Article pubs.acs.org/est
In Situ Spectrophotometric Measurement of Dissolved Inorganic Carbon in Seawater Xuewu Liu,† Robert H. Byrne,*,† Lori Adornato,‡ Kimberly K. Yates,§ Eric Kaltenbacher,‡ Xiaoling Ding,† and Bo Yang† †
College of Marine Science, University of South Florida, 140 7th Avenue S., St. Petersburg, Florida 33701, United States SRI International, 450 8th Avenue S.E., St. Petersburg, Florida 33701, United States § St. Petersburg Coastal and Marine Science Center, U.S. Geological Survey, 600 4th Street South, St. Petersburg, Florida 33701, United States ‡
S Supporting Information *
ABSTRACT: Autonomous in situ sensors are needed to document the effects of today’s rapid ocean uptake of atmospheric carbon dioxide (e.g., ocean acidification). General environmental conditions (e.g., biofouling, turbidity) and carbon-specific conditions (e.g., wide diel variations) present significant challenges to acquiring longterm measurements of dissolved inorganic carbon (DIC) with satisfactory accuracy and resolution. SEAS-DIC is a new in situ instrument designed to provide calibrated, high-frequency, long-term measurements of DIC in marine and fresh waters. Sample water is first acidified to convert all DIC to carbon dioxide (CO2). The sample and a known reagent solution are then equilibrated across a gas-permeable membrane. Spectrophotometric measurement of reagent pH can thereby determine the sample DIC over a wide dynamic range, with inherent calibration provided by the pH indicator’s molecular characteristics. Field trials indicate that SEAS-DIC performs well in biofouling and turbid waters, with a DIC accuracy and precision of ∼2 μmol kg−1 and a measurement rate of approximately once per minute. The acidic reagent protects the sensor cell from biofouling, and the gas-permeable membrane excludes particulates from the optical path. This instrument, the first spectrophotometric system capable of automated in situ DIC measurements, positions DIC to become a key parameter for in situ CO2-system characterizations.
■
INTRODUCTION The characterization of ocean chemical and ecosystem changes related to carbon dioxide uptake or other carbon-relevant processes (e.g., photosynthesis and respiration, and carbonate mineral precipitation and dissolution) requires the examination of CO2-system parameters over a wide range of temporal and spatial scales. Much of our understanding of recent trends in the oceanic CO2 system comes from shipboard analyses conducted during infrequent, long-distance ocean expeditions (e.g., http://cdiac.ornl.gov). For example, a comparison of data collected on two such cruises (1991 and 2006) revealed evidence of cross-basin ocean acidification in the North Pacific Ocean.1 Data from ocean time-series studies2,3 (e.g., the Hawaii Ocean Time-series program and the European Project on Ocean Acidification) have also documented long-term trends in surface ocean CO2 (increasing) and pH (decreasing). The ability to make frequent autonomous measurements over a broad range of spatial scales would greatly augment the current suite of open-ocean and coastal observations. Such a capability would allow, for example, the detailed long-term study of the corrosive-water upwelling events4−6 that significantly decrease carbonate saturation states on short time scales over broad regions of some continental shelves, © 2013 American Chemical Society
particularly those affected by eastern boundary currents. Highresolution in situ measurements are also needed in coastal waters because of their comparatively high spatial and temporal variability,7 their vulnerability to human influence, and their direct provision of ecologically and economically important ecosystem services.8 The carbon dioxide system in natural waters is typically defined by measurement of two or more core parameters: pH, carbon dioxide fugacity (f CO2), total alkalinity (TA), and total dissolved inorganic carbon (DIC). DIC is defined as the sum of the solution concentrations of CO2 plus H2CO3 (CO*2 ), bicarbonate ions (HCO3−), and carbonate ions (CO32−). Shipboard and in situ analyses of these parameters often rely on dissimilar instrumentation and methodologies9 (e.g., ionsensitive field effect transistors, ISFETs;10 infrared detection;11 and titrimetric,12 coulometric,13 spectrophotometric,14 and potentiometric approaches15). Some methodologies produce Received: Revised: Accepted: Published: 11106
April 5, 2013 August 22, 2013 August 30, 2013 August 30, 2013 dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
Figure 1. (a) Main components of the SEAS II instrument (underwater cables not shown). For SEAS-DIC, an underwater power cable connects an external lamp assembly to the optical housing via the external lamp power connector. (b) DIC equilibration/optical cell of the SEAS-DIC instrument. Two end-caps are fitted to the center PEEK rod. A section of thin LCW is inserted into the center channel as an equilibration membrane. Two pieces of optical fiber guide visible light in (from the lamp assembly) and out (to the spectrometer) of the LCW channel.
