Structure and functioning of epipelagic mesozooplankton and response to dust events during the spring PEACETIME cruisein the Mediterranean Sea

The PEACETIME cruise (May-June 2017) was a basin scale study mainly dedicated to the study of different 10 planktonic trophic regimes in the Algerian, Tyrrhenian and Ionian basins and, in particular, focusing on areas impacted by Saharan dust deposition. This paper presents the structural and functioning patterns of the zooplankton component during this survey, including their responses to two major dust events in the Algerian and Tyrrhenian basins. The mesozooplankon was sampled at 12 stations by combining nets with 2 mesh sizes (100 and 200 μm) mounted on a bongo frame for vertical hauls within the upper 300 meter layer. In this general post-bloom situation, total mesozooplankton showed reduced 15 variations in abundance and biomass over the whole area, with a noticeable contribution of the small size fraction (< 500 μm) of up to 50 % in abundance and 25 % in biomass. The taxonomic structure was dominated by copepods, mainly cyclopoids and calanoids, and completed by appendicularians, ostracods and chaetognaths. Distinct zooplankton taxa assemblages in the three main regions were in agreement with recently proposed regional patterns for the Mediterranean Basin, although the assemblages found in the western Ionian stations presented a closer analogy with those of the Tyrrhenian 20 basin than with those of the Ionian basin., probably due to Atlantic water influence. Zooplankton carbon demand, grazing pressure, respiration and excretion rates were estimated using allometric relationships to the mesozooplankton size-spectrum. On average, the daily zooplankton consumption potentially represents 15 % of the phytoplankton stock, almost the whole of the primary production, with a narrow range of variations, and its excretion contributes roughly one quarter of the N and P requirements of phytoplankton production. The small size fractions make a 25 significant contribution to these mesozooplankton estimated fluxes. Whereas in the Algerian basin (long station FAST), the initial impact on the pelagic ecosystem of a tracked dust deposition was studied, the survey of the southern Tyrrhenian basin occurred almost a week after another dust event. The changes in mesozooplankton taxonomic structure appear to be a relevant indicator to study this response, with an initial phase with no real dominance of taxa, then a disturbed state of the community with strong dominance of certain herbivorous taxa and the 30 appearance of carnivorous species, and finally a recovery state towards a more stable system with diversification of the https://doi.org/10.5194/bg-2020-126 Preprint. Discussion started: 4 May 2020 c © Author(s) 2020. CC BY 4.0 License.

phytoplankton stock and production by estimating its ingestion, respiration, ammonium and phosphate excretions using allometric models. 65 2 Material and methods

Study area and environmental variables
The PEACETIME cruise survey was conducted in May/June 2017 in the western Mediterranean Sea (Figure 1) on board R.V. Pourquoi pas?. Among the 12 stations studied, 10 were sampled once for zooplankton (the short stations ST1 to ST9,70 and the long station TYR), whereas two long stations ION and FAST, lasting 3 and 5 days respectively, were sampled three times. The station positions along the transect were planned before the cruise so as to sample the principal ecoregions (see Figure 4 in Guieu et al., submitted), with the exception of FAST, an opportunistic station to monitor a wet dust deposition event which occurred on June 5 a few hours after the first sampling date (Table 1). A quite important dust event occurred over a large area including the southern Tyrrhenian Sea starting on May 10 which could have impacted the samples at ST5, 75 ST6 and TYR which were sampled on May 16, 19 and 22 respectively (pers. comm. C. Guieu).
Hydrological variables (temperature, density, salinity) were measured on vertical profiles using a CTD. Dissolved oxygen was measured using a SBE43 sensor and chlorophyll-a concentration was determined from Niskin bottle samples by HPLC following the protocol of Ras et al. (2008), and with Fluorescence sensor coupled with the CTD. The depth of the mixed layer (MLD) was computed using the density difference criterion ∆ = 0.03 −3 defined in de Boyer Montégut et al. 80 (2004).

