1. Using High-Resolution Forward Model Simulations of Ideal Atmospheric Tracers to Assess the Spatial Information Content of Inverse CO 2 Flux Estimates Steven Pawson and J. Eric Nielsen* Global Modeling and Assimilation Office NASA Goddard Space Flight Center * Science Systems and Applications Inc. Funding: NASA’s Modeling, Analysis and Prediction (MAP) Program Computing: NASA’s High-End Computing resources at NASA GSFC/NCCS AGU 2010 Fall Meeting – San Francisco, CA – Session A54D – December 17, 2010

2. Relating CO 2 Concentrations to Fluxes Surface fluxes (Emissions and Uptake) Atmospheric Observations (Concentrations) Transport Model Global surface fluxes: Inventories (FF, BF) “Constrained” model estimates (biological, biomass burning, ocean uptake) Global Atmospheric Observations: Sparse surface networks More dense UT estimates from satellites Emerging total column from satellites (land bias) This talk is about using a forward transport model to infer information about inverse approaches (top-down flux estimates): specifically, what spatial scales of sources can be resolved

3. Methodology Grid of Idealized Tracers Idealized tracers (constant source strength) are continuously emitted in the shaded grid boxes

4. Methodology Simple Tracers in a Complex Model ComponentConfiguration GEOS-5 AGCM0.5°×0.625°L72 with “full” physics Idealized Tracers48-member grid over Eastern North America Tracer GridEvery 2.5° of longitude & every 2° of latitude EmissionsContinuous emissions (1kgm -2 s -1 ) for 10 days InitializationConcentrations zero at start of each 10-day period Analysis focuses on surface concentrations and layer averages for total column (TC), mid-troposphere (MT: 500-300hPa) and upper troposphere (UT: 300-100hPa)

5. Growth of Plumes with Time Surface Concentration (sum of nine boxes) and mean sea-level pressure in Trial 9 (color scale is logarithmic) Day 2Day 8

6. Growth of Plumes over 10-Day Trial Trial 9: 19-28 February Area with column-integrated mass larger than 10 kgm -2 All 48 tracers shown Color reflects origin – not important here This trial is fairly typical of all 36 cases (summer and winter)

7. Area Overlap Ratio: Definition 0 < R mn = A m&n / A m < 1 The fraction of observations of Tracer ‘m’ that are contaminated by Tracer ‘n’ AmAm AnAn A m&n R mn = 1 An observation contains no unique information about source regions m and n – inverse modeling can infer only the total source of m+n R mn = 0 Observations of Tracer ‘m’ are uncontaminated by Tracer ‘n’

8. Adjacent Sources: Large Overlap More distant sources: small overlap Area Overlap Ratio (R m,n ): Trial 9, Day 1

9. Area Overlap Ratio (R m,n ): Trial 9, Day 8

10. R 20,n : Tracer 20, Trial 9 – after 2 Days

11. R 20,n : Tracer 20, Trial 9 – after 8 Days

12. R 20,n : Tracer 20, Trial 25 – after 8 Days

13. Mid-Troposphere: Four Trials, Day 8 Trial 8 Winter Trial 9 Winter Trial 24 Summer Trial 25 Summer

14. Summary and Further Work At 0.5° model resolution, downstream total-column observations seem unable to isolate sources with less than 10° of separation Flow dependence is large – intra-seasonal variations are as big as inter-seasonal differences, despite contrasting transport mechanisms Need more Trials to build full statistics (e.g., co-variances) of these simple cases – also a bigger grid would be helpful! Planning experimentation with tagged CO 2 sources and pseudo- observations to complement these idealized cases (OSSEs) Intended to enhance understanding of “top-down” flux estimation – “inverse” techniques likely increase smearing of signals Transport uncertainty (advection core and sub-grid) is another aspect that will be investigated