m in depth, the instrument can be deployed on a CTD rosette, stationary platform, or AUV or ROV to collect vertical-profile, time-series, or 3D spatial data. It includes three two-channel pumps, a lamp (FO-6000 from World Precision Instruments), a custom optical cell (design dependent on the analyte of interest), a spectrophotometer (USB2000+, Ocean Optics), and interfaces for power and communications (Figure 1a).25 Optional components include a lamp external to the main pressure vessel, a heater, stirring chambers, valves, a battery canister, and connectors to provide power and collect data from up to four peripheral sensors (e.g., CTD, fluorometer, PAR sensor, or transmissometer). Reagents are normally stored in intravenous (IV) bags attached to the SEAS instrument. Power options include external power (20−24 V DC) or battery power (8 h nominal run time or 16 h with a double battery). The peristaltic pumps and light source are usually the main power draws (pumps, 3 W each; lamp, 1.5 W). If the heater is utilized, then it consumes an additional 30 W. External power offers the ability to collect data over a diel cycle or longer without having to pause data collection to exchange the battery. Details about the spectrophotometric analysis approach used with SEAS are provided in the online Supporting Information. Measurement of DIC with SEAS II and Liquid Core Waveguides. The use of flexible liquid core waveguides (LCWs) to measure seawater DIC in laboratory and shipboard settings has been previously described.14,27 In this work, we extend these principles to autonomous underwater applications by configuring SEAS II instruments18,25 to measure DIC in situ. The new sensors are referred to as SEAS-DIC sensors. Earlier instruments configured to measure pH are referred to as SEASpH sensors. A CTD (conductivity, temperature, and pressure sensor) is connected to each SEAS instrument to provide the ancillary data needed to calculate DIC and pH from measurements of absorbance. Prior SEAS models designed to measure nutrients used a coiled LCW and a light source inside the instrument housing. SEAS-DIC uses an uncoiled LCW sealed inside a hollow polymer rod, with the light source external to the main instrument housing (Figure 1b). The hydrophobic LCW tubing
measurements that are subject to appreciable drift (e.g., infrared f CO2 and potentiometric pH and TA).16 In situ observational capabilities are currently best developed for pH and f CO2,10,17,18 but this pair of measurements (pH f CO2) is generally considered suboptimal relative to the pairing of either parameter with DIC8,19 (i.e., pHDIC or f CO2 DIC). Automated multiparameter measurements based on similar principles and requiring infrequent calibration would be highly beneficial for instrumental integration and long-term deployment. Optical-absorbance (spectrophotometric) approaches, which can be applied to all four core CO2-system parameters,14,20 require infrequent14 or no20 calibration in the field (e.g., DIC and pH, respectively). The Multiparameter Inorganic Carbon Analyzer (MICA)14 can simultaneously measure several core parameters spectrophotometrically (pH, DIC, and air and solution f CO2), but it is limited to laboratory and shipboard settings. The Spectrophotometric Elemental Analysis System (SEAS)18 is an automated in situ instrument potentially suitable for obtaining simultaneous measurements of multiple chemical parameters, but DIC has not been among them. At present, the instrument can be user-configured to measure any one of a variety of chemical constituents, for example, pH, trace metals (copper, chromium, molybdenum, and iron),21−24 or nutrients25,26 (nitrite, nitrate, and phosphate). In this work, we report on an integration of the MICA and SEAS technologies to produce a new instrument (SEAS-DIC) that is capable of autonomously measuring DIC in situ in both laboratory and field settings. The design is inherently selfprotective against signal degradation resulting from biofouling and environmental turbidity. We tested the capabilities of the new SEAS-DIC instrument in turbid and biofouling-prone coastal marine environments.
