Methane inversion from satellite, TRANSCOM workshop, Jena

1
Inverse modelling of methane sources and sinks
using satellite observations

• Jan Fokke Meirink, Henk Eskes, Michiel van Weele,
  Albert Goede

Royal Netherlands Meteorological Institute (KNMI)
2
Overview

• EVERGREEN
• SCIAMACHY
• inverse modelling strategy
• first experiments

3
EVERGREENEnVisat for Environmental Regulation of
GREENhouse gases

• EC 5th framework programme
• Feb. 2003 Feb. 2006
• Objective use ENVISAT (SCIAMACHY and MIPAS)
  measurements for inverse modelling of GHG
  emissions
• Partners KNMI (NL, coordinator), Univ. Bremen
  (DE), Univ. Leicester (GB), Univ. Heidelberg
  (DE), NILU (NO), SRON (NL), MPI-BGC (DE),
  BIRA-IASB (BE), UPMC-SA (FR), RWE-Rheinbraun
  (DE), Univ. Liège (BE), EC-JRC-IES (IT)
• website http//www.knmi.nl/evergreen

4
EVERGREEN tasks

• Retrieval and validation CH4, CO, (CO2), O2
  columns clouds
• Radiation budget modelling use of measured trace
  gas distributions in radiative forcing
  calculations
• (Inverse) modelling CH4, CO, CO2
• emission inventory
• (forward) model intercomparison (222Rn, SF6, ...)
• inverse modelling

5
SCIAMACHY on ENVISAT
CO2
CO
CH4
6
SCIAMACHY measurements

• No useful retrievals yet
• calibration problems (mainly dark current
  correction)
• ice layers on the detectors, channel 7 and 8
• NIR retrieval quality will depend on
• albedo over water low signal-to-noise ratio
• solar zenith angle
• clouds

7
SCIAMACHY example nadir measurements (simulated)
8
Chemistry-transport model TM3

• horizontal resolution 7.5x10 / 3.75x5 / 2.5x2.5
  deg
• vertical resolution 19 / 31 layers up to 10 hPa
• slopes scheme for advection
• ECMWF meteorology
• 37 tracers (22 transported)
• CBM-4 scheme for NMHC chemistry
• nudging at model top to climatology (O3) and UARS
  data (CH4)
• single-tracer version with OH fields from full
  model

9
Planned model setup in Evergreen

• horizontal resolution 2x3 deg
• extension of model in vertical direction use
  subset of ECMWF layers up to 0.1 hPa
• nudging of CH4 to surface observations to have
  model reproduce measured NH-SH gradient
• up-to-date emissions from WP 4100

10
Inversion strategy

• 4D-var method
• adjoint model of TM3 CH4-only version has been
  developed
• optimize surface fluxes and initial CH4 field
• expected in the beginning adjustments to CH4
  field, later to surface fluxes
• time frame 1 week to 1 month

11
4D var (1)
12
4D var (2)

• cost function

• model

13
4D var (3)

• gradient of the cost function

14
Preconditioning

• S diagonal matrix with standard deviations
• LLT symmetric matrix with correlations

15
First experiments

• First week of January 2000
• Sat.obs. taken from perturbed model run
• Optimize emissions only
• Background error covariance
• Standard deviation 50 of emission
• Horizontal correlation function Gaussian
  with length scale of 1000 km
• Observation error covariance
• Diagonal standard deviation 0.5 of column

16
A priori emissions
17
Perturbed run
50 enhanced emissions
18
Effect on CH4 field after one week
19
Pseudo satellite observations
20
Optimized CH4 emissions
21
Resulting CH4 field
22
Setting ?obs to 2
23
Adding 0.5 noise to sat.data
24
Summary

• EU project EVERGREEN for emission estimates of
  CH4, CO, and possibly CO2, using satellite data.
• SCIAMACHY has the potential of measuring CH4 and
  CO columns, but first some calibration problems
  have to be solved.
• Inverse modelling of CH4 using 4D-var has been
  set up at KNMI and first experiments have been
  done.