Green Edge ice camp campaigns: understanding the processes controlling the under-ice Arctic phytoplankton spring bloom

. The Green Edge initiative was developed to investigate the processes controlling the primary productivity and the fate of organic matter produced during the Arctic phytoplankton spring bloom (PSB) and to determine its role in the ecosystem. Two ﬁeld campaigns were conducted in 2015 and 2016 at an ice camp located on landfast sea ice southeast of Qikiqtarjuaq Island in Baﬃn Bay (67.4797N, 63.7895W). During both expeditions, a large suite of physical, chemical and biological variables was measured beneath a consolidated sea ice cover from the surface 5 to the bottom at 360 m depth to better understand the factors driving the PSB. Key variables such as temperature, salinity, radiance, irradiance, nutrient concentrations, chlorophyll-a concentration, bacteria, phytoplankton and zooplankton abundance and taxonomy, carbon stocks and ﬂuxes were routinely measured at the ice camp. Here, we present the results of a joint eﬀort to tidy and standardize the collected data sets that will facilitate their reuse in other Arctic studies. The dataset is available at https://www.seanoe.org/data/00487/59892/ (Massicotte 10 et al., 2019a). the LEFE-CYBER online repository. A comprehensive discussion about nutrient dynamics during the Green Edge missions can be found in Grondin et al. (2019). Joannie Ferland, Marie-Hélène Forget, Erin Reimer, and Pierre-Luc Grondin contributed to the measurements. For ap, Atsushi Matsuoka, Céline Dimier, Léo Lacour, Joséphine Ras, Mathieu Ardyna, Henry Bittig, Blanche St-Béat and Thomas Lacour contributed to the measurements. In 2015, particulate spectral absorption was also done by Lisa Matthes, Christine Quiring and Jens Ehn. Nicole Pogorzelec (who also did snow and ice salinity and overall chl-a ﬁltrations in the ﬁeld lab). Marie-Pier Amyot worked on tidying and uniformizing 345 the data. Martí Galí ran the radiative transfer calculations and compared them to irradiance measurements taken on the ice camps. Lisa Matthes, Simon Lambert-Girard, Bob Hodgson, Jens Ehn, Nicole Pogorzelec and CJ Mundy designed and/or carried out the TriOS and ROV under-ice irradiance measurements Christos Panagiotopoulos and Richard Sempéré coordinated the sampling strategy for sugars/DOC and the analyses. Remi Amiraux collected the samples. Between October 2014 and July 2016, Éric Brossier and France Pinczon du Sel conducted measurements, collected clams, maintained equipment, kept a time-lapse 350 photography record and represented the Greenedge team in Qikiqtarjuaq outside of the sampling season. Debra Christiansen Stowe coordinated logistics in Qikiqtarjuaq, in support of the 2016 ice camp. Makoto Sampei designed and curried copepods incubations to collect fecal pellets out at the ice camp in 2016. Makoto Sampei made microscopic observations on the collected fecal pellets in the laboratory. Sea ice and snow hemispherical directional reﬂectance were measured on the ice camp in 2015 by Sabine Marty and Clémence Goyens. The set-up was designed by Sabine Marty, Edouard Leymarie, Simon Bélanger and 355 Clémence Goyens. They also processed and analyzed the data. Acknowledgements. The GreenEdge project is funded by the following French and Canadian programs and agencies: ANR (Con-tract #111112), CNES (project #131425), IPEV (project #1164), CSA, Fondation Total, ArcticNet, LEFE and the French Arctic Initiative (GreenEdge project). This project would not have been possible without the support of the Hamlet of Qikiqtarjuaq and the mem- 360 bers of the community as well as the Inuksuit School and its Principal Jacqueline Arsenault. The project was conducted under the scientiﬁc coordination of the Canada Excellence Research Chair in Remote Sensing of Canada’s new Arctic frontier and the CNRS & Université Laval Takuvik Joint International laboratory (UMI3376). The ﬁeld campaign was successful thanks to the contribution of A. Wells, M. Benoît-Gagné, and E. Devred from the Takuvik laboratory as well as R. Hodgson from the University of Manitoba. Pascale Bouruet-Aubertot and Yannis Cuypers who provided the SCAMP and contributed to the processing, quality 365 control, analysis and interpretation of the data. We also thank Michel Gosselin, Québec-Océan, the CCGS Amundsen and the Polar Continental Shelf Program for their in-kind contribution to the logistic and scientiﬁc equipment. Thanks to Etienne Ouellet for IT support and data infrastructure management.


