Introduction: Process studies at the air–sea interface after atmospheric deposition in the Mediterranean Sea – objectives and strategy of the PEACETIME oceanographic campaign (May–June 2017)

Abstract. In spring, the Mediterranean Sea, a well-stratified low-nutrient–low-chlorophyll region, receives atmospheric deposition by both desert dust from the Sahara and airborne particles from anthropogenic sources. Such deposition translates into a supply of new nutrients and trace metals for the surface waters that likely impact biogeochemical cycles. However, the relative impacts of the processes involved are still far from being assessed in situ. After summarizing the knowledge on dust deposition and its impact on the Mediterranean Sea biogeochemistry, we present in this context the objectives and strategy of the PEACETIME project and cruise. Atmospheric and marine in situ observations and process studies have been conducted in contrasted areas encountering different atmospheric deposition context, including a dust deposition event that our dedicated “fast-action” strategy allowed us to catch. Process studies also include artificial dust seeding experiments conducted on board in large tanks in three ecoregions of the open waters of the Mediterranean Sea for the first time. This paper summarizes the work performed at sea and the type of data acquired in the atmosphere, at the air–sea interface and in the water column. An overview of the results presented in papers of this special issue (and in some others published elsewhere) is presented.


Introduction 34
Understanding the exchange of energy, gases and particles at the ocean-atmosphere interface 35 is critical for the development of robust predictions of future climate change and its 36 consequences on marine ecosystems and the services they provide to society. Our 37 understanding of such exchanges has advanced rapidly over the past decade but we remain 38 unable to adequately parameterize fundamental controlling processes as identified in the new 39 research strategies of the international Surface Ocean-Lower Atmosphere Study group (Law et 40 al., 2013and SOLAS 2015-2025. A critical bottleneck is the 41 parameterization and representation of the key processes brought into play by atmospheric 42 deposition in Low Nutrient Low Chlorophyll (LNLC) regions. A perfect example of a LNLC 43 region, and of the role of the atmospheric deposition, is the Mediterranean Sea where the 44 ecosystem functioning may be modulated by pulsed atmospheric inputs in particular the 45 deposition of Saharan dust (Guieu et al., 2014a) and nutrients of anthropogenic origin (Richon 46 3 Indeed, the Mediterranean quasi-enclosed basin continuously receives anthropogenic aerosols 48 originating from industrial and domestic activities from all around the basin and other parts of 49 Europe, both in the western (Bergametti et al., 1989;Desboeufs et al., 2018) and eastern 50 (Tsapakis et al., 2006;Moon et al., 2016) basin. In addition to this continuous 'background' 51 inputs, the surface of the Mediterranean Sea episodically receives biomass burning particles 52 (Guieu et al., 2005) and Saharan dust (e.g. Loÿe-Pilot et al., 1986, Vincent et al., 2016. Some 53 deposition events are qualified as 'extreme events', as dust inputs as high as 22 g m -2 (event in 54 Nov. 2001 recorded at Ostriconi-Corsica Island, Guieu et al., 2010;event in Feb. 2002 recorded 55 at Cap Ferrat, Bonnet andGuieu, 2006) can occur on very short time scales (hours to days) 56 representing the main annual dust flux. Associated atmospheric deposition of major macro-57 nutrient (N, P) (Kouvarakis et al., 2001;Markaki et al., 2003Markaki et al., , 2010Guieu et al., 2010), of iron 58 (Bonnet and Guieu, 2006) and of trace metals (Theodesi et al., 2010;Guieu et al., 2010;59 Desboeufs et al., 2018) represents significant inputs likely supporting the primary production 60 in surface waters especially during the stratification period (Richon et al., 2018a(Richon et al., , 2018b. 61 Among the atmospheric deposited nutrients, anthropogenic reactive nitrogen is critical on the 62 fluxes of inorganic and organic N (Markaki et al., 2010, Violaki et al., 2010. Soil dust 63 deposition plays an important role on the fluxes of P and trace metals due to the intense but 64 sporadic inputs (Bergametti et al., 1992;Guieu et al., 2010; Morales-Baquero and Perez-sampling of the ION station, conducting to the decision to start the Fast Action. The sequence 283 of events leading to the Fast Action are described later. 284

