The survey was carried out on board the German R/V Maria S. Merian from March 2nd to April 3rd 2018 (Figure 1). The cruise started at Heraklion, Crete, Greece and ended in Cadiz, Spain (Hainbucher et al., 2020). In total, 136 CTD casts were conducted during the cruise materializing a complete east-to-west Mediterranean transect plus another one in the Ionian Sea. The CTD system used was a Seabird SBE911plus and a Seabird SBE32 carousel equipped with 24/10lt Niskin bottles. Dual sensors for temperature and conductivity were employed. Salinity values were checked against water salinity samples with the help of a Guildline Autosal salinometer in the laboratory. Overall accuracies were within the expected instrument range of salinity. More details for the instruments setup and the sampling strategy can be found in Hainbucher et al. (2020). At 62 stations (with depth <3,000 m) the LISST-Deep was deployed along with the rosette system to obtain optical parameters i.e., beam attenuation coefficient, particle volume concentration, particle median diameter, and the slope of PSD. At 15 stations, seawater samples were also collected for the elemental and stable isotopic analysis of carbon and nitrogen of PM.

Laser diffraction is a method that permits nondestructive in situ measurements of the particle size distribution over a relatively large size range, implemented in Laser in Situ Scattering and Transmissometry (LISST–Deep), an autonomous instrument manufactured by Sequoia Scientific Inc., to measure particle size distribution and the beam attenuation coefficient in the deep sea (Agrawal and Pottsmith, 2000). LISST-Deep derived variables are the beam attenuation coefficient due to particles (c_p at 670 nm in m⁻¹), the total particle volume concentration (VC in μL L⁻¹ or ppm), and the particle median diameter (D₅₀ in μm). A detailed description of the methodology applied is given in the Supplementary Material.

The PSD slope, ξ, provides information on the relative concentration of small to large particles, i.e. steep slopes correspond to higher abundance of small particles and flatter slopes to the opposite; its mathematical description using the power-law model is used most often in marine studies (Buonassissi and Dierssen, 2010; Xi et al., 2014). For the estimation of ξ, we followed Xi et al. (2014). Methodological details are given in the Supplementary Material.

For the determination of POC, PN contents and stable isotopic composition (δ¹³C-POC and δ¹⁵N-PN), seawater samples (5–10 L) were collected close to the sea surface, within the maximum of chlorophyll fluorescence and main water masses from 15 stations across the study area (98 samples in total; Figure 1). The POC and PN contents and their stable isotopic composition were determined at the Stable Isotope Facility at the University of California Davis (Davis, CA, United States), using a PDZ Europa ANCA-GSL elemental analyzer coupled to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Chesire, United Kingdom). Methodological details are given in the Supplementary Material.

To evaluate the sea surface productivity in the study area for the duration of the cruise, monthly mean and 8 day average sea surface chlorophyll-a concentrations (Chl-a) were obtained using the Ocean Color (OC) climate record developed by the European Space Agency (ESA) in the framework of its Climate Change Initiative (CCI) activity. Detailed description is given in the Supplementary Material.

The spatial distribution of the various optical and geochemical parameters considered herein was visualized by using Ocean Data View - ODV (Schlitzer, 2018). The dataset variables are normally distributed (passed Kolmogorov–Smirnov tests) thus, the determined optical and biochemical parameters, including water depth, were tested on their potential statistical relationships by using 2-tailed Pearson correlation analysis (SPSS statistics software v. 17.0); the data set was divided into western and eastern Mediterranean data points and Otranto Strait data points, since those regions exhibited considerable differences in the spatial distribution of the studied parameters (see Discussion below).

The Theta-S diagram of the stations sampled during the MSM72 cruise is presented in Figure 2. The profiles have been assigned into four different groups each with different hydrological characteristics. Two profile groups represent stations in the WMed, one sampled from the Gibraltar Strait up to the Sardinia channel and a second one in the Sardinia channel until the Tyrrhenian Sea. Two profile groups represent stations sampled in the EMed, one sampled in the western part of the EMed (mostly in the Ionian Sea), and a second one sampled in the eastern Ionian Sea, Otranto strait, Cretan Passage and Cretan Sea. The reader is referred to the caption of Figure 2 for information about the group color-coding. For further information related to this section, the reader is referred to Astraldi et al. (1999), Wüst (1961), Millot (1999), Malanotte-Rizzoli (2001), Millot and Taupier-Letage (2005), Tanhua et al. (2013), and references therein.

The Mediterranean Sea is a concentration basin where the amount of water lost to the atmosphere through evaporation is larger than the water added into the basin through precipitation and riverine input. In order to balance this water budget deficit, low salinity Atlantic Water (AW) enters the basin through the Gibraltar Strait in the WMed and follows an eastward path towards the Sicily strait, before advancing into the EMed. The low salinity signal of the entering AW surface mass is much stronger close to the entrance point, and whilst it is still evident along its eastward route in the Tyrrhenian and Ionian Seas, salinity is increased. Finally, in the eastern part of the EMed, surface AW masses are transformed into a high salinity, warm surface water mass known as Levantine Surface Water (LSW), which is the most saline surface mass of the Mediterranean Sea. Levantine Intermediate Water (LIW) formed mainly in the Levantine Sea during winter is a saline water mass with a density not high enough to penetrate the Mediterranean deep layers. Thus, it is usually traced as a local salinity maximum at intermediate depths. In general, intermediate waters in the Mediterranean are defined as those occupying the layer between ∼150 and 600 m. After formation, LIW follows a path opposite to that of AW, and finally exits the Mediterranean through the Gibraltar Strait below the AW layer. This dual flow regime constitutes the upper thermohaline conveyor belt that runs throughout the whole Mediterranean Sea.

