Inversion of Atmospheric CO 2 Concentrations
307
A set of comparisons by Law and Rayner (1999) suggested that the modeling of
the rectifi er could have a signifi cant effect on estimates of land–ocean partitioning of
fl uxes within zones, but a lesser effect on larger (semihemispheric) scales. They also
suggested that estimates of interannual variability were more robust (with respect to
rectifi er uncertainty) than estimates of long-term means. The TransCom control partitions the northern sink as −2.3 ± 0.6 GtC y −1 to land and −1.1 ± 0.5 GtC y −1 to oceans.
SH sink: A relatively low southern hemisphere uptake has been a feature of most
inversion studies. One area of difference is whether there is a net source at latitudes
south of about 50°S. This has been further studied by Roy et al. (2003) based on additional in situ ocean measurements. One of the diffi culties in extracting an overview
is the small latitude range involved, given the differences in spatial discretization of
various modeling groups. One reason for believing the sink to be stronger than previous estimates is that, with only spatially-sparse sampling, Cape Grim, Tasmania
had the lowest annual mean CO 2 concentration. More recently, Easter Island has
been found to have lower concentrations. The TransCom estimate for the southern
hemisphere is −0.8 ± 0.7 GtC y −1 to oceans (and −0.2 ± 1.1 GtC y −1 to land). This is
an area of ongoing scientifi c debate.
11.5 EMERGING TRENDS
In looking to the future, one can foresee changes with analysis of smaller scales, the
incorporation of new data streams, the development of new inversion techniques,
and changing objectives. Some emerging trends are
Smaller spatial scales: The majority of the inversions described in this chapter
apply on the global scale, aiming to resolve fl uxes representative of large regions. In
principle, similar approaches can be applied on smaller scales although details of the
computational techniques change (Enting, 2002, Chapter 17). In analyzing smaller
spatial scales, much of the relevant information is in the concentration variations on
small timescales, the so-called continuous data (to be contrasted with discrete fl ask
sampling and intervals of about a week). A number of “experimental design” studies
(Law et al., 2002, 2003b, 2004) have analyzed the requirements and potential gains
in using high-resolution CO 2 data in global (and regional) inversions. At the time of
writing, this work is being extended as a TransCom project.
Additional/enhanced data streams: There is increasing recognition that improved
understanding of the carbon cycle will involve the interpretation of new types of data.
This expanded framework is sometimes referred to as multiple constraints (Kruijt
et al., 2001; Wang and Barrett, 2003). Of course, the basic approach of the Bayesian
synthesis inversion is built on the use of data other than atmospheric concentrations
as providing the constraints needed to regularize the ill-conditioned inverse problem. The range of carbon cycle data has been reviewed by Canadell et al. (2000)
and the statistical characteristics of these various data sets discussed (qualitatively)
by Raupach et al. (2005). Of the emerging data streams, satellite data will provide
formidable computational challenges. The launch failure of the Orbiting Carbon
Observatory (OCO) represents a major setback. In the process of developing applications involving these new data streams, the use of inversions studies as a tool for
experimental design can be expected to continue.
© 2010 by Taylor and Francis Group, LLC
307
A set of comparisons by Law and Rayner (1999) suggested that the modeling of
the rectifi er could have a signifi cant effect on estimates of land–ocean partitioning of
fl uxes within zones, but a lesser effect on larger (semihemispheric) scales. They also
suggested that estimates of interannual variability were more robust (with respect to
rectifi er uncertainty) than estimates of long-term means. The TransCom control partitions the northern sink as −2.3 ± 0.6 GtC y −1 to land and −1.1 ± 0.5 GtC y −1 to oceans.
SH sink: A relatively low southern hemisphere uptake has been a feature of most
inversion studies. One area of difference is whether there is a net source at latitudes
south of about 50°S. This has been further studied by Roy et al. (2003) based on additional in situ ocean measurements. One of the diffi culties in extracting an overview
is the small latitude range involved, given the differences in spatial discretization of
various modeling groups. One reason for believing the sink to be stronger than previous estimates is that, with only spatially-sparse sampling, Cape Grim, Tasmania
had the lowest annual mean CO 2 concentration. More recently, Easter Island has
been found to have lower concentrations. The TransCom estimate for the southern
hemisphere is −0.8 ± 0.7 GtC y −1 to oceans (and −0.2 ± 1.1 GtC y −1 to land). This is
an area of ongoing scientifi c debate.
11.5 EMERGING TRENDS
In looking to the future, one can foresee changes with analysis of smaller scales, the
incorporation of new data streams, the development of new inversion techniques,
and changing objectives. Some emerging trends are
Smaller spatial scales: The majority of the inversions described in this chapter
apply on the global scale, aiming to resolve fl uxes representative of large regions. In
principle, similar approaches can be applied on smaller scales although details of the
computational techniques change (Enting, 2002, Chapter 17). In analyzing smaller
spatial scales, much of the relevant information is in the concentration variations on
small timescales, the so-called continuous data (to be contrasted with discrete fl ask
sampling and intervals of about a week). A number of “experimental design” studies
(Law et al., 2002, 2003b, 2004) have analyzed the requirements and potential gains
in using high-resolution CO 2 data in global (and regional) inversions. At the time of
writing, this work is being extended as a TransCom project.
Additional/enhanced data streams: There is increasing recognition that improved
understanding of the carbon cycle will involve the interpretation of new types of data.
This expanded framework is sometimes referred to as multiple constraints (Kruijt
et al., 2001; Wang and Barrett, 2003). Of course, the basic approach of the Bayesian
synthesis inversion is built on the use of data other than atmospheric concentrations
as providing the constraints needed to regularize the ill-conditioned inverse problem. The range of carbon cycle data has been reviewed by Canadell et al. (2000)
and the statistical characteristics of these various data sets discussed (qualitatively)
by Raupach et al. (2005). Of the emerging data streams, satellite data will provide
formidable computational challenges. The launch failure of the Orbiting Carbon
Observatory (OCO) represents a major setback. In the process of developing applications involving these new data streams, the use of inversions studies as a tool for
experimental design can be expected to continue.
© 2010 by Taylor and Francis Group, LLC