tdi experiment with nies model and interannually varying ncep winds
DESCRIPTION
TDI experiment with NIES model and interannually varying NCEP winds. S. Maksyutov, P.K. Patra and M. Ishizawa Jena; 13 May 2003. Objectives. Time-dependent inversion Study effect of meteorological fields on inversion Analyse climate impact on CO2 concentration anomaly - PowerPoint PPT PresentationTRANSCRIPT
S. Maksyutov, P.K. Patra S. Maksyutov, P.K. Patra and M. Ishizawaand M. Ishizawa
Jena; 13 May 2003Jena; 13 May 2003
TDI experiment with NIES model TDI experiment with NIES model and interannually varying and interannually varying
NCEP winds NCEP winds
ObjectivesObjectives
Time-dependent inversionTime-dependent inversion
Study effect of meteorological fields on Study effect of meteorological fields on inversioninversion
Analyse climate impact on CO2 Analyse climate impact on CO2 concentration anomalyconcentration anomaly
Link CO2 flux variability to ecosystem Link CO2 flux variability to ecosystem model simulationsmodel simulations
Experiments – Three transport Experiments – Three transport optionsoptions
ECMWF analysis for 1997 (cyclostationary)ECMWF analysis for 1997 (cyclostationary)
NCEP reanalysis for 1997 (followed by 98,99)NCEP reanalysis for 1997 (followed by 98,99)
NCEP reanalysis winds (interannually varying)NCEP reanalysis winds (interannually varying)
Other Issues:Other Issues:
Fossil fuel emission trends: 1996-1999 Fossil fuel emission trends: 1996-1999 (Marland et al., 2002) and 2000-2001 is (Marland et al., 2002) and 2000-2001 is extrapolated, spatial patterns for 90 and 95 are extrapolated, spatial patterns for 90 and 95 are usedused
Preprocessing NCEP windsPreprocessing NCEP winds
NCEP reanalysis (Period: 1988 to 2001), NCEP reanalysis (Period: 1988 to 2001), pressure level data at 2.5 deg resolution. pressure level data at 2.5 deg resolution. Vertical winds are up to 100 mb originally.Vertical winds are up to 100 mb originally.
Diagnostic vertical wind above 100 mb. Vertical Diagnostic vertical wind above 100 mb. Vertical motion along isentropic trajectories is assumed. motion along isentropic trajectories is assumed.
w=(U*dTp/dx+V*dTp/dy)/(dTp/dp)w=(U*dTp/dx+V*dTp/dy)/(dTp/dp)
This simplified approach fails at the poles. Polar This simplified approach fails at the poles. Polar values are smoothed from vicinity (second row values are smoothed from vicinity (second row from the pole).from the pole).
TDI SetupTDI Setup
Basically the same as T3 L2 source code Basically the same as T3 L2 source code (Ft Collins meeting by K. Gurney et al)(Ft Collins meeting by K. Gurney et al)
Originally developed at CSIROOriginally developed at CSIRO
Changes are made to ingest Green’s Changes are made to ingest Green’s function matrices usingfunction matrices using multiple-year multiple-year meteorologymeteorology
Period of source estimation: 1988 – 2001 Period of source estimation: 1988 – 2001 (after 3 years spin up time)(after 3 years spin up time)
COCO22 Observation data Observation data
GLOBALVIEW (August 2002 release)GLOBALVIEW (August 2002 release)
Maximum number of stations: 189Maximum number of stations: 189
Data period: Jan 1979 to Jan 2002 (2001 Data period: Jan 1979 to Jan 2002 (2001 depleted)depleted)
Bare minimum modifications to R. Law’s Bare minimum modifications to R. Law’s programprogram
Transport model re-configuration Transport model re-configuration for Earth Simulatorfor Earth Simulator
Parallelisation idea: array decomposition at tracer Parallelisation idea: array decomposition at tracer dimension (no reaction between tracers), rather dimension (no reaction between tracers), rather than latitude bands etc.than latitude bands etc.Total number of pulses are: (22*12+4) per year; 14 Total number of pulses are: (22*12+4) per year; 14 years processed (88-01)years processed (88-01)Single run uses 72 NEC-SX processorsSingle run uses 72 NEC-SX processors– each running 22 pulses (9 nodes on Earth Simulator)each running 22 pulses (9 nodes on Earth Simulator)– simulating 6 years of monthly-pulses at oncesimulating 6 years of monthly-pulses at once
Asynchronous meteorology (each process is Asynchronous meteorology (each process is allowed to run its own time); i.e., no allowed to run its own time); i.e., no communication between processes communication between processes
Other computer system issuesOther computer system issues
Virtual file system: Virtual file system: – files are copied to each process’s virtual disk space at files are copied to each process’s virtual disk space at
the job preparation stagethe job preparation stage– then disposed after run to disk or tape then disposed after run to disk or tape – process-specific input, output file extensions are added process-specific input, output file extensions are added
(style like .000, .001)(style like .000, .001)
Job script limitations (limited to 256 explicit file Job script limitations (limited to 256 explicit file declarations):declarations):– had to reduce the number of files usedhad to reduce the number of files used– we put the meteorological data for one month in one filewe put the meteorological data for one month in one file
Influence of Atmospheric Influence of Atmospheric transport on CO2 data transport on CO2 data
inversioninversion
Number of observations stations: 69 Number of observations stations: 69 (55% real data in 1988-2001)(55% real data in 1988-2001)
Inversion results: fitting to the dataInversion results: fitting to the data
Annual Annual Mean Mean FluxesFluxes
-reasonably good agreement
Case 1: NCEP-int closer to NCEP-97 than ECMWF-97 (constant offset: L-04, L-06, O-09)Case 2: NCEP-97 and ECMWF-97 are different from NCEP-int (L-05, L-01)
-1998 emission – distributed evenly between tropical land areas
Seasonal Seasonal CyclesCycles
As it comes out of the inversion model calculation!
Well produced for the well constrained regions
Monthly Monthly Flux Flux
anomalyanomaly
Noisy!
Ln-01 & Ln-07 well correlated
1998 emission
Comparison with T3L3 base Comparison with T3L3 base case (L2 by David Baker)case (L2 by David Baker)
Number of observations stations: 76Number of observations stations: 76
Annual Annual Mean Mean FluxesFluxes
Trouble with well constrained regions!
Something is still missing?
Average Average Seasonal Seasonal
CyclesCycles
Matches fairly well
Within ~20%? (haven’t done that precisely)
Reasons??
Monthly Monthly Flux Flux
anomalyanomaly
Too much variation
Still compares quite well
The 1998 emission peak gone missing
Derived Flux and ENSO IndexDerived Flux and ENSO Index
Future OutlookFuture Outlook
High Resolution Inversion (53 regions)High Resolution Inversion (53 regions)
Testing of different ecosystem modelTesting of different ecosystem model
Use ecosystem model output at high time Use ecosystem model output at high time resolution (daily? May be…)resolution (daily? May be…)
ConclusionsConclusionsThree types of transport fields are used in Three types of transport fields are used in interannual inversion which show good interannual inversion which show good agreementagreementSome climate impact on CO2 emission Some climate impact on CO2 emission can be studied; e.g 1998 Indonesian firecan be studied; e.g 1998 Indonesian fireComparison with T3L2 tending to matchComparison with T3L2 tending to match
We wish to contribute to T3L3 by using a We wish to contribute to T3L3 by using a high resolution inverse model and/or with high resolution inverse model and/or with a different ecosystem model resultsa different ecosystem model results