The 2nd Chinese Arctic Research Expedition was carried out from 29 July to 7 September 2003 over a large area including the Chukchi Sea, Chukchi Abyssal Plain, Chukchi Plateau and Canada Basin (Fig. 1). Multi-disciplinary observation was carried out at 37 stations, with six of them (R06–14, P11) investigated twice (Table 1).

At each station, ice coverage was estimated by direct visual observation. Temperature and salinity were measured with a Mark III CTD. Water samples for Chl a concentration were collected with Niskin bottles at depths of 1, 10, 20, 30, 50, 75, 100, 150 and 200 m. A total of 250 mL of natural seawater was taken from each sample and filtered through Whatman GF/F filter. Each filter was extracted with 90% acetone for 24 h at 0°C and fluorescence was measured with a Model 10 Turner Designs fluorometer.

Quantitative zooplankton sampling was conducted at all stations with a conical net (mouth area 0.5 m²; mesh size 330 μm). The net was hauled vertically from within 3 m of the bottom to the surface at 0.5 m/s at stations shallower than 200 m and from 200 m to surface at deep stations. Samples were preserved in buffered 5% formalin seawater.

In the laboratory, copepodite stage compositions were determined with the aid of a dissecting microscope according to morphological characters for Calanus species (Marshall and Orr, 1955; Li and Fang, 1990) and body size for C. hyperboreus (Hirche, 1997; Melle and Skjoldal, 1998). Integrate abundance of C. hyperboreus in the uppermost 200 m were here presented as individuals per m² (ind./m²) taking into account the different sampling depth (Plourde et al., 2003; Ota et al., 2008; Thor et al., 2008; Broms et al., 2009). Prosome length, from the anterior end of the cephalosome to the posterior lateral edge of fifth metasome segments, was measured using a graticule in the microscope eyepiece. Prosome lengths measurement was carried out only for later stages with fully developed prosome somites and making up the vast majority of overwintering population. The original sample was divided with plankton sample splitter, and at least 10% of that was counted for population structure and abundance. For prosome length measurements, no more than 150 individuals from each stage were measured at each station.

In the present study, hydrological conditions and Chl a concentrations were shown in detail only at those stations with C. hyperboreus present. In order to check the possible overlap of cohorts with different geographical and temporal origins, normality of size frequency distribution was checked with one-sample Kolmogorov-Smirnov Test on all measured stages from each station and the entire investigation, respectively. Significance of influence from food availability and temperature was analyzed with Spearman rank correlation on abundance and proportion of each stage, as well as average prosome length of CIII-adult. As diel vertical migration was unknown, food availability was analyzed in terms of column averaged and maximum Chl a concentration, respectively. Their correlation with temperature was tested as both column average and surface value. All the statistics were run with software SPSS 16.0.

The relationships between abundances of each development stage and environmental variables were examined with Canonical Correspondence Analysis (CCA) on software CANOCO 4.5. After test of the length of gradients with Detrended Correspondence Analysis, CCA was performed on log transformed data. The significance of environmental variables to explain the variance of abundances in CCA was tested using forward selection of Monte Carlo simulations.

This investigation covered a large area varying from ice-free shallow neritic to fast ice-cover deep basin areas. Calanus hyperboreus was presented at all stations deeper than 200 m, but was absent from all shallow stations except Stas R16 and S11 close to the basin (Fig. 1). Ice coverage at stations with C. hyperboreus occurrence varied from 10% to 95%. The highest ice coverage was observed at 80°N stations, where sampling was conducted in narrow polynias surrounded by thick sea ice. Ice coverage was lowest in the slope area near Barrow, Alaska. Sea ice retreat was observed at duplicated stations. At Sta. P11, ice coverage decreased from 65% on 9 August to 20% on 7 September (Table 1).

