An autonomous, in situ light-dark bottle device for determining community respiration and net community production
Availability of data and code: A MATLAB script to read in, process, and estimate rates and uncertainties in dissolved oxygen data from the PHORCYS is provided online at https://github.com/jamesrco/DO_Instruments/. The script can be easily adapted to calculate -based estimates of uncertainty in any dissolved oxygen time series. All PHORCYS and Winkler titration data and other scripts required to reproduce the results and figures in this work are available online in the same location.
We describe a new, autonomous, incubation-based instrument that is deployed in situ to determine rates of gross community respiration and net community production in marine and aquatic ecosystems. During deployments at a coastal pier and in the open ocean, the PHORCYS (PHOtosynthesis and Respiration Comparison-Yielding System) captured dissolved oxygen fluxes over hourly timescales that were missed by traditional methods. The instrument uses fluorescence-quenching optodes fitted into separate light and dark chambers; these are opened and closed with piston-like actuators, allowing the instrument to make multiple, independent rate estimates in the course of each deployment. Consistent with other studies in which methods purporting to measure the same metabolic processes have yielded divergent results, respiration rate estimates from the PHORCYS were systematically higher than those calculated for the same waters using a traditional two-point Winkler titration technique. However, PHORCYS estimates of gross respiration agreed generally with separate incubations in bottles fitted with optode sensor spots. An Appendix describes a new method for estimating uncertainties in metabolic rates calculated from continuous dissolved oxygen data. Multiple successful, unattended deployments of the PHORCYS represent a small step toward fully autonomous observations of community metabolism. Yet the persistence of unexplained disagreements among aquatic metabolic rate estimates—such as those we observed between rates calculated with the PHORCYS and two existing, widely accepted bottle-based methods—suggests that a new community intercalibration effort is warranted to address lingering sources of error in these critical measurements.
Accurate, reproducible, and cost-effective estimates of aerobic respiration and primary production in aquatic systems are essential for research across a diverse array of disciplines in the environmental sciences (Del Giorgio and Williams 2005; Volkmar and Dahlgren 2006; Staehr et al. 2012). Rate measurements of these two metabolic parameters can be applied to various problems, including validating the biogeochemical components of global climate models (Denman et al. 2007), determining the trophic status of surface-water planktonic communities in open ocean ecosystems (Williams 1998), measuring rates of biological oxygen demand (BOD) in treated wastewater (Spanjers et al. 1994), and identifying unexpected metabolisms in the deep ocean (Reinthaler et al. 2010).
While an increasing demand for metabolic rate data has encouraged the development of many different methods for estimating rates of photosynthesis (Ducklow and Doney 2013), the number of new methods for measuring aerobic respiration at the community scale has lagged behind considerably (Del Giorgio and Williams 2005). The majority of field-based methods for measuring rates of respiration and primary production in the ocean fall largely into two categories: (1) in situ geochemical tracer techniques that track changes in the concentration or isotopic composition of dissolved oxygen and carbon dioxide within ocean water masses (e.g., the surface mixed layer), and (2) in vitro incubation techniques that track the rates at which plankton exchange oxygen or carbon dioxide in discrete seawater samples (i.e., bottle incubations). The merits and faults of these two categories of approaches have been vigorously debated while significant and often unexplained differences are noted in the rate estimates they yield (Duarte et al. 2013; Ducklow and Doney 2013; Williams et al. 2013). The former category has benefited considerably from recent advances in optical sensor technology (Moore et al. 2009), mass spectrometry (Goldman et al. 2015), and techniques for analysis of optical sensor data from autonomous underwater vehicles (Nicholson et al. 2015). By maximizing the extent to which sensors are integrated into the surrounding environment, low-power instruments increase the spatial and temporal resolution of geochemical tracers in situ and permit increasingly autonomous, long-term deployments (Prien 2007; Riser and Johnson 2008; Porter et al. 2009).
By contrast, the field has seen relatively few technical advances in in vitro incubation techniques. In vitro techniques provide an important complement to in situ methods because they are sensitive to short-term perturbations and are amenable to experimental design. For these reasons, the traditional two-point light and dark bottle incubation technique (Gaarder and Gran 1927) and the 14C incubation method (Steeman Nielsen 1952) continue to dominate incubation-based studies, although a number of other methods based on electron transport (e.g., Kenner and Ahmed 1975) or fluxes of 18O or CO2 (Bender et al. 1987; Robinson and Williams 2005) have been introduced over the course of the last half-century. A number of these methods have been incorporated into modern designs for benthic flux chambers and so-called “benthic landers,” enabling investigators to capture fluxes of oxygen and other gases in situ at the sediment–water interface instead of in core samples aboard ship (Hammond et al. 2004; compare, e.g., Fuchsman et al. 2015; Martens et al. 2016, or Lee et al. 2015 to Kim et al. 2016). However, even the most advanced of these devices can require the use of divers or remotely operated vehicles (ROVs) for deployment, maintenance, or recovery. Additionally, by their nature, few of these designs can be programmed to conduct multiple incubations over the course of a single deployment. Taylor et al. addressed this obstacle with the submersible incubation device (SID), which for the first time allowed multiple, unattended incubations with 14C-bicarbonate to be conducted in situ (Taylor and Doherty 1990; Taylor et al. 1993). The SID represented a significant advance but was limited by its relatively small 400 mL incubation chamber and its reliance on the use of radiolabeled reagents.
Among the advances most consequential for in situ instrumentation was the adaptation to marine applications of optical technologies such as optodes (e.g., Klimant et al. 1995; Tengberg et al. 2006) and optode sensor spots (e.g., Warkentin et al. 2007), which exploit fluorescence (luminescence) quenching to measure dissolved oxygen concentrations non-destructively and without themselves contributing to oxygen consumption in the sample. Integral optodes and sensor spots based on the same technology have now been successfully used in a variety of shipboard configurations to measure rates of gross community respiration in whole, unconcentrated and unfiltered water samples and in water containing particle material from marine and aquatic environments (Edwards et al. 2011; Wikner et al. 2013; Collins et al. 2015). More recently, shipboard in vitro measurements of respiration within individual marine particles were made successfully using oxygen microelectrodes (Belcher et al. 2016, 2016a,b). A significant recent advance was also achieved with the RESPIRE device, which uses an optode fitted into a modified sediment trap to make particle respiration measurements in situ (Boyd et al. 2015; McDonnell et al. 2015).