Close to the formation sites, LIW masses cannot be clearly observed, as the shallower layers in the absence of AW are more saline than LIW (Figure 2). However, along its westward route, local salinity LIW core maxima can easily be discerned. From the Ionian Sea, through the Tyrrhenian basin and further on towards Gibraltar, LIW becomes colder and less saline, thus increasing its density (σ_θ) from ∼29 kg m⁻³ in the Ionian Sea to ∼29.05 kg m⁻³ in the Tyrrhenian and ≤29.10 kg m⁻³ in the western part of the WMed. During this route, LIW core depth increases from 200 to 250 dbar in the Ionian Sea, to 250–300 dbar in the Tyrrhenian, and to 400–500 dbar in the WMed.

The Mediterranean deep water formation areas are different for the west and east sub-basins. The Gulf of Lions in the west and the Adriatic Sea in the east, are the most active formation sources. The deep waters formed are practically isolated by the shallow (∼400 m) strait of Sicily. Hence, East Mediterranean Deep Water (EMDW) is easily distinguishable from West Mediterranean Deep Water (WMDW), as the former is warmer, more saline and denser than the latter. The density (σ_θ) of these two different water masses is ≥29.20 and <29.15 kg m⁻³ respectively. In the lower part of the Theta-S diagram (Figure 2) blue and green WMed profiles merge as WMDW shares almost the same properties throughout the basin. The same applies to the lower part of the yellow and red EMed profiles, with two exceptions. Firstly, the isolated high temperature, salinity and density, hook-shaped profile observed in the red group, is a profile in the Cretan Sea, where deep water density is over σ_θ 29.30 kg m⁻³. This Cretan Deep Water (CDW) mass has stagnated inside the Cretan Sea for the last two decades (Velaoras et al., 2019). Secondly, the profile sampled in the Otranto strait appears to be slightly denser (σ_θ ≥29.24 kg m⁻³) at a depth of ∼760–900 dbar. This marks the winter-formed Adriatic Deep Water (ADW) which plunges through the shallow Otranto strait towards the Ionian abyssal plain. Between the intermediate and bottom layers (approximately from 700 to 1,600 m) of the EMed, a low salinity and oxygen water mass is found called the Transitional Mediterranean Water (TMW). This is the oldest water in the eastern Mediterranean (Stöven and Tanhua, 2014) and is the product of the uplift of deep masses by intrusions of bottom water masses of Adriatic and/or Aegean origin (Velaoras et al., 2019). In the literature (Stöven and Tanhua, 2014; Li and Tanhua, 2020) TMW can be also found as Tracer Minimum Zone (TMZ) due to its low concentration in transient tracers (i.e., CFC-12, SF₆).

Finally, it should be noted that during the last decades both Mediterranean sub-basins have been subjected to severe overturning circulation changes. In the EMed during the early 1990s a shift in the dominant dense water formation source from the Adriatic to the Aegean Sea was observed. This event is known as the Eastern Mediterranean Transient (EMT) and resulted in the massive export of dense, warm and saline deep waters from the Aegean Sea to the EMed abyssal reservoirs. The Adriatic Sea gradually regained its role as the main dense water source of the EMed only after the early 2000s (Roether et al., 2007; Cardin et al., 2015). Between 2004 and 2006 significant dense water formation events were observed in the WMed, producing larger quantities of warmer, saltier and denser deep waters masses than ever before. This event is known as the Western Mediterranean Transition (WMT) and is partly attributed to salinity preconditioning caused by the import from the EMed through the Sicily Strait of intermediate water masses with higher-than-usual salinity due to the preceding EMT event (Schroeder et al., 2016; Schroeder et al., 2017).

The Mediterranean Sea is a semi-enclosed deep basin (maximum depth 5,125 m offshore Peloponnese, Greece, EMed) characterized by oligotrophy, an overall result of low nutrient inputs into the basin (Krom et al., 2014). Nutrient concentrations in the Mediterranean Sea are characterized by a decreasing trend from the oligotrophic WMed to the ultra-oligotrophic EMed (Azov, 1991; Ignatiades et al., 2002). Redfield-deviated (i.e. higher than 16:1) N/P ratios become more pronounced eastwards and are essentially explained by the excess N over P in all nutrient sources arriving in the basin via atmospheric deposition (Mara et al., 2009; Markaki et al., 2010) and river discharge (Ludwig et al., 2009; Krom et al., 2010). The large loss of inorganic nutrients, through the inflow of relatively nutrient-depleted Atlantic water at surface layers and the outflow of the underlying nutrient-rich Mediterranean waters through the Strait of Gibraltar (Béthoux et al., 1998), prevents nutrient accumulation in deep waters, strongly limiting both primary and secondary production and thus particulate matter formation.