Zooplankton sampling and sample processing
A total of 16 zooplankton samples were collected at 12 stations (Table 1) using a Bongo frame (double net ring of 60 cm mouth diameter) equipped with 100 µm and 200 µm mesh size nets (noted N 100 and N 200 below) mounted with filtering codends. At all sampling stations, the Bongo frame was towed from 300 m depth to the surface at a constant speed of 1ms -1 . The 85 sampling was mostly performed during the morning, except for ST7, ST9 and TYR, and night tows were also performed for the long stations FAST and ION. The samples were preserved in 4% borax-buffered formalin immediately after the net was hauled back onto the deck .
The samples were processed using FlowCAM (Fluid Imaging Technologies Inc.) and ZOOSCAN (Gorsky et al., 2010). One of the goals of this study was to achieve the determination of the complete size structure of the zooplankton community by 90 combining different plankton mesh size nets and analysis techniques (FlowCAM and ZOOSCAN) in order to optimize the observed size spectrum. The formalin preserved samples were rinsed with tap water to remove the formalin. For net N 100 , the sample was then split into 3 size fractions: < 200 µm (noted below N 100 F <200 ), 200 µm -1000 µm (noted below https://doi.org/10.5194/bg-2020-126 Preprint. Discussion started: 4 May 2020 c Author(s) 2020. CC BY 4.0 License. N 100 F 200/1000 ), and > 1000 µm (noted below N 100 F >1000 ). For net N 200 , the sample was split into two size fractions:< 1000 µm (noted below N 200 F <1000 ) and > 1000 µm (noted below N 200 F >1000 ). 95 To determine the complete size spectrum, different combinations of size fractions from the two nets and analytical techniques were tested. Taking into account the two mesh sizes, (N 100 , N 200 ), the limits of the size spectrum were defined from the fraction N 100 F <200 for the lower limit and from the fraction N 200 F >1000 for the upper limit. Considering that our FlowCAM does not detect particles larger than 1200 µm of ESD and our ZOOSCAN does not detect particles smaller than 300 µm of ESD, N 100 F <200 was analyzed by FlowCAM and N 200 F >1000 by ZOOSCAN. The intermediate size fractions 100 N 100 F 200/1000 and N 200 F <1000 were both analyzed with ZOOSCAN and FlowCAM. These analyses delivered abundance and biomass values for successive ESD size classes: <200 µm (noted C <200 ); 200-300 µm (C 200-300 ); 300-500 µm (C 300-500 ); 500-1000 µm (C 500-1000 ); 1000-2000 µm (C 1000-2000 ); > 2000 µm (C 200-300 ). The challenge was to choose the best net-analysis technique combination for the intermediate size fractions (C 200-300 , C 300-500 and C  ). The abundance of each class for the two nets and the two treatments was statistically compared. Parts of the spectrum corresponding to fractions C 200-300 and C 300-105 500 from N 100 measured with FlowCAM, and to the fractions C  from N 200 measured with the ZOOSCAN have significantly higher abundances than other net-analysis technique combinations (T test, p<0.000). Consequently, we combined data for N 100 F <200 and N 100 F 200-1000 measured with FlowCAM to compute ESD size classes <500 um ( Figure 2A) and data for N 200 F <1000 and N 200 F >1000 measured with ZOOSCAN to compute ESD size classes >500 um (see Figure 2B). The combination of these data enabled us to compute the final size spectrum ( Figure 2C), that was used to estimate abundance, 110 biomass and metabolic rates for each ESD size class, and then for the whole sample (sum of all the size classes) and for the total mesozooplankton (sum of the size classes C 200-300 , C 300-500 , C 500-1000 and C 1000-2000 ).
For the FlowCAM analyses, the sample was concentrated in a given water volume. Then, an aliquot of each sample was analyzed using FlowCAM in auto-image mode. For the fraction N 100 F <200 , a 4X magnification and 300 µm flow cell were used and the analysis was carried out up to3000 counted particles. For the fraction N 100 F 200-1000 a 2X magnification and 800 115 µm flow cell were used and the analysis was carried out up to 1500 counted particles.
The digitalized images were analyzed using the VisualSpeadsheet® software and classified manually into taxonomic categories. Considered living organism groups for the FlowCAM were copepods, nauplii, crustaceans, appendicularians, gelatinous, chaetognaths and other diverse zooplankton groups (polychaeta, ostracods etc.). Non-organism particles were classified as detritus. Duplicates and bubbles were deleted. 120 To calculate the number of particles in the sample, the following equation was used.

= × ×
Where A is the abundance (ind m -3 ); P a is the number of particles in the analyzed aliquot; V c is the given volume in the concentrated sample and V a is the volume of the analyzed aliquot and V s is the volume of sea water sampled by the zooplankton net (m 3 ). For the ZOOSCAN analyses, the sample was homogenized and split using a Motoda box until a minimum of 1000 particles 125 were obtained. Then, for the digitalization, the subsample was placed on the glass slide of the ZOOSCAN and the organisms were manually separated using a wooden spike to avoid overlapping. After scanning, the images were processed with the ZooProcess using the image analysis software Image J. Particles were classified automatically into taxonomic categories.
Then the classification was manually verified to ensure that every vignette is in the correct category. Considered living groups of organisms for the ZOOSCAN were copepods, nauplii, crustaceans, appendicularians, gelatinous, chaetognaths and 130 diverse zooplankton (polychaeta, ostracods etc.). Non-organism particles were classified as detritus. Blurs and bubbles were deleted.

