like to explore the possibility that NAOI, ONI, or SSN can affect (cause) any of the aerosol parameters, at least those which were found to have a higher r and R² at the 99.9% CI level in Table A3. To implement this statistical procedure, the monthly mean values of all the parameters in Table 5 (those in the headings of columns 3–15) together with those of NAOI, ONI, and SSN were considered. This was performed because the annual values used in the preceding sections can mask any “fine-resolution” connection among them. To resolve this issue, all the monthly time series were deseasonalised using the seasonal decomposition of time series (STL) method in order to remove cyclical (seasonal) effects on the time series of the data [103,104]. It should be reminded here that all the parameters in Table 5 and SSN/NAOI/ONI do have a fixed (for SSN and aerosols) or a variable (for NAOI and ONI) cycle. By performing the STL process on all-time series of the variables, new (deseasonalised) time series were derived and used in the analysis of the current section (see

Table A4. KPSS statistics on the aerosol/atmospheric parameter, q, over the four Mediterranean domains. Unprocessed (non-dimensionalised) monthly mean values of the parameters in the period 1980–2024 were used. The values presented in the Table are KPSS_l, KPSS_t, truncation lag, and stationarity conclusion at the prescribed CI level. Level stationarity exists for KPSS_l < 0.347, 0.463, 0.574, and 0.739, and trend stationarity for KPSS_t < 0.119, 0.146, 0.176, 0.216 at the 90%, 95%, 97.5%, and 99% CI, respectively.

Parameter, q	WestMed	CentMed	EastMed	BalBSea
NAOI	0.2327, 0.1148, 6, level/trend stationarity at 90%/90% CI
ONI	0.1073, 0.0507, 6, level/trend stationarity at 90%/90% CI
SSN	1.2061, 0.1620, 6, level/trend non-stationarity at 99%/96.3% CI (needs deseasonalisation)
TAOD550	1.4767, 0.0558, 6, trend stationarity at 90% CI	2.196, 0.1359, 6, trend stationarity at 93.1% CI	1.4006, 0.1408, 6, trend stationarity at 94% CI	4.8332, 0.4187, 6 (needs deseasonalisation)
AAOD550	0.3933, 0.0480, 6, level/trend stationarity at 92%/90% CI	0.4635, 0.0556, 6, level/trend stationarity at 95%/90% CI	1.2862, 0.1582, 6, (needs deseasonalisation)	0.2097, 0.2160, 6, level stationarity at 99% CI
SAOD550	1.6464, 0.0628, 6, trend stationarity at 90% CI	2.6988, 0.1472, 6 (needs deseasonalisation)	1.5666, 0.1461, 6, trend stationarity at 95% CI	4.9290, 0.4260, 6 (needs deseasonalisation)
DDAOD550	0.0895, 0.0780, 6, level/trend stationarity at 90%/90% CI	0.1610, 0.0790, 6, level/trend stationarity at 90%/93.4% CI	0.2178, 0.0669, 6, level/trend stationarity at 90%/90% CI	0.1313, 0.1017, 6, level stationarity at 90%/90% CI
StAOD550	2.9796, 0.1888, 6 (needs deseasonalisation)	3.9768, 0.1826, 6 (needs deseasonalisation)	2.7110, 0.2011, 6 (needs deseasonalisation)	5.5435, 0.5583, 6 (needs deseasonalisation)
SSAOD550	3.8573, 0.2145, 6 (needs deseasonalisation)	1.6125, 0.4173, 6 (needs deseasonalisation)	3.6875, 0.0992, 6, trend stationarity at 90% CI	3.8994, 0.1983, 6 (needs deseasonalisation)
BCAOD550	3.8221, 0.1895, 6 (needs deseasonalisation)	2.5526, 0.3679, 6 (needs deseasonalisation)	5.4606, 0.2255, 6 (needs deseasonalisation)	0.4781, 0.4447, 6 (needs deseasonalisation)
AE470–870	2.4305, 0.3405, 6 (needs deseasonalisation)	2.5526, 0.3679, 6 (needs deseasonalisation)	2.6150, 0.2978, 6 (needs deseasonalisation)	3.6170, 0.4681, 6 (needs deseasonalisation)
SSA	5.2424, 0.3160, 6 (needs deseasonalisation)	5.2485, 0.3705, 6 (needs deseasonalisation)	4.5733, 0.3567, 6 (needs deseasonalisation)	5.8728, 0.3991, 6 (needs deseasonalisation)
DARF	0.2874, 0.0318, 6, level/trend non-stationarity at 90%/90% CI	0.2355, 0.0266, 6, trend stationarity at 90%/90% CI	0.9382, 0.0671, 6, trend stationarity at 99%/99% CI	0.1046, 0.0553, 6, level/trend stationarity at 90%/90% CI
netSWRBOA,CS,A	0.0037, 0.0037, 6, level/trend stationarity at 90%/90% CI	0.0040, 0.0038, 6, trend stationarity at 90%/90% CI	0.0014, 0.0037, 6, level/trend stationarity at 90%/90% CI	0.0071, 0.0040, 6, level/trend stationarity 90%/90% CI
netLWRBOA,CS,A	0.1023, 0.0138, 6, level/trend stationarity at 90%/90% CI	0.3624, 0.0520, 6, level/trend stationarity at 90.7%/90% CI	2.4677, 0.0491, 6, trend stationarity at 90% CI	1.5676, 0.0808, 6, trend stationarity at 90% CI
COD	0.0935, 0.0832, 6, level/trend stationarity at 90%/90% CI	0.0905, 0.0879, 6, level/trend stationarity at 90%/90% CI	0.0460, 0.0341, 6, level/trend stationarity at 90%/90% CI	0.1542, 0.1033, 6, level/trend stationarity at 90%/90% CI

and

Table A5. KPSS statistics for those aerosol/atmospheric parameters, q, that are mentioned in Table A4. The STL method was applied to only those time series before the KPSS test.

Parameter, q	WestMed	CentMed	EastMed	BalBSea
SSN	1.2061, 0.1620, 6, level/trend non-stationarity at 99%/96.3% CI (needs 1st-order differencing)
TAOD550	no further processing is necessary	no further processing is necessary	no further processing is necessary	5.3726, 0.5543, 6 (needs 1st-order differencing)
AAOD550	no further processing is necessary	no further processing is necessary	2.5479, 0.3662, 6 (needs 1st-order differencing)	no further processing is necessary
SAOD550	no further processing is necessary	3.3870, 0.1956, 6 (needs 1st-order differencing)	no further processing is necessary	5.3992, 0.5460, 6 (needs 1st-order differencing)
StAOD550	2.7938, 0.1749, 6 (needs 1st-order differencing)	3.8396, 0.1733, 6 (needs 1st-order differencing)	2.5888, 0.1616, 6 (needs 1st-order differencing)	5.5774, 0.5635, 6 (needs 1st-order differencing)
SSAOD550	4.6631, 0.2990, 6 (needs deseasonalisation)	1.5834, 0.4151, 6 (needs 1st-order differencing)	no further processing is necessary	4.7182, 0.2903, 6 (needs 1st-order differencing)
BCAOD550	5.5608, 0.4434, 6 (needs 1st-order differencing)	4.2254, 0.5087, 6 (needs 1st-order differencing)	4.9921, 0.5311, 6 (needs 1st-order differencing)	0.8177, 0.7104, 6 (needs 1st-order differencing)
AE470–870	2.5659, 0.3821, 6 (needs 1st-order differencing)	2.6225, 0.3724, 6 (needs 1st-order differencing)	2.7621, 0.3072, 6 (needs 1st-order differencing)	3.8136, 0.5049, 6 (needs 1st-order differencing)
SSA	5.3253, 0.3308, 6 (needs 1st-order differencing)	5.3962, 0.3967, 6 (needs 1st-order differencing)	4.7512, 0.3834, 6 (needs 1st-order differencing)	6.0146, 0.4352, 6 (needs 1st-order differencing)

, Appendix A).

Finding a cause-and-effect link between SSN/NAOI/ONI and the aerosol/atmospheric parameters is the ultimate goal of this study. This can be shown by applying the Granger-causality (GC) test to two time series, i.e., between any of the aerosol parameters as the effect (the dependent variable) and SSN/NAOI/ONI as the cause (the independent variable). The following requirements for the time series in question must apply to the GC test.