Within the Mediterranean, the EMed waters are very transparent and exhibit unique optical characteristics such as very low beam attenuation coefficient (c_p) values at depths >200 m (<0.04 m⁻¹ in the open Ionian Sea), associated with very low particulate matter and particulate organic carbon concentrations (PMC <100 μg L⁻¹, POC <40 μg L⁻¹, respectively) (Karageorgis et al., 2008; Zaccone et al., 2012; Ortega-Retuerta et al., 2019). Overall, the oligotrophy in combination with the dominance of the microbial food web in the pelagic ecosystem results in low food availability for top predators (Turley et al., 2000; Siokou-Frangou et al., 2002) and low levels of carbon sequestration to depth (Gogou et al., 2014).

Satellite chlorophyll-a concentrations during the MSM72 cruise are presented in Figures 3A–C. Figure 3A shows the monthly mean average sea surface chlorophyll-a concentration in the Mediterranean Sea for March 2018, while Figures 3B,C shows the 8 day average sea surface chlorophyll-a concentrations in the Mediterranean Sea for the first (06-13/03/2018) and second (30/03/2018-06/04/2018) legs of the MSM72 cruise, which corresponds to the sampling periods in the EMed and WMed Mediterranean sub-basins, respectively.

According to this, during the MSM72 cruise, a west-to-east, as well as a north-to-south gradient in oligotrophy, was observed in the two sub-basins (Figures 3A–C), which is in accordance with what has previously been described for the different Mediterranean bioregions (D’Ortenzio and Ribera d’Alcalà, 2009; Mayot et al., 2016). In these studies, the biogeography of the Mediterranean Sea and the seasonal cycle of the surface biomass were characterized in different areas of the basin by analyzing 10-year satellite surface Chl-a concentrations, and the results showed that the Mediterranean sub-regions have varying physical, chemical and biological forcing factors. However, it is important to note that surface Chl-a as seen by satellites, is not representative of the total phytoplankton biomass in the Mediterranean waters, i.e. subsurface phytoplankton dynamics, in particular those associated with the Deep Chlorophyll Maximum (DCM) at the base of the upper mixed layer (Barbieux et al., 2019; D’Ortenzio et al., 2021) could be highly underestimated.

A west-east section of c_p, VC, D₅₀, and ξ distribution in the study area is presented in Figure 4, and Table 1 summarizes the range and mean values in the entire Mediterranean as well as the considered sub-regions, i.e., Gibraltar and Algerian Basin, Sardinia and Tyrrhenian Sea, Sicily Strait and western Ionian Sea, Otranto Strait, Ionian Sea (excluding the Otranto Strait) and the Cretan Sea. The beam attenuation coefficient, c_p, clearly exhibits higher values throughout the water column in the WMed, especially in the upper ∼100 m, and progressively the c_p signal decreases towards the east, at first towards the Sicily strait, and then it becomes very weak in the EMed, with values <0.1 m⁻¹. The cruise average c_p was 0.53 ± 0.17 m⁻¹ and the highest values were recorded in the upper ∼200 m of the westernmost part of the section (>1.0 m⁻¹). The maximum value of 1.77 m⁻¹ was recorded near-bottom in the Otranto strait, and the minimum (0.025 m⁻¹) in the intermediate waters east of Crete.

The particle volume concentration, VC, along the section shows similar characteristics to c_p, with higher values in the WMed and progressively lower values towards the central Mediterranean and the EMed. It exhibits an average of 0.89 ± 0.31 μl L⁻¹ in the entire Mediterranean Sea. Similarly to c_p, the maximum value occurs near-bottom in the Otranto strait (3.7 μl L⁻¹), whereas the minimum value appears south of Crete (0.33 μl L⁻¹); overall, we observe a higher abundance of particles in the WMed and a scarcity of particles in the EMed.

Furthermore, the particle median diameter D₅₀, in the Mediterranean shows an average of 27 ± 3 μm (particle size was estimated in the range 1.25–92.6 μm; see Supplementary Material for details). Its evolution from the west to the east is reversed with respect to the previously discussed parameters. D₅₀ exhibits lower values in the WMed up to the Sicily strait, which increases substantially towards the east (>25 μm). The minimum value of D₅₀ is recorded in the Otranto Strait (19 μm), whereas the largest is found in the EMed, south of Crete (37 μm).

Finally, the PSD slope ξ, along the section shows similar characteristics to c_p and VC with high values in WMed and lower values in EMed, thus confirming the predominance of relatively smaller particle populations in the western basin over the larger particles dominating the eastern basin. The minimum value is 2.83 at the mid and deep waters of the Ionian Sea; on the other hand, the maximum value is 3.76 in the deepwater of the Otranto Strait. In the WMed, the average ξ is 3.2 ± 0.1; the average value in the EMed is 3 ± 0.1, whereas the Mediterranean Sea average ξ is estimated at 3.1 ± 0.1.