Normalized biomass size spectrum
The size spectra were computed for each station using combined FlowCAM and ZOOSCAN data, following Suthers et al. (2006). Firstly, the data were classified in size categories of 0.1 µm of ESD from 0.2 to 2.0 µm. Zooplankton biovolume 135 (mm 3 ) was estimated for each category following the equation: With ESD expressed in mm. The X-axis of the normalized biomass sizespectrum (NBSS) was calculated by dividing the biovolume by the abundance of each category and transformed into Log10. For the Y-axis, the biovolume of each category was divided by the difference in biovolume between two consecutive categories and transformed into Log10. NBSS slope and intercept were determined using linear regression model. 140

Zooplankton carbon demand, respiration and excretion rates
The zooplankton carbon demand (ZCD in mg C m -3 d -1 ) was computed based on estimates of biomass from ZOOSCAN and FlowCAM samples and for estimates of growth rate:

= ×
where B zoo is the biomass of zooplankton in mgC m -3 , calculated using the area-weight relationships from Lehette and Hernández-León (2009) and converted to carbon assuming that carbon represent 40% of the total body dry weight (Omori 145 and Ikeda, 1984). Ration (d -1 ) is defined as the amount of food consumed per unit of biomass per day calculated as: where g z is the growth rate, r is the weight specific respiration and A is assimilation efficiency. g z was calculated following Zhou et al. (2010): as a function of sea water temperature (T, °C), food availability (Ca, mgC m -3 ), estimated from Chl-a, and weight of 150 individuals (w, mg C). We consider here that food is phytoplankton following Calbet et al. (1996). Following Alcaraz et al. (2007) and Nival et al. (1975), values of r and A were 0.16 d -1 and 0.7 respectively. ZCD was compared to the phytoplankton stock, converted to carbon assuming a C:Chl/a ratio of 50:1, and to primary production to estimate the potential clearance of phytoplankton by zooplankton.
Ammonium and phosphorus excretion and oxygen consumption rates were estimated using the multiple regression model by 155 Ikeda et al. (1985) with carbon body weight and temperature as independent variables. Contribution to nutrient regeneration by zooplankton was estimated using the values of primary production and converted to nitrogen and phosphorus requirement using Redfield ratio. Respiration was converted to respiratory carbon lost assuming a respiratory quotient for zooplankton of 0.97 following Ikeda et al. (2000) and used as carbon requirement for zooplankton metabolism.

Data analysis 160
Principal component analysis (PCA) was used to explore spatial patterns of the environmental variables temperature, salinity, dissolved oxygen, using mean values of the layer 0-300 m depth, plus the estimated MLD. Averaged Chl-a values from fluorescence sensor coupled with CTD were also included in this analysis. The data were first normalized and then analyzed using Primer 6.0 software.
A taxonomic group-station matrix with the abundance values was created and then square-root transformed to estimate 165 station similarity using Bray Curtis similarity. The similarity matrix was then ordinated using Nonmetric Multidimenstional Scaling (NMDS). These analyses were performed using Primer 6.0 software.
Rank frequency diagrams (RFD) were created using the data from N 200 to see differences in taxonomic composition between the samples. Potential association between zooplankton data and spatial patterns of the environmental variables were tested using Spearman"s rank-correlations. T-test were used to compare mean values between zones. The 100 um sample of TYR 170 station was discarded due to poor state of preservation of the sample.