■
EXPERIMENTAL SECTION
SEAS II. The second-generation Spectrophotometric Elemental Analysis System (SEAS II) is a modular in situ instrument capable of collecting high-resolution chemical concentration data autonomously or manually. Rated to 1000 11107
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
in the fraction of DIC that exists as CO*2 . For mixing-ratio variations that would create pH values between 2.7 and 3.0, variations in the calculated DIC would be less than 1 μM. At the beginning of each SEAS-DIC measurement cycle, baseline solution (of known TA without BCP) is delivered to the LCW for ∼2 min at a pump rate of 1 mL min−1. Baseline absorbance measurements are taken at 432, 589, and 700 nm (wavelengths specific to bromocresol purple). A dye pump is then turned on for 2 min at a pump rate of 1 mL min−1 to deliver measurement solution (of known TA, with BCP) to the LCW. After the external sample solution and the internal measurement solution are equilibrated (∼5 min), the pH of the internal solution is measured spectrophotometrically using the BCP indicator absorbances. DIC is then expressed as27
acts not only as a long-pathlength optical cell but also as a CO2permeable membrane. The 15 cm PEEK (polyetheretherketone) polymer rod is 2 mm in inner diameter (i.d.) and 10 mm in outer diameter (o.d.). At each end of the LCW, 200 μm optical fibers couple light from the lamp to the optical cell and the spectrometer. The space between the LCW tubing and the exterior rod (0.5 mL volume) is filled with continuously flowing sample water (seawater or fresh water) that has been acidified to convert its DIC content to CO2. This flowing solution is referred to as external solution. The interior cavity of the LCW tubing (10 μL volume) holds a solution of known total alkalinity (internal solution). Gas equilibration of the external and internal solutions occurs across the LCW membrane. Each end of the LCW cell is sealed with a piece of PEEK tubing (1/16 in. o.d. and ∼1.5 cm in length). An O-ring seated over the PEEK tubing between the end-cap and the main PEEK cell separates the internal solution from the external sample solution. Two forms of internal solution are used sequentially in the LCW. The first is a clear baseline solution (a solution of known total alkalinity without indicator dye) used to measure baseline absorbances. The second, the measurement solution (baseline solution plus indicator), is used for sensing solution pH. The internal solution in this work has a low ionic strength (approximately 1 mM, prepared using Na2CO3) because our pH measurements rely on prior bromocresol purple (BCP) calibrations28 obtained at low ionic strength. An internal solution of 1000 μM total alkalinity (equivalent to 500 μM sodium carbonate) equilibrated with 2000 μM CO2 in the acidified outer solution will have a pH near 6.0, well within the indicating range of BCP (5.3 ≤ pH ≤ 6.8). The SEAS-DIC instrument is always codeployed with a SEAS-pH instrument, providing a total of six pumps for the two analyses. One of the SEAS-DIC pumps (DIC#1) continuously combines sample (ambient) water with 2.5 M HCl drawn from an IV bag via one of the SEAS-pH pumps (pH#1) to achieve a seawater-to-acid mixing ratio of ∼714:1 and a pH of ∼2.7. The purpose of the acid is to convert all forms of sample DIC to CO2. The sample pump (DIC#1, plumbed with purple/white, 2.79 mm i.d. three-stop tubing; Cole Parmer) draws water at a flow rate of ∼10 mL min−1. The acid pump (pH#1, orange/ blue, 0.25 mm