Accounting for variation in temperature and oxygen availability when quantifying marine ecosystem metabolism

It is critical to understand how human modifications of Earth’s ecosystems are influencing ecosystem functioning, including net and gross community production (NCP and GCP, respectively) and community respiration (CR). These responses are often estimated by measuring oxygen production in the light (NCP) and consumption in the dark (CR), which can then be combined to estimate GCP. However, the method used to create “dark” conditions—either experimental darkening during the day or taking measurements at night—could result in different estimates of respiration and production, potentially affecting our ability to make integrative predictions. We tested this possibility by measuring oxygen concentrations under daytime ambient light conditions, in darkened tide pools during the day, and during nighttime low tides. We made measurements every 1–3 months over one year in southeastern Alaska. Daytime respiration rates were substantially higher than those measured at night, associated with higher temperature and oxygen levels during the day and leading to major differences in estimates of GCP calculated using daytime versus nighttime measurements. Our results highlight the potential importance of measuring respiration rates during both day and night to account for effects of temperature and oxygen—especially in shallow-water, constrained systems—with implications for understanding the impacts of global change on ecosystem metabolism.

measurements conducted in the light (on different days associated with "Day" and "Light" measurements made during "Day-Night" and "Light-Dark" sampling, respectively) showed that oxygen fluxes measured using the two methods were similar across dates (F 1,4 = 4.7, p = 0.096) and that net community production did not change seasonally (by Date; F 7,28 = 1.2, p = 0.330; Fig. 1a). In contrast, rates of oxygen consumption in the dark were much greater when samples were collected during daytime "Dark" incubations than during nighttime "Night" sampling (F 1,4 = 139.1, p < 0.001), and they changed substantially over time (F 7,28 = 8.5, p < 0.001; Fig. 1b). Specifically, seasonal changes in community respiration were more pronounced in daytime "Dark" samples than in those collected at "Night" ('Method × Date' interaction, F 7,28 = 8.5, p < 0.001). Because volume was closely correlated with area in these tide pools (r 2 = 0.97), we observed similar patterns when measurements were expressed on a per-area basis (i.e., mg O 2 m −2 h −1 ; Supplementary Fig. S1a, b).
Because of these differences in respiration between sampling methods, gross community production (GCP)calculated by adding the rate of oxygen consumption in the dark to the rate of oxygen production in the lightwas higher when using the "Light-Dark" method than when using the "Day-Night" method (F 1,4 = 67.3, p < 0.001; Fig. 1c). There was seasonality in GCP (F 7,28 = 9.0, p < 0.001), and seasonal variation depended on the method used ('Method × Date' interaction, F 7,28 = 8.5, p < 0.001). As with the measured changes in oxygen concentrations, we observed similar patterns for GCP when values were quantified on a per-area basis ( Supplementary Fig. S1c).

Figure 1.
Effects of methodology on measured changes in oxygen (O 2 ) concentrations and gross community production (GCP) in tide pools. (a) Measurements made in the light (estimates of net community production) generally indicated net production of O 2 , were similar regardless of method ("Light" vs. "Day"), and did not change over time. (b) Measurements in the dark (estimates of community respiration) typically indicated net consumption of O 2 , differed substantially depending on method ("Dark" vs. "Night"), and changed over time.
(c) Measurements of GCP made consecutively during daytime (Light-Dark) resulted in higher estimates of GCP than those made during day and night (Day-Night). Overall, GCP varied with time, and the difference between methods changed with time. Values are means ± s.e., and "nd" marks months when "no data" were collected. Asterisks (*) indicate differences between methods (p < 0.05).   Values are means ± s.e., and "nd" indicates months when "no data" were collected. Asterisks (*) indicate differences between methods (p < 0.05). www.nature.com/scientificreports/ Integrated estimates of production and respiration. Using "Night" values to estimate O 2 consumption rates (Fig. 4a) resulted in considerably reduced (less negative) estimates for CR relative to the other two methods (Fig. 4b, c), which used daytime "Dark" measurements for one or both estimates of CR. Overall, using only Day-Night values resulted in annual estimates of CR (mg O 2 L -1 yr -1 ) an order of magnitude lower than the other methods ( Fig. 4a-c). Additionally, using daytime (Light-Dark) "Dark" values to estimate nighttime CR rates (Fig. 4b) resulted in estimates of daily O 2 consumption that were qualitatively more rapid (i.e., more negative, indicating O 2 consumption) than those that combined "Dark" and "Night" measurements ( Fig. 4c).
