The CEO is an array of closely colocated subsurface moorings in the northeast Chukchi Sea (71^∘35.976^′ N, 161^∘31.621^′ W; Figs. 2–4; https://www.youtube.com/watch?v=ypmzTuAQ98k&feature=youtu.be, last access: 9 November 2018). The CEO is situated in 45 m of water on the southeastern flank of Hanna Shoal within a productive biological “hotspot” (Grebmeier et al., 2015). The shoal's shallow depths result in deep ice keel groundings (Barrett and Stringer, 1978) and the accumulation of thick ice, which serves as an important habitat for walrus and other animals (Jay et al., 2012). The exact observatory siting and our ecological understanding of the greater region are based on many years of multidisciplinary sampling on the NE Chukchi Shelf, including those of the Chukchi Sea Environmental Studies Program (Day et al., 2013), the Chukchi Offshore Monitoring in Drilling Area (Dunton et al., 2014), and the Distributed Biological Observatory (DBO; Moore and Grebmeier, 2018).

The CEO moorings carry sensors that collectively measure an extensive suite of physical, biogeochemical, and biological variables (Fig. 4). These sensors allow us to observe and understand the phenology and connections within this Arctic marine ecosystem. The sensors capture temporal variations in sea ice cover and thickness, light, currents, waves, water column structure, and concentrations of dissolved oxygen, nitrate, inorganic carbon species, and particulate matter. They document the presence of phytoplankton blooms and export, zooplankton abundance and vertical migration, the presence of Arctic cod and other fishes, and the vocalizations of marine mammals. The CEO is designed to monitor the ecosystem year-round, making it well-suited for studying interactions among ecosystem components, especially during the poorly documented winter months. Although to our knowledge, no other single Arctic monitoring site measures the full suite of variables collected by the CEO, of course many of the individual measurements are also made elsewhere across the Arctic. With time, we expect that insights derived from the CEO observations will be extended to other Arctic shelf ecosystems and trigger new comparative studies.

The artist's depiction of the Hanna Shoal ecosystem (Fig. 3) illustrates the seasonal cycle at the CEO site (Fig. 3A). Moving from left to right, Fig. 3 captures seasonal shifts from the ice-covered winter, into the productive summer, and finally into the stormy and biologically senescent autumn.

Through heat loss, sea ice formation, and brine rejection (Fig. 3B) in late fall and winter, the water column over the Chukchi Shelf becomes more saline and vertically homogenized (Weingartner et al., 2005), deepening the mixed layer depth to a maximum of ∼36 m in March (Peralta-Ferriz and Woodgate, 2015). Although nutrients are in abundant supply from the incoming Anadyr-origin waters, planktonic production remains limited due to the scarcity of light (Fig. 3C). Over the course of winter, continued heat loss to the atmosphere leads to thermodynamic thickening of the sea ice, while convergence of the ice pack leads to mechanical thickening in the form of pressure ridges. Divergence of the ice pack creates open water in the form of leads and polynyas, which during winter will freeze and thicken through the same processes. As the light begins to return in the spring, diatoms and other algae begin to bloom within the ice matrix and at the ice–water interface (Fig. 3D; Ambrose et al., 2005; Gradinger, 2009).

It is not until late May or June, when the days are long, insolation is strong, and warm water moves in from the south that sea ice begins to melt, thin, and recede northwards (Fig. 3E). During this time, the water column stratifies with inputs of fresh meltwater and heat at the surface (Fig. 3G), leading to a shoaling of the mixed layer depth to a minimum of ∼12 m (Peralta-Ferriz and Woodgate, 2015). This is the time when extraordinary phytoplankton blooms occur in the nutrient-rich surface waters (Fig. 3F; Hill et al., 2018). These processes set up strong vertical gradients of inorganic carbon and nutrients across the shallow water column (Fig. 3H; Bates, 2006). These high rates of primary production support large fluxes of sinking particulate organic matter to the seafloor (Fig. 3I, Lalande et al., 2007), thereby sustaining a rich benthic ecosystem (Fig. 3J; Grebmeier et al., 2006, 2015), which attracts large numbers of marine mammals that forage on the benthos (Fig. 3K; Jay et al., 2012; Hannay et al., 2013) or Arctic cod (Fig. 3L).

