Inclusion of uncertainty in the calcium-salinity relationship improves estimates of ocean acidification monitoring data quality
Introduction
The anthropogenically driven increase of atmospheric pCO2 from 280 μatm (pre-industrial) to approximately 410 μatm has had many impacts on the global environment (Le Quéré et al., 2015). However, this increase in the atmosphere represents only part of anthropogenic emissions, since the oceans have absorbed approximately 40% of CO2 emissions since the beginning of the industrial era (Sabine et al., 2004; Zeebe et al., 2008). This oceanic uptake of CO2 has led to a process known as ocean acidification (OA), whereby the chemical balance of the marine carbonate system has shifted towards the production of hydronium (H3O+) and bicarbonate (HCO3−) ions and the reduction of carbonate ions (CO32−) (Dickson et al., 2007). Although the full impact of these changes is still being explored, it is clear that the effects on ocean chemistry, physical properties, ocean ecosystems, and biogeochemical processes will be substantial and long-lasting (Guinotte and Fabry, 2008; Hofmann et al., 2010; Kroeker et al., 2013).
Our ability to monitor and assess the impacts of OA relies on an accurate representation of the marine carbonate system. The marine carbonate system can be characterized with knowledge of any two of the four carbonate parameters of an oceanographic water sample: pH, pCO2, dissolved inorganic carbon concentration (DIC), and total alkalinity (AT), along with the appropriate thermodynamic equilibrium constants and ancillary data, e.g. salinity, temperature, pressure, and total phosphorus and silicate concentrations. When characterising the marine carbonate system, the uncertainty of each output variable will depend on the measurement uncertainty of chemical and physical parameters (e.g. DIC, AT, salinity, temperature), as well as the uncertainty in the thermodynamic equilibrium constants (Millero, 2007).
To estimate the uncertainty in output carbonate variables (e.g. [H+], [CO32−], ΩA, ΩC), Orr et al. (2018) recently expanded several of the open-source carbonate system software packages to allow routine propagation of uncertainty in both measurements and thermodynamic constants. The uncertainty estimation methods implemented in these widely used software packages are based on previous work, which introduced approaches based on Gaussian uncertainty propagation (Dickson and Riley, 1978; Millero, 1995) and Monte Carlo simulations (Fassbender et al., 2017; Williams et al., 2017). Both methods are included in the R-based seacarb package(Jean-Pierre Gattuso et al., 2020), which allows direct comparison between the approaches; the other packages only support Gaussian uncertainty propagation (Orr et al., 2018).
Studies into uncertainty estimates using these revised software packages have focused on the contribution of the carbonate input pair (e.g. AT – DIC, pH – AT) and the thermodynamic constants (Orr et al., 2018). However, there has been little consideration into the effects of uncertainty in salinity measurements and calcium ion concentration upon the standard uncertainty of carbonate system calculations. Salinity is a fundamental property of the marine carbonate system, affecting thermodynamic equilibrium constants and physical properties of seawater (e.g. density). Although uncertainty in salinity is included as an input parameter in the uncertainty packages, the effect of salinity uncertainty has been largely ignored because measurements made with the high quality salinometers or CTD instruments used by experienced laboratories are generally considered to have an uncertainty of only ±0.002 (or 0.005% at S = 35) on the Practical Salinity Scale (PSU) (Le Menn, 2011). However, the measurement uncertainty of the salinometers and conductivity meters which are commonly used in coastal studies and by emerging laboratories can be much higher, e.g. 1% for the YSI Pro30 (YSI, USA) and 2% for HI 98194 (HANNA instruments, Italy). These relative uncertainties eclipse the typical uncertainties associated with the other carbonate parameters (Table 1). Thus, the impact of salinity uncertainty on carbonate system characterisation must be determined in order to properly assess the impacts of OA.