Despite the significant progress represented in these optode-driven systems, incubation-based methods remain prone to a number of sources of error that demand reconciliation. These can be generally divided into two categories: (1) those that result from the preparation for or act of incubating natural microbial populations and (2) errors inherent in the method used to determine the concentration of dissolved oxygen (e.g., Winkler titration, fluorescence-quenching optode, or Clark electrode). The sources of uncertainty associated with bottle/chamber incubations span both categories and include (1) contamination, disruption, or bias introduced through the process of obtaining seawater samples from depth and preparing them for incubation (Tamburini et al. 2013; Suter et al. 2016); (2) unrepresentative incubation conditions that do not faithfully reproduce the variations in temperature, turbulence, and light inherent in natural systems; (3) so-called “bottle effects” associated with low-volume incubations which may limit nutrient availability (Furnas 2002) or induce unnatural changes in community structure (Venrick et al. 1977; Calvo-Díaz et al. 2011); and (4), in the case of metabolic rate measurements extrapolated from Winker (1888) titrations, the lack of temporal resolution inherent in measurements based only on two endpoints. In any study where incubations are used, the choice of incubation methodology places inherent limits on the spatial and temporal resolution of the data collected (Karl et al. 2001). The integration of point measurements—data sparse in time and/or space, whether based on in situ observations or incubations—also creates significant representation error. One solution to this problem is to greatly increase the number of measurements made during data collection using automated technologies.
We describe here the PHOtosynthesis and Respiration Comparison-Yielding System (PHORCYS), a large-volume (i.e., > 2.5 L), light and dark chamber incubation system for autonomous measurement of rates of primary production and respiration at high temporal resolution and under in situ conditions. In designing the instrument, we endeavored to minimize the major hypothesized sources of uncertainty associated with traditional incubation-based methods while constructing a system that functions autonomously and interrogates water samples non-destructively. We also sought to eliminate or reduce the need for repeated wet-chemical field measurements such as Winkler oxygen titrations or reagent-based methods used in other autonomous systems. We first describe design and validation of the PHORCYS using two independent methods, and then present results of several deployments of the instrument in different ecosystem types.
Materials and procedures
Instrument design and operation
The PHORCYS is composed of only a few basic components, making the design highly scalable and cost-effective (Fig. 1a,b). In nearly all instances, “off the shelf” components of different size or capacity can be easily substituted for those we describe here. The PHORCYS consists of two polycarbonate plastic chambers (usable vol. 5.7 L; Table 1), several auxiliary sensors for collection of environmental data in the ambient water mass outside the chambers, and a watertight power supply, control, and data recording module (Fig. 1a). A piston-like, magnetically coupled actuator is programmed to open and close each chamber at a specified interval, allowing users to perform multiple, unattended incubations over the course of a single deployment. The chamber seals are tapered to avoid the use of rubber O-rings that might have introduced a source of organic contamination into the sample water. In experiments, the polycarbonate plastic used for the PHORCYS incubation chambers reduced photosynthetically active radiation (PAR) within the transparent chamber to 83% of incident strength; we present a means of accounting for this attenuation, below. The opaque chamber was darkened by application of a coating to the outside of the cylinder. All PHORCYS components are mounted to a stainless-steel frame, allowing the instrument to operate to a depth of 100 m.
|Bottle or chamber type||Actual usable volume (mL)||Estimated internal surface area (cm2)||Estimated surface area : volume ratio|
|PHORCYS chamber (prototype)||2610||1760||0.67|
|PHORCYS chamber (current model)||5680||2035||0.34|
|Typical 125 mL BOD bottle||149.2 ± 0.3||124.4 ± 5.0||0.83|
|Typical 300 mL BOD bottle||299.2 ± 0.4||229.1 ± 4.3||0.77|
Dissolved oxygen concentrations in the two chambers are monitored with fast-response, fluorescence quenching oxygen optodes (Aanderaa model 4531D; accuracy < 8 μM O2; resolution < 1 μM O2; response time < 30 s; Aanderaa Data Instruments, Bergen, Norway); each optode is fitted into its chamber using a water- and gas-tight flange assembly. The instrument also has several external sensors, including a third optode to monitor dissolved oxygen concentrations in the water mass outside of the two chambers, PAR sensor, beam transmissometer, chlorophyll fluorometer, and CTD. We did not investigate whether the arrangement of these sensors had any effect on the external dissolved oxygen field surrounding the instrument.
With a full sensor suite, the nominal power consumption of the PHORCYS is 50 mA at 12V; standby current is 2 mA. When programmed to sample at a 50% duty cycle, a single 12V primary “D” cell battery pack (20,000 mAh capacity) will power the instrument for up to 30 d in an unattended deployment mode. Data are recorded in an ASCII fixed-field format onto a micro SD card in a DOS-readable format. For attended deployments, a combination communications and external power port provides the ability to observe data in real time, allow program updates, download data, and power the instrument indefinitely. The sampling interval is nominally set to 1 min, though data can be collected as frequently as every 15 s. The acquisition program determines sampling activity by way of a real-time clock. The chambers can thus be programmed to open and close at any time, allowing the investigator to make multiple incubations of any desired length. In the configuration used to acquire the data presented here, (Fig. 2, symbol; e.g., Fig. 3) the chambers were programmed to open at or around sunrise and the same operation was repeated at sunset, providing two incubations in each 24 h period that aligned with the beginning and end of the photoperiod. The chambers are opened and then closed sequentially (i.e., one after the other) to reduce total current draw from the power source. The chambers remain open for 30 min at the outset of each incubation, providing sufficient time for the water to be fully exchanged before closure; we confirmed this flushing time was sufficient in both quiescent and flowing (∼ 1 m s−1) waters using a series of tests with a tracer dye (results not shown). While we used a standard 30 min flush time in the deployments for which data is presented below, any time can be specified in the instrument's control software, which is written in BASIC.
The earlier PHORCYS data from 2012 (Fig. 2, symbols; e.g., Fig. 4) were obtained using a prototype instrument that permitted only one incubation cycle per deployment. This prototype (Fig. 1b) was assembled from two 2.5 L Niskin-style sampling bottles mounted to an aluminum frame (opaque polyvinylchloride and transparent polycarbonate plastic, respectively; actual usable volume, 2.6 L; General Oceanics, Miami, Florida, U.S.A.). Closure of the chambers for incubation was effected using an electrolytic time release (i.e., “burn wire”) system. Prior to deployment, the Niskin bottle endcaps were cocked open and the retaining cable was rigged to a fusible burn wire plug. Once in the water, a sufficient current was applied to the burn wire at a time set by the user, corroding the wire and allowing the bottle endcaps to close. The chambers were then sealed and the incubation began. For the deployments presented here, we programmed the chambers to close approximately 45 min after the PHORCYS had reached the desired depth. In the prototype instrument, Aanderaa model 4330F optodes (accuracy < 8 μM O2; resolution < 1 μM O2) were used to record dissolved oxygen concentrations.
