A potential tool for high-resolution monitoring of ocean acidification

Analytica Chimica Acta 786 (2013) 1–7

Contents lists available at SciVerse ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

A potential tool for high-resolution monitoring of ocean acidification Aron Hakonen a,b,∗ , Leif G. Anderson b , Johan Engelbrektsson c , Stefan Hulth b , Bengt Karlson a a b c

Swedish Meteorological and Hydrological Institute, Oceanographic unit, Sven Källfelts gata 15, SE-426 71 Västra Frölunda, Sweden Department of Chemistry and Molecular Biology, University of Gothenburg, Kemivägen 10, SE-412 96, Sweden SP Technical Research Institute of Sweden, Box 857, SE-501 15 Borås, Sweden

h i g h l i g h t s

g r a p h i c a l

a b s t r a c t

• Highly resolved on-ship pH map• • • •

pings from several months are presented. An economic, low-power and low dye consumption fluorescence method was developed. Unique non-linear salinity and temperature adjusted calibrations were successfully utilized. The sampling eliminated pH perturbation from ambient gasses and provided a filtering mechanism. Excellent measurement robustness and on-ship precision are demonstrated.

a r t i c l e

i n f o

Article history: Received 19 January 2013 Received in revised form 8 April 2013 Accepted 11 April 2013 Available online 28 April 2013 Keywords: Ocean acidification pH monitoring Fluorescence Non-linear calibrations 6,8-Dihydroxypyrene-1,3-disulfonic acid High-flow/low-flow sampling

a b s t r a c t Current anthropogenic carbon dioxide emissions generate besides global warming unprecedented acidification rates of the oceans. Recent evidence indicates the possibility that ocean acidification and low oceanic pH may be a major reason for several mass extinctions in the past. However, a major bottleneck for research on ocean acidification is long-term monitoring and the collection of consistent high-resolution pH measurements. This study presents a low-power (<1 W) small sample volume (25 ␮L) semiconductor based fluorescence method for real-time ship-board pH measurements at high temporal and spatial resolution (approximately 15 s and 100 m between samples). A 405 nm light emitting diode and the blue and green channels from a digital camera was used for swift detection of fluorescence from the pH sensitive dye 6,8-Dihydroxypyrene-1,3-disulfonic acid in real-time. Main principles were demonstrated by automated continuous measurements of pH in the surface water across the Baltic Sea and the Kattegat region with a large range in salinity (∼3–30) and temperature (∼0–25 ◦ C). Ship-board precision of salinity and temperature adjusted pH measurements were estimated as low as 0.0001 pH units. © 2013 Elsevier B.V. All rights reserved.

1. Introduction

∗ Corresponding author at: Department of Chemistry and Molecular Biology, University of Gothenburg, Kemivägen 10, SE-412 96, Sweden. Tel.: +46 704964657. E-mail addresses: [email protected], aron [email protected] (A. Hakonen). 0003-2670/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.aca.2013.04.040

The oceans act as an important sink for CO2 and have so far absorbed approximately one-third of the anthropogenic emissions [1–4]. Since pre-industrial times average ocean pH has decreased approximately 0.1 pH units and is predicted to decrease by an additional 0.4 pH units or more during this century [1–4]. This is likely to influence marine organisms and ecosystems, examples include

