Researchers are deploying a global network of ocean sensors to help quantify carbonate chemistry changes related to ocean acidification. In addition to discrete laboratory-based measurements being taken on research cruises in-situ sensors are deployed to measure the numerous CO2-system parameters. This creates a need to ‘intercalibrate’ or validate the various types of data, especially in the case of overdetermined datasets where more than two CO2-system parameters are measured. Understanding the concentrations and associated uncertainties of the parameters rely on calculations of the CO2-system of seawater computed by solvers such as CO2SYS (Lewis and Wallace, 1998). A key drawback to the publicly available solvers is, despite the ever-increasing number of overdetermined datasets and measurements, the solvers themselves can only calculate and propagate errors in an exactly determined system–wherein two of the measured parameters are input. In this poster we present QUODcarb, a novel approach to solving the CO2-system equations in a way that can handle more than two measured input parameters. QUODcarb formulates a Bayesian problem and finds the most probable CO2-system state given the measurements and their precisions. The novel solver then summarizes the posterior probability distribution by approximating it using a Gaussian distribution, from which it draws uncertainty estimates. Come by our poster as we show how the novel Bayesian approach works with a few examples of up to five measurements input simultaneously (pH, pCO2, TA, DIC, CO3^2- into QUODcarb.

Primary Presenter: Marina Fennell, University of California, Irvine (mfennell@uci.edu)

Authors:

Marina Fennell, University of California, Irvine  (mfennell@uci.edu)

Francois Primeau, University of California, Irvine  (fprimeau@uci.edu)