Saroja PolavarapuEnvironment Canada

DAOS Working group Meeting, 15-16 Aug. 2014

Carbon Flux Data Assimilation

Thanks: D. Jones (U Toronto), D. Chan (EC), A. Jacobson (NOAA)

• The 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.

Pg C/yrhttp://www.scidacreview.org/0703/html/biopilot.html

The Global Carbon Cycle

Earth’s crust 100,000

1 Pg = 1 Gt = 1015 g

Net surface to atmosphere flux for biosphere or ocean is a small difference between two very large numbers

Net perturbations to global carbon budget

LeQuere et al. (2014, ESSD)

• Based on 2002-2012• 50% of emissions remain

in atmosphere• 25% is taken up by

terrestrial biosphere• 25% is taken up by

oceans

Interannual variability: 1870-2013LeQueré et al. (ESSD, 2014)

Atmospheric accumulation has strong variability due to land uptake. This is due to climate variability.

Ocean uptake is not as variable.

What are the key science questions?• The atmospheric burden of the global carbon cycle can be

determined by the difference between the anthropogenic emissions and natural fluxes (biosphere, ocean and land use change). The interannual variability and largest uncertainties are with the terrestrial biospheric uptake.

• What is the observing network needed to be able to reduce uncertainty on biospheric fluxes in order to identify anthropogenic fluxes on policy relevant scales?

– What is the spatial and temporal density of observations we need and what type?

– How do we deal with issues in the model? Inversion methodology needs perfect transport.

– Can we use multiple tracers to better constrain the carbon budget? (e.g. C13, C14, CO, etc.)

• What will be the long term evolution of sources and sinks? Are the nonlinear feedback processes sufficiently well represented in climate-carbon cycle models? (parameter estimation for ecosystem models)

Atmospheric observations give feedback on model forecasts

• If forecast does not match observation, difference could be due to errors in CO2 initial conditions, meteorological analyses, prescribed fluxes, model formulation, representativeness, or observation errors.

CO2 forecast Fluxes

caaf ScxMc ),(Meteorologyanalysis

CO2analysis

Forecast model Model error

Conventional inverse problem setup

World Data Centre for Greenhouse Gases (WDCGG)

http://gaw.kishou.go.jp/cgi-bin/wdcgg/map_search.cgi http://transcom.project.asu.edu

22 TransCom regions

Weekly avg obs

One or more yearsJ F M A M J J A S O N D J F M A M J J

Monthly mean flux

Inversions using surface network

• Inversion methods differ in:

– Methodology– Observations

▪ Sfc: 100 flask + continuous

– A priori fluxes– Transport models

• Interannual variability is similar and due to land

Peylin et al. (2013)

1 5-62 7-83 94 10

11

Mean XCO2 Aug. 2009 GOSAT Greenhouse Gas Observing Satellite

The changing observing system

World Data Centre for Greenhouse Gases

http://gaw.kishou.go.jp/cgi-bin/wdcgg/map_search.cgi

• ~100 highly accurate surface stations with weekly or hourly data

• Regular aircraft obs over Pacific

• Satellites: GOSAT (2009), OCO-2 (2014) +…

v2.0 averaged at 0.9°x0.9°

GOSAT figure courtesy of Ray Nassar, EC

How is data assimilation relevant to Carbon Flux estimation problems?

• In flux inversions, if one solves for fluxes only, the transport model is needed to relate the flux to the observation: model is a strong constraint

• Exact mass conservation in transport model overs years of simulation is needed to attribute model-data mismatch to fluxes.

• Techniques used to solve inverse problem: 4D-Var, EnsKF, Bayesian Inversion, Markov Chain Monte Carlo (MCMC)

• With a CTM for transport, analyses or reanalyses are needed: IFS, ERA-I, MERRA, JRA55, etc.

• Parameter estimation for ecosystem models: Carbon Cycle DAS• Ultimately coupled atmosphere-ocean-land data assimilation could

be envisioned

TfobsTfobsTbTb ScHcScHcSSSSSJ 11 RB21

21)(

flux conc obs

Spatial interpolation Forecast model

Prior flux

The future vision: Comprehensive Carbon Data Assimilation System

Comprehensive carbon assimilation systems are being built by NASA, NOAA, agency-consortiums in Europe, Japan and EC.

GEO Carbon Strategy Report (2010)

How is Carbon Flux estimation relevant to NWP data assimilation?• Need mass conservation over long time scales (years). Global

mass of CO2 is very sensitive to model errors. Can help improve NWP, air quality and climate models.

– Key model processes affecting CO2 forecast distribution:▪ Advection (NWP model)▪ Boundary layer mixing (NWP model)▪ Convective transport of tracers (Air quality model)▪ Ecosystem modeling of biospheric respiration and

photosynthesis (Climate model)• Radiance assimilation (RTTOV) assumes constant CO2 value.

Impact of using 3D CO2 fields is reduction in range of bias corrections, and slight improvement in tropical forecasts near 200 hPa. (Engelen and Bauer 2011)

Future directions Near real time CO2 forecasts• Coupled CO2 and meteorological assimilation

– ECMWF: Real time operational 5-day CO2 forecasts since 2013. No assimilation of CO2 obs. Updated initial conditions from flux inversions every Jan. 1. Plans: Near-real time assimilation of surface obs of CO2 with coupled meteorological/tracer assimilation

– GEOS5: Coupled CO and CO2 assimilation to meteorological assimilation. Weakly couple ocean and land data assimilation systems to atmospheric assimilation system

– Justification▪ Provide boundary conditions for regional modelling and flux inversions.▪ Improve modelling of radiative transfer, evapotranspiration▪ Feedback on modeling of boundary layer, convection, advection▪ Provide a prioris for satellite retrievals of CO2 and CH4

• Coupled CO2, flux and meteorological estimation:– Kalnay group (U Maryland)– EC-CAS (Polavarapu)

Summary

• Although similar DA tools and techniques are being used to estimate carbon cycle dynamics, the problem is different because the system is not chaotic: tracer transport is linear.