2

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

calcifying organisms such as corals, molluscs, echinoderms and coccolithophores [1,2,5–10]. Other predicted adverse effects of ocean acidification (OA) are, for example, extended zones with low oxygen or anoxic conditions [11], decreased nitrification rates [12], declined replenishment of fish populations [13], increased iron bioavailability [14] and reduced biodiversity [15]. Recent studies indicate that ocean acidification [16] may have been one of the main reasons for the end-Cretaceous [16], end-Triassic [17] and end-Permian [18] mass extinctions. As current rates of ocean acidification are likely unprecedented and certainly unsettling [19], the predicament of present emissions of carbon dioxide need to be constrained [20] and carefully monitored [21–24]. Quantification of subtle changes in seawater pH presents significant analytical challenges, and for OA purposes it is appropriate to have a precision better than the predicted annual acidification rate of 0.002 pH units [21,24,25]. Complications during seawater pH measurements include, but are not restricted to, conditions that influence proton activity, mainly temperature and ionic strength (salinity). Especially, large variations in salinity present a major analytical challenge that normally needs to be corrected for, which can be troublesome for high-performance marine and estuarine measurements where extremely high and widely fluctuating salinity are common [25,26]. Another serious and often ignored problem during pH measurements are ambient gases, such as CO2 and NH3 , which affect pH when equilibrating with samples and reference buffers. For example a seawater sample of salinity 35.165 at 25 ◦ C can alter pH from 8.293 to 7.934 when increasing ambient carbon dioxide from normal outdoor conditions (∼380 ppm) to 1000 ppm (common indoor condition) [26]. Natural cycles and oscillations in ocean pH can be on diurnal, seasonal and inter-annual time scales [27,28]. For accurate interpretation and to resolve natural variations from those caused by ocean acidification, precise (precision < 0.002 pH units) long-term high resolution (temporal and spatial) measurements are required [24]. One of the few long-term (almost 8 years) studies with relatively high temporal resolution (30 min intervals from late spring to late summer) based on measurements was demonstrated by Wootton et al. (station based, potentiometric method) [27]. Studies have for more than two decades suggested spectrophotometric methods as the most convenient method for accurate seawater pH measurements at near open ocean conditions (high salinity) [21,29,30]. However, reports of long-term high-resolution pH measurements are still deficient, ship-board as well as stationary. Shorter-term continuous ship-board studies have been shown for the spectrophotometric method, ship-board precisions of ∼0.001 pH units [24,31,32]. Some difficulties have been reported for the spectrophotometric methods, such as variances between suppliers of indicator dye and impurities [33]. Purification of the indicator has been assessed for improvement [33]. The Baltic Sea region is one of the largest areas of brackish water in the world, with a surface area of ∼4.3 × 103 km2 and a volume of ∼23 × 103 km3 , representing about 0.1% and 0.002% of the world’s ocean area and volume, respectively, and maximum and mean depths are 465 and 60 m [34]. The Baltic Sea fauna hosts at least 6065 species, where regional ecosystem/species distribution is highly salinity dependent [34]. Due to the large salinity range (∼3–30), the Baltic Sea region provides a suitable environment to study mechanisms and principles of ocean acidification, including effects on different ecological systems ranging nearly from open ocean conditions to freshwater. However, the large spatial and temporal variations in salinity, temperature, biological activity and organic matter present major analytical challenges for accurate pH measurements. The objective of this study was to develop a lowmaintenance automated system for continuous high-performance and high-resolution pH measurements on-board a ship of opportunity in the Baltic Sea. The fundamental principles of ratiometric

fluorescence measurements rely on similar principles as spectrophotometric techniques, thus minimizing the need for frequent calibrations. Fluorescence techniques normally benefit from several advantages, compared to spectrophotometry, such as lower sensitivity to dye impurities and matrix interferences. Accurate and reliable ship-board or field station based pH measurements will be a powerful tool for continuous high-resolution monitoring of the acidic state of the oceans. 2. Methods 2.1. Sampling vessel FerryBox-systems are automated measurement and water sampling systems mounted on e.g. on ferries and merchant vessels, sometimes referred to as ships of opportunity (SOOP). These systems are becoming increasingly popular in marine science since they provide cost efficient means for sampling surface waters of the oceans and coastal seas [35]. In this study we have integrated a novel pH-measurement system in a FerryBox-system on a merchant ship, M/S TransPaper, which operates the route Gothenburg (Sweden)–Kemi (Finland)–Oulo (Finland)–Lübeck (Germany)–Gothenburg on a weekly basis. A large part of the Baltic Sea and the Kattegat is thus covered twice a week. The FerryBox-system on the ship consists of a water intake at 3 m depth, a water pump, a de-bubbling device, flow-through sensors for temperature, conductivity, chlorophyll fluorescence, phycocyanin fluorescence, Colored dissolved organic matter (CDOM) fluorescence, turbidity and oxygen. In air sensors include irradiation, air temperature and air pressure. Data is stored every 20 s and transmitted in near real time to shore. The pHmeasurement system described here uses water from the same intake as the other sensors. A separate water pump is used and no de-bubbling device is included to avoid gas exchange with the air in the ship. Carbon dioxide measurements will also be included on the non-de-bubbled flow. 2.2. Automated pH measurement system The pH measurements are based on mixing the fluorescent pH sensitive probe 6,8-Dihydroxypyrene-1,3-disulfonic acid (DHPDS; 75 ␮M in MilliQ; pH 7.10) with the sample to a final concentration of 3 ␮M. Mixing was performed in a simple T-cross connecting the sample and dye tubes directly after the peristaltic pump (Fig. 1). The rather high concentration is needed to cancel out auto-fluorescence and emission uptake from the seawater. Despite the high concentration less than 4 mg fluorophore is consumed during a typical week of sampling in the Baltic Sea. Fig. 1 illustrates the set-up of the measurement system. The sample/dye mixture is pumped at a rate of 347 ␮L min−1 (Reglo Digital MS-4/12-100; tubing Tygon® ST (R3603), d = 0.19 mm (dye) and d = 0.95 (sample); Labinett AB, Gothenburg) to a 25 ␮L flow-through fluorescence cuvette (Hellma fluorescence cuvette 25 ␮L 3h320933 Z = 8.5 mm) and a temperature measurement (PT-100 3 mm, UK02) flow-through cell made in house. Fluorescence excitation is provided by a 405 nm LED (at minimum power, <1 W, for maximum lifetime and minimal drift, up to 100 000 h; Thorlabs UV 405 nm Mounted High Power LED) with a collimating lens (LED-optic Luxeon Rebel, 10193) and the emission is recorded by a CMOS color camera with a focusing lens (Thorlabs DCC1645C; N-BK7 Bi-Convex Lens, Ø1 in., f = 30.0 mm, ARC: 350–700 nm). The blue and green channels are providing a real-time ratiometric signal. Measurements are performed at a rate of 5.33 min−1 and washing and rinsing cycles (15 min) are executed (by programmable switching the valves) after every 25 min of