• Stationarity. It is necessary for the time series to be stationary, meaning that its statistical characteristics, such as μ and σ, must remain constant over time. Numerous statistical tests can be used to analyse stationarity. The present study decided to use the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test [105]. The H_o hypothesis makes the assumption that the time series are stationary during the test. The time series are said to be stationary if the corresponding p-values are less than or equal to α (0.05 for 95% CI or 0.01 for 99% CI), which means that H_o applies. The stationarity results for all variables examined for the four Mediterranean subregions between 1980 and 2024 are displayed in Table A4 (see Appendix A). Before moving on to the GC application, the deaseasonalised time series in Table A5 that were determined to be non-stationary received a first-order difference statistical modification to become stationary (see

  Table A6. KPSS statistics for those aerosol/atmospheric parameters, q, that required first-order differencing (see Table A2).

  Parameter, q	WestMed	CentMed	EastMed	BalBSea
  SSN	0.1369, 0.0463, 6, level/trend stationarity at 90%/90% CI
  TAOD550	no further processing is necessary	no further processing is necessary	no further processing is necessary	0.0101, 0.0103, 6 level/trend stationarity at 90%/90% CI
  AAOD550	no further processing is necessary	no further processing is necessary	0.0117, 0.0113, 6 level/trend stationarity at 90%/90% CI	no further processing is necessary
  SAOD550	no further processing is necessary	0.0086, 0.0083, 6 level/trend stationarity at 90%/90% CI	no further processing is necessary	0.0104, 0.0106, 6 level/trend stationarity at 90%/90% CI
  StAOD550	0.0145, 0.0139, 6, level/trend stationarity at 99%/97.4% CI	0.0183, 0.0112, 6 level/trend stationarity at 90%/90% CI	0.0160, 0.0150, 6 level/trend stationarity at 90%/90% CI	0.0157, 0.0159, 6 level/trend stationarity at 90%/90% CI
  SSAOD550	0.0088, 0.0087, 6 level/trend stationarity at 90%/90% CI	0.0369, 0.0131, 6 level/trend stationarity at 90%/90% CI	no further processing is necessary	0.0083, 0.0083, 6 level/trend stationarity at 90%/90% CI
  BCAOD550	0.0160, 0.0127, 6 level/trend stationarity at 90%/90% CI	0.0109, 0.0095, 6 level/trend stationarity at 90%/90% CI	0.0150, 0.0116, 6 level/trend stationarity at 90%/90% CI	0.0185, 0.0120, 6 level/trend stationarity at 90%/90% CI
  AE470–870	0.0220, 0.02171, 6 level/trend stationarity at 90%/90% CI	0.0161, 0.0158, 6 level/trend stationarity at 90%/90% CI	0.0169, 0.0166, 6 level/trend stationarity at 90%/90% CI	0.0158, 0.0160, 6 level/trend stationarity at 90%/90% CI
  SSA	0.0290, 0.0235, 6 level/trend stationarity at 90%/90% CI	0.0254, 0.0253, 6 level/trend stationarity at 90%/90% CI	0.0239, 0.0238, 6 level/trend stationarity at 90%/90% CI	0.0232, 0.0231, 6 level/trend stationarity at 90%/90% CI

  , Appendix A).
• Linearity. There are several statistical measures available to assess whether the independent and dependent time series are linear. The rather low R² values indicate a weak linearity between the paired parameters, although this does not exclude the use of the GC statistic. In this case, the R² values were adopted in the regression analysis between SSN/NAOI/ONI and an aerosol/atmospheric variable given in Table A3.
• Lag length. In the GC analysis, it is essential to select a suitable lag length for the dependent aerosol/atmospheric variables with respect to the independent variables of SSN/NAOI/ONI time series. Six (months) was determined to be the ideal lag period for the independent variables under investigation. This was accomplished by computing the Bayesian Information Criterion (BIC) and the Akaike Information Criterion (AIC); it was discovered that both BIC and AIC values became stabilised following lag 6. The (ideal) lag 6 in Table A4 further supports this.
• No omitted variables. The results may be skewed if important (independent) variables are left out of the analysis; therefore, all pertinent variables should be included. For this reason, all (independent) SSN/NAOI/ONI time series were used in this study with a lag of six months.
• No multi-collinearity. The independent variables should not be perfectly multi-collinear; high correlation coefficients (around +1 or −1) suggest that multi-collinearity may exist. However, the low (near-to-zero) r values among the SSN/NAOI/ONI time series in Table A3 (Appendix A) indicate that there is no such collinearity for them.
• No auto-correlation of the error terms. Any dependent or independent variable was subjected to a regression analysis against time; the auto-correlation plot of the residuals from each fit revealed no discernible trend. Consequently, no auto-correlation was discovered. By using the Durbin–Watson (DW) test on the regression residuals, the same conclusion was reached: all DW values fell between 1 and 2, suggesting that there was little to no auto-correlation in the residuals. It should be noted that DW values range from 0 to 4, and an auto-correlation is minimal or non-existent when the DW value falls between 1.5 and 2.5.

The above criteria for the implementation of the GC test were, therefore, met for the time series of the examined variables. The detailed statistical results of the Granger-causality test are shown in

Table A7. Application of the GC test: linear regression analysis between the (dependent) parameter, q, and the (independent) lagged SSN/NAOI/ONI time series with categorial factor the unlagged ONI time series. The regression expression for each q is in the form as follows: q = a + b₁·SSN_t−1 + b₂·SSN_t−2 + b₃·SSN_t−3 + b₄·SSN_t−4 + b₅·SSN_t−5 + b₆·SSN_t−6 + c₁·NAOI_t−1 + c₂·NAOI_t−2 + c₃·NAOI_t−3 + c₄·NAOI_t−4 + c₅·NAOI_t−5 + c₆·NAOI_t−6 + d₁·ONI_t−1 + d₂·ONI_t−2 + d₃·ONI_t−3 + d₄·ONI_t−4 + d₅·ONI_t−5 + d₆·ONI_t−6 + e·t (t = month number from 1 to 540, i.e., from January 1980 to December 2024, respectively). The H_o hypothesis assumes no Granger-cause–effect of the lagged SSN/NAOI/ONI time series on the parameter q; the alternative hypothesis H₁ is just the opposite. In the regression expression, the significant level for each component is shown in parenthesis. Deseasonalised monthly mean values of the time series in the period 1980–2024 were used in all four Mediterranean domains, and first-order differenced time series for the indicated dependent variables; a deseasonalised SSN time series was used throughout all GC tests.