Beam c_p values from the present study appear to be high when compared to other data sets from the Mediterranean Sea (cruise average 0.53 m⁻¹). Karageorgis et al. (2008) compiled 11 years of transmissometry data from the EMed and reported that the average cp value in the open sea is ∼0.1 m^–1, i.e. five-fold lower than the present case. It should be noted however, that absolute cp values depend highly on the acceptance angle of the transmissometer employed. Boss et al. (2009) concluded that eight different transmissometers’ mean attenuation values differ markedly and in a consistent way with instrument acceptance angle: smaller acceptance angles provide higher beam attenuation values. Later, Boss et al. (2018) tested cp obtained from LISST-100X (5 cm pathlength, acceptance angle 0.069°) against cp from a hyperspectral spectrophotometer and transmissometer (ac-s, Sea-Bird Scientific, Bellevue, WA, United States, formerly WET Labs, 25 cm pathlength, acceptance angle 0.93°) and found particulate beam attenuation coefficients to be well correlated, with LISST values being slightly larger. LISST’s cp values, because of the very small-angle transmissometer employed, would be higher than the ones measured with conventional instruments. In that respect, direct cp comparisons should be avoided, with the exception of cp variation studies conducted in a relative manner.

Data for LISST-Deep derived parameters VC, D50, and occasionally ξ, presented in the current study represent a ∼30% increase relative to that available in the literature (Karageorgis et al., 2012; Karageorgis et al., 2014; Karageorgis et al., 2017b; Parinos et al., 2017; Turner et al., 2017; Leroux et al., 2018). A substantial amount of data have been previously reported with LISST-100X, i.e. the ‘shallow’ version of the instrument, operating up to 300 m depth (Mikkelsen et al., 2006; Barone et al., 2015; White et al., 2015; Yamada et al., 2015; Many et al., 2016; Qiu et al., 2016).

Literature mean VC values from the coastal and shallow (10–12 m depth) Adriatic Sea were two orders of magnitude higher than our measurements (Mikkelsen et al., 2006). In the northwestern Mediterranean, Gulf of Lions, VC average values in the vicinity of the Rhône River plume decreased progressively (Many et al., 2016) in the offshore direction. The latter notion is further strengthened by the observations of Leroux et al. (2018), who reported VC values well below 1 μl L⁻¹ in the open northwestern Mediterranean. In the open Ionian and Levantine Seas, Karageorgis et al. (2012) reported mean VC values in the range 0.29–0.30 μl L⁻¹.

The mean D50 literature values for the coastal Adriatic Sea varied from 112 to 291 μm, much higher than the values reported here. With respect to the Gulf of Lions, in the northwestern Mediterranean Sea, Many et al. (2016) reported PSD in the range of 30–200 μm offshore the Rhône River; it should be noted, however, that those values were derived from a combination of LISST-100X and LISST-HOLO instruments thus covering a wider particle size range, including micro-aggregates or macro-aggregates. In the open northwestern Mediterranean Sea, Leroux et al. (2018) identified particle populations in the range of 2–109 μm that were attributed to river-derived fine silts and microflocs. In the Ionian, Aegean, and Levantine Seas, D50 varied between 86 and 88 μm (note that the latter D50 values have been estimated using different methodological assumptions; Karageorgis et al., 2012). The aforementioned literature excerpts, as well as the MSM72 new data, exemplify the pronounced variety of particle sizes encountered in the marine environment of the Mediterranean Sea, resulting from the interplay of numerous biogeochemical processes, and thus enhancing the complexity of the particle pool and its composition dynamics.

With respect to the PSD slope ξ, very little information exists for the Mediterranean Sea. Many et al. (2016) estimated the spectral PSD slope γ from backscattering measurements and reported values in the range of 0.3–0.7, approximately corresponding to ξ values 3.3–3.7, thus similar to the MSM72 cruise values. Another example of ξ estimates from the north Aegean Sea (Karageorgis et al., 2017a) reports higher values varying from 3.46 to 4.86 which corresponds to smaller particle populations than the present study. In most oceanic observations ξ values range from 2.5 to 5 (Boss et al., 2001; Kostadinov et al., 2012), which, according to our measurements, is the case for the WMed and the EMed as well. Nevertheless, the PSD slope is highly dependent on: 1) the method of calculation; 2) the size range used, and 3) the existence or not of large deviations from the power law distribution. Regardless of methodological differences, the present data set on PSD slope ξ in the Mediterranean Sea indicates that there exists a substantial difference between the small particle abundance in the WMed in comparison to the larger particle populations in the EMed. The potential causes of this consistent pattern will be investigated hereafter using the biochemical data set obtained during the cruise.

A west-east section of POC and PN contents and their isotopic composition in the study area is presented in Figure 5, while as in the case of optical properties above, Table 1 summarizes their ranges and mean values in the entire Mediterranean as well as in the various sub-regions considered here. POC and PN concentrations in the study area, considering the entire water column, ranged from 1.06 to 7.12 µmol L⁻¹, 2.71 ± 1.40 µmol L⁻¹ on average, and from 0.09 to 1.99 µmol L⁻¹, 0.40 ± 0.32 µmol L⁻¹ on average, respectively. Consequently, the POC/PN molar ratio ranged from 1.9 to 13.7, with an average of 8.5 ± 2.6. Higher POC and PN concentrations were consistently recorded within surface waters or the maximum of chlorophyll fluorescence of all stations with the highest values being detected in the Alboran Sea and Algerian Basin while lower concentrations were observed in the eastern Ionian Sea and South Cretan margin (Figure 5; Table 1). Within intermediate and deep water layers of the studied sub-basins, POC and PN concentrations gradually decrease exhibiting a rather uniform distribution. The POC/PN ratio values exhibited a uniform distribution within surface waters or the maximum chlorophyll fluorescence depth of all stations with an average of 6.9 ± 0.9. Within intermediate and deep water layers, POC/PN ratio values increase with higher values, up to 11.5, being detected in the EMed stations and the Otranto Strait (Table 1).