We conducted six unattended deployments of the PHORCYS in three distinct ecosystem types in the North Atlantic basin (Fig. 2; Table 2, Supporting Information Table S1). Open-ocean deployments (2–7 d in length, using the prototype instrument) were conducted during cruises aboard the R/V Knorr; during these deployments, the instrument was suspended at various depths in the euphotic zone from a drifting surface buoy (Fig. 1c). Deployment and recovery were accomplished in 45–60 min from a standard oceanographic research platform (Supporting Information Fig. S1). An Argo satellite beacon (Fig. 1c) allowed us to track the array remotely between deployment and recovery while the ship traveled up to 300 km away to conduct other shipboard scientific operations; we specifically designed both models of the PHORCYS to be wholly autonomous, incurring a minimal burden on other shipboard operations. Pierside deployments were conducted using the present, multiple-incubation version of the PHORCYS at the Iselin Marine Facility, Woods Hole, Massachusetts, U.S.A. (41° 31′ 24″ N 70° 40′ 20″ W); the site adjoins a highly productive coastal embayment. In both cases, oxygen concentrations (μmol L−1 O2), percent saturation, and temperature were then recorded for each chamber at 1 min intervals. Post-acquisition corrections for salinity were applied to both the open-ocean and coastal data using concurrent observations of salinity and manufacturer-supplied correction coefficients. Concurrent salinity data were obtained from the continuous, flow-through CTD system aboard the R/V Knorr (for open-ocean deployments) or from the Seabird CTD unit mounted as an external sensor on the present PHORCYS model (2016 coastal deployments). Due to the limited number of at-sea deployments of the PHORCYS prototype and the challenges we encountered during our initial cruises, we pooled our prototype results with the data we later obtained from the production-model instrument in order to assemble a larger, more robust dataset for subsequent analysis.
|Deployment dates||Incubation period||Incubation duration (h)||Locationa||PHORCYS model†||
Community respiration (GR)
(μmol O2 L–1 d–1 ± uncertainty)
|PHORCYS opaque bottle‡||Shipboard incubations§||Two-point difference of Winkler titrations at t = 0 and recoveryǁ|
|24 Apr 2012–27 Apr 2012||Entire deployment||71.6||QL-1||Prototype||1.8 ± 0.2||3.2 ± 0.7||0.6 ± 0.1|
|30 Apr 2012–03 May 2012||Entire deployment||65.4||QL-2||Prototype||4.2 ± 0.3||1.1 ± 0.2||1.2 ± 0.04|
|17 Jun 2012–19 Jun 2012||Entire deployment||41.2||PS-1||Prototype||2.4 ± 0.3||3.4 ± 0.5||3.5 ± 0.2|
|23 Jun 2012–27 Jun 2012||Entire deployment||77.4||PS-2||Prototype||7.8 ± 0.4||4.0 ± 0.3||—|
|07 Jul 2012–11 Jul 2012||Entire deployment||94.0||PS-4||Prototype||6.0 ± 0.5||7.9 ± 0.6||7.4 ± 0.2|
|07 Nov 2016–08 Nov 2016||17:15–06:00||12.7||Iselin Pier||Present model||18.9 ± 1.9||—||5.9 ± 1.0|
|08 Nov 2016||06:15–16:45||10.5||Iselin Pier||Present model||2.2 ± 1.6||—||−0.8 ± 2.4|
|08 Nov 2016–09 Nov 2016||17:20–06:00||12.7||Iselin Pier||Present model||8.0 ± 1.9||—||4.3 ± 0.9|
|09 Nov 2016–10 Nov 2016||17:30–06:00||12.5||Iselin Pier||Present model||10.5 ± 7.5||—||5.4 ± 1.3|
- a Cruise station or geographical location (Fig. 2); additional metadata for each station are provided in Supporting Information Table S1.
- †See Fig. 1.
- ‡Uncertainty adjusted for effective degrees of freedom, as described in the Appendix.
- §Mean of ≥ 5 replicates; uncertainty derived from standard error of regression slope.
- ǁMean of 3 replicates; uncertainty derived from standard error.
Instrument calibration and choice of deployment depth
Instrument validation by two independent methods
First, to validate the optodes' ability to accurately track respiration, we used a standard analytical method—two-point Winkler titration—to determine dissolved oxygen consumption in triplicate water samples at the beginning and end of each incubation period (present model instrument) or deployment (for data obtained with the prototype). Winkler titrations were conducted in 125 mL BOD bottles according to EPA Method 360.2 as modified for shipboard determination in seawater (Knapp et al. 1989). Initial Winkler titrations were made in samples collected within 15 min of deployment using a Niskin or Go-Flo bottle suspended at the same depth as the instrument. A set of three darkened 125 mL BOD bottles containing water from the same Niskin or Go-Flo bottle was incubated at in situ temperature until the PHORCYS was recovered or the incubation period ended; these samples were then sacrificed according to the same protocol. The BOD bottles used in these incubations were triple-rinsed with 10% HCl and then Milli-Q water prior to sampling. All reagents for Winkler titrations were A.C.S. grade or better; the sodium thiosulfate titrant was standardized daily. Amperometric titration was performed using an autotitrator (Metrohm 904 Titrando; Metrohm U.S.A., Riverview, Florida).
As an additional means of comparison during the open-ocean deployments, we also tracked changes in dissolved oxygen in a series of continuously monitored shipboard bottle incubations. Water from the PHORCYS deployment depth was retrieved for these incubations from a hydrocast made within 1 h of deployment. Incubations were conducted in gas-tight, 300 mL glass BOD bottles; at least five replicates were used for each series of measurements. Determination of dissolved oxygen was made at 3–9 h intervals using optode spot minisensors (PreSens PSt3; response time < 40 s; Precision Sensing GmbH, Regensburg, Germany; Warkentin et al. 2007) that were glued to the inside surfaces of the bottles using food-quality silicone cement. The use of these optode spots eliminated the need for drawing of aliquots from the sample bottles; the bottles had been soaked in Milli-Q water for > 2 months following application of the sensor spots. Incubations were conducted in the dark at in situ temperature as described in Edwards et al. (2011).
PHORCYS rate calculations
To estimate uncertainties in PHORCYS rate estimates, we adapted an effective degrees of freedom technique traditionally applied to time-series data in physical oceanography (Emery and Thomson 2001). The method, which serves as an alternative to the standard error of the regression slope in instances when true replication cannot be achieved, is described in detail in the Appendix.
