towards a multi-species variational assimilation system for surface emissions of ch 4, co, h 2 i....

Towards a multi-species variational assimilation

system for surface emissions of CH4, CO, H2

I. Pison, F. Chevallier, and P. Bousquet

Laboratoire des Sciences du Climat et de l’Environnement (LSCE)

HYMN, 1st annual meeting, Totnes October 22-24, 2007


Outline

• Atmospheric chemistry modelling at LSCE

• First results of PYVAR system

• Update on “traditional” inversions


LMDZ-INCAFull model

LMDZ-INCAGHG

SACASSimplified

Atm. Chemistry Assim. System

CHIMERERegional

model

Forward modelling

Inverse modelling

Traditionalinversion

VariationalInversion

PYVAR

Atmospheric chemistry modelling at LSCE


Inverse method

€

J x( ) =1

2y −Hx( )

TR−1 y −Hx( ) +

1

2x − xb( )

TB−1 x − xb( )

Critical choice of R & B

∂c∂t

+ v.grad(c) +1

ρdiv(ρ ′v ′c ) + k[OH ]c = S


What is PYVAR ?

J cost function


Modular System to combine atmospheric observations and prior information to infer surface fluxes :

– LMDZ transport model inside, air mass fluxes nudged on ECMWF winds (u,v)

– Variational principle (based on 4D-VAR)

– Written in PYTHON language

– 1 year of multi-species inversion ~10 days of calculations on an OPTERON PC (5 iterations)

– Parallel version in development

What is PYVAR ?


Simplified atmospheric chemistry assimilation system

SACAS

Satellite

Satellite

Satellite


• Inversion at model resolution (96x72)

• Weekly time resolution (48 steps per year)

• CH4, CO, CH2O, H2, MCF fluxes in the control vector

• OH scaling factor for 4 regions in the control vector

• Spatial correlation based on correl. length (500 km over lands)

• 1-year inversion (2004) performed so far

• MCF emissions prescribed with tiny error bars to get OH

• Cost function minimized by gradient conjugate (~5 iterations)

• Posterior var.-covar. matrix possible with more iterations

PYVAR inversion

QuickTime™ et undécompresseur TIFF (non compressé)

sont requis pour visionner cette image.


SACAS : Comparison with INCA

CH2OCH4

CO

H2

MCF


0

> 0

< 0

SACAS : Improvements of CO simulation

ABS(PRIOR-MOPITT) - ABS(OPT-MOPITT)

Validation of SACAS against MOPITT retrievals


SACAS : Atmospheric biases

CH4 CO ppm

ppm



MCF H2ppm

ppm


SACAS : First annual increments

CH4 CO

1010xkg/m2/s

• Large TRANSCOM regions

• Several processes on each

• 1979-2006 period

• MCF optimisation to get OH

• CH4 optimisation using optimized OH over 1984-2006

• Monthly emissions & observations. Use of 13CH4 observations

• Iterative procedure to calculate response functions

• 18 different inversions varying set-up

----> Can provide initial conditions for HYMN


Update on “Traditional” inversion


Consistent with

top-down estimates ?

Bousquet et al., ACP, 2005

OH variability ?


Update on methane inversion

Ano

mal

ous

dese

ason

aliz

ed f

lux

(TgC

H4/

yr)

Decreasing natural emissions

v.s.

IncreasingAnthropic emissions

NATURAL

ANTHROPIC

QuickTime™ et undécompresseur TIFF (LZW)

sont requis pour visionner cette image.



Inversionwith 13CH4 data

Inversionwithout13CH4 data

Impact of isotopic observations

Biomass burning CH4 emissions



Ano

mal

ous

dese

ason

aliz

ed f

lux

(TgC

H4/

yr)

Geographical

Europe

Source

Breakdown


First H2 inversions

TgH2/yr This study Hauglustaine 2004

Fossil fuels 20 ± 2 15-20Biomass burning 22 ± 3 10-20Oceans 6 ± 2 0-10Photochemical production 40 ± 3 30-50Nitrogen fixation 8 ± 2 3-10Total source 96 ± 5 70-100

Oxydation by OH not yet 10-30Soil uptake 63 ± 5 40-90Total sink - 60-100

Interannual

variations

Budget (1990-2003 average)

1991 2003

Global

Deposition

Biomass burning

1991 2003

1991 2003


WP5 : Forward simulations with full LMDZ-INCA model

- Outputs for year 2004 given to HYMN- Optimized CH4 emissions available (1985-2006)

WP6 : Inverse modelling

- Multi-species variational system ready- Traditional inversion for CH4 and H2 ready

N2O observations ? Available through NitroEurope IP(NOAA, AGAGE, RAMCES, other ?)

Work in progress …


HYMN WP6 Description of work

Global chemistry-transport models will be used in conjunction with satellite and ground based observations to improve the estimates in sources and sinks of atmospheric CH4, H2, and N2O.

1. Global forward modelling of CH4, N2O, and H2 and systematic evaluation against satellite data (CH4) and ground based measurements (CH4, N2O, H2). Derive error estimates and biases in model results.

2.1 Inverse modelling of surface emissions and sink by OH oxidation of CH4.2.2 Inverse modelling of surface emissions, atmospheric photochemical production, destructionby OH and surface uptake of H2. This work will be done based on available surfacemeasurements and satellite observations of CH4. These inversions will be performed on a regional basis.

3. Derive global distributions of optimized sources and sinks of CH4 and H2 through variational assimilation of methane satellite data and surface measurements.