i.d. three-stop tubing) introduces acid into the flowing stream of sample water at 0.014 mL min−1 via a Yconnector equipped with a flow-check valve (CV-3320, Idex). An inline static mixer (EW-04668-26, Cole Parmer) mixes the sample and acid thoroughly. As noted, both pumps run continuously, refreshing the sample solution external to the LCW at an overall rate of ∼10 mL min−1. The remaining two SEAS-DIC pumps (DIC#2 and DIC#3) periodically deliver baseline solution and then measurement solution to the interior chamber of the LCW. The remaining two SEAS-pH pumps (pH#2 and pH#3) draw sample water and indicator solution for pH analyses. Because the volume of HCl required to acidify the sample is very small relative to the volume of sample, we always have an excess of acid, and minor variations in the mixing ratio do not affect the DIC measurement. The excess acid delivered is such that even factor-of-two variations in the concentration of excess acid would cause only ∼2 μM variations in the calculated DIC. The relatively small impact of mixing-ratio variations on DIC measurements is largely due to the fact that mixing-ratio variations create two effects that act in opposition: (a) variations in the dilution of seawater DIC and (b) variations
⎛ R−e ⎞ ⎛ (K ) ⎞ 1 log DIC = log⎜ 0 a ⎟ + B(T ) − log⎜ ⎟ ⎝ (K 0)i ⎠ ⎝ 1 − Re3/e 2 ⎠
(1)
where (K0)a and (K0)i are Henry’s law constants appropriate to the acidified external solution (a) and the interior solution (i). T is absolute temperature in Kelvin, and B(T) is an empirically determined temperature-dependent term (fully described below). R is the ratio of baseline-corrected absorbances (A) measured in the LCW internal solution at 589 and 432 nm (i.e., R = A589/A432). Absorbance measurements at the nonabsorbing wavelength of 700 nm are used for quality control monitoring and compensation of small baseline shifts. The e1, e2, and e3 terms are BCP molar absorptivity ratios (0.00387, 2.858, and 0.0181, respectively).29 Equation 1 can be rewritten in terms of directly measured variables ⎛ ⎛ T ⎞ ⎟ log DIC = S⎜0.023517 − 0.023655⎜ ⎝ 100 ⎠ ⎝ ⎛ T ⎞2 ⎞ ⎟ ⎟ /2.303 + B(T ) + 0.0047036⎜ ⎝ 100 ⎠ ⎠ ⎛ R − 0.00387 ⎞ ⎟ − log⎜ ⎝ 1 − 0.00633R ⎠
(2)
where S is salinity. Absorbance readings are taken at a rate of one reading per min for 50 readings. At the end of the 50 min, the baseline solution and then the measurement solution are refreshed to begin another cycle. This procedure is repeated autonomously until the mission is manually terminated. The full derivation of eq 2 is given in the Supporting Information. Investigators are encouraged to check their coding of the DIC algorithm (eq 2) by comparing their test calculations against Table S1 of the Supporting Information. The B(T) term in eqs 1 and 2 is an empirically derived temperature-dependent term with the general form of ⎛ KI(T )e 2 ⎞ ⎟⎟ B(T ) = log((TA)i + [H+]i − [I2 −]) + log⎜⎜ ⎝ K ′1(T ) ⎠ (3)
where TAi is total alkalinity of the internal solution, [ ] represents the concentrations of H+ and I2−, KI(T) is the temperature-dependent dissociation constant of HI−, and K′1(T) is the temperature-dependent dissociation constant of CO*2 . The equation describing the temperature dependence of K1′ is30,31 11108
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
Figure 2. Seawater DIC during the Bayboro Harbor (BH) deployment. The blue data points show the in situ SEAS-DIC readings, and the red dots show the DIC concentrations measured in discrete water samples. The water depth is shown in brown.