Parameter, q	WestMed	CentMed	EastMed	BalBSea
TAOD550	overall regression RMSE = 0.0706, R2 = 0.6510, p-value = 0.0002 (≈100% CI)   +0.1815 (96.5%) +0.0002·SSNt−1 (63.9%) −0.0004·SSNt−6 (98%) −0.0072·ΝΑΟΙt (88.8%) −0.0099·ΝΑΟΙt−1 (96.1%) +0.0068·ΝΑΟΙt−3 (86.1%) +0.0062·ΝΑΟΙt−4 (81.2%) +0.0050·ΝΑΟΙt−5 (72%) +0.0082·ΝΑΟΙt−6 (94.3%) −0.0387·ONI t−1 (70.1%) +0.0452·ONI t−2 (72.4%) −9.2806 × 10−5·t (99.7%)	overall regression RMSE = 0.0430, R2 = 0.5714, p-value = 0.1947 (80.5% CI)   +0.0001·SSNt−3 (67.4%) +0.0002·SSNt−4 (96.8%) +0.0037·ΝΑΟΙt (83.4%) +0.0150·ΝΑΟΙt−3 (94%) −0.0075·ΝΑΟΙt−4 (65%) −0.0110·ΝΑΟΙt−5 (81.6%) −0.0008·ONI t−2 (61.8%) −0.0004·ONI t−4 (85%)	overall regression RMSE = 0.0573, R2 = 0.5969, p-value = 0.0440 (95.6% CI)   +0.1216 (93.4%) −0.0002·SSNt−6 (89.5%) +0.0103·ΝΑΟΙt−3 (66.7%) +0.0011·ONI t−2 (65.5%) −0.0007·ONI t−3 (62.1%) −0.0009·ONI t−4 (73.3%) +0.0007·ONI t−6 (73.3%) −8.4831 × 10−5·t (99.9%)	overall regression RMSE = 0.0405, R2 = 0.5710, p-value = 0.1819 (81.8% CI)   −0.0438 (65.4%) +0.0002·SSNt−4 (95.2%) +0.0001·SSNt−5 (76.5%) +0.0103·ΝΑΟΙt−1 (82%) −0.0085·ΝΑΟΙt−4 (74.6%) −0.0160·ΝΑΟΙt−5 (96%) +0.0144·ΝΑΟΙt−6 (94.9%) +0.0006·ONIt−1 (62.6%) −0.0019·ONIt−2 (97.9%) +0.0015·ONIt−3 (99.2%) −0.0011·ONIt−4 (97.8%) +0.0006·ONIt−4 (70.2%) −2.6653 × 10−5·t (86.9%)
AAOD550	overall regression RMSE = 0.0044, R2 = 0.6508, p-value = 0.0002 (≈100% CI)   +0.0089 (90.1%) +1.1025 × 10−5·SSNt (62.8%) +2.5971 × 10−5·SSNt−1 (95.8%) +1.5868 × 10−5·SSNt−2 (73.3%) −2.1539 × 10−5·SSNt−6 (94.1%) −0.0004·SSNt−6 (98%) −0.0006·ΝΑΟΙt (94.1%) −0.0008·ΝΑΟΙt−1 (93.6%) +0.0004·ΝΑΟΙt−3 (84.2%) +0.0004·ΝΑΟΙt−4 (83.6%) +0.0004·ΝΑΟΙt−5 (86.5%) +0.0004·ΝΑΟΙt−6 (89.9%) −0.0022·ONI t−1 (64.4%) +0.0029·ONI t−2 (72.7%) +4.9699 × 10−6·t (98.9%)	overall regression RMSE = 0.0042, R2 = 0.6930, p-value = 0.0575 (94.3% CI)   +1.3261 × 10−5·SSNt (75.2%) +2.1955 × 10−5·SSNt−1 (93.7%) +1.4242 × 10−5·SSNt−2 (71.4%) −2.2672 × 10−5·SSNt−6 (96.3%) −0.0004·ΝΑΟΙt (90.4%) +0.0010·ΝΑΟΙt−1 (80.2%) +0.0011·ΝΑΟΙt−3 (86%) −0.0007·ΝΑΟΙt−5 (61.6%) −7.6620 × 10−5·ONI t−1 (70.4%) +0.0002·ONI t−2 (97.5%)	overall regression RMSE = 0.0024, R2 = 0.5601, p-value = 0.3127 (68.7% CI)   −0.0054 (94.6%) +7.7984 × 10−6·SSNt−3 (68.7%) +1.1722 × 10−6·SSNt−4 (88.7%) −5.4475 × 10−6·SSNt−6 (60.9%) +0.0007·ΝΑΟΙt−3 (86.6%) −4.4589 × 10−5·ONI t−2 (63%) +3.7864 × 10−5·ONI t−5 (75.7%)	overall regression RMSE = 0.0031, R2 = 0.5616, p-value = 0.2955 (70.5% CI)   +0.0064 (92.9%) +7.3445 × 10−6·SSNt−1 (60.2%) +9.5150 × 10−6·SSNt−5 (74.7%) −8.4786 × 10−6·SSNt−6 (71.1%) −0.0004·SSNt−6 (98%) −0.0003·ΝΑΟΙt (89.8%) +0.0012·ΝΑΟΙt−1 (95.6%) +0.0006·ΝΑΟΙt−2 (71.1%) +0.0008·ΝΑΟΙt−3 (82.8%) −0.0008·ΝΑΟΙt−5 (81.8%) −0.0005·ΝΑΟΙt−6 (67.3%)
SAOD550	overall regression RMSE = 0.0666, R2 = 0.6551, p-value = 0.0001 (≈100% CI)   +0.1726 (96.7%) −0.0004·SSNt−6 (94.7%) −0.0066·ΝΑΟΙt (88.1%) −0.0091·ΝΑΟΙt−1 (95.6%) +0.0064·ΝΑΟΙt−3 (86%) +0.0058·ΝΑΟΙt−4 (66.7%) +0.0046·ΝΑΟΙt−5 (70.4%) +0.0078·ΝΑΟΙt−6 (64.8%) −0.0366·ONIt−1 (70.2%) +0.0424·ONIt−2 (72.1%) −9.7801 × 10−5·t (99.9%)	overall regression RMSE = 0.0406, R2 = 0.5699, p-value = 0.2090 (79.1% CI)   +0.0001·SSNt−3 (67.1%) +0.0002·SSNt−4 (94.2%) −0.0066·ΝΑΟΙt (88.1%) +0.0136·ΝΑΟΙt−3 (92.8%) −0.0071·ΝΑΟΙt−4 (65.8%) −0.0106·ΝΑΟΙt−5 (82.6%) +0.0008·ONIt−2 (63.6%) −0.0007·ONIt−4 (85.1%)	overall regression RMSE = 0.0540, R2 = 0.6094, p-value = 0.0167 (98.3% CI)   +0.1173 (94.4%) −0.0001·SSNt (61.4%) −0.0002·SSNt−6 (89.7%) +0.0095·ΝΑΟΙt−3 (65.7%) −0.0366·ON t−1 (70.2%) +0.0010·ONI t−2 (61.2%) −0.0007·ONI t−3 (62.5%) −0.0008·ONI t−4 (80.3%) +0.0007·ONI t−4 (74.3%) −8.9362 × 10−5·t (≈100%)	overall regression RMSE = 0.0389, R2 = 0.5761, p-value = 0.1547 (84.5% CI)   −0.0417 (64.9%) +0.0003·SSNt−4 (97.2%) +0.0001·SSNt−5 (83.1%) +0.0001·SSNt−6 (70.3%) +0.0100·ΝΑΟΙt−1 (82.6%) −0.0077·ΝΑΟΙt−4 (71.2%) −0.0159·ΝΑΟΙt−5 (96.6%) +0.0148·ΝΑΟΙt−6 (96.2%) +0.0007·ONIt−1 (66.8%) −0.0018·ONIt−2 (97.9%) +0.0015·ONIt−3 (99.3%) −0.0011·ONIt−4 (98.2%) +0.0006·ONIt−5 (74.8%) −2.5458 × 10−5·t (86.6%)
DDAOD550	overall regression RMSE = 0.0501, R2 = 0.6388, p-value = 0.0009 (≈100% CI)   +0.0860 (84.3%) +0.0002·SSNt−1 (91.3%) +0.0002·SSNt−2 (67.9%) −0.0002·SSNt−6 (94.5%) −0.0065·ΝΑΟΙt (95.7%) −0.0086·ΝΑΟΙt−1 (98.9%) +0.0054·ΝΑΟΙt−3 (90.3%) +0.0045·ΝΑΟΙt−4 (81.9%) +0.0039·ΝΑΟΙt−5 (75.6%) +0.0038·ΝΑΟΙt−6 (78.1%) −0.0279·ONI t−1 (70.9%) +0.0417·ONI t−2 (84.3%)	overall regression RMSE = 0.0482, R2 = 0.5741, p-value = 0.1713 (82.9% CI)   +0.0001·SSNt−1 (68.7%) −0.0001·SSNt−4 (66.6%) −0.0003·SSNt−6 (97.6%) −0.0041·ΝΑΟΙt (83.7%) +0.0084·ΝΑΟΙt−1 (64.5%) +0.0144·ΝΑΟΙt−3 (89.3%) −0.0009·ONI t−1 (71.1%) +0.0025·ONI t−2 (99%) −0.0006·ONI t−3 (63.4%) −0.0008·ONI t−4 (81.6%)	overall regression RMSE = 0.0516, R2 = 0.4895, p-value = 0.9653 (3.5% CI)	overall regression RMSE = 0.0298, R2 = 0.4923, p-value = 0.9577 (4.2% CI)
StAOD550	overall regression RMSE = 0.0149, R2 = 0.5941, p-value = 0.0534 (94.7% CI)   −6.3900 × 10−5·SSNt−1 (86.7%) +4.4276 × 10−5·SSNt−3 (64.9%) +8.9313 × 10−5·SSNt−4 (95%) +0.0023·ΝΑΟΙt (98.4%) −0.0040·ΝΑΟΙt−1 (99.9%) −0.0034·ΝΑΟΙt−5 (99.9%) +0.0008·ΝΑΟΙt−6 (62.5%) −0.0094·ONI t−1 (77%) −0.0080·ONI t−4 (62.7%) +0.0082·ONI t−5 (67.4%)	overall regression RMSE = 0.0220, R2 = 0.4865, p-value = 0.9722 (2.8% CI)	overall regression RMSE = 0.0170, R2 = 0.5279, p-value = 0.7142 (28.6% CI)	overall regression RMSE = 0.0263, R2 = 0.5944, p-value = 0.0522 (94.8% CI)   +0.0002·SSNt−4 (95.9%) +0.0001·SSNt−5 (95.9%) +0.0001·SSNt−6 (90.8%) +0.0106·ΝΑΟΙt−1 (96.6%) −0.0049·ΝΑΟΙt−4 (68.7%) −0.0077·ΝΑΟΙt−5 (87%) +0.0084·ΝΑΟΙt−6 (91.8%) +0.0005·ONI t−1 (68.5%) −0.0013·ONI t−2 (98.1%) +0.0010·ONI t−3 (99.2%) −0.0006·ONI t−4 (95.3%) +0.0003·ONI t−5 (67.3%)
SSAOD550	overall regression RMSE = 0.0029, R2 = 0.5269, p-value = 0.7252 (27.5% CI)	overall regression RMSE = 0.0053, R2 = 0.6028, p-value = 0.0286 (97.1% CI)   −0.0094 (87.7%) +2.3837 × 10−5·SSNt−3 (84%) −1.5358 × 10−5·SSNt−5 (71.2%) −1.2409 × 10−5·SSNt−6 (62.8%) +0.0006·NAOI t (92.8%) +0.0009·NAOI t−5 (62.1%) +0.0010·NAOI t−6 (68.3%) −0.0001·ONIt−2 (78.6%) +0.0003·ONIt−3 (≈100%) −0.0001·ONIt−4 (96.7%) +0.0002·ONIt−5 (98%) −6.4949 × 10−5·ONIt−6 (74.2%)	overall regression RMSE = 0.0053, R2 = 0.6084, p-value = 0.0182 (98.2% CI)   +0.0182 (99.7%) +1.3741 × 10−6·SSNt−1 (64.4%) +1.3953 × 10−5·SSNt−4 (61.8%) −1.2409 × 10−5·SSNt−6 (62.8%) +0.0002·ONIt−1 (96.5%) −0.0001·ONIt−2 (76.1%) +0.0001·ONIt−3 (94.3%) +9.5472 × 10−6·t (≈100%)	overall regression RMSE = 0.0029, R2 = 0.5386, p-value = 0.5862 (41.4% CI)