Spatial patterns of environmental variables
The Principle Component Analysis (PCA) on environmental data explains 90.3 % of the total variance in the first two axes 175 and delivers three clusters of oceanographic areas plus two distinct stations ( Figure 3). The first axis (62 % of the variance) is mostly influenced by temperature and dissolved oxygen, whereas the second axis (28.3 %) is mostly influenced by MLD, salinity and Chl-a.
The cluster of western stations in the Algerian Basin (AB) includes ST3, ST4, ST9, and FAST which are characterized by low temperature, salinity and MLD values. The cluster located in the Tyrrhenian Basin (TB) comprises (ST5, ST6 and TYR stations) is very close to the first group, but with lower chlorophyll-a concentrations and higher values of temperature and salinity. Eastern stations (ST7, ST8 and ION stations) located in the Ionian Basin (IB) are characterized by the temperature and salinity values and the lowest dissolved oxygen concentrations found during the survey. Stations 1 and 2 on the northsouth transect (NS) do not cluster with any of the other stations due to deeper MLD and higher chlorophyll-a concentrations.
The highest abundances are found in the PB transect and AB, and the highest biomass in the AB region. The averaged total biomass in PB is lower than in AB, due to the very low contribution of the size classes C 1000-2000 and C >2000 , but size classes from C <200 to C 500-1000 present higher biomass values than in AB. In TB, total biomass values decrease between ST4 and ST6, 190 the latter presenting the lowest biomass value of the whole survey. Note that the biomass of TYR is obtained only for the size classes above 500 µm ESD, and the corresponding abundance is comparable to those obtained in ST5 and ST6 for these larger size classes. In IB, total biomass and abundance are lower than in AB and with low variability between stations.
high as at ST2, and between 5 to 13 times higher than the rest of the transect ( Figures 6A and 5). Similarly, Centropages spp. abundance is 10 times higher at ST1 and ST2 than in other stations of the survey. In contrast, abundances of Oithona spp. 215 and Corycaeus spp., are respectively 6 and 10 times lower at ST1 and ST2 than at other stations. The zooplankton community in AB is slightly different from those in TB and IB due to appendicularians and unidentified calanoid copepods being more abundant in AB and to Haloptilus spp. being more abundant in TB and IB. Within TB and IB, the three sampling dates (ION1, ION2, ION3) at ION station form a unique cluster, whereas, ST7 and 8 are grouped with the TB station in another cluster. This differentiation of ST7 and 8 from the ION sampling dates in the NMDS analysis is mainly due to higher 220 relative abundance of small copepods ( higher contributions of small zooplankton compared to large ones, and potentially linked to daily migration of larger forms deeper than 300 m.

Zooplankton community changes at long stations
The RFDs for stations TYR, ST5, ST6, ION and FAST are presented separately in Figures 9A to 9D, and grouped in Figures 230 9E and 9F. As only one sample was done at TYR station, nine days after a large dust deposition event in the Southern Tyrrhenian Sea, RFDs of ST5 and ST6 also sampled in TB (six and twelve days after the dust event, respectively) are added for comparison (Figures 9A and 9B). At all three TB stations, RFDs are characterized by high dominance of herbivorous zooplankton Para/Clausocalanus spp. and Oithona spp. in 1 st and 2 nd position with a strong drop in abundance for the following ranked taxa (undefined calanoid copepods or Corycaeus). Appendicularians drop from the 4 th position at ST5 and 235 TYR to the 10 th position at ST6. The shapes of RFDs change more between ST5 and TYR than between TYR and ST6. In ION station RFD shapes are similar at both sampling dates (ION1 and ION3) with the community dominated by Para/Clausocalanus spp. ( Figure 9C). Corycaeus spp. changes from the 2 nd position to the 4 th , calanoid copepods from 3 rd to 6 th and Oithona spp. from 4 th to 2 nd . Appendicularians occupy a very similar position in both RFDs (6th and 7th rank at ION1 and ION3 respectively). At FAST station, the taxonomic composition is dominated by copepods ( Figure 9D), but the rank 240 order of the most dominant species changes between the two sampling dates (FAST1 and FAST3). Oithona spp. and Para/Clausocalanus spp. have the 1 rst and 2 nd ranks during FAST1, but this order is reversed in FAST 3. The 3 rd place on both days are occupied by calanoid copepods. Appendicularians present one of the most significant changes, with their rank dropping from 4 th to 14 th between the two dates. It is remarkable that the RFDs change from a convex shape at FAST1 to a more concave one at FAST2, influenced by the high dominance of Para/Clausocalanus at the first rank ( Figure 9D). The 245 comparison of the standardized RFDs for all the stations ( Figure 9E) highlights that the greatest change in shape is visible at https://doi.org/10.5194/bg-2020-126 Preprint. Discussion started: 4 May 2020 c Author(s) 2020. CC BY 4.0 License.
FAST, whereas it stays moderate in ION and negligible in TB. Figure 9F is similar to Figure 9E, but without ION, to visualize changes in zooplankton community composition at different time lags after a dust event, and will be commented on in more detail in the Discussion section.
Assuming phytoplankton as the major food source, zooplankton consumption potentially represents15% of the phytoplankton stock on average per day and 97 % of the primary production (see Table 2). ZCD follows the zooplankton biomass pattern with higher values in AB and lower values in TB, and does not increase with primary production (r= -0.18, p>0.05). The average respiration (mean: 83.1 mgC m -2 d -1 and range between 62.9 and 112.2 mgC m -2 d -1 ) corresponds to 255 36.4 % of the integrated primary production. Almost half of this zooplankton respiration is due to organisms smaller than 500 µm of ESD. Mean ammonium excretion is 12.3 mg NH4 m -2 d -1 (range between 9.1 and 17.7 mg NH4 m -2 d -1 ), and mean phosphate excretion 1.7 mg PO4 m -2 d -1 (range between 1.3 to 2.3 PO4 m -2 d -1 ). The potential contributions of excreted nitrogen and phosphorus to primary production are respectively 31.5 % (range between 19.9 to 42.6 %) and 26.3% (range between 19.9 to 42.6%). Zooplankton size classes smaller than 500 µm of ESD contribute 45 % and 47 % of the total 260 ammonium and phosphate excretion respectively. The detailed data is shown in Table 2.