log K1′ = − 3404.71/T + 14.8435 − 0.032786T
provided by Dr. Andrew Dickson of the Scripps Institution of Oceanography. The 2.5 M HCl solution used to acidify sample waters was prepared by diluting concentrated HCl with Milli-Q water. The BCP stock solution (8 mM) was prepared by dissolving BCP salt in Milli-Q water. NaOH or HCl (1 N) was added to adjust the pH to ∼6.0. This mildly acidic condition minimizes acid− base reactions with container surfaces, promotes indicator solubility, and minimizes CO2 uptake in case of contact with the atmosphere. Internal (baseline and measurement) solutions were prepared by adding Na2CO3 (sufficient to establish TA ∼1000 μmol kg−1) to a 20 L glass bottle filled with Milli-Q water. To eliminate the hydrophobic adsorption of the indicator dye onto the Teflon AF tubing,18 10 mL of a 10% SDS solution was added to each 20 L batch. Half of this batch was set aside to serve as the baseline solution. The BCP stock solution was added to the other half to produce the measurement solution (final BCP ∼2 μM and SDS ∼0.005%). These baseline and measurement solutions were stored in IV bags (3 L maximum capacity) for in situ dispensing. The HCl solution was stored and dispensed from a 500 mL iv bag. Field Tests. SEAS-DIC instruments were tested in situ at two locations: Bayboro Harbor (BH) and Florida Bay (FB), both in Florida, USA. BH is a small, restricted basin that receives fresh water from a storm culvert and two small creeks that drain the urban landscape of St. Petersburg (Figure S2a of the Supporting Information). This deployment was from a seawall just west of the University of South Florida (USF) College of Marine Science (27.7608° N, 82.6336° W; November 2011). The water depth was 1 to 2 m deep and highly turbid, with a visibility of ∼20 cm. To block debris from entering the sample channel, a piece of 100 μm nylon mesh was placed at the end of the sample intake tube, which was located ∼20 cm above the seafloor. The instrument was powered continuously from shore. FB is a relatively remote, shallow estuary located between the Florida mainland and the Florida Keys (Figure S2b of the Supporting Information). The waters here are relatively clear with a high primary productivity. The study site was located in Lignumvitae Basin (24.92438° N, 80.74025° W; May 2009;
(4)
and the temperature dependence of pK0I (i.e., of pKI at zero ionic strength)28 is pKI0 = (378.1 ± 10.2)/T + 5.226 ± 0.035
(5)
Each of the terms contributing to B(T) in eq 3 can be calculated or estimated, so it is possible to calculate B(T) from first principles. However, the uncertainty in pK0I alone (eq 5) causes a propagation of error in DIC calculations on the order of ±10%. Additional uncertainties, for example, in e2 and TAi, also propagate through the DIC calculation. As such, it is advisable to perform a calibration to determine B(T) experimentally for each batch of measurement solution. In this work, B(T) was determined experimentally using certified reference material (CRM) and eq 2. Sample DIC and S were known, and R values were measured over a range of temperatures. The Supporting Information document provides a comparison of our experimental B(T) determinations (obtained using eq 2) and theoretical estimates (obtained using eqs 3, 4, and 5). Both approaches show that B(T) can be well represented as a simple polynomial function of 1/T. During SEAS-DIC deployments, temperature and salinity must be measured in situ with an accompanying CTD so that B(T) and DIC (eq 2) can be calculated accurately. Benchtop analyses (e.g., MICA) are typically conducted at a constant temperature (25 °C, controlled by a thermostatting water bath or Peltier device). For in situ analyses, ambient-temperature analyses provide the substantial benefit of eliminating the need for an underwater heater. Even with a heater, maintaining a constant solution temperature would be difficult at the flow rates used by SEAS-DIC. Materials and Reagents. Teflon AF 2400 (DuPont) tubing (0.5 mm o.d. × 0.4 mm i.d.) was obtained from Biogeneral, Inc. The PEEK block and rod were obtained from McMaster Carr. Vitalmix IV bags made of non-DEHP plastic were obtained from Cole-Parmer. The sodium salt of BCP (ACS Certified) was obtained from Sigma-Aldrich. Na2CO3 (ultrapure) and concentrated HCl (ACS certified) were obtained from J.T. Baker. Sodium dodecyl sulfate (SDS) surfactant was obtained from Fisher Scientific. DIC CRMs were 11109
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
Figure 3. Diel cycle of DIC in western Florida Bay (FB). The blue data points show the in situ SEAS-DIC readings, and the red dots show the DIC concentrations measured in discrete seawater samples. (a) In open bay waters before the SHARQ tent was enclosed. (b) Inside the SHARQ enclosure.