BCAOD550	overall regression RMSE = 0.0008, R2 = 0.5239, p-value = 0.7564 (24.4% CI)	overall regression RMSE = 0.0009, R2 = 0.5208, p-value = 0.7876 (21.2% CI)	overall regression RMSE = 0.0008, R2 = 0.5240, p-value = 0.7560 (24.4% CI)	overall regression RMSE = 0.0016, R2 = 0.4916, p-value = 0.9599 (4% CI)
AE470–870	overall regression RMSE = 0.1117, R2 = 0.5461, p-value = 0.4890 (51.1% CI)	overall regression RMSE = 0.1478, R2 = 0.5188, p-value = 0.8056 (19.4% CI)	overall regression RMSE = 0.1482, R2 = 0.5050, p-value = 0.9050 (9.5% CI)	overall regression RMSE = 0.1327, R2 = 0.4475, p-value = 0.9993 (0.1% CI)
SSA	overall regression RMSE = 0.0038, R2 = 0.5187, p-value = 0.8067 (19.3% CI)	overall regression RMSE = 0.0051, R2 = 0.4881, p-value = 0.9687 (3.1% CI)	overall regression RMSE = 0.0052, R2 = 0.5233, p-value = 0.7632 (23.7% CI)	overall regression RMSE = 0.0051, R2 = 0.5758, p-value = 0.1571 (84.3% CI)   −0.0087 (86.3%) +1.4672 × 10−5·SSNt (70.4%) +2.8446 × 10−5·SSNt−1 (95.1%) +2.5693 × 10−5·SSNt−3 (88.8%) +1.7178 × 10−5·SSNt−4 (73.4%) +1.7562 × 10−5·SSNt−5 (79.7%) +3.0049 × 10−5·SSNt−6 (97.6%) +0.0004·NAOIt (81.9%) +0.0015·NAOI t−2 (89%) −0.0013·NAOI t−4 (81.7%) +7.0650 × 10−5·ONIt−5 (70.3%)
DARF	overall regression RMSE = 2.4927, R2 = 0.6433, p-value = 0.0005 (≈100% CI)   +0.0070·SSNt (68.9%) +0.0136·SSNt−1 (94.2%) +0.0100·SSNt−2 (78.8%) −0.0116·SSNt−6 (92.9%) −0.3498·NAOIt (97.2%) −0.3853·ΝΑΟΙt−1 (97.8%) +0.2425·ΝΑΟΙt−3 (86.5%) +0.3482·ΝΑΟΙt−4 (96.4%) +0.3333·ΝΑΟΙt−5 (95.6%) +0.2160·ΝΑΟΙt−6 (84.3%) +1.3390·ONIt−2 (63.9%) +0.0026·t (98.3%)	overall regression RMSE = 2.2501, R2 = 0.5875, p-value = 0.0819 (91.8% CI)   −0.2267 (87.7%) +0.0075·SSNt (77.3%) +0.0101·SSNt−1 (88.6%) +0.0068·SSNt−2 (65.2%) −0.0125·SSNt−6 (96.5%) −0.2632·NAOIt (94.3%) +0.5943·ΝΑΟΙt−1 (83.7%) +0.7397·ΝΑΟΙt−3 (92.4%) +0.3510·ΝΑΟΙt−4 (60%) +0.0714·ONI t−2 (87.8%) +0.0018·t (93.6%)	overall regression RMSE = 1.8977, R2 = 0.5922, p-value = 0.0604 (94% CI)   +0.0092·SSNt (92%) +0.0103·SSNt−1 (94.5%) +0.0080·SSNt−2 (81%) +0.0056·SSNt−3 (64.3%) +0.0053·SSNt−5 (69.2%) −0.0065·SSNt−6 (81%) −0.2250·NAOIt (94.6%) +0.5957·ΝΑΟΙt−1 (90.2%) +0.4686·ΝΑΟΙt−3 (81.7%) +0.3951·ΝΑΟΙt−6 (74.5%) +0.0705·ONIt−2 (92.9%) −0.0258·ONIt−3 (66.9%) +0.0023·t (99.5%)	overall regression RMSE = 1.8606, R2 = 0.5822, p-value = 0.1116 (88.8% CI)   +0.0056·SSNt (72%) +0.0063·SSNt−1 (77%) +0.0041·SSNt−2 (69.8%) +0.0062·SSNt−5 (77.7%) −0.0069·SSNt−6 (84.7%) −0.2535·NAOIt (97.3%) +0.5968·ΝΑΟΙt−1 (90.9%) +0.3334·ΝΑΟΙt−2 (67.5%) +0.4841·ΝΑΟΙt−3 (84%) −0.4100·ΝΑΟΙt−5 (74.7%) −0.3822·ΝΑΟΙt−6 (73.9%)
netSWRBOA,CS,A	overall regression RMSE = 60.8066 W−2, R2 = 0.6209, p-value = 0.0060 (99.4% CI) +130.6428 (92.3%)   +0.2024·SSNt (76.9%) +0.2629·SSNt−1 (86.9%) +0.2234·SSNt−2 (74.6%) −0.1976·SSNt−6 (79.5%) −10.3043·NAOIt (99.2%) −4.8354·ΝΑΟΙt−1 (76.2%) +6.7794·ΝΑΟΙt−3 (91.3%) +10.2190·ΝΑΟΙt−4 (98.8%) +8.8681·ΝΑΟΙt−5 (97.2%)	overall regression RMSE = 67.8601 W−2, R2 = 0.5954, p-value = 0.0488 (95.1% CI)   +98.0516 (79.1%) +0.2576·SSNt (83%) +0.2743·SSNt−1 (84.5%) −0.2643·SSNt−6 (86.3%) −12.0709·NAOIt (99.6%) +19.5626·ΝΑΟΙt−3 (88%) +10.2190·ΝΑΟΙt−4 (98.8%) −0.9205·ONIt−3 (66.8%)	overall regression RMSE = 65.3489 W−2, R2 = 0.5971, p-value = 0.0435 (95.7% CI)   +107.7180 (84.8%) +0.2525·SSNt (83.9%) +0.2743·SSNt−1 (85.2%) −0.2606·SSNt−6 (87.2%) −11.5600·NAOIt (99.6%) +18.9829·ΝΑΟΙt−3 (88.3%) +10.2190·ΝΑΟΙt−4 (98.8%) −0.8794·ONIt−3 (66.4%)	overall regression RMSE = 78.0674 W−2, R2 = 0.5655, p-value = 0.0485 (95.2% CI)   +0.3042·SSNt (84.1%) +0.3203·SSNt−1 (85.1%) +0.2123·SSNt−2 (60.2%) −0.3213·SSNt−6 (88.4%) −14.0072·NAOIt (99.6%) +21.7551·ΝΑΟΙt−3 (86.7%) −1.0693·ONIt−3 (67.2%) +0.0299·t (61.9%)
netLWRBOA,CS,A	overall regression RMSE = 5.9091 W−2, R2 = 0.5927, p-value = 0.0586 (94.1% CI)   −96.7403 (99.9%) −0.0168·SSNt (69.3%) +0.0253·SSNt−1 (86.5%) −0.0234·SSNt−2 (78%) −0.0250·SSNt−3 (81.5%) −0.0161·SSNt−5 (68%) +0.0150·SSNt−6 (67.8%) +0.5868·NAOIt (88%) +0.4046·ΝΑΟΙt−2 (70%) −0.9142·ΝΑΟΙt−4 (98%) −1.0876·ΝΑΟΙt−5 (99.4%) +8.8681·ΝΑΟΙt−5 (97.2%) −3.1520·ONIt−3 (61.9%) +3.5934·ONIt−4 (68.5%) −3.4192·ONIt−5 (69.4%) +1.9677·ONIt−6 (70.6%)	overall regression RMSE = 4.3512 W−2, R2 = 0.5972, p-value = 0.1321 (86.8% CI)   −103.6478 (≈100%) +0.0124·SSNt (69.7%) +0.0147·SSNt−1 (76.6%) −0.0161·SSNt−4 (77.7%) −0.0104·SSNt−6 (64.1%) −0.9157·NAOIt (99.9%) −0.9835·ΝΑΟΙt−1 (76.8%) −1.3072·ΝΑΟΙt−2 (90.1%) −1.0363·ΝΑΟΙt−5 (78.3%) −0.1569·ONIt−1 (95.8%) +0.1914·ONIt−2 (96.7%) −0.1014·ONIt−5 (91.8%) +0.1151·ONIt−6 (98.5%) +0.0027·t (84.2%)	overall regression RMSE = 3.3649 W−2, R2 = 0.5811, p-value = 0.1191 (88.1% CI)   −107.7368 (≈100%) +0.0091·SSNt−1 (65.8%) +0.0093·SSNt−2 (61.1%) −0.5049·NAOIt (98.5%) −0.9429·ΝΑΟΙt−2 (87.6%) −1.3072·ΝΑΟΙt−2 (90.1%) −0.0961·ONIt−1 (89.4%) +0.1211·ONIt−2 (92%) +0.0550·ONIt−3 (75.7%) −0.0732·ONIt−4 (93%) +0.0666·ONIt−6 (93.3%) +0.0058·t (≈100%)	overall regression RMSE = 4.4422 W−2, R2 = 0.5679, p-value = 0.2283 (77.2% CI)   −95.0348 (≈100%) −0.0167·SSNt (82.5%) −1.0330·ΝΑΟΙt−2 (79.8%) +0.8875·ΝΑΟΙt−3 (71.9%) +0.7096·ΝΑΟΙt−6 (61.8%) +0.1469·ONIt−2 (89.3%) −0.0670·ONIt−4 (79.2%) +0.0528·ONIt−6 (72.9%) +0.0085·t (≈100%)
COD	overall regression RMSE = 5.8037, R2 = 0.5935, p-value = 0.0554 (94.5% CI)   +15.6029 (97.3%) −0.0170·SSNt−3 (64.2%) +0.5651·ΝΑΟΙt−1 (85.1%) −0.8495·ΝΑΟΙt−4 (97.2%) −0.8446·ΝΑΟΙt−5 (97.2%) −1.3957·ΝΑΟΙt−6 (≈100%) −3.0681·ONIt−4 (61.8%)	overall regression RMSE = 8.9935, R2 = 0.6002, p-value = 0.0346 (96.5% CI)   +25.36029 (98.5%) −0.0388·SSNt (88%) −0.0307·SSNt−1 (77%) −0.0304·SSNt−2 (71%) +0.0267·SSNt−6 (74.4%) +2.1672·ΝΑΟΙt (≈100%) −1.7244·ΝΑΟΙt−1 (68.9%) −2.3902·ΝΑΟΙt−3 (84.8%) −0.1864·ΝΑΟΙt−4 (84.4%) +1.6426·ΝΑΟΙt−5 (65.7%) +2.9777·ΝΑΟΙt−6 (92.9%) +0.1864·ONIt−3 (86.1%) −0.1703·ONIt−4 (88.6%) +0.1215·ONIt−5 (68.8%)	overall regression RMSE = 10.0918, R2 = 0.6140, p-value = 0.0113 (98.9% CI)   +20.85789 (92.7%) −0.0457·SSNt (89.8%) −0.0286·SSNt−1 (68.1%) +0.0351·SSNt−4 (74.8%) +0.0512·SSNt−6 (94.7%) +3.0228·ΝΑΟΙt (≈100%) −2.8010·ΝΑΟΙt−3 (86.6%) −3.5270·ΝΑΟΙt−4 (94%) +2.4207·ΝΑΟΙt−6 (81%) −0.1072·ONIt−4 (74.4%)	overall regression RMSE = 6.6790, R2 = 0.5492, p-value = 0.4486 (55.1% CI)

