Improved routines to model the ocean carbonate system: mocsy 1.0

Abstract

Abstract. Software used by modelers to compute ocean carbonate chemistry is often based on code from the Ocean Carbon Cycle Model Intercomparison Project (OCMIP), last revised in 2005. As an update, we offer here new publicly available Fortran 95 routines to model the ocean carbonate system (mocsy). Both codes take as input dissolved inorganic carbon CT and total alkalinity AT, the only two tracers of the ocean carbonate system that are unaffected by changes in temperature and salinity and conservative with respect to mixing, properties that make them ideally suited for ocean carbon models. With the same basic thermodynamic equilibria, both codes compute surface-ocean pCO2 in order to simulate air–sea CO2 fluxes. The mocsy package goes beyond the OCMIP code by computing all other carbonate system variables (e.g., pH, CO32−, and CaCO3 saturation states) and by doing so throughout the water column. Moreover, it avoids three common model approximations: that density is constant, that modeled potential temperature is equivalent to in situ temperature, and that depth is equivalent to pressure. These approximations work well at the surface, but total errors in computed variables grow with depth, e.g., reaching −8 μatm in pCO2, +0.010 in pH, and +0.01 in ΩA at 5000 m. Besides the equilibrium constants recommended for best practices, mocsy also offers users three new options: (1) a recent formulation for total boron that increases its ocean content by 4%, (2) an older formulation for KF common to all other such software, and (3) recent formulations for K1 and K2 designed to also include low-salinity waters. More total boron increases borate alkalinity and reduces carbonate alkalinity, which is calculated as a difference from total alkalinity. As a result, the computed surface pCO2 increases by 4 to 6 μatm, while the computed aragonite saturation horizon (ASH) shallows by 60 m in the North Atlantic and by up to 90 m in the Southern Ocean. Changes due to the new formulation for K1 and K2 enhance pCO2 by up to 8 μatm in the deep ocean and in high-latitude surface waters. These changes are comparable in magnitude to errors in the same regions associated with neglecting nutrient contributions to total alkalinity, a common practice in ocean biogeochemical modeling. The mocsy code with the standard options for best practices and none of the 3 approximations agrees with results from the CO2SYS package generally within 0.005%.

Comparison of ten packages that compute ocean carbonate chemistry

“…0.1 µmol kg −1 pH 0.003 0.0003 pK 0 0.002 0.0002 pK 1 0.01 0.001 pK 2 0.02 0.002 pK i (other) 0.01 0.001 at 5000 m, where csys is 0.001 larger than other packages, which agree within 0.0003; for the Revelle factor R f , packages agree within 0.02 throughout the water column except for seacarb, whose discrepancy grows to 0.2 at 5000 m. Although seacarb computes R f with an efficient analytical formula (Frankignoulle, 1994), that approach neglects effects of P T and Si T on total alkalinity, unlike the less efficient numerical approach used in other packages (Orr and Epitalon, 2015). Overall, discrepancies among packages are larger at depth, but they remain negligible (Table 8) except for pH and pCO 2 .…”

“…These differences between packages are at least 70 times smaller than the actual changes in computed variables attributable to alkalinity from P T and Si T . With the A T -C T input pair, this nutrient alkalinity increases computed pCO 2 by 6 µatm for average surface waters in the Southern Ocean and by 12 µatm for average deep waters (below 2000 m); simultaneously, CO 2− 3 is reduced by about 2 µmol kg −1 in the same waters (Orr and Epitalon, 2015).…”

“…Parallel to those developments for the observational and experimental communities, the Ocean Carbon-Cycle Model Intercomparison Project (OCMIP) provided routines to compute surface pCO 2 and air-sea CO 2 fluxes from simulated A T and C T (Orr et al, 1999). Those were adapted to include the full suite of other carbonate system variables (Orr et al, 2005), then later improved and released publicly as the mocsy package (Orr and Epitalon, 2015). To assess the consistency of these packages, we compared results generated by running them with common sets of constants, pH scales, and input data.…”