Inverse emission estimates for Europe using tall tower observations and the COMET inverse model Alex Vermeulen 1, Gerben Pieterse 1,2 1: ECN2: IMAU.

Inverse emission estimates for Europe using tall tower observations and the COMET inverse model

Alex Vermeulen1, Gerben Pieterse1,2

1: ECN 2: IMAU

28th Transcom meetingPurdue Univ.; April 24, 2007

ECN and Cabauw observations


Cabauw – CBW – The Netherlands (ECN)

ÊÚ

TestUrbanUrban sprawlIndustrialRoad/Rail(Air)portsMineral extractionConstruction siteRecreationArableFruit treesPasturesCropsForestNatural grasslandMoors, heathlandDunes, sand, beachMarshesWaterSeaNo Data

ÊÚ Theme1.shp

30 0 30 60 Kilometers

S

N

EW

Gas Method Operational Precision

CO2 LICOR 7000 Nov-04 0.05 ppm

Flask sampler CIO Nov-04

222Rn ANSTO Nov-05 50 mBq.m-3

CH4 GC-FID

Nov-04

2 ppb

CO GC-FID 1 ppb

N2O GC-ECD 0.4 ppb

SF6 GC-ECD 0.2 ppt

Height: 200m AGL

Base: -2m ASL

Lon: 04°56’

Lat: 51°58’

Levels: 20m, 60m, 120m, 200m

LU: Grassland, crops


CBW CO2 trend 2000-2006


Flux and concentration vertical gradients Cabauw

Casso-Torralba et al, 2007 (in prep)


Measurements, Modelling, SRM’s and Inversion

Independent verification of bottom up estimates: UNFCCC, process models

It seems so easy: Subtract the influence of meteorology on the concentration, what remains is the influence of emissions

BUT: the atmosphere is a very efficient mixer, most of the signal is lost in 1-2 days of travel

SO: measure close to the sources at high temporal resolution, extend in the mixed layer to reduce the very local influence

Current global network not sufficient

Not many stations, but high frequency measurements and lots of noise in atmosphere and models

Mathematically: Ill posed problem, we need constraints


The COMET model

•Trajectory model, offline meteorology•COMET model, 0.5o windfields, Flextra trajectories •ECMWF meteorology, MLH Crit. Richardson scheme•CH4 fluxes from METDAT (Berdowski et al, 1998): 3-hourly time-res. at 10’ horiz. resolution•Mixed layer bulk concentration •Hourly 144-hr backward trajectories•CH4 meas vs. model:

R2=0.84 during summer, bias=0 ppb, RMSE=115 ppb (6%)

Full year: R2=0.72Vermeulen et al, ACPD, 8727, 2006


Modeling framework at ECN

ForwardCalculations

Spatial Aggregation

InverseCalculations

dXk=Cklel

Trajectory data

LPDM data

Inventory data

Flux model

Concentration data

Flux data?

Recursive Spatial

Aggregation

LPDM model

(FLEXPART1)

COMETmodel

(FLEXTRA1)

(C

o)

Var

ian

ce

Opt

imiz

atio

n

Concentrations

ConcentrationsFluxes

Synthetic data

1 Stohl, A. (1998): Computation, accuracy and applications of trajectories - a review and bibliography, Atm. Env., 32, 947-966

Deposition fluxes

Concentration enhancements

SRMCalculations


Diurnal variation get lost after 24 hours…

Model: COMET; Background CO2 levels from TM5 (Krol, pers comm).

HUN 48m 2003 CO2 modelled as function of travel time

360

370

380

390

400

410

420

430

440

450

460

1-5 3-5 5-5 7-5 9-5 11-5 13-5

Date time

Co

nc

en

tra

tio

n [

pp

m]

CO2

CO2 backgr.