Methodological concerns and the importance of the small zooplankton fraction
This methodology combining two nets (N 100 and N 200 ) and two sample treatments (FlowCAM and ZOOSCAN) enables us to 265 deliver a more accurate mesozooplankton community size spectrum (200-2000 µm), whereas size classes C <200 and C >2000 at the edges of the spectrum range remain under-sampled and require other equipment for proper sampling (respectively bottles and larger mesh size net). The length:width ratio of mesozooplankton organisms is quite variable, from 1 for the nearly round-shaped organisms such as nauplii or cladoceran, to more than 10 for long organisms such as chaetognaths (Pearre, 1982) or some copepods such as Macrosetella gracilis (Böttger-Schnack, 1989), with an average value between 3 and 4 for 270 copepods (Mauchline, 1998). If we consider that organisms with a length:width ratio of 6 caught by the 200 µm mesh size will present an ESD of at least 490 µm, it is consistent that this net quite correctly samples organisms having an ESD above 500 µm ESD. For these organisms (> 500 µm ESD), ZOOSCAN is the most appropriate tool to deliver the size spectrum.
Similarly, the 100 µm mesh size net allows small organisms of width just below 100 µm to pass through, but most of them might have an ESD up to 200 µm because for these smaller sizes, the length:width ratio is mostly below 4 (Mauchline, 275 1998). Due to the threshold of ZOOSCAN at 300 µm ESD, FlowCAM is the best tool to process organisms in the fraction below 500 µm.
Several authors have already highlighted the limitation of the 200 µm mesh size to catch small zooplankton individuals :comparisons of different zooplankton mesh size nets comprised between 60 and 330 µm have systematically shown a decrease in abundance with increasing mesh size (Turner, 2004;Pasternak et al., 2008;Riccardi, 2010;Makabe et al., 2012;280 Altukhovet al., 2015). When the goal of the study is to achieve a full understanding of the complete mesozooplankton community structure and functioning, the size selectivity of the sampling nets is an important issue: clearly, a large fraction of organisms of ESD between 200 and 500 µm is undersampled using a single 200 µm mesh size net. Pasternak et al.(2008) reported that a 220 µm mesh can lose up to 98% of the abundance of Oithona spp. and 80% of copepodite stages of Calanus spp. Riccardi (2010) found that a classical 200 µm net catches only 11% of the abundance and 54 % of the biomass 285 compared to a 80 µm mesh size, leading also to differences in observed species composition in the Venice lagoon. During the PEACETIME survey, the small size classes (C 200-300 and C 300-500 ) of mesozooplankton have been optimally sampled using a 100 µm mesh size net (N 100 ). Consequently, these size classes represent very large percentages of the total abundance (respectively 52.3 and 34.8 %) and a significant contribution to the total biomass (respectively 14.5 and 25.9 %). These reliable estimations have direct consequences for the estimated fluxes (see below). 290