Introduction
In the Arctic Ocean, the phytoplankton spring bloom (PSB) initiates the period of highest biomass primary production of the year (Sakshaug, 2004;Perrette et al., 2011;Ardyna et al., 2013). Although it was discovered that the PSB may occur more extensively and more frequently beneath a consolidated ice-pack (Arrigo et al., 2012(Arrigo et al., , 2014Assmy 15 et al., 2017), only a small number of research initiatives (e.g., Fortier et al., 2002;Galindo et al., 2014;Mundy et al., 2009Mundy et al., , 2014Wassmann et al., 1999;Gosselin et al., 1997) have been investigating the processes controlling the Arctic PSB in the ice-covered water column. Additionally, ice algal communities play an important role within the Arctic food web and for the carbon export to the benthos during the winter-spring transition (Leu et al., 2015). However, primary production within the Arctic ice-pack is still poorly understood. The Green Edge project was conceived 20 in an effort to better understand the Arctic PSB from the level of fundamental physical, chemical and biological processes to that of their interactions within the ecosystem, and at spatial scales ranging from local to pan-Arctic.
Besides studying each major component of the processes controlling Arctic PSB, another objective of Green Edge was to investigate its impact on the nutrient and carbon dynamics within the ecosystem. A total of three Green Edge campaigns were conducted: two ice camp campaigns on landfast sea ice in 2015 and 2016, and an oceanographic 25 cruise aboard the CCGS Amundsen in Baffin Bay in 2016. In this article, we present an overview of an extensive and comprehensive data set acquired during two surveys conducted at the Green Edge ice camp.

Study area, environmental conditions and sampling strategy
The field campaigns were conducted on landfast sea ice southeast of the Qikiqtarjuaq Island in Baffin Bay (67.4797N, 63.7895W, Fig. 1) in 2015 (April 24 -July 17) and in 2016 (April 20 to July 27). These periods were chosen in order 30 to capture the dynamics of the sea-ice algae and phytoplankton spring blooms, from bloom initiation to termination. The field operations took place at a location (the "ice camp") south of the Qikiqtarjuaq Island where the water depth is 360 m. Continuous records of wind speed and air temperature were made with a meteorological station (Automated Meteo Mat equipped with temperature (HC2S3) and wind (05305-L) sensors (Campbell Scientific) positioned near (< 100 m) the tent (Polarhaven, Weatherhaven) in which water sampling was carried out. During the 35 sampling periods, the study site experienced changes in snow cover and ice thickness (Fig. 2). In 2015, the snow and ice thickness varied between 2-40 cm (mean = 21 cm) and 103-136 cm (mean = 121) respectively. In 2016, the snow and ice thickness varied between 0.3-49 cm (mean = 19 cm) and 106-149 cm (mean = 128 cm) respectively.  3 Data quality control and data processing 45 Different quality control procedures were adopted to ensure the integrity of the data. First, the raw data were visually screened to eliminate errors originating from the measurement devices, including sensors (systematic or random) and errors inherent from measurement procedures and methods. Statistical summaries such as average, standard deviation and range were computed to detect and remove anomalous values in the data. Then, data were checked for duplicates and remaining outliers. Once raw measurements were cleaned, data were structured and 50 regrouped into plain text comma-separated (CSV) files. Each of these files was constructed to gather variables of the same nature (ex.: nutrients). In each of these files, a minimum number of variables (columns) were always included so the different data sets can be easily merged together (Table 1). More than 120 different variables have been measured during the Green Edge landfast-ice expeditions. The complete list of variables is presented in Table   2 and detailed metadata information can be found on the LEFE-CYBER online repository http://www.obs-vlfr.fr/ 55 proof/php/GREENEDGE/greenedge.php. The processed and tidied version of the data is hosted at SEANOE (SEA scieNtific Open data Edition) under the CC-BY license (https://www.seanoe.org/data/00487/59892/, Massicotte et al. (2019a)). In the following sections, we present a subset of these variables along with the methods used to collect  . Temporal evolution of the salinity in the first 100 meters of the water column for both campaigns. Note that for visualization, salinity below 31.5 g kg -1 have been binned to 31.5 g kg -1 . Note that salinity as low as 4 g kg -1 was observed during flushes of freshwater at the ocean surface due to snow/ice melt (dark blue color in the figure).
Ocean current profiles in the water column were measured using a downward-looking 300 kHz Sentinel Workhorse each day, properties (assuming they follow a conservative mixing behaviour) will appear to be vertically displaced.
Therefore, when comparing properties from vertical profiles taken at the ice camp, we suggest that comparisons of profile variables should be made on isopycnal (constant density) coordinates, rather than depth coordinates (Fig.   4).  auger holes that had been drilled at distances of 82 and 113 m from the tent and cleaned of ice chunks. Once the ICE-Pro was underneath the ice layer, fresh clean snow was shovelled back into the hole to avoid, as much as possible, having a bright spot above the sensors (see supplementary Fig. B1 and Table B1). The frame was then 100 manually lowered at a rate of approximately 0.3 m s -1 . The above-surface reference sensor was fixed on a steady tripod installed approximately 2 m above the ice surface and above all neighbouring camp features. Data processing and validation were performed using a protocol inspired by that of Smith1984, which is now used by several space agencies. Measurements were taken between 380 and 875 nm at 19 discrete spectral wavebands. Vertical