(see Appendix A).

Table 6. Summary of the SSN/NAOI/ONI GC test on the time series of the aerosol/atmospheric parameters examined in the present work over the four Mediterranean domains in the period 1980–2024. DARF, netSWR, and netLWR statistics have units in Wm⁻². “Yes” implies a cause–effect of any of the SSN/NAOI/ONI time series on the examined variable. “No” means no cause–effect. All parameters in the time series consist of their monthly mean deseasonalised or deseasonalised/first-order differenced values. A “Yes” under the heading of SSN, NAOI, or ONI refers to any lagged value of the phenomenon between one and six months.

Parameter	WestMed			CentMed			EastMed			BalBSea
	SSN	NAOI	ONI	SSN	NAOI	ONI	SSN	NAOI	ONI	SSN	NAOI	ONI
TAOD550	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
AAOD550	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
SAOD550	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
DDAOD550	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	No	No	No
StAOD550	Yes	Yes	Yes	No	No	No	No	No	No	Yes	Yes	Yes
SSAOD550	No	No	No	Yes	Yes	Yes	Yes	No	Yes	No	No	No
BCAOD550	No	No	No	No	No	No	No	No	No	No	No	No
AE470–870	No	No	No	No	No	No	No	No	No	No	No	No
SSA	No	No	No	No	No	No	No	No	No	Yes	Yes	Yes
DARF	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No
netSWRBOA,CS,A	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
netLWRBOA,CS,A	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
COD	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No

shows a summary of the cause–effect of the SSN/NAOI/ONI variables on the aerosol/atmospheric time series. The Table consists of “Yes”/”No” for a cause–effect/no cause–effect of the SSN/NAOI/ONI time series on the examined aerosol/atmospheric parameter. If a lagged component of any of the SSN/NAOI/ONI time series is present in the regression expression (see Table A7), a “Yes” is marked in Table 6; otherwise, a “No”.