Another source of uncertainty in carbonate system characterisation that has received little consideration is the effect of uncertainty in the calcium ion concentration, [Ca2+], which is required for determination of the saturation state of aragonite (ΩA) and calcite (ΩC), the two common polymorphs of CaCO3 found in marine systems. In most carbonate system characterisation software packages, [Ca2+] is calculated from the conservative [Ca2+]-salinity relationship empirically determined by Riley and Tongudai (1967), which found the [Ca2+]-salinity ratio to be 2.938 × 10−4 mol kg−1 PSU−1 when averaged across all major ocean basins and depths. Two sources of [Ca2+] uncertainty can be derived from the use of this relationship: the direct impact of salinity measurement uncertainty on [Ca2+] calculation, and unknown deviations from the expected ratio. In the open ocean, deviations from this global mean [Ca2+]-salinity relationship are generally small (≈0.2%), so it is deemed unnecessary to consider localised variability in [Ca2+] in Ω uncertainty estimations (Atwood et al., 1973; Besson et al., 2014; Culkin and Cox, 1966; Krumgalz, 1982). However, some studies have demonstrated greater spatial and temporal divergence from the global oceanic mean [Ca2+]-salinity relationship (Rosón et al., 2016; Sharp and Byrne, 2019; Tsunogai et al., 1968). In an extreme case, temporal variations in the [Ca2+]-salinity ratio of up to 10% of the global mean was measured in the Sargasso Sea (Billings et al., 1969). Temporal variations in the chemistry of coastal waters, which are affected by riverine inputs, geological weathering of calcium–rich environments, the production or dissolution of CaCO3, and upwelling of deep water have also been observed (Beckwith et al., 2019; Jiang et al., 2014; Mathis et al., 2011). In coastal waters of the Bay of Bengal the average [Ca2+]-salinity ratio of coastal waters was approximately 5–8% higher than the global open ocean relationship (Das and Sahoo, 1996; Sen Gupta et al., 1978). Nessim et al. (2015) found the [Ca2+]-salinity ratio in waters of the Egyptian Mediterranean coastline varied seasonally, and were on average 18% lower than the global average. Clearly, because of the direct proportionality between [Ca2+] and Ω, consideration and accommodation of variability in the [Ca2+]-salinity ratio in carbonate system characterisation would improve the accuracy and reliability of OA monitoring efforts in coastal waters.
Although the implications of [Ca2+]-salinity variabilities on Ω determinations and carbonate system monitoring have been noted in previous studies (Wanninkhof et al., 2015), [Ca2+] uncertainty is not an included variable in any of the uncertainty propagation packages. This exclusion leads to inaccuracies in Ω calculations and underestimation of relative standard uncertainty of Ω. Here, we examine the effects of uncertainty in salinity and the [Ca2+]-salinity relationship on carbonate system calculations using seacarb's Gaussian uncertainty propagation and Monte Carlo simulations. The published routines were updated to accurately simulate the effect of larger salinity measurement uncertainties and, for the first time, these routines include the effect of [Ca2+] uncertainty. Salinity and [Ca2+] uncertainties of up to 5% were shown to increase the propagated combined uncertainties of [CO32−] and ΩA by approximately 20% and 50%, respectively. These increases in carbonate system output uncertainties have substantial implications on OA monitoring data quality. Incorporation of [Ca2+] uncertainty into the updated routines provides more accurate carbonate system characterisation.