δ¹³C-POC ranged between −28.73 and −25.02‰. Close and Henderson (2020) reported δ¹³C-POC values in open oceanic waters that span over a broad range, varying from −35‰ at high latitudes to −16‰ at low-mid latitudes. Less depleted values of δ¹³C-POC were recorded within surface waters or the maximum chlorophyll fluorescence depth in all stations. Intermediate and deep water δ¹³C-POC values varied between −28.73 and −27.10‰, averaging −27.89‰. Within the considered sub-regions, δ¹³C-POC values in near-surface waters of the WMed are comparable to those recorded within the Ionian Sea in the EMed, while more depleted values are evident within surface waters of the Sicily Strait, Otranto and the Cretan Sea. At deeper water depths, δ¹³C-POC values in the WMed, the Ionian Sea and the Cretan Sea average −27.96‰ while in the Sicily Strait δ¹³C-POC values average −28.47‰. In the Otranto Strait, more enriched δ¹³C-POC values are observed, with the ADW value being −27.36‰ (Table 1; Figure 5).

δ¹⁵N-PN values in our data set spans a range from −1.31 to 8.10‰. Depleted values are recorded within surface waters or the maximum of chlorophyll fluorescence in all stations, ranging from –1.31 to 3.35‰ (average 1.88‰), while enriched δ¹⁵N-PN values are recorded in deeper water layers ranging between 2.67 and 8.10‰ (average 4.91‰). Within the considered sub-regions, δ¹⁵N-PN values in near-surface waters of the WMed averaged 2.40‰, in the Sicily Strait averaged 3.20‰, while in the Ionian Sea, the Otranto Strait and the Cretan Sea δ¹⁵N-PN values averaged 1.10‰, with the minimum value being recorded in the surface waters of the Otranto Strait and maximum δ¹⁵N-PN value at the maximum of Chl-α fluorescence depth in the Cretan Sea. In deeper water layers, δ¹⁵N-PN values in the Gibraltar and the Algerian Basin range from 2.63 to 5.85‰, averaging 4.18‰, while more enriched values are recorded westwards in Sardinia and Sicily Strait, averaging 6.07‰, and the Ionian Sea, 5.64‰ on average with a maximum of 7.79‰ within the EMDW layer in station 55. Less enriched values are evident in the Otranto Strait, where values reaching 1.92‰ for ADW are recorded (Table 1; Figure 5).

The POC and PN contents of PM reported in our study fall within the ranges of previously published data for open Mediterranean Sea waters. POC values up to 16.5 µmol L⁻¹ have been reported for the top 400 m of the water column in open Mediterranean waters (Struck et al., 2001), up to 13.8 µmol L⁻¹ for near surface waters of the open Levantine Basin (Tanaka et al., 2007; Bar-Zeev et al., 2011), up to 7.2 µmol L⁻¹ in epipelagic waters of the open northwestern Mediterranean Sea (Ortega-Retuerta et al., 2017) and up to 9.8 µmol L⁻¹ in near surface waters of the north Aegean Sea (Parinos et al., 2017). Zaccone et al. (2012) has reported mean POC values of 2.29, 0.60, and 0.59 µmol L⁻¹ in epipelagic, mesopelagic and bathypelagic layers of the WMed, with the corresponding values for the EMed being 1.82, 0.77 and 0.59 µmol L⁻¹, respectively. Karageorgis et al. (2018) reported POC values in the range 2.3–11.8 µmol L⁻¹ in canyons and open slopes of the southern Cretan margin (EMed), while Ortega-Retuerta et al. (2019), reported POC values ranging from 0.10 to 8.60 µmol L⁻¹ along an east-west transect across the Mediterranean Sea. Concerning PN contents, Emeis et al. (2010) have reported similar values in vertical profiles of pelagic stations in the Ionian Sea and NW Levantine Sea as has Struck et al. (2001) for the top 400 m of the water column in open Mediterranean waters. Pavlidou et al. (2020) reported mean PN contents values of 3.17, 2.40, and 0.94 µmol L⁻¹ in epipelagic, mesopelagic and bathypelagic layers, in the NE Ionian Sea. It is noteworthy that POC concentrations encountered in the present study are amongst the lowest values observed in the global ocean, as values <10 μmol L⁻¹ were found only in the open Pacific Ocean and Bermuda in the Atlantic Ocean (Gardner et al., 2006).