The following conclusions can be drawn from the above GC statistical analysis.

• Acceptable regression results in the form q = f(SSN,NAOI,ONI) occur for TAOD₅₅₀, AAOD₅₅₀ (all domains except ONI in BalBSea), SAOD₅₅₀ (all domains), DDAOD₅₅₀, (WestMed, CentMed), StAOD₅₅₀ (WestMed, BalBSea), SSAOD₅₅₀ (CentMed, EastMed except NAOI), SSA (BalBSea), DARF (all domains except ONI in BalBSea), netSWR_BOA,CS,A, netLWR_BOA,CS,A (all domains except ONI in WestMed), and COD (all domains).
• Solar activity (through SSN) does not show a cause–effect on (i) SSAOD₅₅₀, BCAOD₅₅₀, AE_470–870, and SSA over WestMed; (ii) StAOD₅₅₀, BCAOD₅₅₀, AE_470–870, and SSA over CentMed; (iii) DDAOD₅₅₀, StAOD₅₅₀, BCAOD₅₅₀, AE_470–870, and SSA over EastMed; and (iv) DDAOD₅₅₀, SSAOD₅₅₀, BCAOD₅₅₀, AE_470–870, and COD over BalBSea.
• From the above statement, it is obvious that solar activity does not possess a statistically significant cause–effect on BCAOD₅₅₀ and AE_470–870 over the entire Mediterranean region.
• The North Atlantic Oscillation (through NAOI) shows a cause–effect on almost the same parameters and Mediterranean areas with those of SSN.
• The El Niño–Southern Oscillation (through ONI) influences all aerosol parameters mentioned above.
• From Table A7, the SSN/NAOI/ONI time series in the regression analyses include a prevailing retarded six-month influence on the Mediterranean atmosphere. This is almost ½ of the 11-year solar activity cycle and exactly half a full year. In the latter case, one may consider a warm and a cold season of six months duration each; perhaps this occurs because nature favours these two main seasons.
• In the analyses for the cause–effect of the SSN/NAOI/ONI time series on the aerosol/atmospheric parameters in Table A7, only the expressions with overall α ≥ 0.4 (a 60% CI) are shown. Nevertheless, the regression expressions with α < 0.4 still contain certain lagged values of the independent variables.
• The last observation comes in agreement with the initial outcome of a rather synchronous fluctuation in DARF and SSN (see periods 1988–1992 and 1998–2004 in Figure 9 of [31]), and of DARF and ONI (see marked rectangles in Figure A3 of [31]). The characterisation “synchronous” was given by Kambezidis [31], who used annual mean values to show an almost synchronous temporal evolution of the DARF/SSN and DARF/ONI time series. To demonstrate this synchronous fluctuation in ONI and DARF over the four Mediterranean subregions in the case of progressively decreasing positive ENSO phases or turning from positive to negative events, Figure 14 was prepared. The ellipses show the above-mentioned events; the red arrows inside the ellipses indicate the effect of a decreasing ENSO phase on DARF, while DARF follows a similar decreasing tendency with that for ENSO with a retardation of about six months, according to the Granger-causality effect shown in this section. It is, therefore, clearly seen that there exists a signalling effect of the ENSO phase on the DARF levels over all four Mediterranean areas.

3.2.5. Normal Distribution of the Examined Variables

The distribution patterns of the parameters under investigation are the focus of this section. P-P (probability–probability) plots were created as a result. With the same μ and σ as the variable’s time series, a P-P plot displays the probability of a variable against its theoretical (normal) one. A normal probability distribution’s fit to the observed data can be evaluated using a P-P plot. To put it another way, a P-P plot indicates which ranges of the parameter values follow a normal distribution curve; data points that fall away from the x = y line indicate that the observed data deviates from the normal distribution, while data points that fall along this line indicate that all or some of the observed data matches the theoretical distribution. In a P-P plot, the x-axis represents the cumulative probabilities of the reference (normal) distribution, and the y-axis the cumulative distribution of the actual data.

In the present study, P-P plots for all parameters in the four Mediterranean subregions were derived using the monthly mean values of their time series in the period 1980–2024. For space saving, selected P-P plots are only shown here. Figure 15 presents the P-P plots for the three physical phenomena (i.e., SSN, NAOI, and ONI) as well as for AAOD₅₅₀ (BalBSea), SAOD₅₅₀ (CentMed), and SSA (WestMed).

The P-P patterns of AAOD₅₅₀ (BalBSea), SAOD₅₅₀ (CentMed), and SSA (WestMed) appear to be significantly influenced by those of SSN, NAOI, and ONI, respectively. The SSA, SAOD₅₅₀, and AAOD₅₅₀ graphs are examples of the three patterns discovered among all analysed parameters in this work and over the four Mediterranean subregions. It is astounding how similar the PDFs are for the cause (impact) of SSN, NAOI, and ONI on SSA, SAOD₅₅₀, or AAOD₅₅₀, respectively. As a consequence, this result has the value of a rule and provides more evidence of how the three physical phenomena affect the atmospheric aerosols in the larger Mediterranean region. This rule is consistent with the study’s findings that the patterns of atmospheric aerosols over the Mediterranean region show a signal (effect), no matter how strong or small of any one of the three natural phenomena or a combination of them.

4. Conclusions and Discussion

The current study was an extension of a former study by the same author [31] that examined the impact of three physical phenomena on atmospheric processes over the Mediterranean: solar activity, the North Atlantic Oscillation, and the El Niňo–Southern Oscillation. The current study divided the Mediterranean region into four smaller areas (West Mediterranean, 6° W–10° E, 30° N–45° N; Central Mediterranean 10° E–20° E, 30° N–45° N; East Mediterranean, 20° E–35° E, 30° N–41° N; Balkans and Black Sea, 20° E–35° E, 41° N–45° N), whereas the earlier work considered the entire region into account. The current study expanded the data to 2024 (45 years), whereas the previous study used data from 1980 to 2022 (43 years). These distinctions were thought to be crucial for a more thorough investigation of the effects of the three physical events on atmospheric aerosols (smaller areas and longer time period).

The present work was articulated in a different way; it was divided into two parts: a physical approach and a statistical approach. In the first part, the effects of SSN, NAOI, and ONI were examined on the physical and optical properties of aerosols, solar radiation, and clouds. The second part was devoted to analysing the above effects in a statistical manner. Both approaches were used to give a more in-depth answer on the significant effects of the physical phenomena on the aerosol processes over the four Mediterranean areas. The main conclusions from the study are deployed below.

• Part A. Physical approach

Volcanic eruptions have been shown to have the ability to alter and/or conceal the optical characteristics of both man-made and natural aerosols. Only worldwide volcanic eruptions with VEIs between 4 and 6 were taken into consideration in the analysis; however, the Etna and Stromboli volcanoes, which are situated in the CentMed subregion and have VEIs of 2 and 3, were also taken into consideration.

Compared to EastMed and BalBSea, a more significant warming effect (higher positive DARF values) was observed over WestMed and CentMed.

Over the wider Mediterranean region, an aerosol scattering profile with SSA values higher than 0.90 was generally observed. This implies that the primary cause of solar radiation attenuation is the scattering aerosol particles.

For the first time, a modelling approach to the estimation of TAOD₅₅₀ was made as TAOD’₅₅₀ = DDAOD₅₅₀ + StAOD₅₅₀ + SSAOD₅₅₀ + COD_sca (the scattering atmospheric components) + BCAOD₅₅₀ + COD_abs (the absorbing atmospheric components), where COD_sca = 0.85·(COD/1000) and COD_abs = 0.15·(COD/1000). The difference between TAOD₅₅₀ and TAOD’₅₅₀ annual values varied in the extreme range ±6% in all four Mediterranean areas.