Whereas there was overlap between integrated annual CR values estimated using Light-Dark and combined methods, the Light-Dark method estimated higher O 2 consumption when applying daytime-derived respiration to the overnight periods as well (Fig. 4b,c). Using only "Dark" values from Light-Dark sampling (Fig. 4e) resulted in much lower rates of 24-h NCP (mg O 2 L −1 day −1 ) than the other two methods, including a period from January to May where estimates of production were less than zero (Fig. 4d,f). Using exclusively the Light-Dark method to estimate annual NCP (mg O 2 L −1 year −1 ) suggested that production did not exceed respiration, as average values were negative, although the 95% confidence interval spanned zero (Fig. 4e). Both the Day-Night method and the combined method indicated positive NCP values, but the estimate using only Day-Night sampling was lower (Fig. 4d).
Using the Day-Night method to estimate daily GCP for daylight hours (mg O 2 L −1 day −1 ; Fig. 5a) resulted in values that were considerably lower than those obtained using the Light-Dark method, especially during the summer (Fig. 5b). Similarly, annual GCP estimates (mg O 2 L −1 year −1 ) based on the Day-Night method were substantially lower than those based on the Light-Dark method (Fig. 5).

Discussion
Our results highlight the importance of accounting for the effects of temperature and oxygen in estimating ecosystem metabolism. The most important factor influencing net community production was O 2 concentration (NCP declined as tide pool O 2 concentrations increased, though this could also reflect reduced concentration gradients as tide pool O 2 levels increased), and the most important factor influencing gross community production was temperature (GCP increased at higher temperatures). Despite our efforts to minimize the potential for www.nature.com/scientificreports/ high O 2 levels to influence respiration during daytime "Dark" measurements, higher O 2 concentrations during the day (Fig. 3a) likely reflected higher O 2 production, potentially resulting in higher photosynthate enrichment and O 2 consumption during those measurements. Similarly, estimates of CR and, consequently, GCP were strongly dependent on whether dark measurements were made during the day ("Dark") or at night, suggesting that the ideal method depends on which aspects of ecosystem metabolism researchers aim to quantify. Collectively, our results indicate that a key assumption of "free-water" estimates of ecosystem metabolism-that respiration rates measured at night are equivalent to those during the day 31 -does not hold in this system. We observed substantially higher rates of O 2 consumption (community respiration, CR) during daytime "Dark" incubations than during nighttime "Night" incubations ( Fig. 1b). This was not entirely unexpected, as Odum noted that earlier researchers 29 had found substantial variation in respiration rates in dark bottles between day and night 22 . However, Odum then dismissed this complication and assumed constant community respiration 22 . The daytime CR rates we observed were higher than those measured at night, but they were consistent with other measurements from temperate rocky shores. For example, on an areal basis, our measurements (73-372 mg O 2 m −2 h −1 ; Supplementary Fig. S1b) were similar to those recorded in a series of trials on the coast of Norway (133-511 mg O 2 m −2 h −1 ) 43,44 . We found that community respiration rates were closely associated with temperature and O 2 concentrations: O 2 consumption increased with increasing temperatures and initial O 2 levels (Fig. 2), and diel variation in temperature and O 2 explained the difference between "Dark" and "Night" incubations. In fact, when O 2 and temperature were taken into account (i.e., included in the statistical model), the differences between respiration rates measured in "Dark" and "Night" incubations were no longer evident ( Supplementary Fig. S3). Given substantial diel variation in temperatures and O 2 levels recorded in a variety of shallow marine and freshwater systems [35][36][37][38] , we suggest that the assumption of "constant community respiration day and night" 22 should be reevaluated in some instances. To assess this possibility, we recommend measuring temperature and O 2 levels to assess differences between day and night conditions. And whereas we did not find an effect of irradiance on respiration rates, we cannot discount the potential for light-enhanced dark respiration 45 , and we suggest measuring irradiance levels.