Rate calculations from Winkler titration samples and sensor spot incubations
PHORCYS metabolic rate estimates
We observed a significant degree of daily and hourly variability in the time series data from each PHORCYS deployment (e.g., Figs. 3, 4). This variability manifested itself in a wide range of metabolic rate estimates (Table 2, Supporting Information Table S1) that reflect the interaction of multiple biological and physical forcings, including diel changes in cellular growth cycle, surface-layer water temperature, and irradiance. Daily rates of GR were estimated from dark chamber data for all PHORCYS deployments (Table 2). For data obtained with the present model instrument, hourly rates were calculated for each incubation segment (Fig. 3); these were extrapolated to daily rates to facilitate comparison with other studies and with data from the prototype instrument. For open-ocean data obtained with the PHORCYS prototype, daily rates were calculated using the entire DO time series from each deployment. Erroneous readings from one of the optodes and a system malfunction prevented us from recovering usable NCP data from the transparent chamber during two of the open-ocean deployments of the prototype instrument. During the 24 April 2012–27 April 2012 deployment, the chosen depth of 29 m provided insufficient PAR (< 3% of surface intensity) to support measurable photosynthesis in the transparent chamber; consequently, we could not calculate NCP or GPP for this station (Supporting Information Table S1). An obstruction prevented the transparent chamber from closing during the November 2016 deployment, allowing us to recover useful data from only the dark chamber.
We captured daily rates of GR ranging from 1.8 ± 0.2 μmol O2 L−1 d−1 at a mid-latitude station in the North Atlantic to 10.5 ± 7.5 and 18.9 ± 1.9 μmol O2 L−1 d−1 in two different water masses at the Woods Hole pier in early November (Figs. 3, 4; Table 2). The wide variation in GR we observed with the PHORCYS covers a significant range of the rates for marine systems compiled by Robinson and Williams (2005). Daily rates of GPP at the PHORCYS deployment depth ranged from essentially zero on several days at a mid-latitude station in the North Atlantic (Supporting Information Table S1) to 3.6 ± 0.5 μmol O2 L−1 d−1 during a coccolithophore bloom (Collins et al. 2015) in the sub-Arctic North Atlantic (Fig. 4). Rates of NCP at the deployment depth ranged from −2.0 ± 0.4 μmol O2 L−1 d−1 to −4.2 ± 0.2 μmol O2 L−1 d−1 (Supporting Information Table S1).
Subdaily variation in rates of metabolism; choices concerning data analysis
Increasing evidence suggests the significant sub- and inter-daily variation in metabolic activity captured by the PHORCYS exists in almost all natural aquatic systems (Caffrey 2004; Staehr et al. 2012; Collins et al. 2013); even in oligotrophic waters, respiration and production rates may change significantly from 1 d (or hour) to the next, even as the system maintains an overall state of near trophic balance (Aranguren-Gassis et al. 2012). The types of fluctuations we observed in the various dissolved oxygen time series obtained from the PHORCYS appear to be characteristic of incubation-based in situ instruments. McDonnell et al. (2015) and Boyd et al. (2015) both observed similar patterns in dissolved oxygen data during recent deployments of an in situ device that measures oxygen consumption rates on marine particles.
Nevertheless, the subtle changes in DO concentration we observed in multi-day incubations with the PHORCYS prototype (e.g., Fig. 4) suggest that a single regression line—fitted, in this case, to align with the closing and opening of the chambers—might have been poorly suited to a time series exhibiting such variation. While one might instead have divided the full time series into shorter segments to compute several different regressions (e.g., according to the photoperiod), we chose to define the interval for rate calculations from the prototype instrument according the chamber opening and closing times. This allowed us to avoid the bias inherent in dividing such a time series into smaller segments. The present PHORCYS model allows users to predefine multiple, shorter incubation periods of length appropriate to the ecosystem (e.g., Fig. 3); this feature eliminates the need for a choice between either a subjective, ex post facto division of data or the application of a single regression that might fail to capture the observed variation.
Evaluation of instrument performance using independent methods
PHORCYS estimates of community respiration were systematically higher than those calculated for the same waters using the two-point Winkler titration method (Fig. 5a; Table 2). While there was a significant correlation between estimates from the two different methods (r2 = 0.42; p = 0.04), the traditional Winkler titration approach appeared to underestimate rates of respiration by nearly one-third. Alternatively, the PHORCYS could have overestimated true rates of respiration through artificial stimulation of the microbial community or inadvertent retention of residual dissolved oxygen by the chambers or optode. In contrast, rate estimates from the PHORCYS generally agreed with those based on our non-destructive optode sensor spot incubations, though we had only five data points on which to evaluate the correlation (Fig. 5b; correlation was not statistically significant).
Wikner et al. (2013) observed close agreement (∼ 3% deviation) between gross community respiration rates derived from continuous optode measurements in a shipboard incubation chamber (1 L, clear glass) and a series of parallel incubations based on a Winkler method in 120 mL glass BOD bottles. In previous work, we observed similar coherence between rates derived from a two-point Winkler method in 300 mL glass BOD bottles and those based on incubations in 125 mL glass BOD bottles fitted with optode sensor spots (surface area : volume ratios as reported in Table 1; Collins et al. 2015). Underlying the agreement among methods in these previous studies is a common reliance on incubations aboard ship. One might therefore conclude that the observed divergence between estimates of respiration from the PHORCYS and those we made with the Winkler titration method are due to inherent differences between the in situ PHORCYS approach and shipboard incubation methods. However, evidence suggests that use of the same method is not even a guarantor of agreement. For example, Robinson et al. (2009) found that vastly different rates were obtained from the same method of measuring primary production when the only the timescale of incubation was varied.
