Presentation on theme: "The EC-Carbon Assimilation System"— Presentation transcript:
1The EC-Carbon Assimilation System Saroja Polavarapu, Ray Nassar, Doug Chan (CCMR/CRD)Dylan Jones, Mike Neish, Shuzhan Ren, Feng Deng (U Toronto)John Lin, Myung Kim (U Waterloo)Meeting on Air Quality Data Assimilation and Fusion R&D, Jan , 2011
2The Global Carbon Cycle 8.6Pg C/yrThe natural carbon cycle involves CO2 exchange between the terrestrial biosphere, oceans/lakes and the atmosphere.Fossil fuel combustion and anthropogenic land use are additional sources of CO2 to the atmosphere.
3The Global Carbon Cycle [IPCC, 2007]Only 50-60% of anthropogenic CO2 emissions remain in the atmosphereThe uncertainty and interannual variability in the global CO2 uptake is mainly attributed to the terrestrial biosphereReducing uncertainties in CO2 sources and sinks is an active area of research and has important implications for climate mitigation policies.
4Greenhouse gas measurement network NOAA-ESRL (US), Environment Canada, CSIRO (Australia), JMA (Japan) ...WMO - World Data Centre for Greenhouse Gases (WDCGG)Numerous research groups over the past ~20 years have combined the highly-accurate but sparse atmospheric CO2 measurements from the ground-based network with models, to give estimates of CO2 sources and sinks on coarse spatial scales.With the coming wealth of satellite observations, more sophisticated methods of data assimilation that can handle the large data volume needed to provide estimates on finer scales.
5Variations in Atmospheric CO2 Modeled CO2 at Park FallsDiurnal variations, linked to surface sources and sinks, are strongly attenuated in the free troposphereDiurnal variations in column CO2 are less than 1 ppmLarge changes in the column reflect the accumulated influence of the surface sources and sinks on timescales of several daysColumn CO2Surface CO2Diurnally varying surface fluxes5-day running mean surface fluxes[Olsen and Randerson, JGR, 2004]
6Satellite observations Surface observations are highly accurate but sparseSatellite observations can complement spatial coverage though vertical resolution of nadir obs is typically poorCurrent missions (nadir): HIRS (1978-), AIRS (2002+), SCIAMACHY (2003+), TES (2006+), IASI (2007+), GOSAT (2009+)Current missions (limb): ACE-FTS (2004+)Future missions: IASI (2012,2016), OCO-2 (2013/4), TanSat (2015), GOSAT-2 (2016), CarbonSat (2018), PCW/PHEMOS-FTS (2018), ASCENDS (2020)
7Strategies for Temporal and Spatial Coverage Slide from Ray Nassar, CCMRSun-synchronous Low Earth Orbit (LEO) missions only measure at a single point in the diurnal cycleIncreasing temporal coverage requires a constellation of LEO satellites or different orbitsCarbonSat constellation of LEO satellites has been proposedGeostationary missions like NASA’s GEO-CAPE could have CO2 capability added for continuous coverage from ~50°S-50°NCanada’s Polar Communications and Weather (PCW) Polar Highly Elliptical Molniya Orbit Science (PHEMOS) Weather, Climate and Air quality (WCA) proposal (currently in phase-A) would use a Highly Elliptical Orbit (HEO) for high latitude quasi-continuous coverage
9Estimating CO2 fluxes from surface Can use all observations over a time window to estimate CO2 fluxes at all timesStandard method is Bayesian Synthesis InversionEC CO2 flux inversion capability:Bayesian Synthesis InversionTransport model: NIES (Japan), TM5 (Europe), GEOS-Chem (US)Ground-based observationsCan standard methods of flux estimation handle coming wealth of observations?dCO2/dt = transport + fluxphoto credit: Matt Rogers,Colorado State University
10The flux estimation problem Estimate monthly averaged fluxes from ground-based obs for 1 yearAll source region amplitudes at all times ~22x12=264previous estimateAll observations at all times~100x12=1200Connection between obs and fluxes at all previous times for all regions. 264 model integrations of 1 yearAs number of source regions increases too many model integrations!
