3. Results

3.1 Marine snow and abiotic aggregation potential

No visible sinking marine snow particles (diameter > 0.5 mm) were found in the tray from any of the MSCs deployments. This observation suggests that the concentration of sinking marine-snow-sized particles was less than 0.02 per L. This finding is consistent with particle size distribution data from the UVP (pers. comm. Andrew McDonnell, 2021). Furthermore, during the four aggregation potential experiments, no visible marine snow-sized aggregates formed in the rolling tanks regardless of depth (55 and 95 m) and sample fraction (T0, top or base ).

3.2 Particle concentrations and size

Most of the imaged particles (were smaller than 16 μm in diameter (ESD), whether suspended or sinking (Figure 1 ). Furthermore, particle concentrations measured in the tray were equal or lower than the ones measured in the base for 20 out of 30 MSC deployments. These observations suggest that the concentration differences betweenbase and tray particle fractions were too small to be detected with this method. We hence summed the concentrations of slow- and fast-sinking particles into a single sinking particle fraction. This decision is additionally supported by the lack of marine snow, which would have contributed appreciably to the fast-sinking particle fraction.
Our results, based on the analysis of particles with diameters (ESD) between 4 and 128 μm, suggest that > 70% of suspended and sinking particles had a diameter of 4–8 μm, and that particles larger than 32 μm were at the detection limit of our method (i.e., too rare to be quantified). The partitioning of particles across the size bins was consistent among particle fractions (paired t-test, p = 0.33–0.63, n = 17), indicating that suspended and sinking particles were similarly sized (Figure 1 and Table S4 ).
Additionally, we calculated the particle size distribution (PSD, 4–128 μm) of suspended particles and of sinking (‘tray ’) particles (uncorrected fraction) to assess differences in slopes among fractions with depth. Overall, throughout the entire water column, PSD slopes of suspended particles were similar (-3.3 ± 0.3) to those of sinking particles (-3.5 ± 0.2; Figure S1 ). Yet below 100 m, PSDs of sinking particles were significantly steeper than those of suspended particles (paired t-test, p = 0.03, n = 5; Figure S1 ). Furthermore, the slope of suspended PSD became significantly shallower with depth between 20 and 500 m (R2 = 0.3, p < 0.02, n = 17), while the slopes of sinking PSDs remained constant (R2 = 0; p = 0.79, n = 17; Figure S1 ).
It is important to note that the determination of particle size using this approach is associated with uncertainties due to samples handling. We cannot exclude the possibility that fragile and larger aggregates (ESD < 0.5 mm) may have broken up due to physical disturbance developed during the sampling and partitioning of the three particle fractions and during injection of samples in the FlowCam flowcell used in combination with the x40 objective (minimum size of 300 μm). Therefore, the size characterization performed with the FlowCam must be taken with caution. Lastly, particle type and shape could not be reliably resolved with the FlowCam due to the small size of the particles.

