Dynamical interpolation of surface pCO2 between lines of observation in the North Atlantic Ocean

• Main Author: Friedrich, Tobias
• Other Authors: Oschlies, Andreas, Wallace, Douglas
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: 2008
• Subjects: ddc:550; thesis; pCO2; CO2; North Atlantic; carbon; VOS; Voluntary Observing Ships; remote sensing
• Online Access: https://nbn-resolving.org/urn:nbn:de:gbv:8-diss-33928; https://macau.uni-kiel.de/receive/diss_mods_00003392; https://macau.uni-kiel.de/servlets/MCRFileNodeServlet/dissertation_derivate_00002665/Dissertation_Tobias_Friedrich.pdf

The present PhD thesis aims to elucidate driving mechanisms of oceanic surface pCO2 variability and to develop and analyze techniques for mapping pCO2 on a basinscale in the North Atlantic. First of all, a number of sensitivity tests are carried out in a coarse resolution coupled ecosystem-circulation model simulating the period 1948-2002. The individual contributions by wind stress and surface heat fluxes to naturally driven interannual-to-decadal variability of air-sea fluxes of CO2 and O2 are examined using different atmospheric forcing fields. The model results reveal a pronounced dominance of wind stress in driving interannual-to-decadal variability of CO2 fluxes in the entire model domain. Although the simulated mean carbon uptake takes place in the subpolar basin, interannual fluctuations are of the same magnitude in the subpolar region, the subtropics and the equatorial Atlantic. For O2, mechanisms causing temporal variations can be separated into a wind-stress driven equatorial and a heat-flux driven subtropical and subpolar basin. Subsequently, the potential of monitoring North Atlantic ocean-surface pCO2 on a basin scale by combining Voluntary Observing Ship (VOS) observations with ARGO float and remote sensing data respectively is explored in the context of an eddy-resolving model. Here, model output is sampled according to realistic VOS-line, ARGO float and satellite coverage of the reference year 2005. The synthetic VOS-line observations form a training data set for a self-organizing neural network which is, in the first case, applied to simulated satellite data of SST and surface chlorophyll in order to derive basinwide monthly maps of surface pCO2. In the second case the trained neural network is used to derive punctual pCO2 estimates from ARGO float SST and salinity data which are extrapolated by objective mapping. For a remote-sensing based mapping the basinwide mean RMS-error amounts to 19.0 ppm when missing data in the satellite coverage due to clouds and low solar irradiation at high ...