1. Estimating regional sources and sinks of CO 2 for North America Using NOAA-CMDL measurements and the TM5 model Wouter Peters, NOAA CMDL TRANSCOM May 13th, 2003

2. Acknowledgements Maarten Krol Pieter Tans Lori Bruhwiler Sander Houweling Peter Bergamaschi Peter van Velthoven Frank Dentener Jan Fokke Meirink John Miller

3. Outline My background My background The TM5 global & regional model The TM5 global & regional model The expanded NOAA-CMDL network The expanded NOAA-CMDL network Plans for inversions Plans for inversions Kalman filter Kalman filter 4d-var 4d-var

4. My background Master’s in meteorology and physical oceanography from University of Utrecht Master’s in meteorology and physical oceanography from University of Utrecht Thesis work on LBA-CLAIRE, airborne campaign in Suriname, South America. Thesis work on LBA-CLAIRE, airborne campaign in Suriname, South America. PhD on tropospheric ozone in the tropics with Jos Lelieveld and Maarten Krol. PhD on tropospheric ozone in the tropics with Jos Lelieveld and Maarten Krol.

5. TM3 global model 5x3.75 degrees, 19 layers 5x3.75 degrees, 19 layers ECMWF meteorology (1979-1993, and 1994- 1999) ECMWF meteorology (1979-1993, and 1994- 1999) Full chemistry (O3-NOx-CO-CH4, lumped NMHC’s) + coupled photolysis Full chemistry (O3-NOx-CO-CH4, lumped NMHC’s) + coupled photolysis Extensive budgets of all processes Extensive budgets of all processes Fast parallel version (<2 CPU hrs/yr) allowing multi-year simulations or sensitivity tests Fast parallel version (<2 CPU hrs/yr) allowing multi-year simulations or sensitivity tests

6. TM5: the next generation New advection algorithm (Berkvens, Botchev, Krol, Verwer, 2001) allows online nesting of fine-scaled grids within a global domain. New advection algorithm (Berkvens, Botchev, Krol, Verwer, 2001) allows online nesting of fine-scaled grids within a global domain.

7. TM5 model – coarse grid

8. TM5 model – medium grid

9. TM5 model – fine grid

11. TM5 model – vertical resolution values for US standard atmosphere ~ 5 layers PBL ~ 12 layers free troposphere ~ 8 layers stratosphere 25 vertical layers Hybrid coordinates

12. TM5 Contributions from IMAU, KNMI, JRC, and in the near future NOAA. Contributions from IMAU, KNMI, JRC, and in the near future NOAA. Most important improvements: meteorological input, coding, chemical boundary conditions. Most important improvements: meteorological input, coding, chemical boundary conditions. Special emphasis: ERA40=> re-analysis of ECMWF for 1957-2001. Special emphasis: ERA40=> re-analysis of ECMWF for 1957-2001. Resolution up to ECMWF operational Resolution up to ECMWF operational

13. TM5 – latest developments- Adjoint version of TM5 under con struction Adjoint version of TM5 under con struction 4d-var scheme from ECMWF implemented and working 4d-var scheme from ECMWF implemented and working Kalman filter (Baker, Bruhwiler, Tans) to be implemented Kalman filter (Baker, Bruhwiler, Tans) to be implemented

14. 222 Rn simulations at different resolutions Courtesy: H. Sartorius IAR-Freiburg

15. My plans Forward modeling of CO 2 + isotopes, SF 6 Forward modeling of CO 2 + isotopes, SF 6 Inverse modeling of CO 2 + isotopes Inverse modeling of CO 2 + isotopes Regional scale sources/sinks over the US Regional scale sources/sinks over the US Using vertical profile/ tall tower data Using vertical profile/ tall tower data Possibly assist in planning and optimizing network design/sampling strategies Possibly assist in planning and optimizing network design/sampling strategies

16. NOAA CMDL goals ??????

17. Current network

18. 2005 network ?

19. 2010 network ??

20. Proposed TM5 grid

21. Aspen Denver Pueblo Durango Grand Junction

22. Regions for inversion

23. Fixed-lag Kalman smoother  ‘Batch’ inversion technique using n-months of information at a time (3 < n < 9)  Calculates new fluxes and error estimates for  x J=0 (where x= fluxes and J = cost function)  Balances prior flux estimates against new information from observations using prior flux and model-data mismatch  Computationally efficient for multi-decadal inversions  Details: David Baker’s talk tomorrow

24. 4d-var o Uses steepest descend and the calculated value of  x J to minimize J o value of  x J calculated by adjoint model o Iterative scheme to converge to  x J = 0 o Useful for short time periods and many observations

25. Seasonal CO 2 Flux Estimates Gt/yr Ocean Land NCEP (1984-2000) ECMWF (1981-1993) PRIOR

26. Annual Average Flux Estimates Ocean Land NCEP (1984-2000) ECMWF (1981-1992) PRIOR

27. Annual Average Flux Estimates with Cyclic Meteorology Ocean Land NCEP (1984-2000) NCEP (1990) PRIOR

28. InterfaceZoom Interface: Mixture between ‘fine’ and ‘coarse’: = +

29. InterfaceZoom XYZV Write BC xyz v vzy x Update parent VZYX vzy x xyz v