Underwater photos and videos of ice bottom
Several vertical profiles to 30 m were performed using a GoPro Hero 4 camera mounted on the ICE-Pro and pointing up, towards the ice bottom (see Fig. B1 and Table B1). Still images were captured every five seconds during descent, as well as a video was taken of the complete descent. These photos and videos were used for a qualitative assessment of the pronounced spatial and temporal heterogeneity of the under-ice environment and the associated 115 water column nekton community between the two profiling locations.

Irradiance measurements with TriOS
To quantify the impact of the heterogeneous radiation field under sea ice on irradiance measurements, replicated spectral irradiance profiles were collected beneath landfast sea ice from 5 May to 8 June 2015 and from 14 June to 4 July 2016. The replicates were made on each sampling day, under different surface conditions. In 2015, measure-120 ments were performed prior to melt onset, under different snow depths. In 2016, measurements began after the onset of snowmelt and were performed beneath sea ice with a wet snow cover, shallow melt ponds and white ice.
The deployed sensor array consisted of a surface reference radiometer, which recorded incident downwelling pla-

Inherent optical properties (IOP)
IOPs measurements were made using an optical frame equipped with the physical and bio-optical sensors that were factory calibrated before each field campaign. A Seabird SBE-9 CTD measured temperature, salinity, and pres-140 sure. A WetLabs AC-S was used for spectral beam attenuation (c, m -1 ) and total absorption (a, m -1 ) between 405 and 740 nm, and a BB9 (WetLabs) and a BB3 (WetLabs) were utilized for backscattering coefficients (bb, m -1 ) between 440 and 870 nm. During both campaigns, pure water calibration was performed for the AC-S sensor on each sampling day and linear regression as a function of time was computed for each wavelength of absorption and attenuation signals. Then, the offset applied during the data processing was taken on this linear regression at the exact date 145 of the measurement. Figure 7 shows two vertical profiles of attenuation coefficients at different wavelengths acquired during pre-bloom and bloom conditions in 2016. One can see that during the bloom, attenuation increased markedly in the 0-50 m surface layer due to higher phytoplankton biomass.

Other optical measurements
Other optical variables measured during both field campaigns included absorbance of particulate matter, ab-150 sorbance of dissolved organic matter, snow and sea-ice transmittance, snow/ice hyperspectral and hyperangular hemispherical-directional-reflectance (Goyens et al., 2018) and surface spectral albedo (Table 2). Downwelling spectral irradiance above the surface (1 • ×1 • ) spatial resolution, daily temporal resolution, interpolated hourly) was also computed based on the radiative transfer model SBDART (Ricchiazzi et al., 1998)