analytical system and a motor-stirred gas equilibration chamber on the houseboat and then back to the underwater tent. This system served to maintain turbulent flow within the SHARQ, discouraging stagnation and the development of concentration gradients. At both field sites, the discrete samples were transferred to 300 mL BOD bottles and then stabilized with saturated mercuric chloride (HgCl2). All bottles were sealed with Apiezon grease. In the lab, subsamples (∼20 mL) were transferred by syringe from a newly opened sample bottle to the stripping chamber of an acidification module (CM5130; UIC, Inc.) coupled to a carbon coulometer (CM5014). The samples were acidified with 2 M phosphoric acid. Instrument calibrations were conducted using CRMs provided by Dr. Andrew Dickson of the Scripps Institution of Oceanography. The BH samples were analyzed at the USF College of Marine Science. The FB samples were analyzed at the U.S. Geological Survey (USGS) Carbon Analysis Laboratory.
further site details are available in the Supporting Information). The water depth was ∼2 m at low tide. Here, SEAS-DIC was used to monitor carbon-system dynamics inside a Submersible Habitat for Analyzing Reef Quality (SHARQ) enclosure (Figure S3).32 To ensure that the reagents and flushing ports would be accessible, the SEAS instruments (one measuring pH and one measuring DIC) were placed on the seafloor outside the chamber, with the sampling tubes running beneath the chamber’s lower edge. Sample water was drawn continuously from ∼20 cm above the seafloor inside the tent; no mesh filter was required. The SEAS instruments were powered by a generator on a nearby houseboat. In both deployments, discrete water samples were collected for laboratory coulometric determinations of DIC. At BH, a 100 mL syringe with a three-way switch was used to draw seawater samples from a line installed next to the SEAS-DIC intake port. At FB, the discrete-sample intake port was located ∼1 m above the SEAS sampling tube (Figure S3b). Seawater was pumped continuously from the top of the SHARQ enclosure into a circulating system that ran to a flow-through 11110
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
Figure 4. Differences between the in situ (SEAS) and laboratory (discrete) measurements of seawater DIC. (a) Bayboro Harbor. (b) Florida Bay before and after closure of the SHARQ tent.
■
SEAS-DIC Time Series. DIC concentrations in Bayboro Harbor generally reflect the superimposed influences of tidal exchange, freshwater input, water-column photosynthesis/ respiration, and benthic respiration.33 The earliest days of the SEAS-DIC deployment (Figure 2) were characterized by waterdepth changes of about half a meter, relatively stable salinity (not shown), and DIC fluctuations of ∼150 μmol kg−1. Later days were characterized by larger water-level changes. Rainfall events and deep salinity dips occurred between November 19 and 22. The DIC fluctuations remained of roughly the same magnitude throughout the deployment, but the average daily DIC increased. DIC generally peaked during low tides, when the inflow of presumably high-DIC freshwater would have likely increased. An average DIC concentration of ∼3000 μmol kg−1 has been reported for major rivers near BH.27 The highest DIC (∼2550 μmol kg−1) corresponded to a time of low tide, high rainfall, and low salinity (November 22). The lowest DIC (∼1500 μmol kg−1) was observed during high tides early in the deployment. At Florida Bay, the dominant processes affecting the DIC concentrations are photosynthesis, respiration, calcification, and dissolution. In the SEAS-DIC record, a strong diel cycle was evident both outside (Figure 3a) and inside the SHARQ tent (Figure 3b), with the open-water cycle slightly muted relative to that inside the chamber. The DIC underwent rapid changes on a time scale of hours. Data collection began at 8 am, with the DIC steadily decreasing because of photosynthesis until about 6 pm. At that time, the trend abruptly reversed, as the effects of respiration began to dominate over photosynthesis. A linear DIC increase was then observed until the next morning when the cycle began again. These diel patterns represent primarily the metabolism of localized benthic communities (e.g., seagrasses). The amplification of the DIC diel cycles inside the chamber relative to ambient-water cycles are due to close coupling of the SHARQ water mass with the benthic community. The effects of a brief breach of the SHARQ’s vinyl sheeting by a passing storm can be seen in the abrupt DIC dip observed on the morning of May 11. Biofouling and Turbidity. Biofouling is perhaps the single greatest challenge to the long-term operation, maintenance, and data quality of in situ sensors. The problem can be especially severe in coastal, estuarine, and highly productive freshwater