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

3

Fig. 1. Schematic illustration of the of the measurement system.

sampling to sustain system integrity. Further, a high-flow/low-flow sample injection method is utilized to provide robust and accurate measurements (details in Section 2.4).

pH(RF1,F2 , T, S) = ˛3(T, S) +

2.3. Calibration procedure The system was calibrated according to NBS/NIST scale. NIST traceable buffers were used (Merck, pH 6.00, 7.00, 8.00, 9.00, 10.00). Standard additions of 30.00 ppt NaCl were performed to determine initial salinity (via conductivity) of the buffers and provide high salinity calibration buffers. The system can be calibrated with any reference system that has consistent and predictable responses within the calibrated range. We chose Merck NIST buffers because they are certified, reliable and not likely to be discontinued. The reference system chosen can be changed at any time and earlier measurements can be recalculated according to another reference system. Calibrations were performed at: low salinity/25 ◦ C, low salinity/15 ◦ C and high salinity/25 ◦ C. This generated three calibration functions that smoothly follow three different four parameter sigmoidal functions (Eq. (1)): log10 [RF1,F2 (pH)] = ˛1 +

For Eqs. (2) and (3) i = 1–4. Quantification of pH can be done with the following equation in combination with Eqs. (2) and (3):

˛2 − ˛1 1 + 10(pH−˛3 )˛4

log10 [((˛2(T, S) − ˛1(T, S))/(log10 (RF1,F2 ) − ˛1(T, S))) − 1] ˛4(T, S) (4)

2.4. High-flow/low-flow sampling principle The sampling cup consists of a plastic falcon test tube (50 mL) within a larger plastic bottle to which the waste is connected (Fig. 2). Fresh sample inlet is in the bottom of the sampling tube, and sampling occurs approximately 20 mm above the bottom. Therefore, the sample is shielded from ambient air by a 90 mm layer of water (non-introduced sample). The cup has a volume of 50 mL

(1)

where RF1,F2 is the averaged blue/green ratio from the CMOS camera and ˛1 –˛4 are measurement system parameters. The ˛1 and ˛2 are app ratiometric minimum and maximum signal, ˛3 is the pKa and ˛4 is related to the slope of the function. Analogous to Hakonen and Hulth’s time dependent correction of the parameters to eliminate signal drift [36], the temperature (T) adjustment of the parameters are linearly transformed between 25 ◦ C and 15 ◦ C (Eq. (2)), while the salinity (S) parameter transformation were found to be log-linear (Eq. (3)). Where the temperature and salinity ranges were chosen according to the actual measurement ranges. ˛i (T ) = ˛i,T 1 +

˛i,T 1 − ˛i,T 2 ·T TT 1,T 2

˛i (S) = ˛i,S1 +

˛i,S1 − ˛i,S2 · ln SS1,S2

(2)

S S0

(3)

Fig. 2. Schematic illustration of the principles of the high-flow/low-flow sample injection. Blue arrows show general water flow, inflow and waste arrow are proportional to tubing size. Note that the inner Falcon tube bottom is intact and the entire flow is turned upwards into the high-flow red arrows. The small black pH system inflow arrow represents the low-flow. Small green arrows demonstrate diffusion of ambient CO2 related species. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

4

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

Fig. 3. Molecular model of 6,8-Dihydroxypyrene-1,3-disulfonic acid (DHPDS) and emission spectra at different pH values (6.51, 7.51 and 8.48) when excited by 405 nm light.