Nutrients
Nitrate, nitrite, phosphate and silicate concentrations were measured from water filtered through 0.7 µm Whatman GF/F filters and through 0.2 µm cellulose acetate membranes. Filtrates were collected into sterile 20 mL polyethylene vials, poisoned with 100 µL of mercuric chloride (60 mg L -1 ) and subsequently stored in the dark prior to analysis. Nutrient concentrations were determined using an automated colorimetric procedure described in Aminot 160 and Kérouel (2007). Figure 8 shows an overview of the dynamics of nitrate which is often the limiting nutrient for phytoplankton growth in the ocean (Tremblay and Gagnon, 2009). It can be seen that the depletion of the nitrates started approximately mid-June for both years, coinciding with the initiation of the phytoplankton bloom. However, the depletion was observed deeper in the water column in 2016 compared to 2015 due to stronger currents and a longer sampling period in 2016 . Other nutrients such as dissolved organic and inorganic car-165 bon (DOC/DIC), particulate organic and inorganic carbon (POC/PIC), total organic carbon (TOC), phosphate (PO4), orthosilicic acid (Si(OH) 4 ), and ammonium (NH 4 ), were also measured during both campaigns (Table 2)

Flow cytometry
The abundances of pico-phytoplankton, nano-phytoplankton and bacteria were measured by flow cytometry. Samples (1.5 mL) were preserved with a mix of glutaraldehyde and Pluronic (Marie et al., 2014) and frozen at -80 • C. Samples were analyzed on a FACS Canto flow cytometer (Becton Dickinson) in the laboratory at the Station Biologique de Roscoff. The abundance (cells mL -1 ) of phytoplankton populations was determined on unstained samples and cells 175 were discriminated by their red chlorophyll autofluorescence. Bacterial abundance was determined based on the fluorescence of SYBR Green-stained DNA (Marie et al., 1997). In both 2015 and 2016, bacteria concentrations were initially low, of the order of 100 000 cells mL -1 , and quite uniform throughout the water column. During the bloom, bacterial abundance increased continuously, reaching values of one million cells mL -1 (Fig. 9). Simultaneously, the distribution of highest abundance became stratified with a higher concentration found near the surface in early July 180 before it moved down to the subsurface (between 10 and 20 m) later in July (Fig. 9). In 2015, the sampling period did not extend long enough to capture the full progression of bacterial community development.

Chlorophyll a
Chl a and accessory pigments concentrations were determined by high-performance liquid chromatography (HPLC) 185 following Ras2008. Concentrations were measured using volumes between 0.1 and 1 L of melted ice and volumes between 1 and 2.5 L of seawater. Water was filtered onto Whatman GF/F 25 mm filters and stored at -80 • C until analysis. Filters were extracted in 100% methanol, disrupted by sonication and clarified by filtration. Pigments were analyzed using an Agilent Technologies 1200 Series system with a narrow reversed-phase C8 Zorbax Eclipse XDB column (150 × 3 mm, 3.5 µm particle size) which was maintained at 60 • C. Figure 10 shows the temporal evolution of Primary production during the phytoplankton bloom was incompletely sampled in 2015, while in 2016 it was monitored from the onset under melting sea ice in May to its termination in July (Fig. 11). During the ice-covered period in 2015, primary production, as well as nitrate assimilation (rNO 3 ), occurred at very low but detectable rates reaching 8 and 0.4 mmol m -2 d -1 , respectively. Phytoplankton production rates were higher in the ice than in the 200 water column, representing approximately 80% and 40% for primary production and rNO 3 , respectively. Estimated assimilated concentrations of total carbon and nitrate within the ice cover were 30-96 and 1.4-4.6 mmol m-2 during this period. The break-up of the sea ice cover was characterized by a rapid increase in primary production and rNO 3 . During this period of high light transmission through the melting ice cover (day 169 to 190), concentrations of assimilated total carbon and rNO 3 reached 60 and 8 mmol m-2, respectively, leading to a complete nitrate depletion. 205 The quantities of total carbon and nitrate assimilated during the 2016 PSB in the water column were 562 and 97 mmol m -2 , respectively.