Possible sources of observed discrepancies between methods
Gas permeability of materials
The materials chosen for construction of the PHORCYS, particularly the polycarbonate plastic we used for the incubation chambers, represent one possible source of bias in our method. We chose polycarbonate material for its durability, low cost, and, compared with other plastics such as polyvinylchloride, minimal biological reactivity. Polycarbonate was selected for its minimal biological reactivity in construction of at least two similar in situ incubation devices (Langdon et al. 1995; Robert 2012). However, polycarbonate, like most plastics, is at least partially permeable to dissolved oxygen; by comparison, the borosilicate glass of which BOD bottles are fabricated is nearly impermeable (Kjeldsen 1993; Robert 2012). While some diffusion of dissolved oxygen across the polycarbonate chamber walls could have occurred during incubations with the PHORCYS, thus biasing our results, we believe the effect would likely have been very modest given the dissolved oxygen gradients and timescales typical of our deployments. First, the gradient necessary to drive any diffusion across the PHORCYS chamber wall (i.e., the difference between the internal and external dissolved oxygen concentrations) did not exceed 25 μmol O2 L−1 during any deployment; in the deployment presented in Fig. 3, the average difference between internal and external concentrations was just 6.1 μmol O2 L−1. In a series of oxygen-diffusion experiments with a device similar to the PHORCYS, Robert (2012) monitored the dissolved oxygen concentration inside a polycarbonate chamber after the water inside was treated with sodium hydrosulfite to render it anoxic. After manipulating the dissolved oxygen concentration in water outside the chamber to create cross-boundary gradients ranging from approx. 50 μmol O2 L−1 to > 350 μmol O2 L−1, the authors observed little change in the dissolved oxygen concentration inside the chamber on timescales similar to those of the PHORCYS deployments (11–12 h for the current version of the instrument; 3–5 d for the PHORCYS prototype). Further, in a comprehensive study of the permeabilities of several gases (O2, N2, CO2, and CH4) in various polymers, Kjeldsen (1993) concluded that bias introduced by permeability should be taken into account primarily when certain materials such as silicone rubber were considered for use in anoxic waters; silicone rubber is more than 250 times as permeable to dissolved oxygen than the polycarbonate plastic we used in construction of the PHORCYS (Kjeldsen 1993).
Handling or sampling bias
Errors or bias introduced during the handling and manipulation of samples for bottle incubations might also explain some of the discrepancy in rate measurements. For example, the PHORCYS minimizes physical disturbances associated with seawater handling: Since the instrument takes seawater samples and then incubates them in place, the planktonic community is not subjected to rapid changes in temperature, pressure, and light associated with bringing water samples to the surface via hydrocast and preparing them for shipboard incubations (Calvo-Díaz et al. 2011). The PHORCYS also minimizes another potential bias that can arise when water containing marine microbes is sampled from Niskin bottles; Suter et al. (2016) found that variation in settling rates among marine particles can lead to an undersampling of microbial communities on faster-sinking particles, which can fall below the bottles' spouts before aliquots can be drawn.
Biofouling and surface colonization
While we did not observe any significant biofouling of the PHORCYS or its components during the deployments for which data are presented here, unwanted biological growth represents a significant challenge and source of potential bias when autonomous sensors are deployed in the surface and mesopelagic ocean (Manov et al. 2004; Delauney et al. 2010; e.g., Collins et al. 2013). We imagine biofouling could represent a significant obstacle during future deployments if the deployment duration were to exceed 5–7 d (the maximum explored in this work; Table 2), or if the instrument were deployed in a productive ecosystem during a period of elevated primary production. Compare, however, Robert (2012), who reported little biofilm growth on the chamber of a similar incubation-based instrument after deploying it with no fouling controls for several months in the Mediterranean Sea.
The lack of visual evidence of fouling in the PHORCYS chambers notwithstanding, there is substantial evidence that microbial colonization of surfaces happens rapidly in the marine environment (Salta et al. 2013; Dang and Lovell 2016). It is therefore almost certain that at least some microbial colonization of the chamber walls would have taken place within the 3–5 d timescale of our deployments. Because we did not make measurements of biofilm activity or species abundance, it is impossible to diagnose what contribution these communities might have made to the observed fluxes in dissolved oxygen. However, compared to the surface area to volume ratios of the 150 mL and 300 mL vessels in which the other rate measurements were made (125 mL bottle, 0.83; 300 mL bottle, 0.77), the lower surface area to volume ratios of the PHORCYS chambers (current model, 0.34; prototype, 0.67) provide fewer opportunities for colonization relative to the volume being incubated (Table 1). There is also some evidence that the small volumes typical of BOD bottles may induce non-representative changes in the planktonic microbial community during incubations on timescales of 24–48 h (Pratt and Berkson 1959); however, compare Fogg and Calvario-Martinez (1989), who found that such bottle size effects were only significant during periods of very high primary productivity. A thorough, systematic assessment of these so-called wall and bottle effects should be part of the community intercalibration we advocate in our conclusion.
Other possible sources of bias
Alternatively, the well-documented dependence of community respiration rates on temperature (Yvon-Durocher et al. 2012) might explain the apparent disagreement between methods. While temperatures within the PHORCYS chambers fluctuated only according to the movement of tidal water masses (Fig. 3) or the natural warming and cooling of the surface layer (Fig. 4), the temperature inside the incubator in which the Winkler samples were kept during the shipboard deployments fluctuated during each series of experiments by ± 2°C from the target. A further, more intriguing explanation for the systematic discrepancy—pointing in this case to overestimation of rates by the PHORCYS—is that dissolved oxygen could have accumulated on the optodes or inside walls of the instrument chambers over the course of each deployment, in spite of the standard 30 min flushing protocol. Wikner et al. (2013) offer convincing evidence for the accumulation of dissolved oxygen on the optode surface and acrylic (polymethylmethacrylate) stopper used in their shipboard incubation apparatus; the application of a correction factor for this bias reduced apparent rates calculated from the optode time series relative to the Winkler method. We did not evaluate oxygen retention by the polycarbonate plastic material of which the PHORCYS chambers were constructed.
The nature of linear regression itself may also play a role in differences observed between the PHORCYS and the Winkler-based method. A least-squares regression line fit to a large set of observations collected at high temporal frequency, such as those obtained from the PHORCYS, is sensitive in some degree to each of those observations. In comparison, a rate calculated from the beginning and ending oxygen concentrations in the two-point Winkler method is necessarily sensitive only to those two observations; if some unrepresentative source of variability is present in just one bottle, the entire measurement can be heavily biased. Regression of data from the PHORCYS thus yields rate estimates which are robust to the sampling bias inherent in point measurements, yet the technique leaves the underlying data intact to be further interrogated for information about natural variability.
The “deep breath” phenomenon
Finally, while not evident in data from the deployments presented in Figs. 3 or 4, we did occasionally observe a sharp initial decrease in dissolved oxygen concentration within the chambers shortly after closure (data not shown). This phenomenon was reported extensively by Robert (2012) during testing of a similar instrument in the Mediterranean Sea, and thus warrants further investigation. The initial “deep breath” observed in these instances could reflect the rapid and preferential utilization by the microbial community of a limited but highly labile fraction of the dissolved organic carbon (DOC) pool within the chamber. Such a phenomenon has been observed frequently in rates of DOC consumption in freshwater systems, where labile carbon is often metabolized at a rapid rate in the initial minutes or hours of an incubation, leading to an apparent decline in metabolic activity once the labile pool has been exhausted and the incubation progresses (del Giorgio and Pace 2008; Guillemette and del Giorgio 2011; Guillemette et al. 2013). While we did not directly observe a pronounced initial change in the rate of dissolved oxygen consumption in any of our shipboard optode sensor spot incubations, it is possible that our sampling interval (3–9 h; see above) was of insufficient resolution to detect it.