1200 mL min−1

(height 110 mm), the sample inlet flow rate is and the sample tube is filled in 2.5 s. Hence an upward mass transport rate of 44 mm s−1 is attained within the sample test tube (eliminating ambient CO2 perturbation). Further, the high-flow/low-flow sampling interface also minimizes inlet of particles, microbes and air bubbles into the measurement system, since the flow rate into the measurement system is very small (0.347 mL min−1 ) the suction force will be too small to inject turbid particles/bubbles into it. The sample inlet is orthogonal to the high sample flow to further minimize these undesirable introductions into the pH system (i.e. the high flow cannot contribute with a force vector in 90◦ , into the system). 2.5. Labview software Software on National Instruments (NI) Labview 10 platform were developed to communicate with (interface within parenthesis): camera (USB), pump (NI-RS232/USB), temperature sensor (NI-DAQ/USB), ferry-box computer (NI-RS232/USB), Seabird salinity measurement system (NI-RS232/USB), receive an analog in-harbor signal (NI-DAQ) and control two valves (NI-DAQ/USB) for washing/rinsing. The in-harbor signal stops the system after wash. Unused parts of the system can be turned off in the program set-up. The software controls the system collects data and performs all calculations, thus a near real-time temperature and salinity corrected pH value is continuously delivered. The software is available in the supplementary information including a runtime version of Labview (that runs the program legally without licence) and a manual. Programming in runtime is not possible, but calibration and camera parameters etc. can be changed in our software. 3. Results 3.1. Measurement principles The system is based on fluorescence of 6,8-Dihydroxypyrene1,3-disulfonic acid (Fig. 3), of which the partial photoacidic and the single excitation - dual emission ratiometric properties provide a unique platform for green and blue imaging detection [37]. Photoacidity expresses that a fluorophore has a significantly lower pKa in the excited state than in the ground state, and dual

Fig. 4. A representative ship-board transect of salinity data for a route from Kemi to Gothenburg (GOT).

emission ratiometry refers to fluorophores having two clearly separated emission maxima. The sensitivity was found to be approximately 5% better using excitation at 405 nm compared to previously demonstrated excitation at 420 nm [37], likely due to reduced background scattering into the blue channel. The emission spectrum displayed (Fig. 3) a convenient redshift in fluorescence from blue emission (Emmax ≈ 450 nm) toward green (Emmax ≈ 500 nm), when going from acidic to alkaline conditions. This enabled near real-time color camera detection, where the CMOS spectral sensitivity is used for to generate a blue/green ratio. The excitation light source (405 nm LED) and fluorescence detector (CMOS camera) are both based on low cost semiconductor technology. The power consumption is less than 1 W for the fluorescence equipment. The pH response of DHPDS follows a smooth sigmoidal function that can be described with four parameters that each has an intrinsic salinity and temperature dependence (Section 2, Eqs. (1)–(4)). Parametric assessment provides gradual transition between calibration curves at different salinities and temperatures. The method demonstrated a wide operating range due to the dual pH sensitive hydroxyl groups with proximate pKa ’s in the pH region 7.3–8.5 [37], app giving a combined pKa (inflection point of the sigmoidal calibration curve) of 8.0063 ± 0.0009 for this particular system at salinity 4 and 25 ◦ C. 3.2. High resolution ship-board pH measurements Fig. 4 illustrates the sampling region, including a high-resolution salinity trace. While the salinity was highly variable (7.0–30) in the Transition area between, the Baltic proper, the Belt area and the Kattegat, stable salinities were observed in the Baltic proper (7.0), Bothnian Sea (5.2) and Bothnian Bay (3.0), respectively. Fig. 5a demonstrates an ship-board mapping of high-resolution pH data for a normal route from Kemi to Gothenburg (11 219 measurements), with approximately 10 and 4 measurements per km and minute, respectively. The results are displayed at 15 ◦ C, which was chosen as this is approximately the yearly average temperature in the measurement region. In Fig. 5a (and normally) an evident pH gradient is displayed from normal ocean pH levels (8.1–8.2) in the

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

5

Fig. 5. (a) A representative ship-board transect of pH data (temperature and salinity corrected NIST scale, T = 15 ◦ C) data for a route from Kemi to Gothenburg 14–18 Oct. 2011. (b) All pH transects from the initial sampling period (1 Oct.–13 Dec. 2011) represented in longitude, latitude, pH 3-D space (colored datapoints according to the pH colorbars). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