Concerning δ¹³C-POC, and to the best of our knowledge, there are no previous published data along the entire water column, from the surface to deep waters, across the open Mediterranean Sea. Çoban-Yildiz et al. (2006) reported an δ¹³C-POC average value of -22.9‰ in the euphotic zone of the Rhodes Gyre in the Levantine Basin, while Struck et al. (2001) reported an δ¹³C-POC range between −28.41 and −24.50‰ for the top 400 m of the water column in open Mediterranean waters. In the case of δ¹⁵N-PN, the values reported herein fall within the range of previously published data for open Mediterranean Sea waters. Emeis et al. (2010) reported δ¹⁵N-PN values averaging 1.3 ± 0.7‰ in the mixed layer and up to 8.1‰ in deeper waters of the EMed. In a transect across the Mediterranean Sea, Pantoja et al. (2002) observed a significant west-to-east decrease in the isotopic composition of PN from the upper 200 m of the water column. In the western basin, δ¹⁵N-PN averaged 2.7 ± 1.2‰, in contrast to −0.2 ± 0.7‰ in the eastern basin. Çoban-Yildiz et al. (2006) reported an δ¹⁵N-PN average value of 2.2‰ in the euphotic zone of the Rhodes Gyre in the Levantine Basin. Koppelmann et al. (2009) reported δ¹⁵N-PN values within a range of 0.43–1.19‰ in surface waters of the Levantine Basin, while slightly higher values were found in the DCM and mesopelagic and bathypelagic depths, ranging between 6.19 and 8.49‰. Struck et al. (2001) reported δ¹⁵N-PN values ranging from 2.64 to 13.7‰ for the top 400 m of the water column in open Mediterranean waters, while finally, Pavlidou et al. (2020) reported mean δ¹⁵N-PN values of −0.57, 2.98 and 4.18‰ in epipelagic, mesopelagic and bathypelagic layers, in the NE Ionian Sea (Nestor site).

In order to gain insights on the relationships between measured optical properties and identified water masses in the Mediterranean Sea, T-S diagrams superimposed with c_p, VC, D₅₀, and ξ information were used (Figure 6). Surface Atlantic waters (AW) are clearly characterized as the most turbid waters in the Mediterranean Sea, with average beam c_p of 0.68 ± 0.01 m⁻¹. On the other hand, TMW constitutes another extreme; one of the lowest observed c_p values averaging 0.20 m⁻¹. AW spreading towards the EMed and particularly in the Ionian Sea exhibits moderate beam c_p values. Intermediate waters masses as LIW show overall c_p values in-between AW and LSW (average 0.57 ± 0.03 m⁻¹), whereas deep water masses have different c_p values according to their location. This pattern is closely followed in the case of VC and is inversed for D₅₀. Average D₅₀ in AW is 25 ± 3 μm, in LSW 35 ± 2 μm, in LIW 28 ± 4 μm, whereas the lowest average D₅₀ value of 19 μm is observed in ADW. AW in the western basin is characterized by low salinities, relatively high turbidity and VC, low particle size, and higher ξ. The highest ξ values (>3.15) are observed within AW in the WMed, which decrease as AW spreads out towards the east. LIW on the other hand, exhibits the smallest ξ values, especially in the EMed.

The differentiation between the WMed and the EMed is also illustrated in the variability of mean differential particle volume V’(D) (see Supplementary Material for definition), D₅₀, and PSD slope ξ (Figure 7). Mean V’(D) and ξ are higher in the WMed than in the EMed, whereas the opposite is recorded for mean D₅₀ values. These observations demonstrate that in the EMed particle concentration is overall lower than the WMed, whilst mean particle size increases with depth, suggesting that aggregation processes are active in the EMed.

The results of the correlation analysis of the determined optical and biochemical parameters considered in this study are shown in Supplementary Table 1. POC is significantly and positively correlated with VC, negatively with D₅₀, and positively with ξ in both the WMed and the EMed. The elevated VC, POC, and PN concentrations recorded within surface waters of the WMed stations, highlight the enhanced surface production therein, compared to the EMed, resulting in a higher abundance of particles within the upper water column. POC concentration and ξ are strongly, significantly, and positively correlated in the Mediterranean Sea, confirming that small particles are POC-rich. Finally, the combined positive correlation of VC with POC concentrations and δ¹³C-POC can be attributed to surface productivity.

The positive and significant relationship between c_p, as a proxy of suspended particle concentration (or turbidity), and POC, has been acknowledged for decades in global ocean studies (Bishop, 1999; Gardner et al., 2006) as well as the Mediterranean Sea (Loisel et al., 2011); this also holds for the EMed, as demonstrated by Karageorgis et al. (2008) who reported a significant positive correlation between beam c_p and POC concentrations. However, it should be mentioned that the level of relationship also depends on the acceptance angle of the transmissometer used, and in the present case, LISST employs a very narrow angle transmissometer, hence correlations may differ substantially. In any case, the observed positive and significant correlation (r = 0.613, p = 0.002) in the WMed highlights the predominance of organic particles in the water column, whereas this relationship weakens in the EMed (r = 0.613, p = 0.002) probably due to: 1) the narrow acceptance angle of LISST’s transmissometer resulting in small sampled water volume (Boss et al., 2009), in combination with the extreme scarcity of particles observed in the EMed, and 2) the presence of generally larger particles in the EMed, greater than the 0.5–20 μm range, where most of the light attenuation and scattering occurs, and is detected by transmissometers (Gardner et al., 2006).