Differences in abundance, biomass and zooplankton community structure in relation to regional environmental characteristics
A review of the most relevant information available on zooplankton biomass and abundance in different regions of the Mediterranean Sea (Table 3) shows a wide range of variation that can be attributed to location, sampling seasons and/or sampling methods (mesh size net, depth of the tow, etc), and in general, the values during PEACETIME survey are in the 295 same order of magnitude, although most of other studies were performed with a 200 µm mesh size net and often over a shallower surface layer. However, during this post-bloom period, no clear regional patterns in abundance and biomass were found, unlike other descriptions showing a north-south and west-east decrease in zooplankton stocks (Dolan et al., 2002, Siokou-Frangou, 2004. In PB, Donoso et al. (2017) and Nival et al. (1975) highlighted a strong variability which is consistent with the strong gradient found between ST1 and ST2 during PEACETIME (see Figure 4). In AB, abundance and 300 biomass values obtained during the survey are similar to those recorded in late spring by Nowaczyk et al. (2011), whereas Riandey et al. (2005) found lower abundance and higher biomass values. However, the latter study focused on high resolution of a mesoscale eddy highlighting an important fine-scale variability of abundance and biomass values. For TB, the data are difficult to compare due to different sampling conditions (net mesh size), depth of tow and sampling season). In IB, all biomass presented in Table 3 are in the same order, but abundances found by Mazzocchi et al. (2003Mazzocchi et al. ( , 2014 are three 305 times lower than those observed during PEACETIME, probably due to a high contribution of C <200 and C 200-300 obtained with N 100 (see Figure 4). In general, the better sampling of small size classes with N 100 should lead to higher abundance values.
However, the comparison of data in Table 3 shows that regional and temporal variability of these values partially masked this benefits.
In PEACETIME, clear regional differences are found both in terms of environmental variables and zooplankton taxonomic 310 composition. ST1 and ST2 are clearly differentiated from all others with deeper MLD, higher chlorophyll-a concentrations and a zooplankton community dominated by typical herbivorous copepods of PB (Centropages, Para/Clausocalanus, Acartia, etc), as mentioned by Gaudy et al.(2003) and Donoso et al. (2017), and characterized by a scarcity of thaliaceans which normally occur in ephemeral and aperiodical patches (Deibel and Paffenhöfer, 2009). AB and TB are very closely related to each other in terms of hydrological features and chlorophyll-a, but slightly differentiated in salinity and 315 zooplankton taxonomy. In AB, 17 days separated the sampling of ST3 and ST4 with that of ST9 and FAST, but despite this time gap, they are very close in terms of hydrological features, chlorophyll-a level and zooplankton community structure. IB is clearly differentiated from these groups in terms of environmental parameters (see Figure 3) due to higher salinity and lower chlorophyll-a, but in terms of zooplankton community the western Ionian stations (ST7 and ST8) present more analogy with TB than with the ION station (see Figure 6). During PEACETIME, the station ION appears clearly separated 320 from ST7 and ST8 located further westwards by a north-south jet ( biological and physical data of the epipelagic zone. For instance, ST1 of PEACETIME characterized by high Chl-a, high zooplankton abundance and dominance of small copepods is clearly located in the "consensual Ligurian Sea Region" sensu Ayata et al. (2018) identified as the most productive of the Mediterranean due to intense deep convection events. Among AB stations, stations 3, 4 and 9 are clearly in the "consensual Algerian region" (Ayata et al., 2018), whereas station FAST 335 corresponds to the "western Algerian heterogeneous region". Among the IB stations, the separation of stations 7 and 8 from the ION stations in terms of zooplankton communities and, to a lesser extent, of environmental variables, also correspond to the distinction between the "consensual North Ionian" region and the western part of the "Ionian Sea region", considered as a heterogeneous region (Ayata et al., 2018).

Estimated zooplankton-mediated fluxes during the PEACETIME survey 340
By using allometric relationships relating zooplankton grazing and metabolic rates to size structure, zooplankton impacts However, estimated grazing rates are in the order of the estimated primary production, which corresponds to the highest 345 range of the values summarized by Siokou-Frangou et al. (2010) for the whole Mediterranean Sea (from 14 to 100 %). Just estimating ZCD on the basis of mesozooplankton alone certainly leads to overestimation of its top-down impact on phytoplankton. In the Mediterranean Sea, the primary production is consumed by a "multivorous web" including microbial and zooplankton components (Siokou-Frangou et al., 2010). Mesozooplankton simultaneously grazes on phytoplankton and heterotrophic prey, such as heterotrophic dinoflagellates (Sherr and Sherr, 2007) or ciliates (Dolan et al., 2002), and might be 350 quite flexible in its feeding strategy depending on composition and size of prey as well as on environmental variables such as turbulence (Kleppel, 1993;Yang en al., 2010). On one hand, a large part of the primary production can be consumed by ciliates (Dolan and Marrasé et al., 1995) but on the other hand, mesozooplankton can consume almost the entire ciliate production (Pitta et al., 2001;Pérez et al., 1997;Zervoudaki et al., 2007) potentially explaining the wide variations of standing stock of ciliates over the Mediterranean Sea (Dolan et al., 1999;Pitta et al., 2001;Dolan et al., 2002). The generally 355 described east-west pattern of decreasing grazing impact (Siokou-Frangou et al., 2010) could not be observed during this study as only one station (ION station) was typical of the Eastern Mediterranean Sea.
Estimated NH 3 and PO 4 excretion rates by mesozooplankton during PEACETIME are consistent with the few observations collected in the Mediterranean Sea (Alcaraz, 1988;Gaudy et al., 2003) and with those obtained at similar latitudes (see review in Hernández-León at al., 2008). From our estimation, zooplankton excretion would contribute 360 respectively to 21 -44 % and 17 -38 % of the N and P requirements for phytoplankton production. In the NWMS,  estimated a zooplankton nitrogen excretion contribution to primary production>40%, whereas Gaudy et al. (2003) reported 31-32 % and 10-100 % N and P contributions. This impact on phytoplankton production can be even greater in proximity to the DCM where zooplankton tends to aggregate fuelling regenerated production (Saiz and Alcaraz, 1990) and enhancing bacterial production (Christaki et al., 1998). Zooplankton grazing impact and nutrient contribution to primary 365 production are higher in the western basin than in the Ionian Sea, mainly linked to variations of zooplankton biomass.
Mean carbon released through zooplankton respiration represents 36 % of the primary production during PEACETIME, which is higher than previous measurements in NWMS (by Alcaraz, 1988 andGaudy et al., 2003) from onboard incubation experiments on zooplankton collected by a 200 µm mesh size net.
Metabolic estimations clearly show that the size fractions < 500 µm (optimally captured with the 100 µm mesh size net) 370 make a significant contribution to the whole mesozooplankton estimated fluxes: 14.9 % of the ZCD is due to organisms <300 µm, and this size class contributes 21 % and 20 % of the total ammonium and phosphate excretion respectively.

