An Assessment of Anthropogenic Carbon in the Pacific Ocean with Specific Application to Diagnostic Inverse Models of Circulation

Paul E. Robbins and Andrew G. Dickson

ABSTRACT

We propose to assemble total marine carbon measurements in the Pacific Ocean and estimate the "anthropogenic" and "natural" components of the observed distribution. In addition to the direct quantification of the observed inventories, a key motivation of the work is to prepare the WOCE/JGOFS carbon measurements for assimilation and use in a diagnostic inverse model of the Pacific Ocean. The results of the later venture will be estimates of oceanic carbon transport and storage within the the Pacific Basin. The interpretation of directly observed oceanic carbon transports is greatly enhanced by a separation of the observed concentrations into "anthropogenic" and "natural" components. Calculating the oceanic transport of these properties in basin-scale box-inverse requires careful and accurate assessment of the magnitude and character of error covariance of each component of the total carbon field. Previous methods to estimate the anthropogenic carbon distribution (Gruber et al, 1996) have not considered the errors arising from assumptions implicit in the methodology. Specifically, how the error in the estimate of the anthropogenic carbon at one point is correlated with error at another geographic location. We propose to use a hierarchy of simple kinematic ventilation models to assess the impact of these overlooked processes on the accuracy and precision of the oceanic anthropogenic carbon calculation. The models will simulate the invasion of anthropogenic carbon and other transient tracers in the ocean. The anthropogenic carbon content will be estimated from the simulated tracer fields and compared with the actual simulated increase in total carbon concentration. The results of the numerical simulations will serve as guides to identify which processes within the ocean (mixing, differing source water properties, geographically variable Redfield ratios, undersaturation of oxygen in the mixed layer) have greatest impact on the method. Additionally, the numerical simulations will aid in assessing the scales of geographic correlation of the error covariance of the estimated fields. The results of this project will be available to the community via an electronic atlas (online and CDROM) currently being assembled for the WOCE Pacific one-time survey.