Fall is characterized by surface cooling and more frequent and intense storm systems with strong winds (Fig. 3M) that erode the highly stratified water column. This process brings remineralized nutrients and inorganic carbon from bottom waters into the surface layer, supporting fall phytoplankton blooms and the outgassing of carbon dioxide into the atmosphere (Else et al., 2012; Hauri et al., 2013). Later, as sunlight fades into the darkness of winter, primary production further slows, the planktonic ecosystem becomes senescent (Fig. 3N), and the benthos continues to thrive on organic matter stored in the sediments (Pirtle-Levy et al., 2009).

Due to the logistical complexities of operating in the region, most of the observational work done in the Pacific sector of the Arctic Ocean takes place during the sea-ice-free summer and early autumn via research vessels (Fig. 3O). Autonomous vehicles such as gliders have also found increasing use in recent years (Fig. 3P; Baumgartner et al., 2014; Martini et al., 2016; Danielson et al., 2017).

Figure 4Illustration of the Chukchi Ecosystem Observatory (CEO), including the (a) biophysical, (b) biogeochemical, and (c) freeze-up moorings. Note that the freeze-up mooring includes a surface package, whereas the other two moorings only reach up to 33 m depth. Sensors are calibrated with as many in situ samples as possible. Calibration samples are always collected upon deployment and recovery of the moorings and, depending on other research activity nearby our site, also at other times of the year. pCO₂ and pH sensors were added in summer 2016.

Download

A starting premise of our effort to improve the understanding of this complex ecosystem and monitor ongoing changes is that it is necessary to extend observations of the ecosystem into the ice-covered winter and employ new observational approaches that are appropriate for this challenging environment.

Given the presence of deep ice keels that regularly exceed 20 m depth – and may occasionally extend as deep as 30 m – we restrict the uppermost sensor package of our observatory to 33 m below the surface, leaving only 12 m of the water column safe for mooring instrumentation and hardware. Although we cannot deploy instruments in the upper 30 m of the water column when sea ice is present, upward looking acoustic instruments in the array provide observations above the top-mooring package. An Acoustic Zooplankton Fish Profiler (AZFP, manufactured by ASL Environmental Sciences) measures the presence and abundance of zooplankton and fish and ice draft, while a TeledyneRDI Acoustic Doppler Current Profiler (ADCP) records current velocity and direction. During sea-ice-free conditions, the ADCP instrument also quantifies the height, period, and direction of surface waves.

One example of purpose-built technology for the CEO is a novel “freeze-up detection mooring”, which measures upper water column stratification and heat content through the fall up to the time of freeze-up in ice-covered seas (Fig. 4c). It was first deployed in fall 2015. The freeze-up detection mooring was outfitted with an expendable surface float that housed a satellite communications package, a tether release, an inductive modem, and a sea surface temperature sensor. The surface float was connected to four Sea-Bird SBE 37 inductive modem conductivity–temperature–depth sensors (CTDs) that transmitted hourly temperature, salinity and pressure to the surface float from four subsurface depths (8, 20, 30, and 40 m), along with a subsurface camera that records and sends digital images of the upper water column. The advance of the fall ice pack was closely monitored with satellite imagery and the surface float provided simultaneous real-time monitoring of the temperature and salinity throughout the water column leading up to ice formation. When the ice edge was within 1 day of overrunning the mooring and sparse ice chunks were already floating by, the surface float was remotely released from the mooring, leaving a mid-depth subsurface float to provide floatation for the portion left behind. The data from this mooring are presented and discussed below (Fig. 5).

Cold seawater (temperatures below 0 ^∘C for most of the year) decreases the capacity of all batteries. Some instruments are powered with lithium batteries that provide a higher power density. Engineering constraints dictate the trade-offs between the various sensor battery packs and the desired sampling rates. For example, due to the large power demand of the Kongsberg Contros HydroC pCO₂ sensor, its sampling rate had to be decreased to once every 12 h but the AZFP instrument has sufficient power and memory to sample every 15 s for the entire year.

A dedicated vessel charter for servicing the remote CEO is also not cost-effective given the 2000 km distance to the nearest deep water, year-round ice-free port Dutch Harbor in Unalaska, Alaska. We thereby rely on vessels of opportunity during the summer months to deploy and recover the CEO as part of other funded shipboard research. By partnering with other oceanographic research teams that are operating in the region, we are also able to collect water samples from the CEO site to provide for in situ calibration of sensors throughout the deployment. Furthermore, ship-based observational efforts such as the Arctic Marine Biodiversity Observatory Network and the DBO programs add spatial context to the CEO data. Conversely, the CEO can help place research cruise data into a broader temporal context, including variability on scales ranging from the synoptic to the interannual (Danielson et al., 2017).