D₅₀ is significantly positively correlated with water column depth in both the WMed and the EMed (excluding the Otranto Strait) (Supplementary Table 1) likely suggesting that the process of aggregation of PM with increasing depth is probably present in the study area. Interestingly, D₅₀ is significantly positively correlated with POC/PN ratio values only within the EMed stations (excluding the Otranto Strait), while a significant correlation of the POC/PN ratio vs. depth is also evident (r = 0.422, p = 0.02, n = 30). Various reasons could account for this result. In oligotrophic systems, such as the EMed, a faster and more efficient decomposition of N versus C has been highlighted by previous studies (Altabet, 1996; Meador et al., 2010), which could potentially account for the observed elevated POC/PN values with depth. Moreover, in the EMS, a significant proportion of C fixed by phytoplankton is released to the dissolved phase (Teira et al., 2001; Lagaria et al., 2011; López-Sandoval et al., 2011), leading to the enhanced extracellular release of organic carbon mainly in the form of carbon-rich acidic polysaccharides (Passow, 2002; Gogou and Repeta, 2010; Parinos et al., 2017). The coagulation of this colloidal dissolved organic matter could lead to the formation of extracellular polymeric substances, mainly transparent exopolymer particles (TEP), bearing a C/N ratio signature well above Redfield, i.e. mean value >20 (Engel and Passow, 2001; Mari et al., 2001). The observed significant positive correlation of D₅₀ vs POC/PN ratio vs depth (Supplementary Table 1) could be indicative of the efficiency of aggregation mechanisms within the EMed, with carbon-rich surface active organic particles (i.e. TEPs) acting as “biological glue” for the formation of larger sized aggregates (Bar-Zeev et al., 2011; Parinos et al., 2017; Ortega-Retuerta et al., 2019). In support of this, Martínez-Pérez et al. (2017), recorded a substantial increase of high molecular weight dissolved organic matter (TEP precursors) in a transect spanning from the Atlantic and Gibraltar to the Levantine Sea (from 40–55% to 68–76%, respectively) thus providing independent evidence of more important aggregation potential in the eastern sub-basin.

The horizontal patterns of δ¹⁵N-PN observed in the mixed layer of the EMed reflect the interplay of isotopically distinct N sources and the degree of N utilization during primary production. Extremely low δ¹⁵N-PN values in NO₃ ^- replete waters (Hainbucher et al., 2020) at stations 47 and 9 give evidence for diazotrophic N₂ fixation or atmospheric NO_x deposition in the ultra-oligotrophic EMed (Pantoja et al., 2002; Emeis et al., 2010; Moon et al., 2016). Moreover, low values at still relatively high nitrate levels in the Alboran Sea point towards preferential assimilation of ¹⁴N, however, the ambient nitrate most probably had a higher δ¹⁵N signature due to its Atlantic origin (Pantoja et al., 2002). Slightly elevated values, in concert with almost depleted NO₃ ^- concentrations in the surface waters of remaining stations, imply the assimilation of ¹⁵N enriched NO₃ ^- at the end of a blooming period, but may, to some extent, also reflect the increasing importance of ¹⁵N enriched sources such as Atlantic water and riverine inputs in the WMed as proposed by Struck et al. (2001) and Pantoja et al. (2002).

The striking segregation of relatively depleted δ¹⁵N-PN in the surface or mixed layer waters and enriched values in deep Mediterranean waters has been also observed by Emeis et al. (2010) and Pavlidou et al. (2020). Both attributed the higher δ¹⁵N-PN values in deep waters to the enrichment of residual ¹⁵N during progressive decay of organic matter, whereas the latter for the first time described a systematic relationship to PN concentrations. In contrast to Pavlidou et al. (2020), our EMed data show several (less enriched) outliers in the δ¹⁵N-PN of deep waters. Together with lower POC/PN ratios, these values imply less degraded organic matter (Figure 8A). On the other hand, particle concentrations and particle mean sizes remain stable, in defiance of signals from the surface waters such as a sinking algal bloom. At the same time, nitrate concentrations reach their maxima and O₂ concentrations their minima (Hainbucher et al., 2020). Interestingly, all these outliers (even in the Sicily Strait) have been sampled between 500 and 1,500 m within the TMW. The TMW contains former bottom waters that have been uplifted during the EMT (Stöven and Tanhua, 2014; Velaoras et al., 2019). We consider it possible that fresh organic matter has been released from larger disaggregating particles settling in the bottom water, which remained in the water mass during uplift. The high apparent oxygen utilization (AOU) rates recorded before in the mesopelagic and bathypelagic waters of the EMed (Meador et al., 2010; Santinelli et al., 2010) are indicative of exhaustive organic matter recycling, but the remaining (semi-refractory to refractory) organic matter may not be readily available to deep-sea microbes (Rahav et al., 2019) or is highly diluted (Arrieta et al., 2015).