CO2-24

CO2-48


Measurement NON GPP NPP NEPR2 - 0.15 0.07 0.45 0.49Variance [ppm] 12.3 4.1 10.9 8.1 15.0Bias [ppm] - -11.0 -18.5 -9.9 -1.0

Measurement NON GPP NPP NEPR2 - 0.63 0.35 0.62 0.64Variance [ppm] 21.6 10.0 11.4 13.7 25.2Bias [ppm] - -12.9 -18.0 -12.0 -1.9

CBW

HUN

COMET model forward results: Mixed layer concentrations CO2 for 2002



The source aggregation scheme for SVD inversion

Calculate Source-receptor relationship (SRM) per hour and per observation point at high resolution of 10’ (~10 km) or multiple of this

Run transport model to determine maximum annual average SRM value ppm/(kg/(m2.s)) or potential contribution (SRM*E) in ppm

Aggregate neighbouring areas by joining until sum of area >= maximum contrib: SRM shrinks from 200*400 to ~200 rows=regions

Rerun transport model to build SRM for aggregated regions

Iteratively perform SVD and aggregate adjacent areas with high covariance in emissions (dipole) against observations

Iteratively remove areas with resulting emission of high variance (e.g. >30%) from SRM

Until stable number of regions or no regions left…

Procedure retrieves the maximal spatial resolution that can be resolved from the combination of model and measurements. Modification for (partial) resolving emissions of source categories, temporal patterns or any combination of these is relatively easy


METDAT prior emissions for 1998


Source area aggregation based on SRM

+

=>


Results: CBW station only

Synthetic inversion:Forward modelledConcentration is input for inversion

200 km

Cabauw measurements 2006

25 areas can be resolved

Fluxes in kg(CH4).km-2.s-1

Areas 24 is United Kingdom, 25 N-Germany, 1-8 are Netherlands

Prior Posterior+s.d.


Results for CBW single years: 2002

Prior

Post 2002


Results for CBW single years: 2003

Worldcountries.shpWorldcountries.shp

Mergeded Grids of Map Calculation 1<0.010.01 - 0.020.02 - 0.030.03 - 0.0410.041 - 0.0510.051 - 0.0610.061 - 0.0710.071 - 0.0810.081 - 0.091No Data


Inverse determined annual emissions for The Netherlands

Year Emission kTon CH4/yr

Prior (METDAT, 1998) 1020

2000 1600

2001 2000

2002 1350

2003 1600

2005 1350

2006 1950

Uncertainty: 20-30%


Improvement forward COMET after emission update

a Posterioriy = 0.8411x - 0.0066

R2 = 0.6426

0

0.5

1

1.5

0 0.5 1 1.5

Meas [ppm]

Mo

del

[p

pm

]

a Prioriy = 0.6767x - 0.0041

R2 = 0.6147

0

0.5

1

1.5

0 0.5 1 1.5

Meas [ppm]

Mo

del

[p

pm

]


Inverse calculations for methane at multiple sites for 2002

Em

issi

on [

mg

m-2 m

in-1]

Region index


Conclusions & Outlook

•Emissions (of methane) can be constrained from atmospheric signal without a priori information on their size

•Atmospheric inversion of area of big emissions needs high resolution in space and time, 10’ and hourly or better: otherwise degradation of signal

•Source Aggregation+SVD form robust combination

•Systematic model errors still problematic: need to get the models right, minimise bias: background concentration, MLH

•CBW CH4 concentration data constrains emissions of an area 400x400 km2

•Multiple years extend the constrained area

•Multiple stations extend the area as well (of course)

•Tall tower data very valuable provided continuous vertical gradients are measured at high frequency

•By 2007 and further 6 tall towers will provide 2006 data for CH4, N2O, SF6 and CO2


Acknowledgements

Climate Changes spatial planning/Klimaat voor ruimte Research Program

EU FP5: CHIOTTO, contract EVK2-2002-0163

National funding agencies:

VROM

Senter/NOVEM

The Transcom continuous experiment: http://www.purdue.edu/transcom/protocol_v5.pdf

Maarten Krol (IMAU,WuR) for background CO2 data from TM5

Sander Houweling (IMAU) for background CH4 data from TM3

Pim van den Bulk

Piet en Mike Jongejan

Han Mols

Inverse emission estimates for Europe using tall tower observations and the COMET inverse model Alex Vermeulen 1, Gerben Pieterse 1,2 1: ECN2: IMAU.

Documents

th transcom meeting

comet model trajectory

imau slide

prep slide

comet model flextra

hours model

comet background

cabauw cbw