Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

<jats:p>This is the second part of a two-part paper considering network design based on a Lagrangian stochastic particle dispersion model (LPDM), aimed at reducing the uncertainty of the flux estimates achievable for the region of interest by the continuous observation of atmospheric CO&lt;sub&gt;2&lt;/sub&gt; concentrations at fixed monitoring stations. The LPDM model, which can be used to derive the sensitivity matrix used in an inversion, was run for each potential site for the months of July (representative of the Southern Hemisphere Winter) and January (Summer). The magnitude of the boundary contributions to each potential observation site was tested to determine its inclusion in the network design, but found to be minimal. Through the use of the Bayesian inverse modelling technique, the sensitivity matrix, together with the prior estimates for the covariance matrices of the observations and surface fluxes were used to calculate the posterior covariance matrix of the estimated fluxes, used for the calculation of the cost function of the optimisation procedure. The optimisation procedure was carried out for South Africa under a standard set of conditions, similar to those applied to the Australian test case in Part 1, for both months and for the combined two months. The conditions were subtly changed, one at a time, and the optimisation routine re-run under each set of modified conditions, and compared to the original optimal network design. The results showed that changing the height of the surface grid cells, including an uncertainty estimate for the oceans, or increasing the night time observational uncertainty did not result in any major changes in the positioning of the stations relative to the basic design, but changing the covariance matrix or increasing the spatial resolution did. The genetic algorithm was able to find a slightly better solution than the incremental optimisation procedure, but did not drastically alter the solution compared to the standard case. Including correlation appeared to increase the spread in the layout of the stations. Increasing the surface resolution tended to clump the stations around areas of high activity. In conclusion, the specification used in an optimal network design should be chosen to best match the conditions under which an inversion would be run for the region of interest. Increasing the spatial resolution beyond that which the given network size could reasonably resolve may lead to a network which would ignore small areas of high activity and reduce the capacity of the network to resolve fluxes for subregions in the domain of interest. Overall the results suggest that a good improvement in knowledge of South African fluxes is available from a feasible atmospheric network and that the general features of this network are robust to many reasonable choices in a network design study.</jats:p>

Original publication

DOI

10.5194/acpd-14-11301-2014

Type

Journal article

Journal

Atmospheric Chemistry and Physics Discussions

Publisher

Copernicus GmbH

Publication Date

2014

Volume

14

Pages

11301 - 11342