Despite design advances and features, limitations and challenges still remain to be overcome. For example, without real-time data communication capabilities, instrument function and data returns can only be assessed annually following the CEO turnaround. This delays data availability and makes the approach poorly suited for adaptive sampling efforts after deployment. Furthermore, throughout the winter, there are currently no ship-based efforts or autonomous vehicles operating in the region. The result is limited spatial context during winter and early spring, and an inability to collect samples for frequent calibrations of the CEO's deployed sensors. Many of the measurements made from the CEO are collected at a single depth within the water column, thus limiting the interpretation of upper water column variables, especially over the winter months. Future innovations and investments in technologies such as profiling winches, direct-to-shore submarine cable communications, or under-ice autonomous assets are a few possibilities that could mitigate some of these challenges.

Figure 5Time series from the NOAA-operated Wiley Post-Will Rogers Memorial Airport weather station and the Chukchi Ecosystem Observatory freeze-up detection mooring deployment in 2015. Shown are (a) wind speed and direction (arrows pointing downwind), (b) temperature (^∘C), and (c) salinity. Conductivity and temperature sensors were moored at 8, 20, 30, and 40 m (dashed lines) from early September until close to freeze-up at the beginning of November.

Download

Figure 6Data from the 2015/16 deployment. Shown are (a) ice draft (m), (b) salinity, (c) temperature (^∘C), (d) photosynthetically active radiation (PAR, $\mathrm{\mu }\mathrm{E}{\mathrm{cm}}^{-\mathrm{2}}{\mathrm{s}}^{-\mathrm{1}}$), (e) nitrate (NO₃, µmol L⁻¹), (f) oxygen (O₂, µmol kg⁻¹), (g) fluorescence (mg m⁻³), Nitzschia frigida flux (million cells m⁻² d⁻¹), (i) diatoms (million cells m⁻² d⁻¹), (j) acoustic zooplankton fish profiler (AZFP, days) 125 KHz, (k) AZFP 38 KHz (days), and (l) acoustic spectra (Hz). In situ NO₃ water samples were collected at times of the CEO deployment and recovery and were analyzed with standard wet chemical determinations of nitrate + nitrite of frozen samples at the Chesapeake Biological Laboratory. Using the calibration samples as anchor points, a drift of 12 µmol L⁻¹ throughout the deployment was found and corrected by linearly detrending the data.

Download

The 2015–2016 data returns from the CEO provide a unique window into the year-round Arctic marine ecosystem (Fig. 6). We also present new data showing water column turnover and cooling processes during the freeze-up period (Figs. 5 and 6).

The physical conditions measured at the CEO include currents, waves, temperature, salinity, ice draft, and light (photosynthetic active radiation, PAR) (Figs. 5 and 6a–d). The temperature and salinity cycles are tied to lateral advection, surface heat fluxes, ice cover, and winds. Sensors on the freeze-up detection buoy show that the upper 20 m of the water column was well mixed and steadily lost heat from around 4 ^∘C at the beginning of September to −1.5 ^∘C shortly before sea ice formed at the beginning of November (Fig. 5). Water temperature at 30 m depth largely followed bottom water (40 m) temperature during September and early October. However, several large departures when 20 and 30 m temperatures briefly warmed and even exceeded those at the surface could indicate the effects of lateral advection and/or the passing influence of intra-pycnal eddies (Lu et al., 2015). The bottom water temperature steadily increased from −1.5 ^∘C at the end of September to 1 ^∘C during freeze-up at the beginning of November, reversing the vertical temperature gradient in mid-October. Between September and November wind speeds in excess of 10 m s⁻¹ were observed during the passage of several low-pressure systems (Fig. 5a). At peak intensity, these storm events did not appear to erode stratification at the CEO site. However, wind direction reversals, from predominantly upwelling-favorable directions (northeasterly) to downwelling-favorable directions (southwesterly), were associated with periodic depressions of the pycnocline. After freeze-up, water temperatures at the 34 and 43 m depths remained near the freezing point (−1.6 to −1.8 ^∘C) through the end of the record in August.