Temperature-salinity characteristics identified various hydrographic regimes among stations where C. hyperboreus was observed. Warm high-salinity water (S>34, T>–1.0°C) was observed in 150–200 m layer at stations in the north Canada Basin and Chukchi Plateau, as well as Stas P11 and R16 in the slope area between the Chukchi Sea and Chukchi Abyssal Plain (CS-slope), but absent at stations in the south Canada Basin and slope areas (Fig. 2). Salinity higher than 34.5 was observed only at 200 m depth at Sta. P11, where temperature was also as high as 0.25°C. At the adjacent Sta. R16, temperature and salinity were –0.53°C and 34.3 respectively. At all stations, surface temperature varied between –1.56 (M03) and –0.56°C (S16), and surface salinity varied between 27.56 (B13) and 29.86 (B80 and P80). High sub-surface temperature (>0°C) was detected only at Stas S25 and P27 in the south Canada Basin. Halocline was commonly observed at all stations in 30–60 m layers.

Chl a concentration was high and more variable above 100 m depth, and lower than 0.10 μg/L at all stations under 100 m depth (Fig. 2). At the northernmost Stas P80 and B80 with highest ice coverage, Chl a concentration was highest (0.21 and 0.23 μg/L) in surface layer, whereas sub-surface maximum was observed at the other stations. Extremely high concentrations of 8.60 and 39.10 μg/L were observed at 20 m and 30 m depth at Stas R16 and S11, respectively. Except these two stations, the maximum Chl a concentration through the water column was highest at Sta. P11, where concentrations of 1.16 and 1.28 μg/L were recorded at 30 m depth during the duplicated observations.

In our study, C. hyperboreus was observed in 21 samples collected from 20 stations located mainly in deep basin and adjacent slope areas (Fig. 1). Calanus hyperboreus is able to reach the south edge of the Chukchi Abyssal Plain and Canada Basin respectively, but its abundance in these two areas differed by 1–2 orders of magnitudes. At Sta. P11, total abundance of C. hyperboreus was 5 815.0 ind./m² in August and 5 475.0 ind./m² in September, and abundance of 1 110.0 ind./m² was observed at the nearby Sta. R16 (Fig. 3). At the other stations, their abundance was no higher than 950.0 ind./m². The lowest abundance of 55.0 and 40.0 ind./m² was observed at Stas P27 and P25, respectively. At shallow Stas S11 and S13 near the Canada Basin, total abundance was also lower than 100 ind./m².

Based on the population structure, late stages including CV and adult females were frequently recorded at deep high-latitude stations (B11, B13, S13, S16, S25 and P25), but populations at shallow slope stations overwhelmingly consisted of early stages (CI–CIV). In samples from Stas R16, P11 and P11-2, CV and AF (adult females) were extremely scarce, and the highest proportion was observed in CIII and CI. Meanwhile, at shallow stations in the Canada Basin, the population completely consisted of CII and CIV at Sta. S11 and CV at Sta. S13. At stations deeper than 1 000 m, the proportion of CV and females ranged from 27.6% to 86.8% and averaged at 59.8%. In the southern Canada Basin, CV and females accounted for proportions higher than 80% at central stations (S25, S16, B11 and B13) and higher than 50% at stations near the Northwind Ridge (P17 and P27). Well-proportioned population structure was observed at the four northernmost stations (B78, B79, B80 and P80), whereas dramatic change was recorded along the transect from the Northwind Ridge to the Chukchi Abyssal Plain. Population at Sta. P25 was comprised of completely CV and adult females, but CIII accounted for the highest proportion at the other four stations (P23, P21, M01 and M02). Population structure also differed between the repeated observations at Sta. P11. The proportions of CI, CII, and CIV increased from 6.0%, 12.5%, and 9.0% on 9 August to 34.9%, 24.0%, and 28.8% on 7 September, whereas that of CIII decreased from 71.4% to 12.3%.

In early stages (CI–CIV), abundance and proportion increased with Chl a concentration, showing significant positive correlation between abundance of CIII–IV and proportion of CIV and Chl a concentration (Table 2). Significant negative correlation was observed in CV and AF. Both abundance and proportion correlated more significantly with column maximum Chl a concentration than with column average Chl a concentration.