3.3 Biogeochemical partitioning by particle sinking fraction

The background suspended POC concentrations decreased significantly with depth in the euphotic zone between 20 and 65 m (R2 = 0.43, p = 0.03, n = 11) and in the upper mesopelagic between 95 and 500 m (R2 = 0.59, p = 0.001, n = 14) (Figure 2A ), whereas background sinking POC concentrations did not significantly attenuate with depth (R2 = 0, p = 0.61, n = 24). The relative contribution of background suspended POC to total background POC decreased significantly in the upper mesopelagic (R2 = 0.70, p < 0.001, n = 13), resulting in a concomitant increase in the contribution of the background sinking POC (R2 = 0.70, p < 0.001, n = 13) (Figure 2A ). Nearly all the background POC was found in suspended particles, whose contribution to the total POC was on average 88 ± 9%. In respect to outliers, in five out of 29 MSC deployments, the measured POC of either suspended or sinking particle fractions was at least twice higher than POC concentrations measured at other stations in the same depth layer (Table S5 ).
Patterns of PON (Figure S2 ) were similar to those of POC, with molar POC-to-PON ratios of the suspended fraction on average slightly higher than the Redfield ratio of 6.6 (average: 7.9 ± 3.9). The molar POC-to-PON ratios did not show any trend with depth (R2 = 0.06, p = 0.25, n = 24) and were in general agreement with molar POC-to-PON ratios measured during the EXPORTS field campaign (Graff et al., in review). Molar POC-to-PON for sinking particles varied between 3 and 16 (average: 7.2 ± 3.8) and did not show any significant trend with depth (R2= 0.11, p = 0.1, n = 26).
PIC was below detection in 20% of the suspended particle fractions, and in 60% of sinking particle fractions analyzed. Suspended PIC concentrations increased significantly with depth (R2= 0.40, p = 0.05, n = 10) (Figure 2B ). Background sinking PIC was on average 0.4 ± 0.3 µg C L-1 and remained constant with depth (R2 = 0.05, p = 0.59, n = 8). Background molar PIC-to-POC ratios calculated for suspended particles increased significantly with depth (R2 = 0.80, p < 0.001, n = 10) and were significantly higher than PIC-to-POC ratios of sinking particles in the mesopelagic (paired t-test, p = 0.01, n = 5) (Figure 3A ).
Background suspended and sinking bSi concentrations and their contribution to total bSi concentrations did not significantly change with depth between 50 and 500 m (R2 = 0.06–0.18, p = 0.17–0.49, n = 8–11) (Figure 2C ). Suspended particles stored on average 83 ± 9% of the total background bSi pool. Molar bSi-to-POC ratios associated with background suspended particles increased significantly with depth between 50 and 500 m (R2 = 0.84, p < 0.001, n = 8), whereas ratios of sinking particles remained constant (R2 = 0.16, p = 0.32, n = 8). Furthermore, molar bSi-to-POC ratios of sinking particles were on average higher than the ratios associated with suspended particles above 95 m (n = 3), and lower between 350 and 500 m (n = 4). Ratios above and below 95 m were statistically different (paired t-test, p = 0.01 and p = 0.008, respectively) (Figure 3B ).
Suspended and sinking lSi concentrations and relative contribution to the total lSi pool did not display a significant trend with depth and showed high variability (R2 = 0.004–0.07 p = 0.39–0.85, n = 11-13) (Figure 2D ). Suspended lSi accounted, on average, for 83 ± 19% of the total lSi pool. Molar lSi-to-POC ratios associated with suspended particles increased significantly with depth between 50 and 350 m depth (R2 = 0.78, p = 0.02, n = 6). For sinking particles, molar lSi-to-POC ratios displayed the tendency of decreasing with depth between 60 and 500 m, although the trend was not statistically significant (R2 = 0.05, p = 0.58, n = 9) (Figure 3C ). Ratios between suspended and sinking particles were statistically different below the euphotic zone (paired t-test, p < 0.01) with suspended particles characterized by consistently higher ratios than sinking particles.
Suspended TEP decreased significantly with depth (R2 = 0.53, p =0.001, n = 16) and like POC, displayed the sharpest decrease between 50 and 95 m. On average, 95 ± 4% of TEP was suspended, and sinking TEP was often at or near detection (average = 0.3 ± 0.2 µg GXeq L-1, n = 15) (Figure 2E ). TEP concentrations measured in the MSCs were in line with the concentrations measured in the water collected with the CTD rosette (Figure S3 ). After converting TEP to carbon concentration of TEP, we estimated, in 11 out of 15 deployments, that 15 ± 5% of suspended POC was TEP and only 3 ± 1% of sinking POC was TEP. In 4 observations collected above 100 m, TEP-C accounted for 13% of the sinking POC (Figure 3D ). TEP-C-to-POC ratios associated with suspended and sinking particles were statistically different (paired t-test, p < 0.001, n = 15).
Background POC, TEP, bSi and lSi concentrations in the suspended and sinking particle pools did not display a significant variability over time at 50–65 m, 95 m or 300–500 m (t-test, 95% confidence). Temporal trend for PIC could not be tested owing to the limited number of observations available.
To provide an overall assessment of the relative changes in the measured biogeochemical composition of suspended and sinking particles with depth, we summed the background cruise-wide averaged concentrations of organic TEP-C, POC and PON and ballast minerals and evaluated the differences in their relative contributions (Figure 4 ). We found that the composition of suspended particles was dominated by POC (64%) in the euphotic zone and by POC (34 ± 11%), lSi (30 ± 7%) and PIC (16 ± 8%) in the upper mesopelagic, whereas sinking particles were on average composed mainly by POC (50 ± 10%), followed by lSi (20 ± 7%), bSi (14 ± 7%) and PON (9 ± 3%) throughout the water column. The relative contribution of TEP-C in suspended particles was on average twice the one in sinking particles (Figure 4B ).

3.4 Constraining the particle sinking velocity

The settling time experiments found no consistent differences in the fraction of sinking particles to the whole particle population (suspended plus sinking) as a function of settling time (t-test, 95% confidence). After 1, 2 and 4 hours of settling, sinking POC made up 10, 11 and 10% and sinking PON 6, 7 and 8% of the whole particle population (Table S6 ). Sinking bSi made up 23, 25 and 25%, whereas sinking TEP made up 6, 3 and 5%. However, our experiment was limited to only two casts within the euphotic zone; hence, we cannot assume that all sinking particles collected had the potential to sink within one hour at 36 m d-1. We therefore present fluxes calculated assuming the standard sinking velocity of 18 m d-1, and we reinforce that these fluxes represent a lower bound as suggested by our settling time experiments (Figure 5 ). Background POC fluxes averaged to 4.2 ± 2.6, 3.1 ± 1.5 and 3.1 ± 1.1 mmol C m-2 d-1 at 50–65 m, 95 m and 300–500 m, respectively. When including the outliers, the estimates increased up to 7 ± 10, 9 ± 18 and 4 ± 2 mmol C m-2 d-1, at these same depth intervals. Fluxes of bSi were on average 0.2, 0.2 and 0.1 mmol Si m-2 d-1 and lSi fluxes were 0.3, 0.4 and 0.2 mmol Si m-2 d-1 at 50–65 m, 95 m and 300–500 m, respectively.