Optimal sampling of the atmosphere for purpose of inverse modeling: A model study

The 66 stations of the GLOBALVIEW-CO2 sampling network (GLOBALVIEW-CO2: Cooperative Atmospheric Data Integration Project - Carbon Dioxide, (1997)) are located primarily remotely from continents where signals of fossil fuel consumption and biospheric exchange are diluted. It is thus not surprising that inversion studies are able to estimate terrestrial sources and sinks only to a very limited extent. The poor constraint on terrestrial fluxes propagates to the oceans and strongly limits estimates of oceanic fluxes as well, at least if no use is made of other information such as isotopic ratios. We analyze here the resolving power of the GLOBALVIEW-CO2 network, compare the efficiency of different measurement strategies, and determine optimal extensions to the present network. We find the following: (1) GLOBALVIEW-CO2 is well suited to characterize the meridional distribution of sources and sinks but is poorly suited to separate terrestrial from oceanic sinks at the same latitude. The most poorly constrained regions are South America, Africa, and southern hemispheric oceans. (2) To improve the network, observing stations need to be positioned on the continents near to the largest biospheric signals despite the large diurnal and seasonal fluctuations associated with biological activity and the dynamics of the PBL. The mixing in the atmosphere is too strong to allow positioning of stations remote from large fluxes. Our optimization results prove to be fairly insensitive to the details of model transport and the inversion model with the addition of 10 optimally positioned stations. (3) The best measurement strategy among surface observations, N-S airplane transects, and vertical profiles proves to be vertical profiles. (4) Approximately 12 optimally positioned vertical profiles or 30 surface stations in addition to GLOBALVIEW-CO2 would reduce estimate uncertainties caused by insufficient data coverage from 1 Pg C yr^-1 per region to 0.2 Pg C yr^-1 per region.