Geographical variation in prosome length differed among the four measured stages (Fig. 4). In Stage CIII, average prosome length was highest at Sta. R16 (2.61 mm), followed by 2.53 mm at Sta. P11 and 2.48 at Sta. P11-2. According to 95% confidential interval in Fig. 4, prosome length at the above three stations was significantly higher than that from any other stations. The difference of average prosome length between the maximum and minimum station was largest in CIII (1.21 mm), followed by 1.10 mm in CV. The ratios in CIV and adult female were 1.06 and 1.04 mm, respectively. Comparatively high prosome length was observed at same stations also in CIV and CV with less significant difference. Female prosome length was highest at Sta. M02. Significant correlation between average prosome length at each station and Chl a concentration was detected only in CIII and CV (Table 2), in which average prosome length increased with maximum Chl a concentration.

According to one-sample Kolmogorov-Smirnov Test, normality was accepted for each developmental stage from all stations (P>0.05; Table 3), except CIV at Sta. P11-2 (P<0.05). However, normality was rejected for all the measured developmental stages as a whole, and asynchronous recruitment in various regions was accepted. The frequency distribution was positively skewed for CIII and negatively skewed for the other stages, indicating that asynchrony in newborns may be resulted from accelerated growth in some individuals and that in overwintered stages was induced by retarded development. The absolute values of skewness increased from CIV to AF, showing that individuals shorter than mode size increased in number with age.

The two first axes in CCA, with eigenvalues of 0.085 and 0.018, explained a total of 20.1% of cumulative variance in abundance data, and 98.2% of the correlations between environmental variables and abundance. After forward selection, four variables (in descending order, average Chl a concentration, surface temperature, ice coverage and surface salinity) played significant role (P<0.05) on the geographical distribution of each development stage. Based on the intersect correlations, Chl a and salinity were largely related to the first axis, while temperature and ice coverage were respectively positively and negatively related to the second axis.

With regard to the ordination (Fig. 5), environmental conditions were less variable at deep basal stations comparing to shelf stations. Most of the basal stations located near the ice coverage arrow, whereas scattered distribution was observed on the CS-slope stations. Abundance of each development stage was associated mainly with Chl a concentration, with CI and CII at higher but CV and female at lower value.

The most important finding in our study is that CIV can be achieved in one growing season in the productive CS-slope area. In the multi-year life cycle of C. hyperboreus, it was commonly accepted that CIII is attained in the first year and descended to the deep layer as the main overwintering stage (Hirche, 1997). Most of the overwintered CIIIs molted to CIV from May to July in the next year (Dawson, 1978; Hirche, 1997). At the SHEBA station, which drifted from the Canada Basin over the Northwind Ridge and Chukchi Plateau and back over the basin, average prosome length of CIII was highest in May, the claimed molting season for overwintered individuals (Ashjian et al., 2003), but it was still lower than that in CS-slope area in our study. In our study, an overwhelming majority of CIIIs in CS-slope area were larger than the median of its size range in previous reports (Table 4). Meanwhile, at the other stations, prosome length of most CIIIs were shorter than the median of reported size range, and on average shorter than the station specific minimum at SHEBA station. We suggest that, only in CS-slope area, CIII can reach molting size in August of the first growing season.

According to temporal variation at the same sampling site from August (Sta. P11) to September (P11-2), molting from CIII to CIV was suggested during this period, as both abundance and proportion decreased in CIII but increased in CIV. Simultaneously, prosome length of CIII decreased on average, the same as the SHEBA station results. Normality was not significant for size frequency distribution of CIV at Sta. P11-2, indicating overlap of different cohorts. In the Resolute Passage (Conover, 1988) and Barrow Strait (Conover and Siferd, 1993), CIV had been recorded as the dominant overwintering stage, although its achievement in one growing season was still doubted. However, its further development to CV was suggested occurring separately in early spring and summer (Conover and Siferd, 1993), showing that the overwintered CIVs contained cohorts with different temporal origin. Thus, at least a part of the newborns of C. hyperboreus can develop to CIV before winter in the shelf waters of the western Arctic Ocean.