Objectives

Estimate the surface emissions of CH4 and H2 on a regional basis base on inverse modelling and their evolution over the period 1980-2000.

Estimate the evolution of global and hemispheric OH in the atmosphere over the period 1980-2000 through inverse modelling.

Derive the global optimized distributions of CH4 and H2 sources and sinks on a pixel by pixel basis based on 4D variational data assimilation.

WP6


WP6Deliverables

D6.1 Report on model performances in terms of forward simulations of CH4, H2, and N2Ousing a-priori distribution of emissions and atmospheric sinks and derived error and biasesagainst model results and observations (month 27, coordinated by partner 5).

D6.2 Report on emissions of CH4 and H2 over the 1980-2000 period derived from inversemodeling of available satellite and ground-based observations on a regional basis. Comparisonwith a priori emissions (month 36, partner 5).

D6.3 Report on fully optimized distributions of emissions, production, and sinks for CH4 andH2 obtained through 4D var data assimilation (month 36, partner 1).

Milestones and expected results

M6.1 Forward simulations of H2, CH4, and N2O using a priori emissions and comparison withavailable observations (planned month 12, finished H2, CH4 simulation, N2O simulation delayedto month 18)

M6.2 Inversion of CH4 surface emissions and atmospheric OH over the 1980-2000 period andcomparison with a priori emissions (month 33).

M6.3 Inversion of H2 surface emission and soil uptake and photochemical production over theperiod 1980-2000 period (month 33).

M6.4 Fully optimized distribution of sources and sinks of CH4 and H2 through variationaldata assimilation (month 35)


Inversion setup (to be discussed) :

1-year inversion for 2004. More ?

Time resolution ?

Common observations : satellite + surface network + FTIR (keep for validation ?)+ aircraft obs (keep for validation ?)

+ satellite over oceans (validation ?)

Common variance scenarios (obs)

Aggregated flux or partition between processes ?

What about flux prior variances ?

Error Correlations for fluxes ? Retrievals ?

WP6 : Planned work for HYMN


CH4 budget with KNMI & LSCE variational systems

H2 budget with LSCE variational system (if available)

Use of traditional inversions for CH4 (1985-2006) & H2 (1990-2006)Test on OH

Sensitivity inversions (variational) : - different datasets (satellite, surface) obs freq.- different prior patterns- different OH (from WP5)- others ? Error correl tests

Strategy for surface elevation in SCIA. vs models

Outputs - to be standardized

Plotting package (IDL) available at LSCE for flux comparison

WEB interface

WP6 : Planned work for HYMN

Time schedule

• Protocol by December– Inverse setup– Surface stations obs.– SCIA treatment

• First results May 08 ?

3/ 2030 simulation


Atmospheric chemistry modelling at LSCE

LMDz-INCA General Circulation Model. INCA: tropospheric gaz phase chemistry, aerosols and long-lived greenhouse gases (CO2, CH4, N2O). 3.75° x 2.5°.

ORCHIDEE dynamical vegetation model used to derive surface properties, vegetation distribution, carbon cycle, biogenic and soil emissions. 40km x 40km.

Real-time chemical weather based on operational meteorology OR reanalysis (ERA40) OR free running GCM.

Chimère regional air quality model nested in LMDz-INCA global model. 50km X 50km.

1/ Global to regional scale modeling platform

PHOTOCOMP intercomparison (IPCC AR4-ACCENT) of future atmospheric composition. 25 state-of-the-art global chemistry transport models. 3 different scenarios for future surface emissions.

Regional model Chimère constrained by LMDz-INCA : relative impact of emissions versus long-range transport of pollution on air quality in Europe.

2/ 1960-2000 long-term simulationRETRO EU project (2003-2006) : reanalysis of the tropospheric chemical composition over the past 40 years.

Best available meteorology (ERA40), new monthly resolved anthropogenic and biomass burning surface emissions, stratospheric ozone climatologies.

Multimodel approach: 2 GCMs with chemistry and 3 CTMs.

http://www.lsce-inca.cea.fr/

3 day (NCEP) and 5 day (ECMWF) forecasts for global tropospheric chemistry,

aerosols and long-lived greenhouse gases (CO2, CH4, N2O)

1/ LMDz-INCA global chemical weather platform



Bousq

uet

et

al.,

2005

, 200

6

∂c∂t

+ v.grad(c) +1

ρdiv(ρ ′v ′c ) + k[OH ]c = S

MCF/CH4 inversion

EmissionsPrior Poste

OH Prior Poste

MCF inversion (CH3CCl3)

Methane Inversion


Hauglustaine et al., JGR, 2004

• Standard horizontal res. 3.75x2.5 • Standard vertical resolution: 19 hybrid

-p levels (surface to 35 km).

• Dynamic: large scale advection of tracers: LMD climate model

• Chemistry: standard version for tropospheric ozone calculation including NMHCs (90 species – 300 reactions); aerosols (mineral, sea-salt, BC, OC, sulfur) : INCA

• Biogenic Emissions from the ORCHIDEE dynamical vegetation model

• Anthropogenic emissions from either Edgar/IIASA/RETRO

• Biomass burning from van der Werf 2006

The general circulation model: LMDz-INCA


SACAS : First annual increments

MCF H2

1010xkg/m2/s

towards a multi-species variational assimilation system for surface emissions of ch 4, co, h 2 i....

Documents

hymn hymn

annual meeting

lsce slide

traditional inversion

s slide

annual increments ch

h2h2 mcf slide

lenvironnement lsce