Daytime "Dark" community respiration rates were substantially higher than those measured at night (Figs. 1b, 4a-c), leading to correspondingly higher estimates of hourly (Fig. 1c) and integrated (daily and annual) gross community production (Fig. 5). For estimating daily and annual rates of community metabolism, we advocate a "combined" method that uses day and night measurements to calculate net community production (NCP; Fig. 4d,f) and that uses daytime "Dark" respiration rates during daytime hours and "Night" respiration rates during nighttime hours to estimate community respiration (CR; Fig. 4c). If daytime and nighttime conditions (e.g., temperature, O 2 ) differ substantially, estimates of daytime gross community production (GCP) should be based on daytime CR measurements whenever possible. We predict that daily estimates of GCP are underestimated by up to an order of magnitude and annual estimates are almost three times lower when "Night" CR measurements are used (Fig. 5). www.nature.com/scientificreports/ One important potential caveat to our suggestion and interpretation is the possibility for light-enhanced respiration. Higher rates of photosynthesis-associated with higher irradiance levels-are associated with higher respiration rates 34,45,46 , leading to higher respiration rates measured during the day in producer-dominated communities 47 . While this is a possible explanation for the observed differences between daytime and nighttime respiration, the fact that GCP was related to temperature, but not to irradiance levels, suggests that temperature was a primary factor. And when tide pool temperatures and O 2 concentrations were taken into account in our statistical analysis of factors influencing CR, differences between daytime and nighttime CR measurements were no longer apparent (Supplementary Fig. S3). Animal respiration-and the factors that influence it-was also an important aspect of these diverse, natural communities, as highlighted by the relationship between CR and abundances of littorine snails. Whereas light-enhanced respiration is clearly important in many systems 34,[45][46][47] , these findings suggest that it does not play a major role in these tide pools relative to other drivers of ecosystem metabolism.
Measurements of community production and respiration are essential for quantifying whole-system responses to anthropogenic changes in environmental conditions 48 . Both O 2 and temperature-which we have identified as strong drivers of ecosystem metabolism (Fig. 2)-are changing at both local and global scales due to climate warming and climate-mediated reductions in O 2 availability 1,6,7,49 . We highlight important mechanisms-temperature and O 2 effects on community respiration (CR) rates (Fig. 2)-by which these changing environmental conditions can modify ecosystem functioning. These effects are, therefore, not only methodological considerations, but they also provide insights into impacts of global change on ecosystems. We found that CR responded strongly to changes in temperature and O 2 levels, which is consistent with a greater response of respiration than production to ocean warming 50,51 and reflects previous work linking respiration to temperature in aquatic systems 50,52,53 . Whole-system warming experiments highlight the role of temperature in increasing respiration relative to production, reducing carbon sequestration 54 .
How system-general are the relationships we observed? The differences between nighttime and daytime respiration rates were highly seasonal, occurring in the spring and early summer, when productivity (enhancing O 2 levels 6 ) and solar insolation (enhancing temperature 55 ) were highest (Fig. 1b). Community respiration rates increased as both temperature and O 2 levels increased (Fig. 2), reflecting higher temperatures and dissolved O 2 concentrations during the day (Fig. 3). Thus, the assumption that nighttime and daytime respiration rates are comparable 22 is likely to be violated in locations where there is substantial seasonal and diel variation in environmental factors (e.g., temperature, O 2 ) that influence respiration. Thus, smaller and shallower bodies of water (e.g., lakes, lagoons, estuaries, tide pools) are more likely to exhibit variation than larger ones 35,37,56,57 .
In conclusion, we found that measuring whole-community respiration rates during day and night-which was perhaps uniquely possible in the contained water of tide pools-resulted in markedly different estimates of respiration and production. These differences were associated with substantial variation in environmental conditions during night and day. Especially in systems characterized by diel variation in factors such as temperature and O 2 , the assumption that daytime and nighttime respiration rates are equivalent may not hold 57 . In these cases, if in-situ daytime respiration rates cannot be accurately measured, there may be a need to follow an alternative method for estimating gross community production and other key components of ecosystem metabolism (e.g., the triple oxygen tracer approach 57,58 ) or to incorporate modeling approaches that relate ecosystem metabolism to temperature 59 and O 2 . More broadly, our work demonstrates that temperature and O 2 concentrations-both of which are changing due to human activities 1,6 -can strongly affect rates of ecosystem metabolism, influencing system-and global-scale processes such as carbon sequestration.