In previous studies carried out in semi-closed coastal regions, no marked temporal pattern was observed in prosome length of C. hyperboreus in Stages CIII–CV during the bloom period, and its somatic growth was detected mainly as an increase in dry weight or carbon content (Plourde et al., 2003; Swalethorp et al., 2011). However, according to our results, the prosome length achieved in various regions might increase with local food availability. Significant correlation between prosome length and Chl a concentration was observed in both CIII (the new-born generation) and CV (the overwintered population). Additionally, prosome length of CIV increased significantly from August to September at the same Sta. P11. Thus, its somatic growth pattern may change with various environmental conditions. In the Barrow Strait, less lipid storage was recorded in CIII and CIV, but they can still successfully overwinter along with the lipid-rich CV stages (Conover and Siferd, 1993).

The recorded abundance of early stages (CI–CIV) in CS-slope area was as high as that observed in other shallow Arctic areas (Hirche et al., 1994; Hirche, 1997; Swalethorp et al., 2011). Cross shelf abundance gradients of C. hyperboreus have been observed along the margins of the deep Nansen, Amundsen and Makarov Basins, and an elevated biomass was observed in the margin off the Laptev Sea (Kosobokova and Hirche, 2009). It was indicated that, though C. hyperboreus is absent from shallow shelf areas of the western Arctic Ocean, the CS-slope area can serve as a hotspot of population recruitment, through acceleration in both development and recruitment rates.

Food limitation to C. hyperboreus was commonly observed in natural environments. Estimated in situ grazing rates for the large copepod species were less than 10% of their maximum rates in a melt water influenced Greenlandic fjord (Tang et al., 2011) and in situ rates for C. hyperboreus and C. glacialis in the central Arctic Ocean were only about 3% and 20% of the expected food-saturated ingestion rates (Olli et al., 2007). However, C. hyperboreus has great potentiality to take advantage of the elevated food supply, as non-selective and non-saturated feeding was observed. It removed all types of phytoplankton in direct proportion to their abundance at Chl a concentrations ranging from 0.53 to 12.1 mg/m³ (Huntley, 1981). Thus, recruitment success in early copepodites (CI–CIV) of C. hyperboreus increased primarily with food availability once the recruitment had started (Ringuette et al., 2002).

In the western Arctic Ocean, alleviated primary production was induced mainly by intrusion of nutrient-rich pacific water (Nishino et al., 2008). However, in agreement with previous report (Hopcroft et al., 2010), C. hyperboreus cannot arrive the productive shelf waters of the Chukchi Sea. High food availability is also expected in the CS-slope area, as mesoscale eddies can bring episodic pulses of nutrients into the euphotic zone. During an investigation on a warm-core eddy in the southwestern Canada Basin, ∼30% higher biomass of pico-phytoplankton was sustained than in the surrounding water (Nishino et al., 2011). In another study, elevated concentrations of most zooplankton taxa were observed in eddies (Llinás et al., 2009). Though Chl a concentration at Sta. P11 was low on average, subsurface maximum was observed in both August and September, indicating high food availability.

Besides higher Chl a concentration in the CS-slope area, earlier onset of spring bloom may also contribute to the flexible life cycle. Under-ice phytoplankton blooms were observed widely in polar regions, including the Canadian Beaufort Sea and north part of Chukchi Sea (Mundy et al., 2009; Arrigo et al., 2012). Spawning of C. hyperboreus is fueled by lipid storage; the spring ascent was prior to or in association with the break-up of sea ice and the development of the spring phytoplankton bloom (Madsen et al., 2001). In the eastern sector of the North Water, the recruitment of the first cohort of copepodites of C. hyperboreus coincided with the onset of the bloom and occurred before any significant increase in temperature (Ringuette et al., 2002).

As our investigation was carried out in summer, positive correlation between abundance and Chl a concentration and accelerated development were observed mainly in CIII. Reproduction of C. hyperboreus starts before May, and most of the new generation has developed to CIII in August, so that correlation between abundance and Chl a concentration was not significant in CI and CII. Besides simultaneous food availability, body size and abundance of late stages (CIV-adult) can be influenced by conditions in the past growing seasons and physical advection. Well-developed individuals can be transported from productive shelf waters to barren basin areas, but we still cannot explain the significant negative correlation in CV.