Section snippets
Uncertainty terminology
We have adopted the technical terminology of the Guide to the Expression of Uncertainty in Measurement (ISO, 1993), whereby error and uncertainty are not synonymous. Error is the difference between an input variable's measured value and its true value, which can be either positive or negative. Although some measurement errors can be corrected for (e.g. dark spectrum subtraction in spectroscopy), the total error cannot be known precisely because all sources of error are not known. Uncertainty is
Correction to original Monte Carlo method
Previous work has validated the application of Gaussian uncertainty propagation method to the carbonate system by comparison with the Monte Carlo method at uS = 0 (Orr et al., 2018). However, upon comparison of seacarb's original Gaussian and Monte Carlo routines using more realistic %uS values, the Gaussian approach showed 20% and 8% increases in %uc[CO32−] and %ucΩA, respectively (Table S2), as uS increased from 0 to 5%. In contrast, the original Monte Carlo routine exhibited no sensitivity
Conclusion
The Gaussian and Monte Carlo uncertainty estimation methods within the R-based seacarb package were extended to include the uncertainty in [Ca2+] on propagated uncertainties in saturation state determinations. These updated routines were used to demonstrate how underestimation of uncertainty in [Ca2+] leads to inadvertent violation of GOA-ON's guidelines for both Climate- and Weather-quality OA monitoring data. Underestimation of uncertainty in [Ca2+] is of primary concern in coastal regions,
Declaration of Competing Interest
None.
References (43)
- et al.
Temporal variation of alkaline earth element/chlorinity ratios in the sargasso sea
Earth Planet. Sci. Lett.
(1969) - et al.
Sodium, potassium, magnesium, calcium and strontium in sea water
Deep-Sea Res. Oceanogr. Abstr.
(1966) Standard potential of the reaction: AgCl(s) + 12H2(g) = Ag(s) + HCl(aq), and and the standard acidity constant of the ion HSO4− in synthetic sea water from 273.15 to 318.15 K
J. Chem. Thermodyn.
(1990)- et al.
The effect of analytical error on the evaluation of the components of the aquatic carbon-dioxide system
Mar. Chem.
(1978) - et al.
Ocean pCO2 calculated from dissolved inorganic carbon, alkalinity, and equations for K1 and K2: validation based on laboratory measurements of CO2 in gas and seawater at equilibrium
Mar. Chem.
(2000) Thermodynamics of the carbon dioxide system in the oceans
Geochim. Cosmochim. Acta
(1995)- et al.
Chemistry of the Egyptian Mediterranean coastal waters
Egypt. J. Aquat. Res.
(2015) - et al.
Routine uncertainty propagation for the marine carbon dioxide system
Mar. Chem.
(2018) - et al.
Association constant of fluoride and hydrogen ions in seawater
Mar. Chem.
(1987) - et al.
The major cation/chlorinity ratios in sea water
Chem. Geol.
(1967)
Calcium distribution in the subtropical Atlantic Ocean: implications for calcium excess and saturation horizons
J. Mar. Syst.
A study of fluoride, calcium and magnesium in the northern Indian Ocean
Mar. Chem.
Carbonate ion concentrations in seawater: spectrophotometric determination at ambient temperatures and evaluation of propagated calculation uncertainties
Mar. Chem.
The boron/chlorinity ratio of deep-sea water from the Pacific Ocean
Deep-Sea Res. Oceanogr. Abstr.
Ocean acidification along the Gulf Coast and East Coast of the USA
Cont. Shelf Res.
Contributions of organic alkalinity to total alkalinity in coastal waters: a spectrophotometric approach
Mar. Chem.
Temporal and depth study of alkaline earth chlorinity ratios in seawater at a single station south of Puerto Rico
Limnol. Oceanogr.
Riverine calcium end-members improve coastal saturation state calculations and reveal regionally variable calcification potential
Front. Mar. Sci.
Calcium, Na, K and mg concentrations in seawater by inductively coupled plasma-atomic emission spectrometry: applications to IAPSO seawater reference material, hydrothermal fluids and synthetic seawater solutions
Geostand. Geoanal. Res.
Influence of fresh water influx on the ratio of Ca, Mg and F with respect to cl in the coastal water of Orissa, east coast of India
Indian J. Mar. Sci.
Guide to best practices for ocean CO2 measurements
Cited by (5)
Multivariate regression studies for investigating and setting the action levels for the system water quality parameters of a nuclear plant
2022, Water and Environment JournalMonitoring Ocean Acidification within State Borders: Lessons from Washington State (USA)
2021, Coastal Management