Continuous daylight in the high-Arctic summer supports high plankton respiration rates compared to those supported in the dark

Plankton respiration rate is a major component of global CO2 production and is forecasted to increase rapidly in the Arctic with warming. Yet, existing assessments in the Arctic evaluated plankton respiration in the dark. Evidence that plankton respiration may be stimulated in the light is particularly relevant for the high Arctic where plankton communities experience continuous daylight in spring and summer. Here we demonstrate that plankton community respiration evaluated under the continuous daylight conditions present in situ, tends to be higher than that evaluated in the dark. The ratio between community respiration measured in the light (Rlight) and in the dark (Rdark) increased as the 2/3 power of Rlight so that the Rlight:Rdark ratio increased from an average value of 1.37 at the median Rlight measured here (3.62 µmol O2 L−1 d−1) to an average value of 17.56 at the highest Rlight measured here (15.8 µmol O2 L−1 d−1). The role of respiratory processes as a source of CO2 in the Arctic has, therefore, been underestimated and is far more important than previously believed, particularly in the late spring, with 24 h photoperiods, when community respiration rates are highest.


Results
Plankton community respiration varied three orders of magnitude among communities, and was significantly higher in the communities sampled in the Svalbard region compared to those sampled in Young Sound (Kruskal-Wallis test, P < 0.01), both when measured in the light and in the dark (Table 1). Mean monthly community respiration rates in the Svalbard region were highest in April and lowest in August (Fig. 1), although these differences were only significant for respiration rates measured in the light (Kruskal-Wallis test, P = 0.013), when rates measured in August were significantly lower than those measured in April and May (Dunn's test, P < 0.05), but not June (Dunn's test, P = 0.27). The statistical significance of seasonal differences could not be tested for the communities examined in Young Sound (Fig. 1), due to the limited number of estimates available and the fact that respiration rates in Young Sound were not evaluated in the spring, when the area is still fully covered by sea ice.
Community respiration rates in the light differed with depth (Kruskal-Wallis test, P = 0.0015, Fig. 2A,B), with the respiration rate in the light in communities sampled at the depth receiving 20% of PAR (photosynthetically active radiation) being significantly higher (Dunn's test, P = 0.0014) than those sampled at the depth of chlorophyll maximum, DCM, and surface samples having the minimum mean respiration rate in the light among the three depths (Table 1). In contrast, community respiration in the dark did not differ with depth (Kruskal-Wallis test, P = 0.53), with comparable mean values across depths (Table 1 and Fig. 2C,D).
Community respiration rates evaluated in the light and in the dark differed consistently (Wilcoxon signed rank test, p < 0.0001), with respiration rates in the light tending to be greater than those measured in the dark (Fig. 3A). The difference between R light and R dark did not differ significantly with depth (Kruskal-Wallis test, P = 0.19), but was greatest in May, when R light tended to be much higher than R dark , compared to a smaller absolute difference in June and August (Fig. 1, Kruskal-Wallis test, P = 0.0085; Dunn's test, P < 0.05). Closer examination showed that community respiration rates evaluated in the light and in the dark differed significantly for the communities evaluated in the Svalbard region (Wilcoxon signed rank test, p < 0.001), but not so for those in Young Sound (Wilcoxon signed rank test, p = 0.22), where community respiration rates were consistently low. The ratio R light :R dark varied three orders-of-magnitude across communities (Table 1), and increased significantly (R 2 = 0.50, P < 0.001) in communities showing high respiration in the light (Fig. 3B). The fitted regression equation showed that the ratio R light :R dark was scaled to the 2/3 power of R light (Fig. 3B), so that the R light and R dark were similar for R light of the order of 1 µmol O 2 L −1 d −1 , but R light was four-fold greater than R dark for high R light rates of the order of 10 µmol O 2 L −1 d −1 (Fig. 3B).
The difference between community respiration rates evaluated in the light and in the dark increased significantly with increasing GPP 18 O rates (Fig. 4), with no significant difference in community respiration rates at GPP 18 (Fig. 4).