Study site and measurement schedule. Our study was conducted on rocky intertidal reefs at John
Brown's Beach on Japonski Island, Sitka, Alaska, USA (57.06° N, 135.37° W) 60 between September 2018 and September 2019. Measurements spanned a year to capture effects of season and temperature on ecosystem metabolism 61 . Tide pools (n = 5, randomly selected from pools at the site) ranged from 3 to 28 L in volume, averaging 9.4 ± 4.7 L (mean ± s.e.m.), and were located from 2.3 to 3.0 m in elevation above mean lower-low water, averaging 2.7 ± 0.1 m. Pool water temperatures ranged from a minimum of 4.6 °C in January to a maximum of 23.1 °C in July. Daylight hours (sunrise-to-sunset) during measurements ranged from 7.5 h in January to 17 www.nature.com/scientificreports/ Bracken, unpublished data). Nighttime samples were collected on nights when the pools were isolated by the tide after dark (measured irradiance levels of 0.0 µmol photons m −2 s −1 ). The second set of measurements consisted of light and dark (Light-Dark) sampling during the daytime. Protocols were similar to those used to calculate light-dark bottle estimates of productivity in oceanographic and limnological applications 12,20 , modified for in-situ measurements of oxygen fluxes in tide pools to estimate net community production, community respiration, and gross community production 19,39 . Measurements were started shortly after pools were isolated by the receding tide, when irradiance levels were sufficient to saturate photosynthesis by N. oregona (see above). Immediately after initial measurements of O 2 , irradiance, and temperature (see above), pools were covered with opaque black plastic sheeting (6-mil black polyurethane plastic sheeting, Film Gard, Berry Plastics, Evansville, Indiana, USA), which was anchored to the substratum with cobbles to ensure that no light entered the pool during the "Dark" incubation. "Dark" incubations were conducted prior to "Light" incubations to minimize the potential for high initial O 2 concentrations, which could lead to photorespiration 63 and reduce the concentration gradient between the primary producers and the tide pool water. Time of day did not affect initial dissolved O 2 concentrations (F 1,23 = 1.3, p = 0.259), likely because measurements were made as soon as possible after the receding tide isolated the tide pools. Irradiance levels beneath the plastic sheeting were uniformly 0.0 µmol photons m −2 s −1 , and the sheeting did not alter tide pool temperatures during the "Dark" incubation (mean ± s.e.m. = − 0.04 ± 0.16 °C; t = 0.2, d.f. = 4, p = 0.832). After an incubation period of at least 30 min (median = 45 min, maximum = 80 min), we made a second set of measurements and removed the plastic sheeting, initiating a "Light" incubation of at least 30 min. At the end of the "Light" incubation, a final set of measurements was made.
We made additional measurements to ensure that covering tide pools did not affect rates of production and respiration. During June and July, when differences between daytime and nighttime conditions were likely to be greatest, we measured daytime changes in O 2 concentrations with and without transparent plastic sheeting (8-mil Crystal Clear vinyl sheeting, Frost King, Thermwell Products Co., Inc., Matwah, New Jersey, USA) and found that rates of O 2 accumulation were not different in transparent-covered and uncovered pools (paired t-test: t = 0.5, d.f. = 4, p = 0.761; Supplementary Fig. S4). Light levels did not differ between uncovered and transparentcovered pools (paired t-test: t = 2.2, d.f. = 4, p = 0.248). Note that some UV radiation was likely blocked by the vinyl sheeting, but the lack of a difference in O 2 accumulation rates suggests that effects were minimal. Similarly, we conducted nighttime incubations with and without opaque plastic sheeting and found that rates of O 2 depletion were not different in opaque-covered and uncovered pools (paired t-test: t = 1.2, d.f. = 4, p = 0.505; Supplementary  Fig. S4). Furthermore, median wind velocities (measured at the adjacent Rocky Gutierrez Airport) were low enough during sampling (median = 2.8 m s −1 ) that O 2 diffusion across the air-water interface should not have affected measurements 19 , and thus covering tide pools at night should not have been necessary. Wind speeds during measurements were typical of velocities measured throughout the year (median = 2.6 m s −1 ).
