Sleep walking copepods? Calanus diapausing in hypoxic waters adjust their vertical position during winter

Abstract While hypoxia is generally associated with negative connotations, some animals may also take advantage of reduced oxygen concentrations. However, the dynamics of such processes for zooplankton are poorly understood. We made continuous acoustic studies of Calanus helgolandicus overwintering in hypoxic waters (Oslofjorden, Norway). Their apparent minimum oxygen tolerance was 0.2–0.3 mL O2 L−1 at 8°C. The copepods adjusted their vertical distribution in concert with the upward progression of hypoxia as oxygen contents declined in the course of winter. The hypoxic overwintering habitat largely excluded potential predators and mortality appeared low in early winter. As the copepod distribution shallowed in phase with declining oxygen contents at depth, mortality increased. In contrast to recent predictions, C. helgolandicus had sufficient energy reserves to sustain long-term overwintering. Termination of the overwintering phase in spring was gradual but appeared to accelerate during the development of the spring bloom. Enhanced oceanic deoxygenation with climate change may affect seasonally migrating copepods in unpredictable ways.

While hypoxia is generally associated with negative connotations, some animals may also take advantage of reduced oxygen concentrations. However, the dynamics of such processes for zooplankton are poorly understood. We made continuous acoustic studies of Calanus helgolandicus overwintering in hypoxic waters (Oslofjorden, Norway). Their apparent minimum oxygen tolerance was 0.2-0.3 mL O 2 L −1 at 8 • C. The copepods adjusted their vertical distribution in concert with the upward progression of hypoxia as oxygen contents declined in the course of winter. The hypoxic overwintering habitat largely excluded potential predators and mortality appeared low in early winter. As the copepod distribution shallowed in phase with declining oxygen contents at depth, mortality increased. In contrast to recent predictions, C. helgolandicus had sufficient energy reserves to sustain long-term overwintering. Termination of the overwintering phase in spring was gradual but appeared to accelerate during the development of the spring bloom. Enhanced oceanic deoxygenation with climate change may affect seasonally migrating copepods in unpredictable ways. KEYWORDS: Calanus helgolandicus; overwintering; dormancy; ocean deoxygenation INTRODUCTION Climate change affects pelagic ecosystems through various mechanisms. For example, higher temperatures cause poleward movements of biogeographic boundaries, also for numerous copepod species (Reygondeau and Beaugrand, 2011;Chivers et al., 2017). Oxygen concentrations in the open ocean and coastal waters are declining, and both coastal hypoxia and oceanic deoxygenation are predicted to worsen with increasing global temperatures (Levin and Breitburg, 2015;Breitburg et al., 2018). Ocean deoxygenation may cause both losers and winners among plankton (Wishner et al., 2020). Some may be excluded from previous habitats, while others may exploit low oxygen waters. Wishner et al. (2020) termed organisms being specially adapted to live in oxygen minimum zones for "hypoxiphilic." Yet, many others may tolerate low oxygen and take advantage of hypoxic habitats on different time scales, e.g. for predator avoidance during either a daily cycle or during diapause (Nolan et al., 2019;Wishner et al., 2020).
The life cycle of temperate and high latitude copepods often includes a dormant overwintering phase in deep waters. Understanding the dormant life-phase of calanoid copepods has long been an aim of biological oceanography (Pond, 2012). Calanus spp. are key copepods in the Northern Atlantic (Conover, 1988). Generally, research has centered on C. finmarchicus, but the ongoing northward distribution shift of the congeneric C. helgolandicus in response to warming (Fromentin and Planque, 1996;Beaugrand et al., 2003), warrants more focus on the less studied C. helgolandicus (Bonnet et al., 2005).
The dormancy phase of Calanus spp. varies among species and regions (Conover, 1988;Kvile et al., 2019) and is less understood for C. helgolandicus than for its northern congeneric (Bonnet et al., 2005;Wilson et al., 2015). In some habitats, C. helgolandicus seems active in upper waters throughout winter (Conover, 1988), while populations in other locations may overwinter in diapause (Hallberg and Hirche, 1980;Hirche, 1983Hirche, , 1984. Wilson et al., (2015) argued that energy reserves of C. helgolandicus would sustain diapause for < 60 days in most of its geographic range, where temperatures exceed 10 • C. They reasoned that the potential inability to diapause for longer periods might largely restrict C. helgolandicus to continental shelf regions (Wilson et al., 2015).
Calanus helgolandicus is the prevailing Calanus-species in the Oslofjord, as established with both morphological and molecular methods (Beyer et al., 1967;Bagøien et al., 2000;Bucklin et al., 1999). The deepwater of the inner fjord often becomes hypoxic. Bagøien et al. (2000) studied the seasonal migration of Calanus spp. in the fjord but did not sample in the inner, hypoxic part. We are unaware of studies addressing the tolerance of C. helgolandicus to hypoxia. However, Calanus euxinus in the Black Sea enters diapause and aggregates near the lower boundary of the oxygen zone (Vinogradov et al., 1992;Sakinan and Gücü, 2017). This species is closely related to-and originally was classified as-C. helgolandicus (Unal et al., 2006). Similarly, Osgood and Checkley (1997a, b) found high concentrations of resting Calanus pacificus just above oxygen deficient bottom waters in Santa Barbara Basin. There, the copepods adjusted their distribution upwards as oxygen concentrations declined, but still inhabited waters with very low oxygen levels (Osgood and Checkley, 1997a, b).
We made continuous acoustic measurements for ∼4 months of C. helgolandicus overwintering in hypoxic, waters of the inner Oslofjord. The acoustic target identity was confirmed with intermittent net sampling during the study period. The acoustic observations documented apparent patterns in mortality, adjustments of vertical distribution reflecting limits of tolerance to hypoxia, as well as duration of the overwintering phase. The acoustic data moreover unveiled small-scale vertical patchiness of the overwintering population in the enclosed basin water below sill depth. While the study presents novel data with an unsurpassed temporal and vertical resolution on this particular Calanus species, it more generally represents a model of a diapausing copepod population responding to a changing physical and biological environment during the course of dormancy.