Atmospheric conditions 285
Several near-real time remote sensing products and model forecasts were used. In terms of 286 aerosol remote sensing, we mainly relied on two products. The first one was the aerosol optical 287 depth at 550 nm (AOD550) distribution over the sea, as produced in near-real time by the ICARE 288 data and service centre, Lille, France (product SEV_AER-OC-L2; http://www.icare.univ-289 lille1.fr/projects/seviri-aerosols; last access 9 Feb., 2020). Data from the Spinning Enhanced  Thieuleux et al. (2005). The MSG satellite position at 0° longitude allows a 295 good coverage for aerosol climatologies and case studies of aerosol transport over the 296 Mediterranean basin (e.g. figure I.19 in Lionello et al., 2012;Chazette et al., 2016 and2019) 297 and surrounding continental regions (Carrer et al., 2014) as well as of desert dust source regions 298 in Africa (e.g. Gonzales and Briottet, 2017). In addition to the quick-look from the level-2 299 product (SEV_AER-OC-L2) available between 4:30 and 18:00 UT at the maximum in mid-300 June in our area of interest, a daily mean level-3 (SEV_AER-OC-D3) is produced every night 301 by averaging all available time slots during the previous day between 4:00 and 19:45 UT. Figure  302 6 illustrates this product for the 3rd of June when an African dust plume from North Africa 303 associated to a cloudy air mass invaded the westernmost Mediterranean basin atmosphere. The 304 horizontal resolution of the product is of 3 x 3 km 2 at nadir, of the order of 12.5 km 2 in the 305 Alboran Sea, 15 km 2 in the North of the Gulf of Genova, and 18 km 2 in the northeasternmost 306 https://doi.org /10.5194/bg-2020-44 Preprint. Discussion started: 28 February 2020 c Author(s) 2020. CC BY 4.0 License. basin (about 13.07,13.64,and 13.96 at the FAST, ION, and TYR station, respectively). 307 Although less accurate than AOD from MODIS when compared to AERONET data, the high 308 temporal resolution of MSG/SEVIRI-derived AOD offers a much better daily coverage of the 309 area than any orbiting satellite (Bréon et al., 2011), especially when partial cloud coverage can 310 be compensated thanks to successive images, as illustrated in figure 6. 311 The second useful remote sensing product was the North African Sand Storm Survey 312 (NASCube) also produced from MSG/SEVIRI, at the Laboratoire d'Optique Atmosphérique, 313 Lille, France (http://nascube.univ-lille1.fr; last access, 9 Feb. 2020). It generates continuous 314 day and night images of desert dust plumes over the northern African continent and Arabian 315 Peninsula, using an artificial neural network methodology producing colour composite images 316 by processing 8 visible, near-infrared and thermal infrared bands of SEVIRI (Gonzales and 317 Briottet, 2017). Figure 7 shows a window of this product for the 1 st June 2017, showing the 318 probable dust source regions (south of Morocco and western Algeria) of the plume found the 319 following days over the westernmost Mediterranean basin as seen in figure 6. 320 During the campaign, we also used on a regular basis air mass trajectories computed with the 321 Hysplit tool of the Air Resources Laboratory of the National Ocean and Atmosphere 322 Administration (NOAA/ARL; https://ready.arl.noaa.gov/HYSPLIT_traj.php; last access 9 Feb. 323 2020; Stein et al., 2015;Rolph et al., 2017) based on global meteorological 192-h forecasts 324 from the Global Forecasting System (GFS) model (1-deg, 3-h resolution) operated by the 325 National Centers for Environmental Prediction (NCEP; Yang et al., 2006). It could be used both 326 in forward mode to forecast the transport over the western Mediterranean of dust plumes 327 detected over Africa by NASCube, and in backward mode to identify the origin of air masses 328 over the ship position. 329 https://doi.org /10.5194/bg-2020-44 Preprint. Discussion started: 28 February 2020 c Author(s) 2020. CC BY 4.0 License.