Discussion
The respiration of plankton communities is a major component of the carbon budget of the oceans 2 . Yet, estimates of community respiration rates are much less frequent than those of primary production, particularly in the Arctic Ocean where community respiration rates had thus far been evaluated only in the dark [11][12][13][14][15] . We found that R light tended to be significantly higher than R dark across the Arctic plankton communities tested. This is consistent with the majority of reports concluding that respiration in the light tends to be greater than that in the dark 6, 7, 16-18 , involving all, except two 19,20 , published reports comparing such rates. However, the underestimation of respiration rates derived from measuring respiration rates in the dark may be particularly acute for Arctic plankton communities, which experience a 24-hour photoperiod during much of the year. The estimates of R light provided here represent the first assessment of respiration under ambient solar radiation for Arctic plankton communities. Previous comparison of R light and R dark for polar plankton communities derived from the Southern Ocean, where two studies had been conducted 6,18 . These studies also concluded that respiration in the light tends to be greater than that in the dark. The mean vertically-integrated R light :R dark ratio was reported to be 1.95 in a summer cruise around 76 °S in the Ross Sea 6 ; and to range between 1.2 and 2 for spring and summer, respectively, in a transect from 52 to 70 °S across the Antarctic Polar Front region 18 . The median R light :R dark ratio in our study was 1.57, within the range of values reported for Southern Ocean plankton communities 6,18 . We found, however, that the R light :R dark ratio increased as the 2/3 power of R light so that the R light :R dark increased from an average value of 1.37 at the median R light measured here (3.62 µmol O 2 L −1 d −1 ) to an average value of 17.56 at the highest R light measured here (15.8 Estimates of gross primary production obtained directly using the 18 O method tend to be greater than those calculated as the difference between NCP and R dark , which comprise all of the estimated gross primary production thus far available for the Arctic Ocean [11][12][13][14][15] . This was interpreted to indicate that R light tends to be higher than R dark 21 , as confirmed by our results. Indeed, when the estimates of R dark obtained here are corrected for the underestimation derived from estimating this rate in the dark by multiplying them by the R ligth :R dark ratio predicted from the regression equation in Fig. 3, the NCP predicted as the difference between GPP 18 O and this corrected R estimate is strongly consistent with the observed NCP (Fig. 5). Hence, whereas reported NCP for plankton communities in the Arctic Ocean 11-15 are robust, previous estimates of gross primary production and respiration rates are underestimates. The reason for this is that the assumption, rejected by our experimental results, that R light equals R dark is particularly inadequate for the high Arctic, where plankton communities do not experience darkness within the photic zone during the 24 h photoperiods in spring and summer.
It has been suggested that R light rates are higher than those in the dark due to the contribution of autotrophic metabolic processes, such as photoenhanced mitochondrial respiration, chlororespiration, photorespiration and/ or the Mehler reaction 22 . Autotrophic respiration has also been proposed to dominate community R during the pre-bloom and bloom phases of the seasonal cycle in the Southern Ocean 23 . These observations are consistent with the observation that the difference between R light and R dark estimates increased with increasing gross primary production (Fig. 4). Figure 4 also shows that for GPP 18 O < 10 µmol O 2 L −1 d −1 , most R dark tend to be higher than R light , reflecting that metabolic processes supporting R light may be specially enhanced over a GPP 18 O threshold of 10 µmol O 2 L −1 d −1 , below which dark processes prevail.
In conclusion, the results presented show that respiration in the light tends to be much higher than that in the dark in productive communities, whereas both values are low in communities with low productivity. Periods of high production, particularly the spring bloom, contribute disproportionately to the annual metabolic budget of the Arctic Ocean 11 . Estimates of net community production in the Arctic 11,15 , which are derived from incubations in the light, are not affected by the bias introduced by dark incubations to estimate respiration rates. However,  these procedures would have led to underestimate the gross primary production of Arctic communities in the summer, where this was derived as the difference between NCP and respiration rates.

Methods
Plankton community respiration (R) in the dark and under ambient irradiance conditions was evaluated in both sides of the Greenland Sea, the western margin of Svalbard region and Young Sound fjord, in NE Greenland (Fig. 6) In the cruises conducted in the Svalbard region, water samples were collected using a Rosette sampler system fitted with 5 L Niskin bottles and a calibrated CTD, at three different depths: surface (2.12 ± 0.13 m), DCM (24.56 ± 1.63 m), which receives, on average, 1% of the incident irradiance, and at an intermediate depth (13.56 ± 0.93 m) between surface and DCM, receiving 20% of the incident radiation on the surface. Only two depths (surface and DCM) were sampled in Young Sound, where temperature and salinity were collected from a CTD cast, and water samples were collected with 5 L Niskin bottles.
