Assimilation of satellite data on atmospheric composition: effective collaboration between academicr

1
Assimilation of satellite data on atmospheric
composition effective collaboration between
academic/research institutions and operational
centres

• Stefano Migliorini
• NCEO, University of Reading
• in collaboration with R. Dragani (ECMWF)
• Thanks to A. ONeill (NCEO), Zofia Stott and Jon
  Styles (Assimila)

2
Contents

• Overview of the ESA-funded GlobModel project
• Summary of the results of the Study
• Experiments set up
• Assimilated dataset
• Assimilation results
• Conclusions

3
The GLOBMODEL project

• ESA Data User Element (DUE) programme aims to
  strengthen the use of EO data by the Earth system
  modelling community and fully exploit investments
  in EO infrastructure
• The objective of the GlobModel project are
• To determine and prioritise the steps to be taken
  in model-EO data fusion
• To stimulate knowledge transfer between science
  and operational applications and encourage ESA
  Member States to continue their support to EO
  missions

4
GlobModel structure

• Two parts Study and Demonstration
• The Study investigates the current and potential
  use of EO data within the Earth system modelling
  community, through a consultation
• Its outcome is a set of recommendations to ESA
  for better exploiting EO data
• The Demo illustrates issues raised in the Study
• Focused on assimilation of SCIA and OMI ozone in
  the stratosphere and tropospheric ozone and NO2
  for air quality applications

5
Summary of GlobModel Study objectives

• How can EO data play a more prominent role in
  Earth system modelling for scientific and
  operational applications ?
• What are the priority applications for future
  efforts ?
• What future role should ESA and other European
  institutions play?
• How should ESAs activities be targeted ?

6
GlobModel Study results priorities for action

• Making operational systems and data more readily
  available for research for the mutual benefit of
  scientists and the operational agencies.
• Pay attention to early development of observation
  operators to link observations and models
• Improve ocean data assimilation, in particular to
  ascertain the impact of different data sources
• Improve chemical data assimilation with focus on
  air pollution forecasts
• Develop the use of EO data in the land component
  of Earth system models
• Providing long term funding for re-analysis
  projects
• Expand the use of Observing System Experiments
  and Observing System Simulation Experiments to
  improve the quantification of the benefits of EO
  data in modelling

7
Scope of the Demonstration

• Carry out assimilation experiments with an
  operational NWP model and remote sensing data
  that are not routinely assimilated by NWP centres
• Focus on stratospheric ozone and tropospheric
  ozone and NO2
• Show that a collaboration between operational
  Centres and research institutions enhances the
  impact of satellite data

8
Demonstration components
9
The DARC/ECMWF experiments set up

• Use of the current operational ECMWF suite
  (CY32R3) at T255 truncation ( 100 km), and L60
  from the surface up to 0.1 hPa.
• The data assimilation is performed using the
  4D-Var scheme, with two main 12-hour analysis and
  first-guess forecast cycles for 00 (2100-0900)
  and 12 (0900-2100) Z.
• Case study between 1 July 2006 00 UTC and 31
  August 2006 18 UTC

10
Experiments configuration

• Control experiment (CTRL) assimilation of
  standard meteorological data and ozone total
  column data from SCIAMACHY
• First perturbation experiment (OMIT)
  assimilation of standard meteorological data and
  ozone total column data from SCIAMACHY and OMI
• Second perturbation experiment (OMIP)
  assimilation of standard meteorological data,
  ozone total column data from SCIAMACHY and ozone
  profiles from OMI

11
Standard assimilation data set

• Routinely assimilated data set in a 12-hour
  assimilation window

12
SCIAMACHY total ozone product

• Algorithm DOAS (Differential Optical Absorption
  Spectroscopy)
• algorithm based on the OMI algorithm combined
  with cloud information from FRESCO
• Version TOSOMI v0.42
• Product - Total ozone column in near-real time
  (within 3 hours)
• - NRT overpass files of about 128 ground
  stations
• Coverage Global coverage in 6 days
• Resolution 80x40 km
• Delivery via FTP and web-access in near-real
  time
• Archive Archive of images and data sets for
  July 2002-present

13
OMI ozone column product

• Retrieval method DOAS
• Resolution 13x24 km (at nadir)
• Global coverage in one day
• Known issues
• Across track 30 (0-based) is noisier than other
  pixels for inhomogeneous scenes.
• Since the end of June 2007, across track pixels
  53 and 54 (0-based) show anomalous behavior.