Comments and recommendations
Through autonomous collection of biogeochemical observations at uniquely high temporal frequency, the PHORCYS yields estimates of community metabolic activity while simultaneously freeing the analyst from the logistical constraints of attended water column sampling and preparation of shipboard incubations. While we could not determine the origin of the systematic discrepancy between the PHORCYS rate estimates and those based on the traditional two-point Winkler method, the instrument's design allows investigators to avoid many of the potential biases associated with bottle incubations that have been previously documented in the literature. The PHORCYS offers a further advantage in that it can be used to collect information over multiple timescales about the metabolic state of marine and aquatic ecosystems at minimal cost and burden to the user. While the mixed-layer deployments we have presented here provided volumetric rates of ecosystem metabolism at a single depth, multiple PHORCYS units could be deployed simultaneously at different depths in the water column as a means of making depth-integrated rate measurements; one could compare these rates to estimates obtained from in situ geochemical tracer studies. Future work could also include direct, side-by-side comparison of metabolic rates obtained from the PHORCYS with those from the classic, bottle-based 14C or 13C tracer methods.
In this spirit, we believe a new, community intercalibration effort is warranted to systematically evaluate the many sources of uncertainty in incubation-based measurements of community metabolism. The workshops of the Group for Aquatic Primary Productivity (GAP; Figueroa et al. 2014) could serve as a model for such an effort, which should include comprehensive evaluation of various combinations of incubation chamber materials, types of oxygen sensors, chamber sizes, and incubation durations. Genetic and biogeochemical tools for characterizing the extent, mechanisms, and effects of surface colonization within the incubation chambers would be critical to the success of any such endeavor. We are optimistic such an effort might reveal the causes of longstanding discrepancies such as those we observed between our PHORCYS rate estimates and those we obtained with bottle-based methods.
A new method for calculation of uncertainties in metabolic rate estimates
The ideal means of estimating uncertainties in PHORCYS rates would have been true biological replication, i.e., the simultaneous deployment of several identical instruments in the same water mass. One could then have used the standard deviation of the rate measurements in each different instrument as an estimate of the overall uncertainty. Because we had only one instrument—an exceedingly common situation in oceanographic work—such true replication was not possible. The standard error of the regression slope provides one possible estimate of uncertainty in time-series dissolved oxygen data; for example, this common approach was recently applied to data from in situ chamber incubations of sinking marine particle material (McDonnell et al. 2015). However, we assumed that the standard error of regression would significantly underestimate the true uncertainty in our estimates since it does take into account the reduced number of degrees of freedom in such a time series. Because the data points in such a dissolved oxygen time series are not independent of one another, there are almost always far fewer effective degrees of freedom in such data than the number of observations (i.e., data points) N (Emery and Thomson 2001). (We represent the effective degrees of freedom by the notation in lieu of the notation used by Emery and Thompson.)
This method of estimating uncertainties in PHORCYS rates produced values of , the number of effective degrees of freedom, which were typically << N, the number of observations in the given dissolved oxygen time series (Table A1). Estimates of the integral time scale T ranged from 0.5 h to 7.2 h; at station PS-2, the 77.4 h deployment for which data are presented in Fig. 4, we estimated T to be 7.2 h (Table A1). Using the derived from these time scales, we obtained adjusted uncertainty estimates for our PHORCYS rates ( ) which were much greater in each case than the standard error of the regression slope, (compare mean precision of 24.8% and 3.4%, respectively; Table A1). While more robust than the corresponding , these still reflect a fundamental limitation of linear regression: Both methods yield estimates of uncertainty which are inversely proportional to the number of data points (i.e., the length of the underlying data series) and the range of values spanned by the independent variable.
|Deployment dates||Location*||PHORCYS community respiration (GR) (μmol O2 L−1 d−1)||No. observations (N)||Incubation duration (h)||Est. integral time scale (h)||Effective degrees of freedom ( )||Estimated uncertainty (μmol O2 L−1 d−1)||Method precision (est. uncertainty as percent of rate measurement)|
|Standard error of regression slope ( )||Adjusted estimate based on||(%)||(%)|
|24–27 Apr 2012||QL-1||1.8||2150||71.6||1.1||66.5||0.03||0.18||1.6||9.9|
|30 Apr 2012–03 May 2012||QL-2||4.2||1964||65.4||3.1||21.0||0.03||0.28||0.7||6.7|
|17–19 Jun 2012||PS-1||2.4||1238||41.2||0.5||76.4||0.08||0.32||3.3||13.1|
|23–27 Jun 2012||PS-2||7.8||2323||77.4||7.2||10.7||0.03||0.43||0.4||5.5|
|07–11 Jul 2012||PS-4||6.0||2820||94.0||1.9||49.8||0.07||0.52||1.2||8.6|
|07–08 Nov 2016||Iselin Pier||18.9||765||12.7||1.3||19.8||0.29||1.87||1.5||9.9|
|08 Nov 2016||Iselin Pier||2.2||627||10.5||1.2||17.0||0.25||1.60||11.6||74.4|
|08–09 Nov 2016||Iselin Pier||8.0||760||12.7||1.6||16.2||0.26||1.89||3.2||23.6|
|09–10 Nov 2016||Iselin Pier||10.5||750||12.5||2.9||8.7||0.71||7.51||6.7||71.4|
- *Cruise station or geographical location (Fig. 2); additional metadata for each station are provided in Supporting Information Table 1.
Additional Supporting Information may be found in the online version of this article.
Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
1994. Redfield ratios of remineralization determined by nutrient data-analysis. Global Biogeochem. Cycles 8: 65–80. doi:10.1029/93GB03318
2012. Balanced plankton net community metabolism in the oligotrophic North Atlantic subtropical gyre from Lagrangian observations. Deep-Sea Res. Part I Oceanogr. Res. Pap. 68: 116–122. doi:10.1016/j.dsr.2012.06.004
2016a. Depth-resolved particle-associated microbial respiration in the northeast Atlantic. Biogeosciences 13: 4927–4943. doi:10.5194/bg-13-4927-2016
2016b. The role of particle associated microbes in remineralization of fecal pellets in the upper mesopelagic of the Scotia Sea, Antarctica. Limnol. Oceanogr. 61: 1049–1064. doi:10.1002/lno.10269
1987. A comparison of 4 methods for determining planktonic community production. Limnol. Oceanogr. 32: 1085–1098. doi:10.4319/lo.1922.214.171.1245
2015. RESPIRE: An in situ particle interceptor to conduct particle remineralization and microbial dynamics studies in the oceans' Twilight Zone. Limnol. Oceanogr.: Methods 13: 494–508. doi:10.1002/lom3.10043
2004. Factors controlling net ecosystem metabolism in U.S. estuaries. Estuaries Coast. 27: 90–101. doi:10.1007/BF02803563
2011. Decrease in the autotrophic-to-heterotrophic biomass ratio of picoplankton in oligotrophic marine waters due to bottle enclosure. Appl. Environ. Microbiol. 77: 5739–5746. doi:10.1128/AEM.00066-11
2013. Estimates of new and total productivity in Central Long Island Sound from in situ measurements of nitrate and dissolved oxygen. Estuaries Coast. 36: 74–97. doi:10.1007/s12237-012-9560-5
2015. The multiple fates of sinking particles in the North Atlantic Ocean. Global Biogeochem. Cycles 29: 1471–1494. doi:10.1002/2014GB005037
2016. Microbial surface colonization and biofilm development in marine environments. Microbiol. Mol. Biol. Rev. 80: 91–138. doi:10.1128/MMBR.00037-15
2005. Respiration in aquatic ecosystems. Oxford Univ. Press.
2008. Relative independence of organic carbon transport and processing in a large temperate river: The Hudson River as both pipe and reactor. Limnol. Oceanogr. 53: 185–197. doi:10.4319/lo.2008.53.1.0185
2010. Biofouling protection for marine environmental sensors. Ocean Sci. 6: 503–511. doi:10.5194/os-6-503-2010
2007. Couplings between changes in the climate system and biogeochemistry, 89 p. In S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, and H. L. Miller [eds.], Climate change 2007: The physical science basis. Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge Univ. Press.
2013. The oligotrophic ocean is heterotrophic. Ann. Rev. Mar. Sci. 5: 18.11–18.19. doi:10.1146/annurev-marine-121211-172337
2013. What is the metabolic state of the oligotrophic ocean? A debate. Ann. Rev. Mar. Sci. 5: 15.11–15.19. doi:10.1146/annurev-marine-121211-172331
2011. Rapid microbial respiration of oil from the Deepwater Horizon spill in offshore surface waters of the Gulf of Mexico. Environ. Res. Lett. 6: 035301. doi:10.1088/1748–9326/6/3/035301 doi:10.1088/1748-9326/6/3/035301
2001. Data analysis methods in physical oceanography. Elsevier Science.
2014. Introduction. Aquat. Biol. 22: 1–4. doi:10.3354/ab00613
1989. Effects of bottle size in determinations of primary productivity by phytoplankton. Hydrobiologia 173: 89–94. doi:10.1007/BF00015518
2015. Benthic fluxes on the Oregon shelf. Estuar. Coast. Shelf Sci. 163 Part B: 156–166. doi:10.1016/j.ecss.2015.06.001
2002. Measuring the growth rates of phytoplankton in natural populations, p. 221–249. In D. V. S. Rao [ed.], Pelagic ecology methodologyA. A. Balkema Publishers.
1927. Investigations of the production of plankton in the Oslo Fjord, p. 1–48. Rapports Et Procès-Verbaux Des Réunions Du Conseil Permanent International Pour L'Exploration De La Mer 42.
2015. Gross and net production during the spring bloom along the Western Antarctic Peninsula. New Phytol. 205: 182–191. doi:10.1111/nph.13125
2011. Reconstructing the various facets of dissolved organic carbon bioavailability in freshwater ecosystems. Limnol. Oceanogr. 56: 734–748. doi:10.4319/lo.2011.56.2.0734
2013. Differentiating the degradation dynamics of algal and terrestrial carbon within complex natural dissolved organic carbon in temperate lakes. J. Geophys. Res. Biogeosci. 118: 963–973. doi:10.1002/jgrg.20077
2004. Methods for measuring benthic nutrient flux on the California Margin: Comparing shipboard core incubations to in situ lander results. Limnol. Oceanogr.: Methods 2: 146–159. doi:10.4319/lom.2004.2.146
2001. Microbial ecology at sea: Sampling, subsampling and incubation considerations, p. 13–39. In J. H. Paul [ed.], Methods in microbiologyAcademic Press.
1975. Measurements of electron transport activities in marine phytoplankton. Mar. Biol. 33: 119–127. doi:10.1007/BF00390716
2016. Low benthic respiration and nutrient flux at the highly productive Amundsen Sea Polynya, Antarctica. Deep-Sea Res. Part II Top. Stud. Oceanogr. 123: 92–101. doi:10.1016/j.dsr2.2015.10.004
1993. Evaluation of gas diffusion through plastic materials used in experimental and sampling equipment. Water Res. 27: 121–131. doi:10.1016/0043-1354(93)90202-S
1995. Fiber-optic oxygen microsensors, a new tool in aquatic biology. Limnol. Oceanogr. 40: 1159–1165. doi:10.4319/lo.19126.96.36.1999
1989. Dissolved oxygen measurements in sea water at the Woods Hole Oceanographic Institution, 18 p. Report WHOI-89-23. Woods Hole Oceanographic Institution.
1995. Measurements of net and gross O2 production, dark O2 respiration, and 14C assimilation at the Marine Light-Mixed Layers site (59°N, 21°W) in the northeast Atlantic Ocean. J. Geophys. Res. Oceans 100: 6645–6653. doi:10.1029/94JC02286
2015. Rates of total oxygen uptake of sediments and benthic nutrient fluxes measured using an in situ autonomous benthic chamber in the sediment of the slope off the southwestern part of Ulleung Basin, East Sea. Ocean Sci. J. 50: 581–588. doi:10.1007/s12601-015-0053-x
2014. lmodel2: Model II Regression. R package, version 1. 7–2.
