New insights into CO2 fluxes from space?

1
New insights into CO2 fluxes from space?

• Michael Barkley, Rhian Evans, Alan Hewitt,
  Hartmut Boesch, Gennaro Cappelutti and Paul Monks
• EOS Group, University of Leicester

2
(No Transcript)
3
(No Transcript)
4
The FSI algorithm Overview

• How do we measure atmospheric CO2?
• WFM-DOAS retrieval technique (Buchwitz et al.,
  JGR, 2000) designed to retrieve the total columns
  of CH4,CO, CO2, H2O and N2O from spectral
  measurements in NIR made by SCIAMACHY
• Least squares fit of model spectrum weighting
  functions to observed sun-normalised radiance
• We use WFM-DOAS to derive CO2 total columns from
  absorption at 1.56 µm
• Key difference to Buchwitzs approach
• No look-up table
• Calculate a reference spectrum for every single
  SCIAMACHY observation i.e. to obtain best
  linearization point no iterations
• See Measuring atmospheric CO2 using Full
  Spectral Initiation (FSI) WFM-DOAS , Barkley et
  al., ACP, 6, 3517-3534 ,2006
• Computationally expensive ?

SCIAMACHY
SCIAMACHY, on ENVISAT, is a passive
hyper-spectral grating spectrometer covering in 8
channels the spectral range 240-2040 nm at a
resolution of 0.2-1.4 nm Typical pixel size 60
x 30 km2
5
Cloud Filter SPICI (SRON) (Krijger et al, ACP,
2005)
A priori Data CO2 profiles taken from
climatology (Remedios, ULeic) ECMWF temperature,
pressure and water vapour profiles A priori
albedo - inferred from SCIAMACHY radiance as a
f(SZA) A priori aerosol (maritime/rural/urban)
SCIAMACHY Spectra, geolocation, viewing geometry,
time
Raw Spectra
Process only if cloud free, forward scan, SZA
75?
SCIATRAN (Courtesy of IUP/IFE Bremen) LBL mode,
HITRAN 2004
Calibration Non-linearity, dark current, ppg
etlaon
I - Calibrated Spectra
I0 Frerick (ESA)
Reference Spectrum weighting functions (CO2,
H2O and temperature)
SCIAMACHY Spectrum (I/I0)
WFM-DOAS fit
CO2 Column (Normalise with ECMWF Surface
Pressure) Accept only Errors lt5, Range340-400
ppmv
Note No scaling of FSI data
6
FSI Spectral fits

• Good fit residuals
• Retrievals errors
• Over land 1-4
• Over oceans 1-25

7
(No Transcript)
8
(No Transcript)
9
North-South Gradients
10
NH vs. SH
11
(No Transcript)
12
Validation Summary

• FTIR
• Park Falls ? -2
• Egbert ? -4
• TM3
• Bias -2
• SCIAMACHY overestimates seasonal cycle by factor
  2-3 with respect to the TM3 reason?
• Bias of TM3 w.r.t Egbert FTIR data -2
• Aircraft collocated observations in time
  space
• Sites over Siberia (r2 gt 0.72-0.9)
• Best at 1.5 km
• Surface Sites - monthly averages
• Time series comparisons (inc. aircraft)
• Out of 17, 11 have r2 gt 0.7

13
(No Transcript)
14
(No Transcript)
15
Can we learn anything?

• Greater CO2 uptake by forests compared to crops
  grass plains?
• Identification of sub-continental CO2
  sources/sinks ?

16
Vegetation Indices vs. CO2
17
GPP (or Carbon Tracker/NAEI/JULES/ ECOSSE
18
(No Transcript)
19
(No Transcript)
20
?CO2 over target area
CO2 mole fractions ending up into the release
domain. These concentrations describe the overall
CO2 collected by the air from the Earth's surface
through its way to the release areas.
21
CO2 over target area
absolute CO2 concentration of the release
area exchanged CO2 concentration data
comparable with those from satellite
22
Background CO2 (A)
CO2 at release location is obtained from weighted
background CO2 plus a weighted flux
contribution. In simplified diagram, the thick
arrow from direction C represents greatest
surface contact time, and the greatest
contribution of the background CO2.
Land type A
Release Location
Land type B
Land type C
Background CO2 (C)
23
Land area is divided into a small grid. For each
grid square the surface influence is multiplied
by the strength of flux (from e.g. carbon
tracker). Summing up all of the grid squares
produces a delta CO2. Adding these to the
weighted background, gives an atmospheric
concentration, that is compared to the satellite.
Strong absorber
Compare to satellite
Weak emitter
Weak absorber
24
Monthly mean (left) Standard Deviation (right)
of CO2 fluxes from Carbon Tracker, over North
America July 2003.
25
Surface influence (time spent at lowest altitude
grid) of air released at 40-50N, 90-100W on 14th
July 2003.
26
Retrieved FSI CO2 columns over North America on
the 2nd July 2003.
27
(No Transcript)
28
Summary

• Encouraging first results from the FSI algorithm
• Good agreement with FT stations at Egbert Park
  Falls (bias -2 to -4)
• Good agreement with TM3 model (more uptake in
  summer)
• Good agreement with aircraft data over Siberia
• Good agreement with AIRS CO2 (accounting for
  different vertical sensitivity)
• Good agreement with surface network
• Good correlation with vegetation spatial
  distribution
• Key point SCIAMACHY can track changes in near
  surface CO2
• Comparison of vegetation indices vs. CO2 indicate
  SCIAMACHY can track biological signal at regional
  level (though at limit of sensitivity)
• Can we measure fluxes
• It is clear that there is information in the
  SCIAMACHY data on the biospheric fluxes
• Preliminary estimates seem to show a quantitative
  correlation with carbon tracker when weighted for
  surface contact time
• Further work is required and underway
• Outlook - Very encouraging for GOSAT and OCO
• Measuring CO2 with a non-ideal instrument
  primitive algorithm

29
Thanks also to

• John Burrows and V. Rozanov
• IUP/IFE University of Bremen
• Alastair Manning and Claire Witham
• Met Office, UK

30
Flux datasets

• Carbon Tracker biosphere, sea, fossil fuel, fire
  and overall
• JULES vegetation (Joerg Kaduk, Geography,
  Leicester)
• ECOSSE soil (Jo Smith, Aberdeen)
• NAEI anthropogenic emissions

JULES Gross Primary Production (GPP) rate at
which an ecosystem produces, captures and stores
chemical energy as biomass. Average over 5 years
1999-2003. Units kgC/m2/s.