In addition to aerosol remote sensing observations we also used near real time rainfall remote 330 sensing produced by the Meteo Company, an international weather network 331 (https://meteoradar.co.uk; last access 9 Feb. 2020) providing every 15 mn real time weather 332 radar-and satellite-derived maps of precipitation, clouds, and lightning on a European window 333 covering most of the Mediterranean basin (north of 32°N or 35.5°N, depending on products). 334 The satellite infrared images from SEVIRI are filtered to show the thicker clouds, and 335 observations from 45 European rain radar are integrated. Figure 8 illustrates the combined 336 SEVIRI satellite and radar product showing both clouds, precipitation and lightning for two 337 time slots on 3 June 2017. They show the beginning and the end, respectively, of a convective 338 rainfall of low intensity (<2 mm h-1) between Algeria and Spain in the dusty and cloudy area 339 visible in Figure 6 west of the ship. 340 A number of operational forecast models were also used, both for weather forecast and aerosol 341 transport. In order to understand the synoptic circulation, we especially considered surface 342 pressure (P) and 500-hPa (about 5.5-km altitude) geopotential (Z500) maps over the European 343 domain covering the whole Mediterranean basin and northern Atlantic from the global 344 numerical weather prediction model ARPEGE (Courtier and Geleyn, 1988), developed and 345 maintained at Météo-France. Its horizontal resolution varies from 7.5 km in France to 37 km 346 over Southern Pacific, and four daily forecasts including data assimilation are carried out every 347 day (available by http://www.meteociel.fr, last access 9 Feb. 2020). Because we were 348 especially targeting possible aerosol deposition events, we also analysed daily a set of up to 5-349 days, 1-, 3-, or 6-hourly depending on models, precipitation forecasts from several models 350 including those made available by meteociel.fr including global weather forecast models such 351 as ARPEGE, IFS (the model developed at ECMWF; Barros et al. 1995), the Canadian CMC-352 MRB GEM model (Côté and Gravel, 1998), the GFS atmospheric model from NCEP 353 https://doi.org/10.5194/bg-2020-44 Preprint. Discussion started: 28 February 2020 c Author(s) 2020. CC BY 4.0 License. (Kanamitsu, 1989) and its ensemble GEFS, but also the regional non-hydrostatic model 354 AROME (Seity et al., 2011) for the NW Mediterranean only at 1.3 km resolution. 355 Three regional dust transport models have also been considered, namely SKIRON operated by In terms of dust transport modeling, we mainly relied on 6-hourly dust optical depth and dry 364 and wet dust deposition fluxes forecasted daily from 12 UTC over the next 72 h by the NMMB-365 BSC-Dust and BSC-DREAM8b v2.0 models and over the next 180 h (5.5 d) by SKIRON. 366 Because of its longer temporal range of forecast, the wet dust deposition product by SKIRON 367 was particularly useful to issue an early pre-alert for the Fast Action during the cruise. Figure  368 9 compares the forecast maps of 6-h accumulated dust deposition flux at 4 time steps from 3 rd 369 June 2017 12 UTC to 5 June 00 UTC, from the 2nd June runs of those 3 models. This period 370 corresponds to the scavenging of the dust plume shown in Figure 6 that was targeted for the 371 Fast Action (see below). 372 We also used a set of forecast of aerosol or dust optical depth from a series of models: (i) 60-h,

Ocean conditions 394
Concerning the surface ocean, several remote-sensing datasets were exploited using the 395 SPASSO (Software Package for an Adaptive Satellite-based Sampling for Ocean campaigns 396 https://spasso.mio.osupytheas.fr/; last access: 9 Feb. 2020) in order to guide the cruise through 397 a Lagrangian adaptive sampling-strategy aiming at avoiding region of complex circulation and 398 dynamics (fronts, small scale eddies). The idea behind this approach was to aim at a situation 399 where the air-sea exchanges dominate and lateral advection and diffusion can be neglected. 