Study site
We studied overwintering C. helgolandicus in the 150 m deep Bunnefjorden, the inner branch of the Oslofjord (59.792171 • N, 10.726776 • E; Fig. 1 in Klevjer and Kaartvedt, 2011). This is a very sheltered location separated from the outer parts of the fjord by a 57 m deep sill. The residence time of the deep waters is about 3-4 years (Gade, 1973) and waters below the sill commonly become hypoxic (Solberg et al., 2015).

Acoustics
We made acoustic measurements at 200 kHz for 4 months during winter and spring 2007-2008, supplemented by intermittent field campaigns. The virtual absence of other macrofauna in the low oxygen water (e.g. Solberg and Kaartvedt, 2017) facilitated the acoustic target identification and enabled the use of instrument settings that made the echosounder sufficiently sensitive to record the small and weak copepod targets (see below). Mutlu, (2003) and Sakinan and Gücü, (2017) have previously documented the feasibility of using acoustic data at 200 kHz to assess distribution and biomass of dense concentrations of Calanus (euxinus) entering oxygen minimum zones of the Black Sea. De Leo et al., (2018) applied acoustic data (2 mHz ADCP) from cabled observatories at the west coast of Canada in assessing overwintering among Neocalanus spp.
Details on methods and procedures are given in Røstad and Kaartvedt (2013); Solberg et al. (2015) and Solberg and Kaartvedt (2017) who studied krill and fish in this fjord branch. In short, we deployed a bottom-mounted upward-facing 200 kHz echosounder (Simrad EK60) at 150 m depth from 6 December 2007 to 17 April 2008. The acoustic transceiver (GPT) was mounted in a pressure-proof casing and connected to land with a cable that both provided electricity and transmitted digitized signals to a computer onshore. The echosounder recorded data for the whole winter with a temporal resolution (ping rate) of 1-2 registrations s −1 , except during equipment failure 14-20 December 2007 and between 28 December −10 January 2008, due to theft of the land-based logging computer.
The acoustic data were plotted in MATLAB. We present a diel echogram covering the whole water column using an acoustic threshold of −80 dB to illustrate how an oxycline acted as a "false bottom" excluding most macrofauna. We subsequently focus on the deeper hypoxic part, using lower thresholds (−90 and −100 dB) better unveiling the weak copepod targets. We present acoustic data on time scales ranging from 24 h to a composite of the whole study period. In comparison of acoustic data with net catches, we made average echograms for months, as in Staby et al. (2011). To get an acoustic estimate of Calanus total biomass, we for the first part of the study period integrated the total backscatter between 90 m and 149 m depth (1 m above the seafloor) and averaged every hour. After 1 January 2008, there was noise close to the seafloor, due to gas bubbles from the echosounder, and Calanus had moved higher in the water column. The integration was, therefore, shifted upwards from 80 to 140 m depth). This upper limit may have missed some Calanus at the end of the study period when the copepods stayed shallower but were necessary to avoid the inclusion of stronger targets that appeared at shallower depths. To eliminate noise and any stronger targets than Calanus, the upper integration threshold was set to a backscatter of −70 dB, with a lower threshold of −90 dB. The integrated acoustic backscatter is presented as Nautical Area Scattering Coefficient (m 2 nmi −2 , NASC).

