1. Evaluation of land model simulations across multiple sites and multiple models: Results from the NACP site-level synthesis effort Peter Thornton 1, Gautam Bisht 1, Dan Ricciuto 1, NACP Site-Level Synthesis Participants 1 Oak Ridge National Laboratory, Environmental Sciences Division and ORNL Climate Change Science Institute

2. Sponsors NASA Terrestrial Ecology Program DOE, Office of Biological and Environmental Research, Climate and Environmental Sciences Division, Terrestrial Ecosystem Science Program

3. Premise Models can and should serve as tools for the integration and synthesis of our best understanding and knowledge Models can and should provide testable (falsifiable) hypotheses Through model-data synthesis efforts, those hypotheses can and should be tested, and discarded or improved when confidence is shown to be low

4. Analysis setting Subset of sites and models from full NACP site-level synthesis effort Forest sites (evergreen and deciduous) Range of climates Models that include diurnal cycle Carbon, sensible heat, latent heat fluxes Diurnal cycle, seasonal cycle, interannual variability, long-term mean Influence of steady-state vs. transient forcings

5. 12 Models and 13 Sites CAN-IBIS CNCLASS CLM-CN ECOSYS ED2 ISOLSM LOTEC ORCHIDEE SIB SIBCASA SSIB2 TECO CA-Ca1 Campbell River CA-Oas Old aspen CA-Obs Old black spruce CA-Ojp Old jack pine CA-Qfo Mature black spruce CA-TP4 Turkey Point US-Dk3 Duke Forest pine US-Ha1 Harvard Forest main US-Ho1 Howland main US-Me2 Metolius intermediate US-MOz Missouri Ozark US-NR1 Niwot Ridge US-UMB U Michigan Bio Stn

6. Diurnal cycle of GPP: US-Dk3 Mean diurnal cycle for June-July-August, y-axis units = umol/m2/s, x-axis is half- hour time step. Results from steady-state simulations

7. Diurnal cycle of GPP: CA-Obs

8. Diurnal cycle of GPP: US-UMB

9. Diurnal cycle of NEE: CA-Oas

10. Diurnal cycle of NEE: US-Ha1

11. Diurnal cycle of NEE: US-Dk3

12. Diurnal cycle of NEE: CLM-CN

13. Seasonal cycle of CLM-CN: US-Ha1

14. Findings: 1 Time-scale of N-limitation mechanism in CLM-CN is wrong. –Evident at both diurnal and seasonal –Original hypothesis that plants respond to N availability on sub-daily time scale should be rejected –Introducing new mechanism to buffer N availability in time

15. Findings: 2 Evaluation of LE suggests that current basis for estimation of stomatal conductance in CLM-CN is reasonable –This result should be revisited once new N storage mechanism is added

16. Findings: 3 CLM-CN is very sensitive to fine root : leaf allocation patterns –Difficult measurement –Likely candidate parameter for data assimilation –Evidence emerging from global-scale studies and comparison to root turnover data that model fine root longevity needs to be modified Other models sensitive to this as well?

17. Findings: 4 (underway) Introducing transient forcing (disturbance, rising atmospheric CO 2, changing N deposition) seems to improve estimate of decadal-scale NEE –Doesn’t seem to change conclusions obtained from steady-state simulations –This is the most critical flux for evaluation of long-term climate-carbon cycle feedbacks

18. Conclusions Approach has proved very useful in identifying strengths and weaknesses in CLM-CN This kind of critical evaluation across multiple models provides a path forward for improved future model generations Improving modelers’ ability to know what to ask for from observationalists and experimentalists.