Reconstructing upper ocean carbon variability using ARGO profiles and CMIP6 models

<p>Historically, ocean carbon content has been poorly sampled due to the logistical difficulties inherent in carbonate chemistry measurements. &#160;As a result, global products of ocean carbon content observations have been restricted to calculate climatologies or long-term trends. Recent innovations with machine learning have provided for observational reconstructions of multidecadal and interannual carbon variability. In this work, we create a complementary method for reconstructing historical carbon variability by drawing upon the Ensemble Optimal Interpolation method used for reconstructing historical ocean heat and salinity ^[1-3]. Ensemble Optimal Interpolation draws upon first-order relationships between variables and use covariances from model ensembles to propagate information from data-rich to data-sparse regions.</p><p>We test our method by conducting synthetic reconstructions of upper ocean carbon content using ARGO-style sampling distributions with CMIP6 ensemble covariance fields. Sensitivity tests of local carbon reconstructions suggest that around 50% of ocean carbon variability can be reconstructed using temperature and salinity measurements. Expanding the synthetic reconstructions to include irregular sampling consistent with ARGO profile locations results in a similar capacity to reconstruct ocean carbon variability, as the increased information provided from multiple sampling locations compensates for the propagation of errors within the CMIP6 covariance fields.&#160; Our initial results indicate that the first-order relationships between temperature, salinity, and carbon can be used to describe a substantial proportion of historical carbon variability. In addition to showing promise for a new historical reconstruction complementary to current products, our work emphasises the important links between hydrographic and carbon variability for much of the global ocean.</p><p>&#160;</p><p>References</p><p>^[1]D. M. Smith and J. M. Murphy, 2007. "An objective ocean temperature and salinity analysis using covariances from a global climate model," <em>JGR Oceans</em>.</p><p>^[2]L. Cheng, K. E. Trenberth, J. T. Fasullo, T. Boyer, J. T. Abraham and J. Zhu, 2017. "Improved estimates of ocean heat content from 1960 to 2015," <em>Science Advances</em>.</p><p>^[3]L. Cheng, K. E. Trenberth, N. Gruber, J. P. Abraham, J. T. Fasullo, G. Li, M. E. Mann, X. Zhao and J. Zhu, 2020. "Improved Estimates of Changes in Upper Ocean Salinity and the Hydrological Cycle," <em>Journal of Climate</em>.</p>