400 Such an approach was already successfully adopted during several previous cruises such as 401 LATEX (Nencioli et al., 2011;Doglioli et al., 2013, Petrenko et al., 2017, KEOPS2 (d'Ovidio 402 et al., 2015), OUTPACE (Moutin et al., 2017;de Verneil et al., 2018) and OSCAHR (Rousselet 403 et al., 2019). During PEACETIME, we used the following datasets: (1) altimetry data from the 404 AVISO Mediterranean regional product (https://www.aviso.altimetry.fr/data/products/sea-405 surface-height-products/regional/mediterranean-sea-gridded-sea-level-heights-and-derived-406 variables.html); the altimetry-derived currents were then processed by SPASSO to derive 407 Eulerian and Lagrangian diagnostics of ocean circulation: Okubo-Weiss parameter, particle 408 retention time and advection, Finite Size Lyapunov Exposant (e.g. figure 10); (2) (Table 1) and 460 the ocean and atmospheric sampling started immediately. 461 Although cloudy, only from the 3 rd of June rain conditions were observed in the neighbouring 462 area (see rain radar composite images in figure 11). The SEVIRI AOD remote sensing 463 confirmed the export of a dust plume from North Africa south of the Balearic Islands with high 464 AOD (>0.8; Figure 6) and NASCube confirmed new dust emissions in the night from 3 to 4 465 June. The dust plume was transported to the NE up to Sardinia on June 4, with AOD <0.5 in all 466 the area and clear sky with low AOD was left west of 4°E on June 5 (Desboeufs et al.,in 467 preparation, this issue). Wet deposition of dust for the 4 th and early 5 th June in the FAST station 468 area were confirmed by the deposition maps from the 3 regional dust transport model forecast 469 runs of June 2, although with decreasing intensity compared to the previous runs (except for 470 BSC-DREAM8b that did not forecast dust wet deposition in earlier run), from a very small flux 471 of a few mg m -2 (BSC-DREAM8b) to about 100 mg m -2 (NMMB-BSC) and up to more than 1 472 its below-cloud deposition during the rain event of early 5th June (Desboeufs et al.,in 478 preparation, this issue). The chemical composition of this rain sample confirmed wet deposition 479 of dust reaching a total particulate flux of 12 mg.m -2 (Fu et al., in preparation, this issue), which 480 is among the lowest most intense dust deposition fluxes recorded in this area from long time-481 series of deposition network (Vincent et al., 2016). 482

22
The post-cruise comparison of the hull-mounted ADCP data combined with the SVP drifters 497 trajectories and the altimery-derived currents shows a good agreement all along the cruise and 498 in particular at the FAST station ( Figure 12). Moreover, the agreement between SVP and 499 numerical particle trajectories has been slightly improved when we also took into account the 500 Ekman drift calculated with wind data from the high resolution regional model WRF 3.7. 501 This allowed us to calculate backward trajectories of the surface water masses using the 502 ARIANE Lagrangian tool (Blanke and Raynaud, 1997;Blanke et al., 1999) in order to estimate 503 the origins of the sampled surface water at the LD stations. As seen from the repeated CTDs 504 (see previous section), at the FAST station a southward current associated to the large Algerian 505 eddy was present. We estimated that over the whole station duration, a mean value of 57(26)% 506 of water remained in the station zone after 1(2) day(s). Moreover, combining the particle 507 trajectories and the precipitation data from the WRF 3.7 model we concluded that the rain, 508 which fell slightly upstream the LD-FAST station in the previous days, likely impacted the 509 sampled water mass ( figure 14). 510 All along the cruise, the work at sea was divided between short (~8 hours) and long (up to five 573 days) stations to allow both a good description of the different ecoregions crossed and to 574 perform process studies. The number of short stations was the best compromise in order to (1)  575 allow at least 8 hours of transect at 9 knots in between 2 short stations, necessary for a good 576 continuous monitoring of both low atmosphere and surface waters while cruising and (2)