1. DETECTION OF ANTHROPOGENIC DIC IN THE OCEAN Keith Rodgers (Princeton) Jorge Sarmiento (Princeton) Anand Gnanadesikan (GFDL) Laurent Bopp (LSCE, France) Olivier Aumont (IRD, France) Yasuhiro Yamanaka (U. Hokkaido, Japan) Akio Ishida (FRCGC, Japan) Bob Key (Princeton) Nicolas Metzl (OCEAN, France) Masao Ishii (MRI, Tsukuba, Japan)

2. Repeat Hydrography: task would be far simpler if ocean circulation were steady (not time-varying)!

3. SCIENTIFIC QUESTIONS TO ADDRESS WITH MODELS: (1) What is amplitude of natural variability on DIC in the ocean (on seasonal, interannual, and decadal timescales), and how does this compare to amplitude of anthropogenic signal? (2) eMLR method of Friis et al.[2005] for identification of Anthropogenic DIC from Repeat Hydrography,and subsequent variations of that method: Do they account for dynamical component of signal?

4. Consider linear decomposition for tracer into time- mean and perturbation Can then expand advection/diffusion equation for anomalies of tracer concentration Questions regarding DIC: (1) Which terms are important where? (2) How large is highlighted term near “fronts”?

5. Three Models considered: all participated in European NOCES (Northern Ocean Carbon Exchange Study) Project (1)Japan: FRCGC COCO-NEMURO (~1° res.) (2)Germany: MPI MPIOM1-HAMMOC (~3° res.) (3)France: IPSL ORCA2-PISCES (~2° res.) All models were forced with NCEP reanalysis fluxes over 1948-2003; Two runs conducted with each model: “REF” uses anthropogenic transient in atmospheric CO2 as boundary condition, and “CTL” maintains 280ppm boundary condition for atmospheric CO2 Anthropogenic DIC component: ANT = REF-CTL

8. DIFFERENCE IN INVENTORIES BETWEEN WOCE P17 (May/June 1993) AND CLIVAR P17 (July/August 2001)

11. For continuous monitoring of DIC along a transect, natural variability (ENSO, PDO, NAO, etc.) component can be larger than anthropogenic perturbation component Implications: Need to identify mechanisms responsible for this natural variability in order to include correction term in e-MLR or double-MLR

12. Vertical Inventories of Oceanic DIC (moles/m^2)

13. Mechanisms? Relationship between variability of Natural DIC and other variables (SSH, O2) for ORCA2-PISCES

14. For sea surface height (vertically integrated quantity), widely known that “highlighted” term for Density variations is dominant: High correlation between SSH anomalies and natural DIC inventory anomalies strongly suggests that “highlighted” term is making first order contribution to DIC variability

15. Conclusions (1)Large “natural” variability in DIC introduces significant aliasing problem over much of global ocean for detection of anthropogenic DIC (signal- to-noise problem) [increases with resolution] (2)Natural variability in circulation (i.e. 1997/98 El Nino) can give changes in DIC inventories along section which are much larger than anthropogenic component. (3)Need to use models to test whether skill of eMLR and other related methods can be improved (or errors can be reduced) through use of dynamical information

16. Implications/future work: (1)Why can’t models can’t skillfully simulate changes in DIC at 135°W?: Could be due to limitations of NCEP windstress forcing fields, ocean circulation models, or biogeochemistry models? (2)Test model-based result that DIC/O2 inventories and sea level height (SSH) are correlated (3)Develop correction to eMLR method of Friis et al. [2005] which includes ocean dynamical component to improve detection skill: DIC ssh_eMLR = DIC eMLR + f(TOPEX SSH)

17. Japanese Repeat measurements along 165°E: (in collaboration with Ishii Masao) Modeled DIC on  0 =26.6 along 165°E over 1990-2003 Work in Progress: