330 likes | 466 Views
An evaluation of an ecosystem model for studying CO2 seasonal cycle TransCom-3 (Level-1) related activities at FRSGC. Prabir K. Patra, Shamil Maksyutov, A. Ito and TransCom-3 modellers Jena; 13 May 2003. Goals…. To configure optimal observation system
E N D
An evaluation of an ecosystem model for studying CO2 seasonal cycleTransCom-3 (Level-1) related activities at FRSGC Prabir K. Patra, Shamil Maksyutov, A. Ito and TransCom-3 modellers Jena; 13 May 2003
Goals… To configure optimal observation system • Measurement network optimisation (surface) • Estimate benefits of satellite data in inversion • Evaluate of theirrelative performance
Inverse Modelling Least squares fitting of observed data and model simulations Matrix multiplication and SVD TransCom-3 setup for 11 land and 11 ocean regions HiRes setup for 42 land and 11 ocean regions Forward Modelling 16 global transport models of TransCom-3 Advection, PBL, Convection etc. are treated differently ECMWF, NCEP, GCM meteorological fields Simulation of monthly-mean source/basis functions Tools
Network Optimization CD=RSD2 Patra and Maksyutov, GRL, 29, 28 May 2002
Incremental Optimization of Surface Network (Case 1) O basic [] Model Ensemble
Average uncertainty for TransCom-3 models Total Source Covar C = CS; Average Unc = C/ No. of Region
Model Dependent Uncertainty Reduction 1:UCB 2:UCI 3:UCI:s 4:UCI:b 5:JMA 6:MATCH:b 7:MATCH:c 8:MATCH:l 9:NIES:FRSGC 10:NIRE-CTM 11:RPN:SEF 12:SKIHI 13:TM2 14:TM3 15:CSU Patra et al., Tellus, 55B(2), 2003
Occultation based satellite measurements (Case 2) CD=RSD2 +Inst. Err. 2
Regional flux uncertainty at several satellite data precision
Satellite vs Surface data inversion (inst err=0)
Ecosystem production distribution: a justification for high resolution inverse model The fossil fuel emission do not have seasonality. Oceanic sources and sinks are weaker compared to the land and less heterogeneous.
Comparison of average flux uncertainty C_D=RSD^2
Satellite vs Surface Observations TransCom-3 HiRes setup C_D=RSD^2 + P^2
Multimodel Inversion of SOFIS data Three model groups: 1. High, Low and Intermediate signal in the “global” middle-upper troposphere High C_Ds compared to the signal – flat flux unc. with precision
Multimodel Inversion (no RSDs) Is the use of RSDs (derived from NIES model only) in satellite data inversion justified?
Conclusions • Flux uncertainty reduction with surface network extension depends on vertical profiles near the surface • Diving the Tracom-3 region into four smaller regions do seem to pose a severe aggregation problem • The use to different ATM simulations effect the pseudo-satellite inversion results
An evaluation of an ecosystem model for studying CO2 seasonal cycle
Tests with an Ecosystem Model Outputs • Optimisation of SimCYCLE model parameters: • 1. Q10 for respiration change with temperature • 2. Leaf-level Photosynthetic Capacity (PC) • Both parameters were changed by -20%, -10%, -5%, -3%, -1%, +1%, +3%, +5%, +10%, and +20% SimCYCLE: SIMulation model of the Carbon cYCle in Land Ecosystem (Ito and Oikawa, Eco. Mod., 2002)
Flowchart of SimCYCLE model Source: A. Ito
Light-photosynthesis relationship with different maximum rate Source: A. Ito
Temperature-respiration relationship with different Q10 Source: A. Ito
Procedure • Monthly-mean SimCYCLE outputs are transported using NIES/FRSGC model • Signals are sampled at 8 background stations in NH high latitude: • Alert, Greenland 82.45 297.48 210. • Zeppelin St., Norway 78.90 11.88 474. • Mould Bay, Canada 76.25 240.65 58. • Barrow, Alaska 71.32 203.40 11. • Atlantic Ocean, Norway 66.00 2.00 7. • Storhofdi, Iceland 63.25 339.85 100. • Baltic Sea, Poland 55.50 16.67 7. • Cold Bay, Alaska 55.20 197.28 25. • Mace Head 53.33 350.10 26. • Shemya Island, Alaska 52.72 174.10 40. • The simulations are then fitted to the Observed seasonal cycles of CO2
Fitting at Alert Q10 is not so sensitive Best fit at PSR=-10%
(Bad)Fitting at Baltic Sea Best fit at PSR=-10%
(Good)Fitting at Mace Head Good fit at PSR=-5%
Summary Recommended: 5 to 10% Q10 & -5 to -10% PSR
Thanks for your attention TransCom-3 Modellers: D. Baker (NCAR), P. Bousquet (LSCE), L. Bruhwiler (CMDL), Y-H. Chen (MIT), P. Ciais (LSCE), A. S. Denning (CSU), S. Fan (PU), I. Y. Fung (UCB), M. Gloor (MPI), K. R. Gurney (CSU), M. Heimann (MPI), K. Higuchi (MSC), J. John (UCB), R. M.Law (CSIRO), T. Maki (JMA), P. Peylin (LSCE), M. Prather (UCI), B. Pak (UCI), P. J. Rayner (CSIRO), J. L. Sarmiento (PU), S. Taguchi (NIAIST), T. Takahashi (LDEO), C-W. Yuen (MSC)