11Bayesian Synthesis Inversions Advantages:Uses all obs at all times (months) to determine all monthly fluxes over 1-3 yearsIf assumptions are correct, this is the best, most general solutionProblem 1:Want to use more obs (e.g. continuous, aircraft, satellite) so we can capture finer time and space scales in fluxesSolution 1: Use 4D-Var (e.g. GEOS-Chem, Chevallier et al. (2007,9))~100 forward+ADJ runsNeed to develop and maintain TLM and ADJ modelsSolution 2: Use Kalman smoother (e.g CarbonTracker)Sequential estimation means using obs only over a short time period, then marching forward. Smoother means improving estimate based on future obsDoes not use all obs to estimate all sources
12The flux estimation problem Estimate monthly averaged fluxes from ground-based obs for 1 yearRandom errors due to:Initial conditions in CO2Driving wind analysesModel formulationrepresentativenessInstrument error1200x1200Random errors in source amplitudes264x264All error sources convolved into 1 error estimate (R).In practice only obs and rep errors are accounted for.Often, no correlations are assumed.
13The flux estimation problem Estimate monthly averaged fluxes from ground-based obs for 1 yearRandom errors due to:Initial conditions in CO2Driving wind analysesModel formulationrepresentativenessInstrument error1200x1200Random errors in source amplitudes264x264Estimation errorIf B and R are incorrect, then uncertainty estimates are wrong
14Relaxing the assumptions Problem 2: Assumptions made in practice are not correct,e.g. no errors for analysed wind fields, initial CO2 field, model formulation, source region definitions. Often no error correlations.Because assumptions are not valid, we cannot believe uncertaintiesSolution: Use data assimilation to estimate concentrations, simultaneously inferring fluxes as a “model parameter or forcing”Use ensemble of forecasts to explicitly account for initial state, meteorology, model, representativeness, obs, source region errors.Fully evolve covariances in time, producing full spatial correlationsEnsemble Kalman Smoother used by Japanese (Miyazaki 2011, JGR) for CO2 fluxes.
15EC-CAS Carbon Assimilation System Analysis stepFlask, continuous, aircraft, satellitePerturb initial conc., met fields, fluxesPerturb obsforecast step
16EC-CAS: Carbon Assimilation System New EC-CAS (Carbon Assimilation System) proposed for monitoring carbon and policy/verification purposesProject started in April EC/UT/UW collaboration.Can be used to answer questions on observing system needs (space-based, and EC’s ground-based obs)Will be run routinely but behind real time since it takes time for flux to reach measurement locations.EC-CAS is based on EnsKF with GEM-MACH but will be a Kalman smoother for estimating surface fluxesParameters for EnsKF not clear yet: update frequency, and data window (6h normally)The future vision includes the coupling to ocean and Ecosystem models
17The future vision: Comprehensive Carbon Data Assimilation System EC-CAS will form the basis of a comprehensive carbon assimilation system, comparable to those of NASA, NOAA and agency-consortiums in Europe and Japan.
18Starting point with GEM Figure from D. Chan, CCMRD.Chan/M.Ishizawa had CO2 version with GEM v3.2.0 to see if GEM can capture synoptic scale variability. It does seem to do thisTime series of CO2 at FraserdaleThe minimum CO2 concentration during these two months was subtracted so the time series start from a zero value.Complete time series (top)Daily variability was removed by plotting afternoon mean values only (bottom)
19Early issues with model choice Our development uses MAESTRO which is used to run the EnsKF (CMC uses this for operational EPS)Choice of GEM version for EC-CAS:EnsKF uses GEM v4.2.0 and is not backward compatible so Doug Chan’s GEM v3.2.0 with CO2 tracer v3.2.0 not feasible.Decided to choose GEM-MACH because it already handles emissions, tracers, vertical diffusion and they will move to v Also this permits future interaction and collaboration with AQRD.Model testing with GEM-MACH (v3.3.3) but EnsKF development needs v4.4.0 which is under beta testing.