eastern parts of the Transition area (Kattegatt and Belt Sea) down to below pH 7 in the northern part of Bothnian bay. The pH gradient is relatively slow and continuous from the Transition area through the Baltic proper and the Bothnian sea, with a pH change of approximately 0.2 units over 1000 km (pH 8.0–7.8). However, at the edge to the Bothnian bay there is a quite steep gradient, approximately 0.6 pH units over 100 km (pH 7.8–7.2), which likely can be explained by the absence of halocline in the shallow gulf and pronounced river input. Fig. 5b shows all pH data from one of the initial sampling periods in 2011 (Oct 1–Dec 13; pause 9 Nov–22 Nov). More than 160 000 high-quality pH measurements were executed. The highest pH values (8.2–8.4) were throughout measured in the Skagerrak/Kattegat area, while the lowest pH values (7.1–7.4) were recorded in the northern part of the Bothnian Bay (Fig. 5b). Closer to the coast and near the harbors of Kemi, Oulo, Lübeck and Gothenburg, pH was normally significantly lower (<7.2). The lowest pH values were found close to Oulo (<6.8). 3.3. Ship-board analytical performance For a system which continuously samples a steady flow of water true replicates are typically not accessible. However, to estimate precision of measurements the most non-directional (stable) pH sequence of each sampling cycle (25 min, 100 samples) were identified from the lowest derivative (ıpH/ıt). Here, the standard deviation of three consecutive measurements was determined. A pooled standard deviation (IUPAC, ntot = 4182; nsample = 3) over

Fig. 6. Precision of measurements was estimated at the lowest derivative (ıpH/ıt) of the sample sequence. The standard deviation of three consecutive measurements of pH is indicated in a–c as a function of pH (a), salinity (b) and temperature (c).

the sampling cycles throughout measurements was determined to 0.0001 pH units (full range was 3 × 10−6 to 3 × 10−3 ). 99% of these measurements showed a standard deviation better than 0.0005 pH units, with only 3 high values (above 0.001). Fig. 6a displays the S.D. as a function of pH value to determine trends and correlation of precision with measured pH. The correlation coefficient for a linear trendline was R2 = 0.0583 demonstrating a very small pH dependence of the precision within the pH interval (commonly 7–8.4). The precisions dependence of salinity is shown in Fig. 6b, displaying an even lower correlation (R2 = 0.0021). Three quite sharp lines at specific salinities are displayed in Fig. 6b, corresponding to the three salinity stable regions of the Baltic proper (7.0), Bothnian Sea (5.2) and Bothnian Bay (3.0). A slight overrepresentation of higher standard deviations (S.D. > 0.0004) can possibly be noticed for salinities at 3 and below. Plots of S.D. versus measured temperature (Fig. 4c) and time (during all measurements) revealed similarly low correlation coefficients, R2 = 0.0136 and 0.0036, respectively.

6

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

Fig. 8. Calibration functions during a route from Gothenburg to Oulu (salinity/temperature: 30.10/24.33, 20.13/24.88, 15.66/24.82, 10.03/24.82, 7.00/23.56, 5.68/21.68, 4.00/21.69, 3.00/21.63).

Fig. 7. Continuous laboratory measurements of pH and temperature in a NIST buffer solution (Merck, pH 8.00 at 20 ◦ C). There were 7289 pH and temperature measurements corresponding to 20 h of measurements. The red line denotes pH values measured by the fluorescence system, while the green line corresponds to pH values calculated from the pH and temperature given for the buffer solution. The downward peaks in measured pH are errors likely associated with air bubbles in the sample cuvette. In total, 99.998% of samples were considered reliable. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)

observed in the laboratory measurements (with an artificial steep gradient). For the temperature constant (approximately) sequence, around measurement 3000–5000, an average absolute measurement error was determined to 0.0014 pH units (T = 15.4831 ± 0.0256 ◦ C; ntot = 2125). The systematic error for these measurements were 0.0013, and the random bias (precision) was 0.0001. From our measurements the pH of that specific buffer was 8.0013 ± 0.0001 (T = 20 ◦ C). 3.5. Salinity assessment

The performance of the pH system was considered homogeneous throughout the measured intervals of pH (7–8.4), salinity (3–31), measured temperature (17–28), time (1 Oct–13 Dec 2011). Further, the accuracy of the system was verified before the measurement period with NIST buffers (pHs 7 and 8) of 4 different salinities (see salinity assessment and methods) and a few weeks later with NIST buffer (pH 8) at salinity 4 and 34. The second affirmation of the accuracy was made with only 2 solutions (instead of 8) because it needed to be done while the ship was in Gothenburg harbor. The accuracy was within the limiting accuracy (S.D. ± 0.01) of the NIST buffers. 3.4. Temperature assessment To ensure that proper temperature adjustment of the samples are made a 20 h experiment performing 7289 measurements, on a pH 8.00 (±0.01) at 20 ◦ C NIST buffer (Merck), were made with various temperatures between 23 and 15 ◦ C (Fig. 5). Analysis of standard deviations (S.D.) for three consecutive measurements and set limit of 0.01 pH units as maximum tolerance showed that twelve measurements were above that limit. The measurement success rate for that tolerance would thus be 99.998%. The measurements showed good concurrence with calculated, from temperature according to Merck, buffer pH values. However, with the steep downward temperature gradient in the beginning of the experiment a significant difference is exposed, showing a noticeable positive error of measurements (∼0.01 pH units). This error was likely caused by a slight temperature shift between the fluorescence cuvette and the temperature measurement cell due to the fact that the water (sample) temperature will affect the temperature of the measurement cuvette (Fig. 7). A pronounced gradient in temperature may thus cause temporary erratic measurements of temperature in relation to measured fluorescence leading to errors up to 0.01 pH units. On board measurements, however, the temperature gradient over time is significantly lower and would normally not cause these artifacts