Maximum TAOD₅₀₀ values at the Mauna Loa Observatory, Hawaii, were found to be 0.28 and 0.23 from the El Chichón (1982) and Pinatubo (1991) eruptions, respectively. These values are comparable to the maximum values found in the present study, i.e., 0.289 and 0.319 (WestMed), 0.319 and 0.335 (CentMed), 0.274 and 0.303 (EastMed), and 0.354 and 0.354 (BalBSea), from the El Chichón and Pinatubo eruptions, respectively.

For the first time, it was demonstrated that the TAOD_550,WVE time series, i.e., atmospheric aerosols without volcanic materials but including anthropogenic activities, showed rather constant levels (no significant trends) over any of the four Mediterranean subregions.

The desert-dust and black-carbon aerosols account for half of the aerosol loading in the broader area of the Mediterranean region, especially over the Mediterranean Sea and the surrounding countries, except the Balkans and the Black Sea.

It was discovered that the DDAOD₅₅₀ and SAOD₅₅₀ time series consistently occurred during cold La Niňa occurrences. Additionally, the levels of DDAOD₅₅₀ approached those of AAOD₅₅₀ as one moved from WestMed to BalBSea, suggesting that the absorbing “face” of DDAOD₅₅₀ may be taking a central role. Accordingly, from WestMed to BalBSea, the absorption aerosols’ contribution to the overall solar radiation attenuation decreases significantly as the scattering effect gradually replaces it. Furthermore, in the whole Mediterranean, scatterers such as nitrates, sea-salt, and sulphate particles are essentially the only elements reducing solar radiation.

In agreement with observations made by Sannino et al. [73] in Naples, Italy, who discovered that the AOD₅₀₀ levels due to Saharan dust transport over the area on 25 February 2021 increased because of a coincident arrival of volcanic debris from Etna on the same day, the volcanic-debris-free aerosols showed lower (almost half) AOD values than those when volcanic debris was present over the Mediterranean.

Compared to the other three Mediterranean subregions, the Balkans and Black Sea appear to have been more affected by global volcanism between 1980 and 2024, and they also saw the least amount of aerosol loading in the absence of volcanoes. Accordingly, over the Mediterranean, the free-of-volcanic-debris aerosols displayed lower (almost half) AOD values than those with volcanic debris.

The dependence of TAOD₅₅₀ on the ENSO and NAO phases (see Figure 6), particularly over the BalBSea subregion (see Figure 6d), may be largely explained by the foregoing divergence among the four Mediterranean subregions.

Several patterns demonstrated how each of the factors being studied are influenced by ENSO and NAO.

Table 7. Phase of the NAO- and ENSO-circulation phenomena favouring maximum values of the atmospheric parameter investigated. Red colour = positive phase; green colour = neutral phase; blue colour = negative phase. DARF, netSWR_BOA,CS,A, and netLWR_BOA,CS,A have units in Wm⁻².

Parameter	WestMed	CentMed	EastMed	BalBSea
TAOD550	NAO/ENSO	NAO/ENSO	NAO/ENSO	NAO/ENSO
AAOD550	NAO/ENSO	NAO/ENSO	NAO/ENSO	NAO/ENSO NAO/ENSO
SAOD550	NAO/ENSO	NAO/ENSO	NAO/ENSO	NAO/ENSO
DDAOD550	NAO/ENSO	NAO/ENSO	NAO/ENSO NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO
BCAOD550	NAO/ENSO	NAO/ENSO	NAO/ENSO	NAO/ENSO NAO/ENSO
AE470–870	NAO/ENSO	NAO/ENSO	NAO/ENSO	NAO/ENSO NAO/ENSO
SSA	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO
DARF	NAO/ENSO	NAO/ENSO	NAO/ENSO	various peaks
netSWRBOA,CS,A	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO NAO/ENSO
netLWRBOA,CS,A	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO
COD	NAO/ENSO NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO NAO/ENSO NAO/ENSO	NAO/ENSO NAO/ENSO	NAO/ENSO

summarises the NAO and ENSO phases that have supported the highest parameter values in the four Mediterranean subregions between 1980 and 2024. The highest value of an atmospheric parameter occurred across adjacent subregions for the same NAO/ENSO phase pair if it occurred over a subregion when NAO and ENSO were in the same phase (same colour in Table 7). The same is true when the ENSO and NAO phenomena are at different stages (seen in Table 7 as distinct colours). For instance, over WestMed, neutral and/or positive ENSO phases may be present when the highest COD values occurred under negative NAO. This can be linked to the occurrence of COD during the transition from positive to neutral ENSO phase, while NAO is in a negative phase. Figure 14 shows such instances for DARF. Both observational and model data suggest that there are more instances of negative NAOI occurrences during warm ENSO and positive NAOI occurrences during cold ENSO events, according to Li and Lau’s [106] study on the effects of ENSO on atmospheric parameters (precipitation and surface air temperature) over the North Atlantic during late winter. Their findings occasionally coincide with those of the current study, albeit being for various parameters and seasons (cf. AAOD₅₅₀, BCAOD₅₅₀, and AE_470–870 in BalBSea; SSA in all four subregions; COD in WestMed and CentMed).

Negative and neutral ENSO periods were found to benefit most of the annual TAOD₅₅₀ values.

A TAOD₅₅₀ trend of −0.017/decade (WestMed), −0.024/decade (CentMed), −0.013/decade (EastMed), and −0.037/decade (BalBSea) was discovered.

Nearly an order of magnitude lower than that of TAOD₅₅₀, the trends in TAOD_550,WVE were found to be +0.003/decade (WestMed), +0.003/decade (CentMed), +0.006/decade (EastMed), and +0.002/decade (BalBSea).

In the absence of worldwide volcanic eruptions, the atmospheric aerosol loading over the whole Mediterranean region from 1980 to 2024 was shown to exhibit a very slight but negligible trend. Additionally, it was discovered that in most circumstances, the aerosol loading over the larger Mediterranean region “prefers” neutral ENSO phases.

The stratospheric/sea-salt/cloud effect and the desert-dust/black-carbon aerosols were shown to have a nearly 50%/50% balance on the overall aerosol loading in WestMed, CentMed, and EastMed.

In the four subregions (TAOD₅₅₀ ≈ 17·AAOD₅₅₀, WestMed; TAOD₅₅₀ ≈ 19·AAOD₅₅₀, CentMed, EastMed; and TAOD₅₅₀ ≈ 23·AAOD₅₅₀, BalBSea), TAOD550 was expressed as a function of AAOD₅₅₀ after [38].

All of the AE_470–870 results showed coarse-mode particles throughout the Mediterranean on an annual basis, regardless of the ENSO phase. Only the years 1983 and 1992, which corresponded to the volcanic eruptions of Mt. El Chichón and Mt. Pinatubo, respectively, were associated with fine-mode aerosols. Furthermore, in all four Mediterranean subregions, fine-mode aerosols were preferred by positive ONI and positive/negative NAOI phases.

Over the Mediterranean region, the majority of aerosol types are classified as MTA or UIA. The BalBSea region appears to be less impacted by desert-dust particles.

• Part B. Statistical approach

The Pearson’s correlation coefficient, r, between any of the studied parameters and any of the three physical phenomena was examined for |r| ≥ 0.2. It is interesting to observe that the correlations occurred at higher CI over WestMed and CentMed for the parameter pairs indicated, thus dictating a greater accuracy in these relationships. Nevertheless, it is seen that SSN, NAOI, and ONI influence EastMed more than the other three subregions, regardless of the CI level (see column 4, Table A8). Moreover, SSN also influences CentMed and BalBSea. These observations may result in a general conclusion that all three physical phenomena “prefer” the broader Eastern Mediterranean region. On the other hand, these r values cannot be considered high; therefore, the indicated relationships are rather weak.

The μ, σ, and m values of TAOD₅₅₀, AAOD₅₅₀, SAOD₅₅₀, SSAOD₅₅₀, BCAOD₅₅₀, and SSA across the four Mediterranean subregions have comparable magnitudes. This indicates the existence of similar aerosol loading patterns over the four areas. On the contrary, DDAOD₅₅₀, StAOD₅₅₀, and AE_470–870 show respective decreasing, increasing, and increasing trends from WestMed to BalBSea.