Salinity followed a more cyclical progression of freshening between June and November and salinization in the other half of the year. Ice cover persisted for nearly 9 months (November through August), and the thermodynamic thickening and thinning of the ice can be seen in the overall shape of the ice draft time series (Fig. 6a). Some ice keels extended to deeper than 10 m below the surface, although an ice draft of 1–3 m was more typical. The absence of deeper keels may be due to the upwind proximity of Hanna Shoal, which would likely block or deflect deep-keeled ridges from the northeast. At the same time, the absence of extended midwinter periods of open water demonstrates that the CEO lies outside any polynya formation zone due to the proximity of the shallow Hanna Shoal.

Nitrate is the limiting nutrient in the Chukchi Sea (Walsh et al., 1989) and therefore is an important bottom–up control for the ecosystem. We deployed a SUNA V2 nitrate sensor (Sea-Bird Scientific) on the upper instrument package (34 m below the surface; Fig. 6e). Over 26–28 August 2015, shortly after deployment, nitrate values dropped from above 15 µM down to between 5 and 7.5 µM during a strong and prolonged storm. Nearby shipboard wind measurements exceeded 10 m s⁻¹ for part of each day from 25 to 29 August 2015. The drop in nitrate concentrations was simultaneous with a sharp increase in bottom water temperatures (Fig. 6c) and an increase in dissolved oxygen (Fig. 6f), likely indicating that strong mixing of surface waters with warm, nutrient-depleted, and oxygen-rich waters down to the depth of the sensors. Nitrate and temperature remained relatively constant for a couple of weeks before returning to higher and lower values, respectively, in the middle of September. While dissolved oxygen concentrations increased slowly from mid-September until freeze-up, nitrate concentrations declined as water column stratification weakened and overturning was initiated (Fig. 5) due to the strong heat and buoyancy losses from the surface ocean. These coincident changes indicate that the decline in nitrate during this time period was driven in part by the dilution of bottom water nitrate with low-nitrate surface waters, although some fall production also occurred at this time: chlorophyll a fluorescence at the CEO was measurable through at least early November (Fig. 6g). Following freeze-up, nitrate concentrations slowly increased from a low of approximately 5 µM in early November to 12 µM in early May. These increases and the anticorrelated decrease in dissolved oxygen reflect the ongoing remineralization of organic matter in sediments and the water column throughout the winter.

Chlorophyll  a fluorescence and sediment trap collections reveal a seasonal cycle of phytoplankton and ice algae export from surface waters. Large chlorophyll a peaks were observed in August and September. In June, chlorophyll a fluorescence increased from a low wintertime background level and remained elevated with intermittent peaks throughout the summer. Because the sensor is located below the summer mixed layer, it is not well situated to record the phytoplankton that resides within the upper water column or within/under the ice. However, the identification of sediment trap contents indicates that Nitzschia frigida, an ice-obligate pennate diatom, began sinking from the ice as early as April, with large pulses in May and June (Fig. 6h). During these times, sea ice and surface snowmelt would have begun, a process that would flush diatoms out of the ice matrix. Large quantities of phytoplankton were also collected in the traps in September/October and June/July, correlating with the chlorophyll a peaks, suggesting blooms in spring and fall.

Wavelet analysis of the AZFP acoustic backscatter at 125 kHz indicated a strong diurnal (1-day period) signal (Fig. 6j). This diurnal signal was found in open-water conditions in fall when the zooplankton undergo daily migrations up and down in the water column. More surprisingly, there were also strong indications of diurnal migrations in midwinter (January to March) under ice cover. In the spring, backscatter at 125 kHz was present but not strongly associated with diurnal migration.

The 38 kHz active acoustic data suggest a strong diurnal migration of fish with swim bladders from August into October, diminished backscatter from December to February, and then higher background levels with intermittently strong returns in April through July (Fig. 6k).

The passive acoustic spectra reveal the timing and source of underwater sounds, notably bearded seal vocalizations in spring and early summer (Fig. 6l).

Post-processed data are available through http://www.chukchiecosystemobservatory.org/ (last access: 9 November 2018) and http://www.aoos.org (last access: 9 November 2018). Data from the ice detection buoy are available at https://portal.aoos.org/old/arctic#metadata/75373/station (last access: 9 November 2018). This is NPRB publication no. 678.

SD developed and designed the CEO. CH, SD, AMPM, RRH, PW, PS, CL, KMS, HS, and JKH maintain instrumentation on the observatory. PS built and deployed the observatory. KM created the artwork in collaboration with CH, SD, and AMPM. CH prepared the paper with contributions from all coauthors.

The authors declare that they have no conflict of interest.