Buffers of pH 7.00 and 8.00 were prepared to intermediate salinities (6.79 and 18.63), and 50 measurements of each (ntot = 200) were made and all measurements were within the ±0.01 pH units stated on the original buffers (when buffer values were salinity and temperature adjusted). This was also confirmed by another (identical) system, where more than 1400 measurements on all pH (6, 7, 8, 9, 10) buffers at multiple salinities (in salinity range 3–35, T = 23–25 ◦ C) showed an average error of 0.008 pH units. The sigmoidal calibration function has a salinity dependence that is accounted for with a logarithmic parameter correction (see Section 2). Typical functions during a trip from Gothenburg to Kemi are shown in Fig. 8, for salinities of 3–30 (normal range). 4. Discussion One account for this systems high analytical performance is the imaging approach and the large statistical replication it provides. The standard deviation () of n samples decrease according √ to:  n =  0 / (n). Normally, sample replication rests at about 10 for practical reasons and low improvement in standard deviation above that point (by moderate increase in replication). However, using an imaging detector the n is easily increased by orders of magnitude, therefore, yielding the potential of high-precision measurements to a somewhat hampered method at pixel level. In this case a precision at single pixel level of approximately 0.1 pH unit is enhanced to below 0.0001 pH units at full device level (1.3 Mpixels) and highly stable conditions. When sampling for pH measurements careful protection of samples from ambient air containing pH affecting gases such as CO2 , NH3 and H2 S, where at least carbon dioxide is readily abundant and fluctuating in the machine room of a large ship (location of the measurement system). As mentioned above a seawater sample of salinity 35 at 25 ◦ C can alter pH from 8.293 to 7.934 when increasing ambient carbon dioxide from normal outdoor conditions (∼380 ppm) to 1000 ppm [26], and the difference will normally

A. Hakonen et al. / Analytica Chimica Acta 786 (2013) 1–7

be much more severe in low salinity/alkalinity samples. Indoor air usually contains significantly higher amounts of CO2 than the atmosphere in general, due to insufficient ventilation and people or equipment generating carbon dioxide. As an example of air quality regulations, The United States Occupational Safety and Health Administration (OSHA) limit carbon dioxide concentration in the workplace to 5000 ppm for prolonged periods, and 35,000 ppm for periods up to 15 min. Consequently, pH measurements at indoor air conditions can lead to vast measurement errors. The present sampling design (Fig. 2) is protecting the actual injected sample with: a thick diffusion layer and mass transport of non-injected sample orders of magnitude faster than diffusion of CO2 . The upward mass transport rate is 44 mm s−1 while the diffusion constant of CO2 in water is 0.0016 mm2 s−1 [38], and consequently the one dimensional (downward into the sample tube) diffusion length for 1 s √ √ is 2 (Dt) = 2 (0.0016 × 1) = 0.08 mm. Therefore, diffusion of arbitrary CO2 (or other pH affecting gases) cannot compete with mass transport of the sample due to the significant difference in transport rates (550 times greater). Additionally, this sampling method also introduced a highflow/low-flow liquid interface (Fig. 2), at the injection point, which minimized insertion of particles, air bubbles and micro-organisms into the system by a sort of filter-less “filtering” mechanism, hence minimizing spectroscopic interferences, clogging and biofilm formation in the measurement system. Other potential methods for high-resolution monitoring are electrodes and optical chemical sensors (optodes) [39–42]. However, the necessary precision of below 0.002 pH units needed for ocean acidification monitoring are normally not realized with these methods, and signal drift over time and response times may compromise performance for these methods [43–45]. Our new method may also have advantages compared to spectrophotometric methods based on m-cresol purple due to lower sensitivity to interferences and dye impurities and minimal dye consumption. 5. Conclusions A convenient, economic and low power system for highperformance pH measurements was developed. The sampling volume needed was small (25 ␮L) leading to a low consumption of fluorophore (<4 mg on a typical sampling week; ∼20 000 measurements). The novel pH-measurement method was shown to work on a merchant vessel for months in challenging natural environments with high variability in salinity, organic matter, biological activity, pH and temperature. High precision (commonly down to S.D. 0.0001 pH units) results were obtained in a vast natural salinity range of 3–31. This makes it possible to measure pH at a precision high enough to observe subtle long-term changes (ocean acidification) caused by emission from burning of fossil fuels and other sources. The automated method provides a means to make observations over large sea areas with high temporal and spatial resolution. The system is versatile and applicable in basically any natural aqueous solution where high-performance pH measurements are required. Perspectives for the system are miniaturization and autonomous wireless monitoring. This can be realized by transference toward a microfluidic device with a smartphone as control and detection appliance and solar cells as power supply. Acknowledgements We thank the Transatlantic shipping company and the friendly and helpful crew on M/S Transpaper. Funding was provided by the Swedish Environmental Protection Agency.