Trends, α (units/decade) > |0.2| (see Table 5) were found to be exhibited by DARF and netLWR_BOA,CS,A over all four Mediterranean subregions, netSWR_BOA,CS,A over WestMed, CentMed, BalBSea, and COD over EastMed and BalBSea. The other parameters showed negligible trends in the period 1980–2024.

The Balkan Peninsula and Black Sea region appear to be more affected by stratospheric aerosols, whereas West Mediterranean is mostly impacted by desert-dust outbreaks from the Sahara Desert. As indicated by increasing AE_470–870 readings, fine-mode aerosols gradually take over from WestMed to BalBSea.

It was shown that any one of the three physical phenomena, or a combination of them, had a signalling influence on the patterns of the atmospheric aerosols over the Mediterranean region, regardless of how powerful or weak.

In the Mediterranean atmosphere, a Granger-causality test of the SSN/NAOI/ONI time series on those of the aerosol parameters revealed a predominantly, albeit delayed, six-month influence of the former on the latter. This delay is precisely half a year and nearly ½ of the 11-year solar activity cycle. This result is shown in Figure 14.

P-P plots between any of the three physical phenomena and one of the parameters under study demonstrated that the latter had an immediate impact on the former. It is true that the distribution of the corresponding values of the parameters is dominated by the monthly values of SSN, NAOI, and ONI. Regardless of the strength of the relationship, this fact determines the dominance of the physical phenomena on the parameters.

The absolute impact of the volcanic material released into the atmosphere on the optical characteristics of the aerosols across the larger Mediterranean region was one of the work’s key findings. This finding was validated by Figure 5. The current inquiry concerns the extent to which volcanoes impact the world’s climate. Volcanic eruptions were not taken into account in the most recent Intergovernmental Panel on Climate Change (IPCC, Sixth Assessment Report—IPCC (https://www.ipcc.ch/assessment-report/ar6/ (accessed on 10 February 2025))) 6th Assessment Report (AR6), which was partially released between August 2021 and April 2022. Instead, AR6 makes the assumption that volcanic activity between 2015 and 2100 will be comparable to that which was recorded between 1850 and 2014. This assumption is vague in the sense that, as shown by comparing Figure 3 and Figure 5, the volcanic activity is crucial in modulating TAOD₅₅₀ levels and trends; different TAOD₅₅₀ levels in the atmosphere may result in differing scattering and absorption mechanisms of solar radiation and, consequently, in reforming DARF. This can be demonstrated with basic computations. Examine the four Mediterranean subregions’ inter-period average values of TAOD₅₅₀ and DARF in Table 5 and TAOD_550,WVE in

Table 8. Estimation of DARF without volcanic activity in the atmosphere, DARF_WVE, from simple calculations taking into account TAOD₅₅₀, TAOD_550,WVE, and DARF as average values over the four Mediterranean subregions in the period 1980–2024.

Parameter	WestMed	CentMed	EastMed	BalBSea
TAOD550	0.203	0.228	0.205	0.234
TAOD550,WVE	0.135	0.140	0.117	0.084
DARF (Wm−2)	4.587	3.980	3.258	3.286
DARFWVE (Wm−2)	6.897	6.482	5.708	9.154

. The formula DARF_WVE = DARF·(TAOD_550,WVE/TAOD₅₅₀) can be used to predict DARF_WVE (DARF without volcanic eruptions). Table 8 shows that DARF_WVE levels can rise dramatically, suggesting that aerosols play a more significant influence in the global climate. The following is the explanation: a decrease in (stratospheric) aerosols in the atmosphere implies a rise in the downward flux of solar radiation on Earth’s surface, which raises netSWR_BOA. Thus, DARF_WVE rises in accordance with Equations (1)–(4), so netLWR_BOA may be regarded as nearly constant in the equations.

The dispersion of the different types of aerosols throughout the larger Mediterranean region was the second result of this study. Due to the region’s closeness to the African deserts, desert-dust aerosols are either non-existent or extremely uncommon over the BalBSea region; however, they are more common over WestMed. Because WestMed is bordered by the Atlantic Ocean and the Mediterranean, westerlies have the ability to carry sea-spray aerosols from the Atlantic to the region. This is also true for aerosols of marine origin. However, CentMed has the disadvantage of having a higher concentration of aerosols of urban and industrial origin due to its urbanisation and industrialisation. However, a number of factors (coal-fired plants and low-quality solid fuels, obsolete industrial facilities, an ageing fleet of vehicles, inefficient heating systems, and complex topography; see www.blue-europe.eu and https://westernbalkans-infohub.eu) contribute to the higher levels of UIA over the Balkans and Black Sea.

The influence of solar activity, the North Atlantic Oscillation, and the El Niňo–Southern Oscillation on atmospheric aerosols across the larger Mediterranean region, particularly with their six-month delayed effect, can be summarised as the study’s final significant finding. These postponed impacts of the physical phenomena, either separately or collectively, in the climate models could provide fresh perspectives on the projections of climate change. The interconnectedness of these three phenomena has been examined by numerous researchers. According to Vencloviene et al. [55], the La Niňa phase has a statistically significant effect on NAOI. Leamon [37] discovered a very statistically significant association between the occurrence of the last five solar cycle termination and the Pacific Ocean’s shift from El Niño to La Niña. The impact of solar modulation on the relationship between NAO and ENSO has been investigated by Huo et al. [38].

As regards the research questions posed in the end of Section 1, the analysis of the present study can answer them quite confidently. They are given below.

• Are the effects of solar activity, NAO, and ENSO on the atmospheric processes found in [31] valid in smaller areas of the Mediterranean too? The answer to this question is positive. Table A2 and Table A3 (see Appendix A) provide sufficient correlations (positive or negative) at high CIs over almost all four Mediterranean domains considered.
• How do volcanic eruptions affect the status of atmospheric aerosols over the Mediterranean? Extensive analysis provided in Section 3.1.2 showed that the absence of volcanic activity on Earth would result in rather constant levels of the atmospheric processes over all four Mediterranean domains.
• A third significant conclusion related to the first research question was that El Niňo–Southern Oscillation, North Atlantic Oscillation, and solar activity all have an impact on the atmospheric processes over the larger Mediterranean region, either separately or in combination, and this effect can last up to six months. The three events have a weak but significant cause-and-effect relationship that drives several aerosol characteristics, even at a 99.9% significance level.

From the above conclusions, some areas that may require more research and are deserving of future study are, therefore, brought to light.

• Global volcanism.
  • Look closely at how global volcanism affects the characteristics of aerosols in regions other than the Mediterranean.
  • In the absence of volcanic eruptions, study again the Earth’s climate. Forecast the future and model the past and present world climates devoid of volcanic debris. Since the current study at least showed that large volcanoes have an impact on global aerosol levels and their concentrations as well as on the climate, and that they should not be ignored, there should be alternatives to the “as was in the past” scenario for the years 2015–2100 in light of the IPCC’s upcoming AR7. Thus, the following volcanic scenarios should be considered: (i) unchanged; (ii) absence of dramatic volcanoes; (iii) presence of drastic volcanoes by including extreme volcanism of VEIs of 5 or more (cataclysmic category). Furthermore, these scenarios could potentially be broken down and examined into smaller timeframes within the future timeframe of 2015–2100.
• TAOD₅₅₀ levels.
  • In BalBSea, they are higher for ONI < 0 than for ONI > 0 compared to the other three Mediterranean domains. Then, examine whether this phenomenon holds true for other bordering regions worldwide.
  • Examine why they prefer to occur during neutral ENSO phases.
• ENSO, NAO effects.
  • Examine how ENSO regulates the circulation of atmospheric aerosols outside of the Mediterranean region.
  • Examine how solar activity, NAO, and ENSO affect atmospheric processes in various regions of the planet.
  • Examine why black-carbon aerosols have a greater impact over BalBSea than volcanic debris. The sporadic occurrence of the latter type of aerosols across the area could be one factor.