20GEM-MACH-GHG version GEM-MACH was developed for CO2 simulation by Started from global version (based on v3.3.3) used for stratospheric ozone and developed by Jean deGrandpre (ARQI)Reduced resolution to 400x200 (roughly 1 degree), 80 levelsAdding 6 CO2 tracers, one for each emission source plus a total CO2 and a background CO2 (with no emissions)Coupled tracers to emissions fieldsObtained monthly emissions from Doug Chan, and regridded these to Z grid, 400x200 (preserving total mass)Uses GEM-MACH emissions preprocessor with global fields
21Model validation runHow well can GEM-MACH simulate Carbon? Key concern: mass conservation over multiyear runs. Diagnostics: Seasonal cycle, hemispheric gradients, mass conservation. Comparison against obs and other models (CarbonTracker, GEOS-Chem)Simulation for January 1, 2009 – Jan. 2012?Dates related to GOSAT launch (Jan. 2009) and GEM-strato analyses availability (Operational implementation on June 22, 2009)Initial condition from CarbonTracker for Jan. 1, 2009Meteorology: surface fields (archived surface analyses), 3D winds (prelim cycle, parallel run, operations)Emissions:Every 3 hours (area type) though GEM-MACH set up for monthly fields with diurnal variationbiosphere (CarbonTracker a posteriori)ocean (CarbonTracker a posteriori)Fossil Fuel (CDIAC)Biomass burning (GFED v3)
22EC-CAS development priorities ModelGEM-MACH based on v4.4.0 beta-9 runs in CO2 mode w/o emissions. Need to add emissions. Reconnect vertical diffusion.Repeat model validation run with GEM-MACH-GHG v4.4.0Assimilation (EnsKF)Allow EnsKF and MAESTRO to use GEM-MACH instead of GEMChange control vector change from meteorology to tracers/species + fluxesDevelop observation operators for all new obs to be assimilated or monitoredComplete EnsKF and test with surface obsExtend EnsKF to a Kalman Smoother (use future obs to estimate current flux)Observationsconvert surface obs to BURP for ingestion by data assimilation codes.examine GOSAT data, determine biases, quality control procedures, bias correction procedures.EmissionsIncorporate diurnal/weekly scaling factors developed by Ray Nassar
23GHG and AQ assimilation synergies GEM-MACH development can be coordinated, e.g. vertical diffusion, mass conservationEnsKF development by EC-CAS will be usable (but not tested) with reactive chemistryPrimary/Initial fociiGHG flux assimilationAir Quality assimilationAssimilation needsInverse problem (source estimation)SmootherForecasting problemFilterModel needsTransportEmissionsMass conservationReactive chemistryTime scalesMonths to yearsDaysSpace scalesGlobal, regionalRegional
25Observations from a Three-Apogee Orbit Slide from Ray Nassar, CCMR2 satellites, each with 16 h orbitapogee = kmperigee = 8100 kmImages 16 h / 48 h per region8 (60x60) arrays wide6 (60x60) arrays tall10 x 10 km2 footprintNIR-TIR FTS similar to GOSAT TANSO-FTS (ABB Group) could measure CO2 and CH over ice-free land surfacesNassar et al. (in prep.)Various pointing scenarios for PCW-PHEMOS are currently under consideration
26Canadian Greenhouse Gas Measurement Program Figure from Elton Chan
27Global Greenhouse Gas Measurement Network World Data Centre for Greenhouse GasesNOAA-ESRL (US), Environment Canada, CSIRO (Australia), JMA (Japan) ...WMO - World Data Centre for Greenhouse Gases (WDCGG)
28Present satellite instruments All are nadir except ACE which is occultation (limb)InstrumentDataavailLatitudinal coverageVertical sensitivityHIRS1978-20S-20NUpper trop ~10 kmAIRS2002-80S-80NUpper tropSCIAMACHY2003-60S-80N landTotal columnACE-FTS2004-82S-82N sparse5-100 km, D3 kmTES2006-40S-40NMid trop ~5 kmIASI2007-Upper trop, ~12 kmTANSO-FTS(GOSAT)2009-25S-25N oceanTotal column,
29CO2 Flux Inversion with Regional Focus on North America Deng et al. (2007)30 small regions in North America, 20 large regions for the rest of the globe, and 88 CO2 stations (GlobalView-2005)
30Annual Result for 2003Deng et al. (2007)North American biosphere is a sink of −0.97 ± 0.21 Pg C, Canada’s sink is −0.34 ± 0.14 Pg C.