Analyses. We measured changes in O 2 concentrations in the light (Light and Day) to estimate net community production (NCP) and changes in O 2 concentrations in the dark (Dark and Night) to estimate community respiration (CR). In both cases, we calculated the difference between the final (f) and initial (i) concentrations and divided by the elapsed time (t) in hours: Values were calculated in mg O 2 L −1 h −1 , as those units reflected the concentrations measured on our optical DO meter. However, tide pool volume and area were strongly correlated (r 2 = 0.97), we used estimates of volume (L) and area (m 2 ) to calculate O 2 changes in mg m −2 h −1 , and we include those values in the Supplementary Information ( Supplementary Fig. S1). Estimates of NCP were typically positive, due to increases in O 2 concentrations in the light, whereas estimates of CR were typically negative, due to declines in O 2 concentrations in the dark. Independent estimates of NCP and CR were made using each set of measurements: Day-Night (daytime NCP and nighttime CR) and Light-Dark (Light NCP and Dark CR).
Gross community production (GCP) was estimated by adding together net community production (NCP) and the absolute value of community respiration (CR): Independent estimates of GCP were made using the Day-Night (daytime NCP + nighttime CR) measurements and the Light-Dark (light NCP + dark CR) measurements.
To evaluate differences between methods, we used repeated-measures analyses of variance in SAS v. 9.4 (SAS Institute, Inc., Cary, North Carolina, USA), after verifying both univariate and multivariate assumptions of normality and homogeneity of variances. Analyses were conducted using proc mixed, with 'Date' as the withinsubjects variable (i.e., measurements were repeated at each pool over time) and 'Pool' designated as the subject. Comparisons were made between methods (e.g., oxygen fluxes measured in daytime "Dark" incubations vs. oxygen fluxes measured at "Night"), including changes through time (i.e., 'Method × Date' interactions). We also explored the role of environmental factors (temperature, O 2 , light) in driving variation in productivity using the same analytical structure. www.nature.com/scientificreports/ Daily and annual productivity estimates. We generated estimates of 24-h, integrated community respiration (CR) and net community production (NCP), and daylight gross community production (GCP) for a 1-yr period from 18 Sep 2018 to 17 Sep 2019 that fell within the range of our sampling dates. We made the simplifying assumption that O 2 fluxes would be the same in submerged pools as in isolated pools. This is a reasonable assumption, as these pools are high on the shore (2.3 to 3.0 m above mean lower-low water) relative to the maximum high water level during the year (3.8 m) and, based on tidal predictions, were submerged for only 10.5 ± 3.2% of the time over the year. We also made the assumption that for the primary producers in the tide pools, irradiance would be saturating immediately at sunrise and until sunset, without attempting to account for individual patterns of incident irradiance in each pool. This assumption is reasonable given the low irradiance levels (< 50 µmol photons m −2 s −1 ) required to saturate photosynthesis of the most abundant seaweed species in the tide pools (see above). Using the values measured in the pools on the 8 sampling dates, we linearly interpolated values of CR, NCP, and GCP for both the Light-Dark and Day-Night determination methods, for each intervening day. For each day in the 365 d dataset, we calculated daily CR (mg O 2 L −1 d −1 ) using CR estimates from the (1) daytime "Dark" measurements only, (2) Night measurements only, and (3) Dark measurements during daylight hours and Night measurements during nighttime hours. Similarly, we used three methods to calculate integrated estimates of daily NCP (mg O 2 L −1 d −1 ) over each 24 h period. (1) For the Light-Dark method, we used the "Light" NCP measurement multiplied by the number of daylight hours, and then subtracted the "Dark" CR under the opaque plastic sheet, multiplied by the number of nighttime hours to estimate the total 24 h NCP. (2) For the Day-Night method we used the "Day" NCP estimate multiplied by the number of daylight hours, and the "Night" CR value multiplied by the number of nighttime hours to calculate a 24 h NCP value. (3) We also combined the two methods, using the Light-Dark daytime ("Light") NCP estimate multiplied by the number of daylight hours, and the Day-Night nighttime ("Night") CR estimate multiplied by the number of nighttime hours to calculate an integrated 24 h value.
Integrated estimates of gross community production (GCP, mg O 2 L −1 d −1 )-which only occurs during daylight hours when oxygen is produced and consumed-were calculated from the NCP value plus the CR value for the Light-Dark and Day-Night methods, multiplied by the number of daylight hours on each date.

Data availability
The datasets generated during and/or analyzed during the current study are available in the Dryad repository, https:// doi. org/ 10. 7280/ D1M39B 62 .