14
Ozone profile algorithm

• Optimal estimation (Rodgers)
• A-priori Fortuin Kelder climatology
• Forward model on-line radiative transfer (new
  Labos code)
• Non-linear gt iterative solution (start with
  a-priori)
• Use O2-O2 cloud product for cloud pressure and
  (initial) fraction
• Wavelength region 270 330 nm (UV1 UV2)
• Output ozone layer column for 18 layers,
  a-priori,averaging kernel, DFS, error
  covariance, diagnostics, meta-data

15
First-iteration experiments

• CTRL, OMIT and first version of OMIP
  (assimilation of standard meteorological
  observations, ozone total columns from SCIAMACHY
  and ozone partial columns from OMI over 18 levels)

16
Second-iteration experiments

• The two-month long OMIP experiment was repeated
  with OMI profile data aggregated over 6 layers
• More consistent observation operator especially
  between 70 and 5 hPa

17
Assimilation of OMI total column ozone
18
OMI TCO contribution on the O3 analyses
Daily zonal mean O3 difference (CTRL-OMIT)
Daily TCO difference (CTRL-OMIT)
19
Comparisons against ozone sondes

• Negligible differences were found in the
  comparisons with ozone sondes over the
  midlatitudes.

20
Comparison with MLS ozone profiles
Fit of the ECMWF O3 analyses to MLS ozone
profiles improved of 2-3 in OMIT
21
Comparison with HIRDLS ozone profiles
90S-60S
60S-30S
30S-30N
90N-60N
60N-30N
22
Assimilation of OMI ozone profiles
23
OMI O3 profile contribution on the O3 analyses
Daily zonal mean O3 difference (CTRL-OMIP)
Daily TCO difference (CTRL-OMIP)
24
Comparisons with ozone sondes
High latitudes (SH)
Tropics
Midlatitudes
25
Global mean comparisons with MLS and HIRDLS O3
profiles
HIRDLS
MLS
26
Comparisons with HIRDLS O3 profiles
a)
c)
b)
f)
90S-60S
60S-30S
30S-30N
90N-60N
60N-30N
27
Summary of Demonstration results

• We have shown so far that working together the
  scientific and operational communities can
  achieve important scientific benefits, especially
  in the exploitation of observations that have not
  been assimilated before.
• Comparison between the DARC and ECMWF results
  shows that
• The assimilation of O3 profiles leads to larger
  differences between perturbation and control
  experiments than the assimilation of TCO.
• OMI and SCIAMACHY mutually biased (
  OMIltSCIAMACHY, Bias 5DU)
• Problems were found at high latitudes in the SH
  when assimilating profiles
• Approximation of the observation operator to a
  box-car when assimilating profiles.
• OMI is a UV sensor ? cannot provide measurements
  in the winter hemisphere
• The two sets of experiments performed by DARC and
  ECMWF showed that OMI ozone data can provide
  useful information to improve the quality of the
  ozone analyses in the tropical lower stratosphere
  and troposphere, as well as at high latitudes in
  SH when assimilating TCO.

28
Conclusions

• Collaboration between research and operational
  centres has proved to be successful in
• Accelerating scientific and operational
  utilization of new EO data (e.g., OMI ozone
  profiles)
• Showing that assimilation of OMI ozone total
  column data leads to a better knowledge of the
  ozone field
• Providing feedbacks to data providers on the
  quality of their products before their public
  release (e.g. over polar regions)