• The challenge is to estimate highly variable surface fluxes with too few data, as opposed to the NWP problem of finding an accurate analysis (and uncertainty) with which to make forecasts.

EXTRA SLIDES

Variations in Atmospheric CO2

• Diurnal variations, linked to surface sources and sinks, are strongly attenuated in the free troposphere

• Diurnal variations in column CO2are less than 1 ppm

• Large changes in the column reflect the accumulated influence of the surface sources and sinks on timescales of several days

Olsen and Randerson (2004, JGR)

Surface CO2

Column CO2

Diurnally varying surface fluxes

Modeled CO2 at Park Falls

5-day running mean surface fluxes

But what is the spatial distribution of the fluxes and how is it changing?

Figure courtesy of Elton Chan, CCMR

With the increased coverage from new satellite data, can we get flux estimates at higher spatial resolution?

Mean XCO2 Aug. 2009 GOSAT Greenhouse Gas Observing Satellite

v2.0 averaged at 0.9°x0.9°

Figure courtesy of Ray Nassar, CCMR

OCO-2 launch July 2014

http://oco.jpl.nasa.gov/

Atmospheric model is not perfect

Even with the same sources/sinks, different models give different CO2.

Largest discrepancy between model transport is N.Hemisphere biospheric exchange.

Model errror is not random but systematic and will lead to bias in flux estimates.

Gurney et al. (2003, Tellus)

Zonal mean annual mean CO2

Meteorological winds are not perfectLiu et al. (2011, GRL)

Using same sources/sinks, same model, same initial condition, CO2forecasts are still different due to errors in wind fields.

Forecast spread due to uncertainty in winds creates CO2 spread where gradients are large (near sources/sinks), not where wind uncertainty is large.

Imperfect model transport impacts source/sink estimates

Modelled vertical gradients don’t match aircraft obs

Too strong gradient too little vertical mixing larger emissions are needed to match obs in NH

Verticalgradienttoo weak

Pos

t inv

ersi

on fl

ux (P

gCyr

-1)

Larger gradient produces larger NH uptake

Stephens (2007, Science)

gradientis too strong

Spatial information

• Group 1 all solve for fluxes on grid scale and use obs at sampled time instead of monthly means

Peylin et al. (2013)

If we want to know the spatial distribution of fluxes, then methodology matters!

25N-25S

ECMWF

• Real time operational 5-day CO2 forecasts since 2013– No assimilation of CO2 obs– Updated initial conditions from flux inversions every Jan. 1– Online parameterized ecosystem model (CTESSEL)– Fluxes: GFAS v1.0 (Fire, Kaiser et al., 2012), ocean (Takahashi et

al. 2009), anthropogenic (EDGAR version 4.2)

• Justification– Provide boundary conditions for regional modelling and flux

inversions.– Improve modelling of radiative transfer, evapotranspiration– Feedback on modeling of boundary layer, convection, advection– Provide a prioris for satellite retrievals of CO2 and CH4

• Next: Near-real time assimilation of surface obs of CO2– Coupled meteorological/tracer assimilation– Needs near-real time obs provision

European Centre for Medium Range Weather Forecasting

USA

• GEOS-5– Couple CO and CO2 assimilation to meteorological assimilation– Weakly couple ocean and land data assimilation systems to

atmospheric assimilation system– Dynamic vegetation

Key Future GHG missions with surface sensitivity Slide from Ray Nassar (EC)

• Lots more observations are expected soon, from satellite-based platforms• Can the new space-based measurements help fill in data-gaps in ground-

based network? Can we then get regional flux estimates over Canada?

Key Future International GHG Satellites

OCO launch photo by Matt Rogers,Colorado State University

• Orbiting Carbon Observatory 2is scheduled to launch in July 2014• CO2 mission to replace a 2009 mission lost to a launch mishap

• TanSat is China’s CO2 mission scheduled to launch in mid-2015• Has a very similar design to OCO-2

Chinese Academy of Sciences (CAS), Ministry of Science and Technology (MOST), Chinese Meteorological Agency (CMA)

• CarbonSat (candidate for launch in 2019)• Design optimized for monitoring CO2 and CH4 emissions including point sources like power plants, oil sands

• GOSAT-2 potential launch ~2017• Will measure CO2 and CH4 with more coverage and smaller pixel size than GOSAT

25

Slide from Ray Nassar (EC)

Orbiting Carbon Observatory 2

• OCO-2 launched on 2014-07-02 at 2:56 PDT by a Delta-II• OCO-2 will measure reflected sunlight in the 0.76, 1.61 and

2.06 m bands• Footprints (nadir): GOSAT 10.5 km (d), OCO-2 ≤1.29x2.25 km2

• GOSAT: 4 sec / obs, OCO-2: 8 obs x 3 times per second• OCO-2 will have ~200 times as many measurements as

GOSAT • Glint range from sub-solar latitude: GOSAT ±20°, OCO-2 ±80°

OCO-2

OCO-2 GOSATWashington DC

OCO-2

Sun glint over water

Slide from Ray Nassar (EC)

OCO-2 and beyondNature (June 26, 2014) 510,451–452. doi:10.1038/510451a