Inversion of Atmospheric CO 2 Concentrations
303
The effects of neglecting CO and its precursors depend on what is required:
1. Neglect of a free atmosphere source in the set of basis functions means that
other sources will correspondingly increase. Other things being equal, this
error will have a fairly uniform distribution, somewhat peaked toward tropics. However, a Bayesian inversion will preferentially distribute the error
into the regions with the loosest priors. A correction for this term will lead
to a surface CO 2 inversion that is formally correct.
2. More seriously, many CO 2 inversions are presented in terms of nonfossil
fl uxes by subtracting the full fossil carbon emissions, rather than fossil CO 2
emissions. To correct such a budget to produce nonfossil CO 2 emissions
requires adding back on the non-CO 2 fossil emissions. This produces a
surface CO 2 fl ux budget for nonfossil CO 2 fl uxes.
3. If a surface carbon fl ux budget is required, then one needs to add the
surface biotic emissions of non-CO 2 carbon (excluding components that are
rapidly oxidized to products that return to the ecosystem via rain-out).
4. A carbon storage budget would require an additional correction for carbon cycled from biosphere to oceans via rivers before returning to the
atmosphere.
There are three ways of evaluating these corrections:
From the 2-D analysis by Enting and Mansbridge (1991)
•
From the 3-D study by Suntharalingam et al. (2005)
•
By running synthesis inversions with and without proper treatment of the
•
various effects
Note that all but item (1) are, in principle, off-line corrections, and really constitute
a change in what is being estimated from the inversion. The point where failure to
be consistent in the treatment of (2), (3), and (4) will affect the inversion calculation
(as opposed to its interpretation) is if priors (especially tight priors) are based on the
wrong sort of budget.
As noted above, the analysis by Sarmiento and Sundquist (1992) showed that
a number of the discrepancies in published results arose from the use of different
forms of atmospheric budget. They also noted the effect on inversion estimates of a
skin effect (Robertson and Watson, 1992) that implied a correction to fl ux estimates
derived from ocean carbon measurements.
Around 1990, there were confl icting inversion estimates (Keeling et al., 1989;
Tans et al., 1990) about the size of the ocean sink, with a relatively sharp distinction
between estimates of 1 GtC y −1 or less and those that were nearer to 2 GtC y −1 . Each
of these inversions was based on ad hoc fi ts, chosen to be consistent with various
other data—they were undertaken prior to the introduction of systematic Bayesian
statistical inversions into CO 2 studies.
The inversions by Keeling et al. (1989) indicated a large northern ocean sink. In
part, this was forced by the application of a constraint of a large total ocean sink, a
result derived from double-deconvolution of the long-term CO 2 and 13 CO 2 records.
© 2010 by Taylor and Francis Group, LLC
303
The effects of neglecting CO and its precursors depend on what is required:
1. Neglect of a free atmosphere source in the set of basis functions means that
other sources will correspondingly increase. Other things being equal, this
error will have a fairly uniform distribution, somewhat peaked toward tropics. However, a Bayesian inversion will preferentially distribute the error
into the regions with the loosest priors. A correction for this term will lead
to a surface CO 2 inversion that is formally correct.
2. More seriously, many CO 2 inversions are presented in terms of nonfossil
fl uxes by subtracting the full fossil carbon emissions, rather than fossil CO 2
emissions. To correct such a budget to produce nonfossil CO 2 emissions
requires adding back on the non-CO 2 fossil emissions. This produces a
surface CO 2 fl ux budget for nonfossil CO 2 fl uxes.
3. If a surface carbon fl ux budget is required, then one needs to add the
surface biotic emissions of non-CO 2 carbon (excluding components that are
rapidly oxidized to products that return to the ecosystem via rain-out).
4. A carbon storage budget would require an additional correction for carbon cycled from biosphere to oceans via rivers before returning to the
atmosphere.
There are three ways of evaluating these corrections:
From the 2-D analysis by Enting and Mansbridge (1991)
•
From the 3-D study by Suntharalingam et al. (2005)
•
By running synthesis inversions with and without proper treatment of the
•
various effects
Note that all but item (1) are, in principle, off-line corrections, and really constitute
a change in what is being estimated from the inversion. The point where failure to
be consistent in the treatment of (2), (3), and (4) will affect the inversion calculation
(as opposed to its interpretation) is if priors (especially tight priors) are based on the
wrong sort of budget.
As noted above, the analysis by Sarmiento and Sundquist (1992) showed that
a number of the discrepancies in published results arose from the use of different
forms of atmospheric budget. They also noted the effect on inversion estimates of a
skin effect (Robertson and Watson, 1992) that implied a correction to fl ux estimates
derived from ocean carbon measurements.
Around 1990, there were confl icting inversion estimates (Keeling et al., 1989;
Tans et al., 1990) about the size of the ocean sink, with a relatively sharp distinction
between estimates of 1 GtC y −1 or less and those that were nearer to 2 GtC y −1 . Each
of these inversions was based on ad hoc fi ts, chosen to be consistent with various
other data—they were undertaken prior to the introduction of systematic Bayesian
statistical inversions into CO 2 studies.
The inversions by Keeling et al. (1989) indicated a large northern ocean sink. In
part, this was forced by the application of a constraint of a large total ocean sink, a
result derived from double-deconvolution of the long-term CO 2 and 13 CO 2 records.
© 2010 by Taylor and Francis Group, LLC