The mismatch between distribution centers of early (CI–IV) and late (CV–adult) stages demonstrates the importance of advection processes to population recruitment of C. hyperboreus. Prevalence of late stages (CV–adult) had been observed in the central Arctic Basin, the Norwegian Sea and the Chukchi Plateau (Rudyakov, 1983; Ashjian et al., 2003; Broms et al., 2009). Accordingly, different hypotheses of cross shelf-basin replenishment and insufficient sampling of individuals beneath the ice were put forward as explanations. As our investigation was carried out in August and September and significant ice melting was observed at most stations, underestimation of nauplii and early copepodites was unlikely. According to the reported reproduction season, C. hyperboreus spawns from January to May in the central Arctic Ocean (Brodskii and Nikitin, 1955; Johnson, 1963), and CII was commonly observed in the basin area after July (Ashjian et al., 2003). It was also suggested by numerical simulation that some of the new cohort cannot develop to CIII in one growing season (Ji et al., 2012). Considering the shorter prosome length, the early stages observed in the high latitude area in our study might be recruited locally, but the abundance of CIII was much lower than that of CIV–AF. On the other hand, net emigration was indicated by high abundance of early stages and scarcity of late ones in the CS-slope area. Thereafter, it is more likely that the C. hyperboreus population in the high latitude area is at least partially replenished by advection, and the CS-slope area may be an important potential source.

As early generations can be advected into deep water by the offshore current, backward supplement may also play an important role in the slope-basin interaction, as the CS-slope area tends to be an unfavorable overwintering site for C. hyperboreus. In previous reports, copepodite stages later than CIII were observed only at deeper than 300 m in winter, whereas both early and late stages entered upper layers in summer (Hirche, 1997; Ashjian et al., 2003). The two CS-slope stations with high early stage abundance observed in our study are shallower than 300 m. In our results, adult female were extremely scarce at Stas R16 and P11, and their prosome length showed no significant difference with basin stations.

According to previous reports on current regimes, advection between the central Arctic basin and CS-slope area is possible on both back and forth directions. Currents in deep areas in the western Arctic Ocean, such as the Barents Sea Branches of the Arctic Circumpolar Boundary Current flowing continuously along the Siberian Shelf, can reach as far as the Chukchi Plateau (Aksenov et al., 2011). Meanwhile, C. hyperboreus can be transported from overwintering depths in the deep basin to the Chukchi Shelf by upwelling (Lane et al., 2008; Llinás et al., 2009). Elevated abundances of Arctic-origin copepods, including C. hyperboreus, were recorded in a cold-core eddy in slope area between the Beaufort Sea and Chukchi Sea in summer (Llinás et al., 2009), even though most such eddies probably originated from the edges of the Chukchi and Beaufort shelves and contained water of Pacific origin (Muench et al., 2000). In our study, deep water with higher temperature and salinity reached shallower depth in the northern Chukchi Sea, indicating upwelling of deep waters. On the other hand, its absence in slope area between the Beaufort Sea and Chukchi Sea may be a reason for scarcity of C. hyperboreus. Shelfbreak eddies in the western Arctic Ocean last from weeks to more than a year before moving from the Chukchi Shelf into the interior Canada Basin (Manley and Hunkins, 1985), which can act as a possible pathway carrying C. hyperboreus from CS-slope into central basin.

Although this is the first time that early stages (CI–CIV) of C. hyperboreus was recorded in high abundance in the CS-slope area, its prosperity is always expected whenever overwintered cohort was transported into productive regions influenced by the Pacific waters. In previous studies, both intrusion of C. hyperboreus and elevated primary production had been recorded separately in slope regions in the western Arctic Ocean (Llinás et al., 2009; Nishino et al., 2011). Along the margins of the deep Nansen, Amundsen and Makarov Basins, the Arctic Ocean Boundary Current can transport the Atlantic pelagic copepod population into the Siberian shelf waters, where its abundance was further elevated by higher food availability (Kosobokova and Hirche, 2009). Furthermore, based on food-dependent development, accelerated population recruitment was expected in any other regions with elevated primary production or reduced ice coverage, supporting an increase in biomass of C. hyperboreus even in the Canada Basin with low phytoplankton biomass through advection (Hunt et al., 2014).