Present Status of Global CModels

Recently, Rayner et al (1999) developed a 3D time-dependent inversion model to determine interannual variability in the regional terrestrial and oceanic uptake of fossil-fuel C02 over the last two decades. The Rayner et al. study is used here as a benchmark against which the potential for improved precision and spatial resolution of flux estimates from atmospheric composition measurements is explored.

In the Rayner study, extended records of monthly average concentrations of C02 in background air measured at 12 or 25 selected sites in the NOAA/CMDL (Climate Monitoring and Diagnostics Laboratory) global flask-sampling network were employed. To determine partitioning of carbon between oceanic and terrestrial reservoirs, l3C/l2C in CO, from one site and a new determination at the same site (Cape Grim) of the trend in 02/N2 over two decades (Langenfelds et al, 1999a, b) were used. The small number of selected sites for CO,, can be compared to the current number of global sampling sites that approaches 100, many with records that are decadal or longer. For 13C/12C, at least two global sampling networks have made measurements from several sites since the early 1980s (see Francey et al, 1995 and Keeling et al, 1995), yet only one site-record was used. In the case of 02/N2 there is no other reliable information available over this giobai mocmamncAi. cyci rs in ti tr. cum atf systtm

Text Copyright v 2001 by Academic Press. Ail rights ot' reproduction in .my form reserved. Figures 2-3 Copyright • ' CSIRO and K. !. Francey, 1'. I. Rayner. and C. E. Allison time frame. The limited site selections for the Rayner et al. study reflect a very real concern about the quality and intercalibration of records from different measurement laboratories.

Figure 1 is adopted from Rayner et al. and illustrates the uncertainty ascribed to prior flux estimates (Fig. la) and the modified uncertainties resulting from the inversion of the atmospheric measurements (Fig. lb). The grid-scale of the inversion model has dimensions of 8° latitude X 10° longitude. The numbers refer to flux uncertainties (in Gt C year-1) representing over 25 larger aggregated areas selected as characteristic source regions for the prior source estimates. A reduction in uncertainty in a region from Fig. la to lb indicates that effective constraints are imposed by the atmospheric measurements, and it is no coincidence that the larger improvements occur in regions best represented by atmospheric sampling sites (for example, North America compared to South America). There are still regions of the globe where uncertainties are relatively large (~ ± 1 Gt C year-1, compared to global fossil-fuel emissions of around 6 Gt C year-1). Even where uncertainties appear to be relatively small, e.g., North America at

Prior uncertainty [GtC yr1

Prior uncertainty [GtC yr1

Predicted uncertainty [GtC yr
FIGURE 1 Prior and predicted estimates of uncertainty in air-surface fluxes of CO, as the result of a 3D Bayesian synthesis inversion of atmospheric CO,, <5I!C, and 02/N, data from selected sites for the period 1980-1995 (Rayner et al., 1999).

±0.5 Gt C year-1, this should be viewed against the net derived sink in this study of 0.3 Gt C year-1 and the total fossil source of ~1.6 Gt C year-1. Uncertainties are often reduced when regions are aggregated, but even regions in Fig. 1 are too large for many policy needs.

The potential advantages of the atmospheric inversion approach compared to more conventional on-the-surface carbon accounting methods are that, first, flux estimates are firmly bounded by the global growth rate of atmospheric C02, perhaps the best determined of all inputs to a global carbon budget. Second, if the regional uncertainties can be reduced to levels small enough to detect important changes in net continental emissions and uptakes, then atmospheric monitoring provides an opportunity for continuous, relatively low-cost, globally consistent monitoring. The Kyoto Protocol, 1997 and more recently, COP4 of the UN Framework Convention on Climate Change, Buenos Aires, November 1998, present a new and urgent challenge to the atmospheric science community to provide and monitor regional carbon fluxes for verification and/or regulatory purposes.

Of the three broad inputs to the Bayesian synthesis inversion, namely, atmospheric transport models, surface flux constraints, and atmospheric measurements, all have experienced rapid progress over the last few years. In the atmospheric transport area, the problem of estimating GCM model error is significant. However, progress has been made (a) with increasing availability of analyzed wind fields (Trenberth, 1992; e.g., permitting an examination of the impact of interannual variation in transport on measured parameters) and (b) with identification of model differences in the on-going series of TRANSCOM model comparisons (e.g., Law et al., 1996; Denning et al., 1999). Considerable research effort is now focussed on "bridging the scale gap" between the typical grid cells of the transport models and the volume of atmosphere represented by the atmospheric measurement at surface sites. This research introduces boundary-layer and regional-transport models, direct flux measurement campaigns, and vertical profiling of C02 and related trace species. Significant uncertainties are perceived to remain, for example, in the representation of mass transport in the tropical areas.

A large volume of new information is also emerging on the interaction between terrestrial ecosystems and the atmosphere with process-oriented campaigns focused on major ecosystems such as the Amazon and Siberia. Recent perspectives and a summary of the advances in knowledge of the understanding of the role of the terrestrial biosphere in the global carbon cycle is provided by Schimel (1995) and Lloyd (1999). The situation is similar for the interaction with the world's oceans. Extensive on-going surveys of ocean parameters are elucidating air-sea gas exchange constraints on carbon uptake by the oceans (e.g., Takahashi et al., 1997; Heimann and Maier-Reimer, 1996), while similar constraints are emerging from the development of ocean general circulation models, e.g., Orr (1999) and Sarmiento et al. (1998). The formal integration of the information on terrestrial and oceanic fluxes as additional constraints in the Bayesian inversion framework is in its infancy. Even with these various streams of information, the carbon cycle remains an underdetermined system that requires more and better-calibrated measurements.

0 0

Post a comment