2004. Methods for reducing biofouling of moored optical sensors. J. Atmos. Ocean. Technol. 21: 958–968. doi:10.1175/1520-0426(2004)021<0958:MFRBOM>2.0.CO;2
2016. Sustained in situ measurements of dissolved oxygen, methane and water transport processes in the benthic boundary layer at MC118, northern Gulf of Mexico. Deep-Sea Res. Part II Top. Stud. Oceanogr. 129: 41–52. doi:10.1016/j.dsr2.2015.11.012
2015. Effects of sinking velocities and microbial respiration rates on the attenuation of particulate carbon fluxes through the mesopelagic zone. Global Biogeochem. Cycles 29: 175. doi:10.1002/2014GB004935 doi:10.1002/2014GB004935
2009. Optical tools for ocean monitoring and research. Ocean Sci. 5: 661–684. doi:10.5194/os-5-661-2009
2015. Quantifying subtropical North Pacific gyre mixed layer primary productivity from Seaglider observations of diel oxygen cycles. Geophys. Res. Lett. 42: 4032–4039. doi:10.1002/2015GL063065
2009. New eyes on the world: Advanced sensors for ecology. Bioscience 59: 385–397. doi:10.1525/bio.2009.59.5.6
1959. Two sources of error in the oxygen light and dark bottle method. Limnol. Oceanogr. 4: 328–334. doi:10.4319/lo.1959.4.3.0328
2007. The future of chemical in situ sensors. Mar. Chem. 107: 422–432. doi:10.1016/j.marchem.2007.01.014
1934. On the proportions of organic derivatives in sea water and their relation to the composition of plankton, p. 176–192. In R. J. Daniel [ed.], James Johnstone memorial volumeUniv. Press of Liverpool.
2010. Major contribution of autotrophy to microbial carbon cycling in the deep North Atlantic's interior. Deep-Sea Res. Part II Top. Stud. Oceanogr. 57: 1572–1580. doi:10.1016/j.dsr2.2010.02.023
2008. Net production of oxygen in the subtropical ocean. Nature 451: 323–U325. doi:10.1038/nature06441
2012. Mineralisation in situ de la matière organique le long de la colonne d'eau: Application sur une station eulerienne. Dissertation. Université d'Aix-Marseille.
2005. Respiration and its measurement in surface marine waters. In P. Del Giorgio and P. J. Williams [eds.], Respiration in aquatic ecosystems. Oxford Univ. Press.
2009. Comparison of in vitro and in situ plankton production determinations. Aquat. Microb. Ecol. 54: 13–34. doi:10.3354/ame01250
2013. Marine biofilms on artificial surfaces: Structure and dynamics. Environ. Microbiol. 15: 2879–2893. doi:10.1111/1462-2920.12186
1994. Determining short-term biochemical oxygen demand and respiration rate in an aeration tank by using respirometry and estimation. Water Res. 28: 1571–1583. doi:10.1016/0043-1354(94)90224-0
2012. The metabolism of aquatic ecosystems: History, applications, and future challenges. Aquat. Sci. 74: 15–29. doi:10.1007/s00027-011-0199-2
1952. The use of radioactive carbon (14C) for measuring organic production in the sea. J. Du Conseil 18: 117–140. doi:10.1093/icesjms/18.2.117
1973. Aquatic chemistry: Chemical equilibria and rates in natural waters. 3rd ed., Wiley.
2016. Niskin bottle sample collection aliases microbial community composition and biogeochemical interpretation. Limnol. Oceanogr. 62: 606–617. doi:10.1002/lno.10447
2011. Data analysis concepts and observational methods, p. 147–186. In L. D. Talley, G. L. Pickard, W. J. Emery, and J. H. Swif [Eds.] Descriptive physical oceanography, 6th ed. Academic Press.
2013. Prokaryotic responses to hydrostatic pressure in the ocean – a review. Environ. Microbiol. 15: 1262–1274. doi:10.1111/1462-2920.12084
1990. Submersible Incubation Device (SID), autonomous instrumentation for the in situ measurement of primary production and other microbial rate processes. Deep-Sea Res. Part A Oceanogr. Res. Pap. 37: 343–358. doi:10.1016/0198-0149(90)90132-F
1993. Automated instrumentation for time-series measurement of primary production and nutrient status in production platform-accessible environments. Mar. Technol. Soc. J. 27: 32–44.
2006. Evaluation of a lifetime-based optode to measure oxygen in aquatic systems. Limnol. Oceanogr.: Methods 4: 7–17. doi:10.4319/lom.2006.4.7
1977. Possible consequences of containing microplankton for physiological rate measurements. J. Exp. Mar. Biol. Ecol. 26: 55–76. doi:10.1016/0022-0981(77)90080-6
2006. Biological oxygen demand dynamics in the Lower San Joaquin River, California. Environ. Sci. Technol. 40: 5653–5660. doi:10.1021/es0525399
2007. New and fast method to quantify respiration rates of bacterial and plankton communities in freshwater ecosystems by using optical oxygen sensor spots. Appl. Environ. Microbiol. 73: 6722–6729. doi:10.1128/AEM.00405-07
2013. Precise continuous measurements of pelagic respiration in coastal waters with Oxygen Optodes. Limnol. Oceanogr.: Methods 11: 1–15. doi:10.4319/lom.2013.11.1
1998. The balance of plankton respiration and photosynthesis in the open oceans. Nature 394: 55–57. doi:10.1038/27878
2013. The oligotrophic ocean is autotrophic. Ann. Rev. Mar. Sci. 5: 16.11–16.15. doi:10.1146/annurev-marine-121211-172335
2012. Reconciling the temperature dependence of respiration across timescales and ecosystem types. Nature 487: 472–476. doi:10.1038/nature11205
We thank the captains and crews of the R/V Knorr and R/V Clifford Barnes, Anton Zafereo, Kay Bidle, Bethanie Edwards, Filipa Carvalho, Jared Schwartz, Fiona Hopewell, Gabriel Roy Liguori, Richard Payne, Jason C. Smith, Sujata Murthy, Dave Fisichella, Ed O'Brien, Craig Marquette, Erik Smith, Shawn Sneddon, Richard Butler, Helen Fredricks, David Glover, Oliver Newman, Emily Peacock, Leah Houghton, Matthew Bogard, Olivia De Meo, and Joe Salisbury. Sheri White contributed significantly to early development of the PHORCYS. The comments of two reviewers significantly improved our original manuscript. This research was supported by the U.S. National Science Foundation (awards OCE-1155438 to B.A.S.V.M., J.R.V., and R.G.K., and OCE-1059884 to B.A.S.V.M.), the Woods Hole Oceanographic Institution through a Cecil and Ida Green Foundation Innovative Technology Award and an Interdisciplinary Science Award, and a U.S. Environmental Protection Agency (EPA) STAR Graduate Fellowship to J.R.C. under Fellowship Assistance Agreement no. FP-91744301-0. The publication has not been formally reviewed by EPA. The views expressed in this publication are solely those of the authors, and EPA does not endorse any products or commercial services mentioned in this publication.
Conflict of Interest