Field campaigns
We conducted field campaigns in December 2007 and in January, February and April 2008. Vertical profiles of temperature and salinity were measured by CTD (Conductivity, Temperature and Depth) equipped with Niskin bottles. Water samples from 13 depths between the surface and 150 m (bottom) were analyzed for oxygen content using the standard Winkler method (except for April). Vertical profiles of temperature, salinity and oxygen from the entire water column are given in Solberg and Kaartvedt (2017;their Fig. 2). We therefore here only provide data on density and oxygen below 60 m. Daytime vertical net tows for mesozooplankton (except for January) were made over six depth intervals (bottom-100 m, 100-80 m, 80-60 m, 60-40 m, 40-20 m, 20-0 m) using a WP-2 net. Species identification of Calanus was made based on microscopically checking the coxapodite curvature of the fifth pair of swimming legs (cf. Sars, 1903;Fleminger and Hulsemann, 1977) on 20 individuals in samples from the bottom to 100 m (December), 100-80 (February) and 80-60 (April). Only C. helgolandicus occurred in these subsamples (mostly stage CV). We, therefore, assigned Calanus sp. in the remaining samples to C. helgolandicus.

Diel echogram at threshold −80 dB
A diel echogram (February) illustrates how hypoxic waters (see below) excluded much of the macrofauna (Fig. 1). Echoes of vertically migrating krill and fish prevailed in the upper 60 m. There was a sharp boundary toward waters below, which were largely devoid of backscatter at a −80 dB threshold, apart from some weak backscatter at ∼ 80-100 m (Fig. 1). There was no diel pattern in the hypoxic waters.

Hydrography, net tows and low-threshold acoustic (−90 dB)
The temperature was about 8 • C and salinity 33 in the deeper part of the water column all winter, with only slight changes with depth in the basin water (data presented in Solberg and Kaartvedt, 2017). The density increased slightly toward the bottom, most sharply above about 80 m. Profiles were very similar throughout the study period, though bottom waters appeared slightly less dense at the end of the study (Fig. 2). Oxygen concentrations were < 0.8 mL O 2 L −1 below 60 m and declined both with depth and time (Fig. 2). Oxygen concentrations C. helgolandicus CV made up nearly 90% in terms of numbers of the zooplankton captured in the hypoxic waters below 60 m, with the small copepods Oithona and Oncea dominating the remainder. In December and early winter, abundances were high in deeper parts of the water column with an average of ∼500 individuals m −3 in the 50 m depth interval from100 m to the bottom. Abundance declined with shoaling depth (Fig. 2). In February, the copepods stayed shallower, with peak concentrations of ∼800 copepods m −3 at 100-80 m, while concentrations below had decreased markedly. The integrated abundance of Calanus in net catches declined from 31 540 individuals m −2 in December to 23 050 individuals m −2 in mid-February. In April, the copepods had virtually disappeared below 80 m. Stage CVI females now constituted about 50% of the Calanus-population and dominated in catches above 60 m (not shown).
The low-threshold acoustic records of the weak targets in the hypoxic waters agreed well with the distribution of Calanus in the net samples (Fig. 2). The higher vertical resolution in the acoustic data (averages for whole months) indicates that the majority of Calanus captured at 150-100 m in December stems from the lower half of this depth interval. In February, the weak acoustic record in the deepest part of the water column and the marked acoustic maximum at 80-90 m mirrored the distribution of Calanus in the net catches. In April, neither catches nor acoustic records documented much Calanus below 80 m. There then was a slight discrepancy with somewhat deeper acoustic records compared with the net tows. The acoustic data were averaged for the April records, while the net tows were taken at the end of the study period (Fig. 2).