Phytoplankton taxonomy
The phytoplankton community species composition was determined using an Imaging FlowCytobot (IFCB, Woods Hole Oceanographic Institute, Sosik and Olson (2007), Olson and Sosik (2007)). The size range targeted was be-210 tween 1 and 150 µm, while the image resolution of approximately 3.4 pixels µm -1 limited the identification of cell < 10 µm to broad functional groups. A 150 µm Nitex mesh was used to avoid clogging of the fluidics system by large particles, although this might have induced a bias in the results by preventing large cells to be sampled.
For each melted ice and seawater sample, 5 mL were analyzed and Milli-Q water was run between samples with high biomass in order to prevent contamination between samples. Image acquisition was triggered by chl a in vivo 215 fluorescence, with excitation and emission wavelengths of 635 and 680 nm, respectively. Grayscale images were processed to extract regions of interest (ROIs) and their associated features (e.g.: geometry, shape, symmetry, texture, etc.), using a custom made MATLAB (2013b) code , Olson and Sosik (2007); processing codes are available at https://github.com/hsosik/ifcb-analysis). A total of 231 features (see the full list and description at https://github.com/hsosik/ifcb-analysis/wiki/feature-file-documentation) were derived on the resulting ROIs 220 and were used for automatic classification using random forest algorithms with the EcoTaxa application (Picheral et al., 2017). A learning set was manually prepared for each year, with ca. 20 000 images annotated and used for  (Fig. 12). As it was impossible to count the 225 number of cells in each image, we assumed one cell per image. To account for potential underestimations of cell abundance when colonies or chains were imaged, the biovolume of each living protist on images was computed during image processing according to Moberg and Sosik (2012). Using carbon to volume ratios from Menden-Deuer and Lessard (2000), biovolume was converted into carbon estimates, as described in Laney and Sosik (2014). Detailed information about sea ice algae and phytoplankton community composition can be found in Grondin et al.

Physiology of the phytoplankton community
The photosynthetic potential of microalgae was assessed by measuring F v/F m, namely the maximum photochemical efficiency of Photosystem II (PSII), via dynamic chl a fluorescence: In addition to the photosynthetic potential of microalgae, photosynthetic parameters were measured from seawater incubated at different irradiance levels in the presence of 14 C labelled sodium bicarbonate. The light saturation parameter, E k , is an indication of the physiological state of the phytoplankton community. Figure 13B shows

Other data
An exhaustive list of all measured variables is presented in Table 2 along with contact information of principal 285 investigators associated with each measured parameter.

Recommendations and lessons learned
As with any Arctic surveys, a large number of measurements were acquired during the Green Edge project. Although initial recommendations on good practices about collection, processing and storage of collected data were communicated to all scientists, extensive efforts, such as data standardization, had to be performed to assemble the data. 290 It is important for reducing possible errors, that a uniformized data management plan should be prepared and distributed prior to each mission. Furthermore, dedicated data management specialists should be involved from the beginning of the project to ensure the data are adequately collected, tidied, stored, backed up and archived.

Conclusions
The comprehensive data set assembled during both Green Edge ice-camp campaigns allowed us to study the fun-295 damental physical, chemical and biological processes controlling the Arctic PSB. In this paper, only a handful of variables have been presented. The reader can find the complete list of measured variables in Table 2, all of which are also fully available in the data repository. Furthermore, a collection of scientific research papers is currently being submitted to a special issue of the Elementa journal entitled Green Edge -The phytoplankton spring bloom in the Arctic Ocean: past, present and future response to climate variations, and impact on carbon fluxes and the marine 300 food web. The uniqueness and comprehensiveness of this data set offer more opportunities to reuse it for other applications.

Code and data availability
The raw data provided by all the researchers, as well as metadata, are available on the LEFE-CYBER repository (http://www.obs-vlfr.fr/proof/php/GREENEDGE/greenedge.php). The data presented in this paper and in Table 2 Massicotte et al. (2019a)). Detailed metadata are associated with each file including the principal investigator's contact information. For specific questions, please contact the principal investigator associated with the data (see Table   2). Latitude of the sampling location (degree decimals). longitude Longitude of the sampling location (degree decimals). sample_type Origin of the water ("water", "ice", "meltpond"). sample_source Source of the water ("niskin", "underice" "0-1 cm", "0-3 cm", "3-10 cm", "rosette" ). depth_m Depth at which measurement was made. snow_thickness Qualitative value describing the snow cover under which measurement was made ("thin_snow", "thick_snow"). mission Mission identifier ("ice_camp_2015", "ice_camp_2016") pi Name(s) of the principal investigator(s) responsible of the measured variable.       Hours of the day Tide height (m)