Plankton community respiration rates were estimated using two methods: (1) respiration in the dark (R dark ) was assessed by evaluating oxygen consumption after incubation of samples in the dark, by high-precission Winkler titration 24,25 in Svalbard cruises and by visual end-point detection 26 in Young Sound, and (2) respiration in the light (R light ) was assessed as the difference between gross primary production (GPP 18 O), evaluated using H 2 18 O additions 27 , and net community production (NCP), evaluated from oxygen changes resolved using high-precision Winkler titration 24,25 in Svalbard and using visual end-point detection 26 in Young Sound, of  samples incubated under the incident solar radiation. Daily R light rates were corrected for those communities that were exposed to less than 24 hours of light (only five communities in September and October in Young Sound). The rates determined based on disolved oxygen changes in Young Sound, 12 out of 147 respiration rates reported here, carry considerable error, as the titration end point was determined visually, as a titrator was not available. The precision obtained (expressed as SD of average in %) for O 2 concentration measurements with this procedure was 0.15%.
Per each depth, a set of seven replicated 100-mL narrow-mouth Winkler bottles was fixed immediately to evaluate the initial oxygen content, and two sets were incubated under light and dark conditions for 24 hours. Incubations were done in water baths on deck (maintained at the in situ temperature of the surface water, ±1 °C, through continuous water flow from the surface) in Svalbard; and in situ in Young Sound. Neutral screens were used to reduce incident irradiance as to mimic the light environment in situ. Dissolved oxygen concentrations were determined by automated high-precision Winkler titration with a potentiometric end-point Metrohm 808 Titrando 28 in the Svalbard communities and using starch as indicator for end-point detection in the Young Sound communities. R dark and NCP were calculated from changes in dissolved oxygen concentrations, before and after incubation of samples under "dark" and "light" conditions, respectively, for 24 h in Svalbard and 48 h in Young Sound. As a consequence of the low rates and low precision of dissolved oxygen determination in Young Sound, the communities were incubated for 48 h, thereby experiencing changes that could be resolved with the techniques used there. On the other hand, long incubations may increase the risk of artifacts derived from bottle effects. Rates are reported in µmol O 2 L −1 d −1 and standard errors were calculated using error propagation. In order to compare the R light :R dark ratios obtained here with those observed in the past, we surveyed the literature for results reported in the past 6,7,[16][17][18]27 . An extreme value reported by one of the studies 7 (ratio R light :R dark = 19), 8-fold higher than the rest, was excluded from the comparison.
For evaluation of GPP 18 O, eight 12-ml glass vials were filled per depth. Four replicate samples were immediately fixed (biological activity stopped) with 80 µl of saturated HgCl 2 solution for later analysis of initial δ 18 O(O 2 ) values, and stored upside down in darkness. The other four vials, containing beads inside to ensure mixing, were spiked with 80 µl and 200 µl of 98% H 2 18 O in Svalbard and Young Sound communities, respectively. After being closed, these spiked vials were immediately agitated, to ensure that H 2 18 O was homogeneously distributed inside the vial. The spiked samples were incubated together with the Winkler bottles under "light" conditions. After the 24-hour incubation, vials were fixed with 80 µl of saturated HgCl 2 solution and stored upside down in darkness.
At the stable isotope laboratory, a 4-ml headspace was generated in each vial, by flushing with a helium flow. The vials were left for equilibration at room temperature for 24 hours. The δ 18 O of dissolved oxygen in the headspace was measured in a Finnigan GasBench II attached to a Finnigan DeltaPlusXP isotope ratio mass spectrometer, with precision better than 0.1‰. δ 18 O-H 2 O of spiked samples was measured in a liquid water isotope analyzer (Los Gatos Research), with precision of 0.1‰, and GPP 18 O was calculated 22 .
Statistical analysis were based on non-parametric tests (Wilcoxon signed rank test, Kruskal-Wallis test and Dunn's test), as the data were skewed and not normally distributed, or log-transformed to homogenize the variance prior to fitting least squares linear regression equations.