RESULTS AND DISCUSSION Comparison of In Situ and Discrete DIC Measurements. Figure 2 shows the results of the SEAS-DIC deployment off the Bayboro Harbor seawall. Excellent agreement was observed between the in situ measurements and the discrete DIC values measured coulometrically in the lab. The differences were typically 1 to 2 μmol kg−1 over the DIC range of ∼2250−2550 μmol kg−1 (average ΔDIC = 0.6 ± 1.6 μmol kg−1; n = 6). The results from the Florida Bay deployment are shown in Figure 3. In open bay waters (Figure 3a), the in situ SEAS-DIC data were in poor agreement with DIC measured in the discrete samples (average ΔDIC = −28.1 ± 15.2 μmol kg−1; n = 11). For samples collected from within the SHARQ tent (Figure 3b), the in situ and discrete data were in better agreement (average ΔDIC = −1.0 ± 4.3 μmol kg−1; n = 10). These average differences are approximately 30× and 2× those observed for BH (Figure 2). A plot of the differences between the SEAS-DIC (in situ) and discrete (laboratory) DIC measurements (Figure 4) shows the tight correspondence achieved when discrete samples were collected in close proximity to the SEAS-DIC sample intake port. In BH (Figure 4a), seawater for the coulometric analyses was drawn from a sample volume immediately adjacent to the SEAS-DIC intake port. In FB (Figure 4b), discrete samples were drawn from ∼1 m above the SEAS intake port (Figure S2b). The open waters of FB were apparently somewhat more vertically stratified than SHARQ-tent waters, which were intentionally mixed to prevent concentration gradients. The large ΔDIC values obtained in FB are likely due to spatial separation between the in situ and discrete sampling ports as well the scale of spatial variability in the local environment. Small, semiregular gaps (∼10 min every hour) in the SEASDIC data reflect the time required to start a new data-collection cycle (i.e., to introduce new baseline solution, collect baseline absorbance measurements, introduce new measurement solution, and equilibrate the external (sample) and internal (measurement) solutions). Less frequent, longer gaps indicate times when the analyses were stopped (∼ every few hours) to flush the SEAS-pH optical cell with acid to reduce biofouling. The SEAS-DIC cell did not require such treatment. 11111
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
sizable. A fast sampling rate is also advantageous for towed or floating sensors to capture small-scale spatial variability. In open ocean waters, which are relatively homogeneous, a single DIC reading every 7 min is generally adequate to resolve significant features and trends. For coastal environments, however, a 1 min measurement interval is more appropriate. Our high-resolution data sets indicate that coastal processes can result in very smallscale (submeter) measurable DIC gradients and that significant/resolvable variations can occur over time scales of minutes. The SEAS-DIC instrument can deliver rapid in situ measurements (∼1 reading per minute) over 50 min intervals separated by short gaps (∼10 min) required for reagent replenishment and equilibration. For future SEAS-DIC deployments, it would be useful to minimize the potential influence of the (K0)a/(K0)i term on measurement precision and accuracy. If the ionic strength of the inner measurement solution is matched with the ionic strength of the ambient (sample) water, the (K0)a/(K0)i term in eq 1 becomes nearly equal to unity, whereby log (K0)a/(K0)i ≈ 0 and the influence of CO2 solubility characterizations on the DIC analyses becomes insignificant. To achieve this match-up for seawater, the dissociation constant characteristics of BCP at seawater ionic strength28 would need to be determined over a range of temperature and pressure. The experimental and theoretical behaviors of B(T) (Figure S1) can be expected to be brought into greater conformity through such work, potentially eliminating the need for measurement calibrations (analogous to the case for spectrophotometric pH measurements).20,37 Future designs of spectrophotometric DIC instrumentation will likely include simpler and less expensive components such as light-emitting diodes and diode photometers. It is critically important that absorbance linearity is always observed over the full range of conditions encountered in the measurement solution. In other words, at each wavelength utilized in determination of the function log[(R − e1)/(1 − Re3/e2)], it is essential to demonstrate that the absorbance measurements conform to Beer’s Law. Routine quality control for marine carbon system measurements entails (a) measuring a given parameter by more than one analytical approach, as was done here, or (b) measuring three or more of the core CO2-system parameters (DIC, TA, f CO2, and pH) and then evaluating the combined data set for thermodynamic consistency. Our experience with