306
Air Pollution and Turbulence: Modeling and Applications
inversions (Tans et al., 1989; Enting and Mansbridge, 1991) found a small low latitude (12°S–12°N) source, comparable to (or slightly less than) the expected oceanic
source, implying a near-zero net fl ux from terrestrial biota. The TransCom estimate
of 1.1 ± 1.3 GtC y −1 suggest less of an apparent discrepancy, but insuffi cient data to
resolve the fl uxes.
Interannual variability: Two studies (Francey et al., 1995; Keeling et al., 1995)
used joint globally aggregated CO 2 / 13 CO 2 budgets to partition the interannual variability between atmosphere and oceans. Bousquet et al. (2000) described a timedependent synthesis of the period 1980–1998. They found that in the 1980s, most
interannual variability came from tropical land regions while in the 1990s northern
land ecosystems contributed most variability. In particular, they found a large but
short-lived carbon uptake over North America over the period 1992–1993, associated with colder than average temperatures. Various studies, including Law and
Rayner (1999) described below, and most recently from TransCom (Baker et al.,
2006; Gurney et al., 2008), have noted that estimates of interannual variability are
more robust than estimates of annual means.
Rayner et al. (1999b) compared CO 2 inversion results to the Southern Oscillation
Index (SOI), as a characterization of the ENSO phenomenon. The SOI-CO 2 connection has been known since the 1976 study by Bacastow (1976) and has been studied
by many other workers (Thompson et al., 1986; Elliott and Angell, 1987; Elliott
et al., 1991). In order to help ensure robustness of the results, Rayner et al. used three
different inversion techniques: data assimilation (Dargaville and Simmonds, 2000),
mass balance, and synthesis. They found an apparent close correlation when the CO 2
fl ux led the SOI by some months, in spite of a lack of any physical mechanism. Since
the ESNO is quasi-periodic, with preferential phasing relative to the annual cycle,
such correlation techniques, designed for stationary time series, will not necessarily be appropriate. A more detailed analysis of the results showed an early-ENSO
CO 2 decline closely synchronized with the SOI (and apparently affecting the tropical
oceans), while the CO 2 increase in the later ENSO phase seemed to lag the SOI and
(in two of the three analyses) be associated with land biota.
NH sink: As noted above, an early inversion (Keeling et al., 1989) indicated a
large northern ocean sink. In part, this was forced by the application of the constraint
of a large total ocean sink—a biased constraint due to the failure to distinguish
between fl ux budgets and storage budgets (as described in Section 11.4.2). Ciais et al.
(1995b) identifi ed the northern sink as terrestrial on the basis of 13 C data. Bousquet
et al. (1999) found, for the 1985–1995 average, northern sinks (in GtC y −1 ) of 0.7 ±
0.7 for North America, 0.2 ± 0.3 for North Pacifi c, 0.5 ± 0.8 for Europe, 0.7 ± 0.3 for
North Atlantic, and 1.2 ± 0.8 for North Asia. Note that most of these estimates seem
to have been produced by subtracting fossil carbon sources, rather than fossil CO 2
sources and so are likely to estimate too large a sink, especially in North America.
Rectifi er: It has long been appreciated that covariance between seasonal variations
in fl uxes and transport would create mean spatial gradients. In particular, it is expected
that there was a mean interhemispheric CO 2 gradient in preindustrial times. The effect
was apparent in Phase 1 of TransCom with a “split” into two groups, depending on how
models treated the boundary layer. The issue was emphasized by Denning et al. (1995)
who also noted the effect of covariance and concentration over the diurnal cycle.
© 2010 by Taylor and Francis Group, LLC
Air Pollution and Turbulence: Modeling and Applications
inversions (Tans et al., 1989; Enting and Mansbridge, 1991) found a small low latitude (12°S–12°N) source, comparable to (or slightly less than) the expected oceanic
source, implying a near-zero net fl ux from terrestrial biota. The TransCom estimate
of 1.1 ± 1.3 GtC y −1 suggest less of an apparent discrepancy, but insuffi cient data to
resolve the fl uxes.
Interannual variability: Two studies (Francey et al., 1995; Keeling et al., 1995)
used joint globally aggregated CO 2 / 13 CO 2 budgets to partition the interannual variability between atmosphere and oceans. Bousquet et al. (2000) described a timedependent synthesis of the period 1980–1998. They found that in the 1980s, most
interannual variability came from tropical land regions while in the 1990s northern
land ecosystems contributed most variability. In particular, they found a large but
short-lived carbon uptake over North America over the period 1992–1993, associated with colder than average temperatures. Various studies, including Law and
Rayner (1999) described below, and most recently from TransCom (Baker et al.,
2006; Gurney et al., 2008), have noted that estimates of interannual variability are
more robust than estimates of annual means.
Rayner et al. (1999b) compared CO 2 inversion results to the Southern Oscillation
Index (SOI), as a characterization of the ENSO phenomenon. The SOI-CO 2 connection has been known since the 1976 study by Bacastow (1976) and has been studied
by many other workers (Thompson et al., 1986; Elliott and Angell, 1987; Elliott
et al., 1991). In order to help ensure robustness of the results, Rayner et al. used three
different inversion techniques: data assimilation (Dargaville and Simmonds, 2000),
mass balance, and synthesis. They found an apparent close correlation when the CO 2
fl ux led the SOI by some months, in spite of a lack of any physical mechanism. Since
the ESNO is quasi-periodic, with preferential phasing relative to the annual cycle,
such correlation techniques, designed for stationary time series, will not necessarily be appropriate. A more detailed analysis of the results showed an early-ENSO
CO 2 decline closely synchronized with the SOI (and apparently affecting the tropical
oceans), while the CO 2 increase in the later ENSO phase seemed to lag the SOI and
(in two of the three analyses) be associated with land biota.
NH sink: As noted above, an early inversion (Keeling et al., 1989) indicated a
large northern ocean sink. In part, this was forced by the application of the constraint
of a large total ocean sink—a biased constraint due to the failure to distinguish
between fl ux budgets and storage budgets (as described in Section 11.4.2). Ciais et al.
(1995b) identifi ed the northern sink as terrestrial on the basis of 13 C data. Bousquet
et al. (1999) found, for the 1985–1995 average, northern sinks (in GtC y −1 ) of 0.7 ±
0.7 for North America, 0.2 ± 0.3 for North Pacifi c, 0.5 ± 0.8 for Europe, 0.7 ± 0.3 for
North Atlantic, and 1.2 ± 0.8 for North Asia. Note that most of these estimates seem
to have been produced by subtracting fossil carbon sources, rather than fossil CO 2
sources and so are likely to estimate too large a sink, especially in North America.
Rectifi er: It has long been appreciated that covariance between seasonal variations
in fl uxes and transport would create mean spatial gradients. In particular, it is expected
that there was a mean interhemispheric CO 2 gradient in preindustrial times. The effect
was apparent in Phase 1 of TransCom with a “split” into two groups, depending on how
models treated the boundary layer. The issue was emphasized by Denning et al. (1995)
who also noted the effect of covariance and concentration over the diurnal cycle.
© 2010 by Taylor and Francis Group, LLC