Continuous long-term acoustics records in hypoxic waters
At the start of winter, backscatter below the strong echoes of krill and fish in the in the more oxygenated part of the water column increased with depth, apart for a narrow zone just above the bottom, which was devoid of backscatter (Fig. 3). From mid-January, a band of stronger echoes appeared at ∼90 m. This band developed into a progressively stronger maximum in February and had disappeared by mid-March. Parallel to the strengthening of this thin layer, echoes became weaker below, until virtually disappearing at ∼10 February. Waters below Abundance, as proxied by integrated backscatter (NASC) in the hypoxic waters, initially remained stable with time. There was a sudden decline from early February, with an apparent subsequent accelerated decline from early/mid-March (Fig. 3).

Patchiness/vertical stratification
Resolving the deep acoustic data on a diel scale revealed repeated dense layers alternating with depth intervals without backscatter in the hypoxic water (Fig. 4). The vertical positions of the thin acoustic layers varied somewhat on short-time scales, but not always in synchrony (Fig. 4).

DISCUSSION
We have presented results on seasonally migrating copepods, applying an echosounder cabled to shore providing data of supreme temporal and vertical resolution. As the backscatter followed a seasonal development similar to that of C. helgolandicus in net samples and because there were no alternative acoustic targets in the hypoxic waters, we used the continuous long-termed acoustic records to assess the abundance and vertical distribution throughout winter.
While Calanus finmarchicus generally diapause, C. helgolandicus appears more variable (Bonnet et al., 2005). The Calanus overwintering strategy depends on lipid energy storage. Wilson et al. (2015) argued that the period of dormancy would be less than 60 days in C. helgolanidcus at temperatures comparable to those in our study. However, our direct observations revealed a much longer overwintering period. In Oslofjorden C. helgolandicus initiates overwintering and descends during September (Bagøien et al., 2000). Regardless of any inter-annual variations, the overwintering population was well established at depth when we initiated acoustic sampling in early December. An accelerated decline in the acoustic backscatter ascribed to Calanus during March (cf. Fig. 3) was likely influenced by the initiation of the seasonal ascent associated with the termination of overwintering. This timing matched with the development of the spring bloom (Røstad and Kaartvedt, 2013). While most individuals had terminated overwintering by April, some C. helgolandicus remained at depth after the bloom. Similar, and probably statedependent, long stretched periods of diapause termination over several months occur in other, better oxygenated parts of Oslofjorden (Bagøien et al., 2000). Hypoxia tolerant zooplankton living near their physiological limits may respond to very slight changes in oxygen (Wishner et al., 2018). As oxygen concentrations declined at depth during winter, the overwintering C. helgolandicus adjusted their vertical distribution upwards (Figs 2 and 3). From the vertical relocation relative to the ambient oxygen concentrations, we derive that their lower tolerance is between 0.2 and 0.3 mL O 2 L −1 at 8 • C. Such values compare well with those at which Calanus pacificus may diapause, also occurring in waters of 0.2-0.3 mL O 2 L −1 (Alldredge et al., 1984;Osgood and Checkley, (1997a, b). Also in other settings, zooplankton abundance and biomass appear to decrease sharply when oxygen levels fall below 0.2 mL O 2 L −1 (Longhurst, 1967;Böttger-Schnack, 1996;Ekau et al., 2010). There was no change in water density explaining the shallowing distribution. The enclosed basin water was resident throughout winter as evidenced by the continuous decline in oxygen contents and without signs of intrusion of denser waters. However, we cannot exclude the passive upward movement of animals with differences in lipid composition.
By overwintering in hypoxic waters, the copepods were partly, but not totally protected from predators. The net samples from 13 December and 12 February suggest some (∼30%) reduction in surface integrated abundance in the course of that 2 month period. While the nets only represent two sampling dates without replicates, the integrated backscatter ascribed to Calanus provides a continuous picture. Backscatter remained stable until suddenly higher rates of decline appeared from mid-February, concurring with the formation of a dense copepod layer at 80-100 m (Figs 2 and 3).
