Caldeira Lab Research:Ocean acidification and ocean carbon cycle

Evaluation of ocean carbon cycle models with data-based metrics

K. Matsumoto, J.L. Sarmiento, R.M. Key, O. Aumont, J.L. Bullister, K. Caldeira, J.M. Campin, S.C. Doney, H. Dragne, J.C. Dutay, M. Follows, Y. Gao, A. Gnanadesikan, N. Gruber, A. Ishida, F. Joos, K. Lindsay, E. Maier Reimer, J.C. Marshall, R.J. Matear, P. Monfray, A. Mouchet, R. Najjar, G.K. Plattner, R. Schlitzer, R. Slater, P.S. Swathi, I.K. Totterdell, M.F. Weirif, Y. Yamanaka, A. Yool, J.C. Orr

Based on new data regarding radiocarbon and chlorofluorocarbon-11, the credibility of several different ocean carbon cycle models is analyzed. The results are a red flag to those who use the models: only around a quarter of those analyzed are consistent with new data.

Matsumoto K, Sarmiento JL, Key RM, Aumont O, Bullister JL, Caldeira, K., Campin JM, Doney SC, Drange H, Dutay JC, Follows M, Gao Y, Gnanadesikan A, Gruber N, Ishida A, Joos F, Lindsay K, Maier-Reimer E, Marshall JC, Matear RJ, Monfray P, Mouchet A, Najjar R, Plattner GK, Schlitzer R, Slater R, Swathi PS, Totterdell IJ, Weirig MF, Yamanaka Y, Yool A, Orr JC, Evaluation of ocean carbon cycle models with data-based metrics, Geophysical Research Letters 31, Art No. L007303, 2004.

Observed radiocarbon versus radiocarbon in the models: Observed radiocarbon levels are compared with those in models. The points that fall inside the cross are the models that have an acceptable margin of error for radiocarbon distribution. Very few models agree with the observed data, which means it is not uncommon for them to over or under ventilate the ocean.

Observed and modeled CFC-11: CFC-11 is a chemical whose distribution can be used to determine ocean ventilation on a decadal time scale. Above, observed data regarding CFC-11 distribution is compared to data in models. Again, those that fall inside the cross are models that have an acceptable margin of error in their distribution. Very few models agree well with the observed data.


New radiocarbon and chlorofluorocarbon-11 data from the World Ocean Circulation Experiment are used to assess a suite of 19 ocean carbon cycle models. We use the distributions and inventories of these tracers as quantitative metrics of model skill and find that only about a quarter of the suite is consistent with the new databased metrics. This should serve as a warning bell to the larger community that not all is well with current generation of ocean carbon cycle models. At the same time, this highlights the danger in simply using the available models to represent the state-of-the-art modeling without considering the credibility of each model.