7

References [1] O. Hoegh-Guldberg, P.J. Mumby, A.J. Hooten, R.S. Steneck, P. Greenfield, E. Gomez, C.D. Harvell, P.F. Sale, A.J. Edwards, K. Caldeira, N. Knowlton, C.M. Eakin, R. Iglesias-Prieto, N. Muthiga, R.H. Bradbury, A. Dubi, M.E. Hatziolos, Science 318 (2007) 1737. [2] J.C. Orr, V.J. Fabry, O. Aumont, L. Bopp, S.C. Doney, R.A. Feely, A. Gnanadesikan, N. Gruber, A. Ishida, F. Joos, R.M. Key, K. Lindsay, E. Maier-Reimer, R. Matear, P. Monfray, A. Mouchet, R.G. Najjar, G.-K. Plattner, K.B. Rodgers, C.L. Sabine, J.L. Sarmiento, R. Schlitzer, R.D. Slater, I.J. Totterdell, M.-F. Weirig, Y. Yamanaka, A. Yool, Nature 437 (2005) 681. [3] U. Riebesell, K.G. Schulz, R.G.J. Bellerby, M. Botros, P. Fritsche, M. Meyerhofer, C. Neill, G. Nondal, A. Oschlies, J. Wohlers, E. Zollner, Nature 450 (2007) 545. [4] K. Caldeira, M.E. Wickett, Nature 425 (2003) 365. [5] K.P. Helmle, R.E. Dodge, P.K. Swart, D.K. Gledhill, C.M. Eakin, Nat. Commun. 2 (2011) 215. [6] J.A. Kleypas, R.W. Buddemeier, D. Archer, J.-P. Gattuso, C. Langdon, B.N. Opdyke, Science 284 (1999) 118. [7] R. Rodolfo-Metalpa, F. Houlbreque, E. Tambutte, F. Boisson, C. Baggini, F.P. Patti, R. Jeffree, M. Fine, A. Foggo, J.P. Gattuso, J.M. Hall-Spencer, Nat. Clim. Change 1 (2011) 308. [8] S. Dupont, O. Ortega-Martínez, M. Thorndyke, Ecotoxicology 19 (2010) 449. [9] P. Kaniewska, P.R. Campbell, D.I. Kline, M. Rodriguez-Lanetty, D.J. Miller, S. Dove, O. Hoegh-Guldberg, PLoS ONE 7 (2012) e34659. [10] S. Dupont, A. Moya, X. Bailly, PLoS ONE 7 (2012) e29568. [11] R.J. Diaz, R. Rosenberg, Science 321 (2008) 926. [12] J.M. Beman, C.-E. Chow, A.L. King, Y. Feng, J.A. Fuhrman, A. Andersson, N.R. Bates, B.N. Popp, D.A. Hutchins, Proc. Natl. Acad. Sci. U.S.A. 108 (2011) 208. [13] P.L. Munday, D.L. Dixson, M.I. McCormick, M. Meekan, M.C.O. Ferrari, D.P. Chivers, Proc. Natl. Acad. Sci. U.S.A. 107 (2010) 12930. [14] D. Shi, Y. Xu, B.M. Hopkinson, F.M.M. Morel, Science 327 (2010) 676. [15] M.R.W. Rands, W.M. Adams, L. Bennun, S.H.M. Butchart, A. Clements, D. Coomes, A. Entwistle, I. Hodge, V. Kapos, J.P.W. Scharlemann, W.J. Sutherland, B. Vira, Science 329 (2010) 1298. [16] L. Alegret, E. Thomas, K.C. Lohmann, Proc. Natl. Acad. Sci. U.S.A. 109 (2012) 728. [17] M. Ruhl, N.R. Bonis, G.-J. Reichart, J.S.S. Damsté, W.M. Kürschner, Science 333 (2011) 430. [18] D.E. Ogden, N.H. Sleep, Proc. Natl. Acad. Sci. U.S.A. 109 (2012) 59. [19] R.A. Kerr, Science 328 (2010) 1500. [20] D. Woolf, J.E. Amonette, F.A. Street-Perrott, J. Lehmann, S. Joseph, Nat. Commun. 