highprecision, high-frequency sampling in strongly dynamic environments affirms that particular attention must be paid to careful colocation of the samples in both time and space. Otherwise, temporal and spatial offsets between ostensibly coincident samples can result in poor measurement agreement and apparent thermodynamic inconsistencies. Integrated sampling would be beneficial. At present, commercial CO2system instruments measure only single parameters,10,20,38−40 and combining instruments with widely different sensing modalities increases complexity and sensor size. For SEAS, simultaneous measurement of more than one parameter currently requires the deployment of multiple instruments.18 The spectrophotometric approach described in this work can be extended to measure all four core CO2-system parameters in a single, integrated, in situ instrument. Work to develop a single SEAS instrument with multiple detectors is underway. SEAS-DIC capabilities are particularly well suited for studies of coastal benthic ecosystems, including coral reefs, where shallow water depths and relatively long residence times enhance the impact of benthic metabolism on seawater
environments. Various approaches to combating biofouling organisms have been adopted,34 including mechanical wiping, toxic painting, and copper screening. Nevertheless, biofouling of optical windows used for spectrophotometric measurements remains a challenge. Warm, productive waters such as those of FB are especially problematic. During the SHARQ deployment, the SEAS-pH optical window showed evidence of biofouling after just a few hours. This optical cell therefore required periodic acid washing (as evidenced by the data gaps in Figure 3). The SEAS-DIC encountered no such problems. Continuous acidification of the water sample (external solution) prevents biofouling build-up on the outside of the LCW membrane. Interior LCW surfaces never come into contact with sample (ambient) water or biofouling organisms. Water-column turbidity can also present a challenge for in situ sensors that rely on optical sensing. Light-scattering particles can render spectrophotometric absorbance measurements impossible. BH waters (Figure 2) are hospitable to biofouling organisms and also carry a large particulate load. Nevertheless, the SEAS-DIC instrument, outfitted with a mesh screen prefilter, performed well despite the high turbidity because the optical absorbance was measured not in the water sample but in an interior solution that was CO2-equilibrated with the sample water.
■
IMPLICATIONS The SEAS-DIC sensor is the first in situ spectrophotometric instrument demonstrated to provide automated high-quality, high-frequency measurements of total dissolved inorganic carbon in seawater. DIC now joins pH and f CO2 as a strong candidate parameter for long-term in situ monitoring of ocean acidification and other carbon-relevant processes. When paired with existing pH or f CO2 sensors, SEAS-DIC can contribute to complete characterization of the carbon dioxide system in natural and manipulated aqueous environments. The SEAS-DIC instrument includes several design features important for long-term deployments. The spectrophotometric approach provides freedom from requirements for calibration during deployments. Equation 2 is in principle similar to the equation that describes spectrophotometric pH measurements.17 As long as the known TA of the paired (baseline and measurement) internal solutions remains constant, the SEAS-DIC measurements should not be subject to the drift that is characteristic of infrared or potentiometric measurements.35,36 The use of an acidified outer (sample) solution and a protected inner (synthetic) solution enables DIC determinations in waters that would otherwise be too turbid for optical measurements and also eliminates issues of longterm biofouling. Low-frequency replacement of baseline and measurement solutions conserves reagent and minimizes the data gaps that result from taking frequent baseline measurements. Long deployment times are essential not only for characterizing long-term variability and trends but also for producing reliably integrated quantities. For example, scaling up from carbon system data collected over short time intervals in a highly variable environment can yield significantly different budget estimates depending on the timing of data collection within the diel cycle. SEAS-DIC also provides autonomous, rapid sampling (one per minute). High-frequency measurements are necessary in dynamic coastal, freshwater, and laboratory systems, where even hourly changes of dissolved inorganic carbon can be 11112
dx.doi.org/10.1021/es4014807 | Environ. Sci. Technol. 2013, 47, 11106−11114
Environmental Science & Technology
Article
chemistry.41,42 Fine-scale oxygen gradients, for example, have been measured over vertical distances