In support of this hypothesis, Sisma-Ventura et al. (2016) reported depleted values of δ¹³C-DIC in the LIW and the EMDW of Aegean origin, claiming that isotopically light anthropogenic carbon penetrates the Levantine interior during pronounced convection events such as the EMT. Alternatively, we cannot exclude that lower δ¹⁵N-PN values result from surface waters that have downwelled in the Cretan Sea and the EMed during more recent, extreme winter mixing events (i.e. in 2012 and 2014–2015) (Gačić et al., 2014; Pedrosa-Pàmies et al., 2016; Velaoras et al., 2017; Pedrosa-Pamies et al., 2021).

Similar processes could account for the less enriched values of δ¹⁵N-PN being observed in some cases at the subsurface layers of the WMed, as a western Mediterranean Transient after the EMT, and extensive downwelling events, have been reported in this sub-basin (Durrieu De Madron et al., 2013; Stöven and Tanhua, 2014; Testor et al., 2018). Finally, we should also mention the possibility that the outliers could give evidence for in-situ aphotic diazotrophic N₂ fixation as earlier described for the Mediterranean Sea (Rahav et al., 2013; Benavides et al., 2016; Moisander et al., 2017), although the isotopic fractionation of the processes and sources involved are not yet delineated.

δ¹³C-POC values show a surface enrichment-depth depletion profile in all stations. A plausible explanation for the relatively depleted δ¹³C-POC values in the marine organic pool of the open Mediterranean Sea can be the contribution of terrestrial organic matter, mainly via atmospheric dry and wet deposition (Gogou et al., 1996; Guerzoni et al., 1999; Pedrosa-Pamies et al., 2021; Stavrakakis et al., 2013). Vertical differences in organic matter quality and quantity are also reflected in lower N concentrations and the increase of the POC/PN ratio in deep waters of all studied sites. Following the approach of Pavlidou et al. (2020), we assume a Rayleigh type isotope fractionation discriminating ¹⁵N during the decomposition of sinking and suspended PN (Möbius, 2013). Evidenced by their data, basic assumptions of Pavlidou et al. (2020) are: 1) PN distribution in the water column is almost solely driven by vertical transport (sinking) from sea surface to bottom waters and 2) additional sources and sinks of PN are negligible. In this case, the strength of fractionation (fractionation factor ε¹⁵) is expressed by the slope of the logarithmic function of δ¹⁵N-PN vs PN in depth profiles. In our study, fractionation factors calculated vary from 1.34 to 4.66‰ (Figures 8B,C) and thus are similar to those published in Pavlidou et al. (2020). In contrast to Pavlidou et al. (2020) excellent correlations that cover the full range between both endmembers, our data show stronger scatter and concentrate on the endmembers with only few exceptions. We explain the scattering by both, the above-mentioned outliers in the TMW and by the fact that our sampling sites in the open and deep Mediterranean Sea may be biased by horizontal transport. As correlations in our data set are weaker, we consider the absolute values of fractionation factors with care—nevertheless, a tendency towards stronger fractionation in the EMed can be noted (Figures 8B,C). This could either simply reflect the stronger vertical gradients in δ¹⁵N-PN in the EMed (extremely low in surface waters) or it is plausible as organic matter decomposition is expected to be more complete in the nutrient starving ultra-oligotrophic EMed in comparison to the oligotrophic WMed.

Apart from biogeochemical properties, local mesoscale features’ variability and dynamics seem to play an important role in regulating the optical and biochemical characteristics of particulate matter in the study area. Supplementary Figure 1, shows the variability of salinity and beam attenuation coefficient in the Alboran Sea and the Algerian basin. Therein, a sinusoidal variation of salinity is observed, possibly caused by the presence of anticyclonic eddies (Millot, 1999). A similar variation is also observed in the distribution of the beam attenuation coefficient, suggesting that the inherent optical properties of the water masses, and in particular the spatial variability of the beam attenuation coefficient, are largely controlled by mesoscale dynamics.

Interesting optical and biochemical features were also observed in the Otranto Strait which separates the Adriatic and the Ionian Seas (Supplementary Figures 1, 2). Dense waters formed locally in the Adriatic during winter, exit through the bottom of the Otranto Strait, while LIW enters the basin at intermediate layers. The result is a cyclonic flow pattern, with water entry at the eastern part of the Strait and exit through the western part (Kovačević et al., 1999; Yari et al., 2012). Indeed, in the studied section, we observed that waters of Adriatic origin occupy the western Otranto Strait and are characterized by higher c_p and VC, and inverse D₅₀ values when compared to the LIW waters prevailing in the eastern Otranto Strait. This is accompanied by elevated POC concentrations along with enriched δ¹³C-POC, and depleted values for δ¹⁵N-PN at intermediate and near bottom-waters in the western part of the section, likely attributed to the newly formed dense water-bearing the biochemical signature of surface waters that they originated from. Finally, maximum values of c_p and VC are observed near the bottom layers of the central part of the section. This could likely be attributed to surface sediment resuspension due to currents, a commonly observed feature in the world ocean and the Mediterranean Sea as well (Karageorgis and Anagnostou, 2003; Karageorgis et al., 2008; Karageorgis et al., 2012). In support to this, minimum values for D₅₀ are observed in the same area, probably due to the fact that resuspension favors the detachment of smaller particles from the seabed.