1 (2010) 56. [21] J.E. Dore, R. Lukas, D.W. Sadler, M.J. Church, D.M. Karl, Proc. Natl. Acad. Sci. U.S.A. 106 (2009) 12235. [22] G.E. Hofmann, J.E. Smith, K.S. Johnson, U. Send, L.A. Levin, F. Micheli, A. Paytan, N.N. Price, B. Peterson, Y. Takeshita, P.G. Matson, E.D. Crook, K.J. Kroeker, M.C. Gambi, E.B. Rivest, C.A. Frieder, P.C. Yu, T.R. Martz, PLoS ONE 6 (2011) e28983. [23] D.I. Kline, L. Teneva, K. Schneider, T. Miard, A. Chai, M. Marker, K. Headley, B. Opdyke, M. Nash, M. Valetich, J.K. Caves, B.D. Russell, S.D. Connell, B.J. Kirkwood, P. Brewer, E. Peltzer, J. Silverman, K. Caldeira, R.B. Dunbar, J.R. Koseff, S.G. Monismith, B.G. Mitchell, S. Dove, O. Hoegh-Guldberg, Sci. Rep. 2 (2012), http://dx.doi.org/10.1038/srep00413. [24] V.M.C. Rérolle, C.F.A. Floquet, M.C. Mowlem, R.R.G.J. Bellerby, D.P. Connelly, E.P. Achterberg, TrAC – Trends Anal. Chem. 40 (2012) 146. [25] J.D. Hardege, J.M. Rotchell, J. Terschak, G.M. Greenway, TrAC – Trends Anal. Chem. 30 (2011) 1320. [26] G.M. Marion, F.J. Millero, M.F. Camões, P. Spitzer, R. Feistel, C.T.A. Chen, Mar. Chem. 126 (2011) 89. [27] J.T. Wootton, C.A. Pfister, J.D. Forester, Proc. Natl. Acad. Sci. U.S.A. 105 (2008) 18848. [28] A. Ulfsbo, S. Hulth, L.G. Anderson, Mar. Chem. 127 (2011) 20. [29] T.D. Clayton, R.H. Byrne, Deep Sea Res. I: Oceanogr. Res. Papers 40 (1993) 2115. [30] G.L. Robert-Baldo, M.J. Morris, R.H. Byrne, Anal. Chem. 57 (1985) 2564. [31] S. Aßmann, C. Frank, A. Körtzinger, Ocean Sci. 7 (2011) 597. [32] R.G.J. Bellerby, A. Olsen, T. Johannessen, P. Croot, Talanta 56 (2002) 61. [33] X. Liu, M.C. Patsavas, R.H. Byrne, Environ. Sci. Technol. 45 (2011) 4862. [34] H. Ojaveer, A. Jaanus, B.R. MacKenzie, G. Martin, S. Olenin, T. Radziejewska, I. Telesh, M.L. Zettler, A. Zaiko, PLoS ONE 5 (2010) e12467. [35] C. Ainsworth, Science 322 (2008) 1627. [36] A. Hakonen, S. Hulth, Anal. Chim. Acta 606 (2008) 63. [37] A. Hakonen, S. Hulth, Talanta 80 (2010) 1964. [38] Handbook of Chemistry and Physics, CRC Press/Taylor and Francis, Boca Raton, FL, 2012. [39] D. Aigner, S.M. Borisov, F.J.O. Fernandez, J.F.F. Sanchez, R. Saf, I. Klimant, Talanta 99 (2012) 194. [40] N. Stromberg, A. Hakonen, Anal. Chim. Acta 704 (2011) 139. [41] H.M.R. Goncalves, C.D. Maule, P.A.S. Jorge, J. da Silva, Anal. Chim. Acta 626 (2008) 62. [42] I. Kasik, J. Mrazek, T. Martan, M. Pospisilova, O. Podrazky, V. Matejec, K. Hoyerova, M. Kaminek, Anal. Bioanal. Chem. 398 (2010) 1883. [43] N. Strömberg, E. Mattsson, A. Hakonen, Anal. Chim. Acta 636 (2009) 89. [44] A. Hakonen, S. Hulth, S. Dufour, Talanta 81 (2010) 1393. [45] D. Aigner, S.M. Borisov, I. Klimant, Anal. Bioanal. Chem. 400 (2011) 2475.