Some invertebrate predators like chaetognaths can tolerate low oxygen values and may co-occur with copepods in low oxygen waters (Sakinan and Gücü, 2017), but none were captured below 100 m during the current study (Røstad and Kaartvedt, 2013). In this fjord branch fish and krill generally avoid the lower part of the water column when oxygen deficit (e.g. Fig. 1; Kaartvedt et al., 2009;Solberg et al., 2015). However, planktivorous fish (sprat; Sprattus sprattus) made occasional short excursions into the low oxygen water, apparently foraging on the overwintering Calanus during this winter (Solberg and Kaartvedt, 2017). In early winter the bulk of Calanus inhabited near-bottom waters (Fig. 2) and was out of reach for such short-range and short-time forays, but sprat dived into the shallowing Calanus-layer later in winter (Solberg and Kaartvedt, 2017). Calanus would become more accessible to the planktivorous fish when moving upwards and concentrating closer to the oxycline that defines the lower depth of the main sprat distribution. Calanus indeed dominated the prey in sprat stomachs this particular winter (Solberg and Kaartvedt, 2017). While we cannot exclude other causes of mortality (cf. Daase et al., 2014) predation appears to be the most plausible explanation for the enhanced decline in Calanus numbers in mid-winter when vertical distribution became progressively shallower. Advection is an unlikely explanation for the observed population decline, as the water properties attested to an enclosed water mass below the sill depth of Bunnefjorden throughout the study.
Similarly to our interpretation, studies elsewhere have shown that hypoxia can lead to a decoupling of predator-prey interactions (Taylor and Rand, 2003), with the spatial extent of the suitable habitat for fish declining during hypoxic events, while zooplankton may find refuge from predation. Thus, hypoxic conditions may reduce the predation risk of zooplankton from pelagic fish (Taylor and Rand, 2003), although examples of fish exploiting high zooplankton concentrations in hypoxic waters also exist. For example, anchovy may make short-range/term excursions into hypoxic waters with concentrated zooplankton prey (Taylor et al., 2007). Alternatively, if enhanced hypoxia would force the copepods even further upwards, low oxygen might instead be detrimental by making the copepods more vulnerable to visual predators, as suggested for other predator/prey relations with expanding oxygen minimum zones (e.g. Netburn and Koslow, 2015). In this way, hypoxia-related to coastal eutrophication and climate change may alter trophic fluxes through food webs in non-linear ways. The effects of long-term climate-driven changes remain hard to predict, also because marine food webs are more complex than simply considering the spatial overlap between plankton and fish as focused on here (Ekau et al. 2010).
The backscatter ascribed to overwintering Calanus displayed a strikingly stratified vertical structure. We have not resolved the role of passive versus active accumulation. Pond (2012) argued that achieving neutral buoyancy is essential for copepods that overwinter in diapause since active swimming will both attract predators, and deplete metabolic energy reserves. Zooplankton may accumulate passively in response to density gradients in the environment (Tiselius et al., 1994). Some slight apparent vertical variations on short-time scales (Fig. 4) probably reflected physical displacements of the water rather than individuals actively relocating. However, layers and aggregations may also form actively in response to chemical (e.g. Heuschele and Selander, 2014) or physical conditions, invoked both directly by the environment per se (Buskey et al., 1995) or indirectly via for example cues emitted from predators or conspecifics (Heuschele and Selander, 2014;Leising and Yen, 1997). Copepods have a whole suite of sensory abilities and can detect and process at least chemical, hydrodynamic and light signals and alter their behavior and distribution in response, so sensory and behavioral mechanisms for maintaining such layers are in place.

CONCLUSION
C. helgolandicus in the Oslofjord sustained long-term overwintering in hypoxic waters. Their apparent minimum oxygen tolerance was 0.2-0.3 mL O 2 L −1 at 8 • C. The copepods adjusted their vertical distribution in concert with the upward progression of hypoxia as oxygen contents declined in the course of winter. Such apparent active shifts in vertical distribution during diapause have implications not only for our understanding of altered distribution patterns in response to future climate scenarios but also for predictions of population mortality during winter.

DATA ARCHIVING
Data are available from the corresponding author upon request.

ACKNOWLEDGEMENT
Thor A. Klevjer was instrumental during the fieldwork. Rita Amundsen did the species identification of Calanus helgolandicus.