Impact of dust deposition on the zooplankton community
In the past years, responses to Saharan dust inputs in marine systems have been mostly studied in microcosm and mesocosm experiments, but more rarely observed in situ. Most studied responses to dust are focused on the microbial biota and are generally marked by an increase in metabolic rates rather than by standing stock changes (probably due to trophic transfer along the food-web) (Ternon et al., 2011;Guieu et al., 2014;Ridame et al., 2014;Herut et al., 2016). In mesocosms, changes in zooplankton stocks are strongly dependent of the initial conditions, and cannot really reflect what could occur in natural waters within the Mediterranean "multivorous planktonic food-web" (Siokou-Frangou et al., 2010). Pitta et al. (2017) found an increase in mesozooplankton biomass 9 days after the beginning of a mesocosm experiment, probably as a result of an 380 earlier increase of prey (flagellates, ciliates and dinoflagellates). Tsagaraki et al. (2017) described an increase in productivity after an artificial dust deposition that was transferred to higher trophic levels by the classical food web, resulting in an increase of copepod egg production 5 days after the beginning of the experiment. Very few in situ studies have documented mesozooplankton responses to Saharan dust. Abundance increase was observed by Thingstad et al. (2005) in the Eastern Mediterranean Sea, and by Hernández-León et al. (2004) in Atlantic waters close to the Canary Islands one week after the 385 deposition. In this latter area, Franchy et al. (2013) detected increases of zooplankton grazing and zooplankton biomass after another event. Thus, the PEACETIME survey dedicated to the tracking of such events was an opportunity to observe real in situ responses.
At station FAST (an opportunistic station after a Saharan dust deposition event), an increase in nitrate (from 50 nM to 120 nM) and phosphate concentrations (from 8 nM to 16 nM) occurred in the mixed layer (pers. comm. C. Guieu), which led to 390 an increase primary production from FAST1 to FAST3, but with no visible changes in phytoplankton biomass or species composition (see Table 2). For zooplankton, the total abundance slightly decreases but the community composition presents obvious changes, mainly a decrease of appendicularians and an increase of Para/Clausocalanus spp. and of carnivorous taxa (Candacia spp., chaetognaths, siphonophores) (see Figure 9D). The sharp decrease of appendicularian abundance (four-fold decrease) and rank position (see Figure 9D) could potentially be linked either to food limitation or to predation. Size and 395 species composition of the phytoplankton community in FAST did not show any change after the dust (pers. comm. J. Uitz), but there were potential increases in food competition with Para/Clausocalanus spp. (Lombard et al., 2010) and/or in predation by chaetognaths and siphonophores (Purcell et al., 2005). Although total zooplankton biomass remains relatively stable at FAST, the contribution of the size classes C 500-1000 and C 1000-2000 increase relative to the smaller size classes (see Figure 4B) inducing variations on the NBSS slope from -0.76 to -0.63 (see Figure 8). This 15% increase in biomass is 400 mainly due to large migrating taxa such as copepods Eucalanus spp., Rhincalanus spp. and Candacia spp., chaetognaths and siphonophores. The daily observation of sediment traps at 200 and 500 meters over five days between FAST1 and FAST3 (pers.comm. C. Guieu) shows a relative increase of swimmers collected at 500 m versus those collected at 200 m, also suggesting increasing numbers of migrants. An obvious planktonic transition occurred during this period but it is difficult to conclude which of the bottom-up (changes in primary producers) or top-down (increase of carnivorous migrants) effects was 405 dominant. The change in the RFDs (Figure 9D), from a convex shape at FAST1, indicating a more stable system with no dominance of the first taxonomic groups, to a more concave shape at FAST3 influenced by the high dominance of Para/Clausocalanus at the first rank, could reflect a disturbance effect (sensu Pinca and Dallot, 1997) of the dust deposition on the zooplankton community.
A synoptic analysis of the RFDs linked to the dust events observed in the Tyrrhenian basin and at FAST station offers a basis 410 for proposing a conceptual model of a virtual time series of zooplankton community responses after a dust deposition event ( Figure 9F): the first sampling occurs before the event (FAST1), and several other samplings are realized with time-lag of five days (FAST3), six days (ST5), nine days (TYR) and twelve days (ST6) after the event. FAST1 represents an initial steady state (state 1) with no dominance in the first taxa ranks, meanwhile FAST3 and ST5 represent a disturbed state of the community (state 2) with strong dominance of the first taxa and the collapse of the following ones. TYR and ST6 represent 415 the beginning of recovery towards a stable system (state 3) with the move up of the second rank. State 1 before the dust event is characterized by oligothropic conditions with low nutrients, low phytoplankton concentration dominated by smallsize cells and their typical zooplankton grazers (e.g. appendicularians and thaliaceans), leading to a convex RFD shape (like FAST1 Figure 9F) reflecting a mature community (sensu Frontier, 1976). State 2 is characterized by a nutrient input linked to the dust event stimulating larger phytoplankton cells and their herbivorous grazers (copepods) and attracting carnivorous 420 migrants leading to a more concave RFD shape (like FAST3, ST5 and TYR Figure 9F) typical of a disturbed community (sensu Frontier, 1976). State 3 is characterized by the diversification of herbivorous taxa leading to changes in RFD towards a convex shape (like ST6 Figure 9F).

Conclusion 425
To our knowledge, PEACETIME was the first study in the Mediterranean Sea that managed to collect zooplankton samples before and soon after natural Saharan dust deposition events and to highlight in situ zooplankton responses in terms of community composition and size structure. Our study suggests that a complete understanding of the mesozooplankton community response to a single massive dust event would require continuous observation over two to three weeks, from an initial state just before the event to a complete process of zooplankton community succession after the event. To identify 430 such a succession, the rank-frequency diagrams of the zooplankton taxonomic structure appears to be a more practical and sensitive index than observable changes in stock (abundance and biomass) or in metabolic rates, and should be further tested.
This approach requires a complete overview of mesozooplankton size spectrum and community composition which was achieved in our study by combining data from two mesh size nets (100 and 200 µm) and two analytical techniques (FlowCAM and ZOOSCAN). In our study, this strategy also enabled us to show the importance of small forms (< 500 µm of 435 ESD) both in terms of stocks and fluxes.

Acknowledgments
This study is a contribution to the PEACETIME project (http://peacetime-project.org), a joint initiative of the MERMEX and ChArMEx components supported by CNRS-INSU, IFREMER, CEA, and Météo-France as part of the programme 440 (LOV) and Karine Desboeufs (LISA). We thank the PEACETIME project coordinators and scientists on board, especially Nagib Bhairy and Cécile Guieu who did the zooplankton sampling. Thanks to E. Maranon and M. Perez-Lorenzo for the PP data and to Julia Uitz, Céline Dimier and the SAPIGH analytical service at the Institut de la Mer de Villefranche (IMEV) for onboard sampling and HPLC analysis. Thanks to Cécile Guieu, Elvira Pullido, France Van Wambeke, and Julia Uitz for 445 critical reading and advice on a first version of the draft, and to Michael Paul for correcting the language. G. Feliú was supported by a Becas-Chile PhD scholarship (CONICYT Government of Chile).

Data availability
All data and metadata will be made available at the French INSU/CNRS LEFE CYBER database (scientific coordinator: Hervé Claustre; data manager, webmaster: Catherine Schmechtig). INSU/CNRSLEFE CYBER (2020) 450

Authors contribution
GF, MP and FC wrote the paper with contributionsby PH. GF participated in the sample treatment. GF, FC, MP and PH participated in the data analysis