1. Enza Di Tomaso* and Niels Bormann ECMWF *EUMETSAT fellow
Assimilation of ATOVS radiances at ECMWF: Bias correction and impact in NWP
Enza Di Tomaso* and Niels Bormann
ECMWF
*EUMETSAT fellow

2. Enza Di Tomaso* and Niels Bormann ECMWF *EUMETSAT fellow
Assimilation of ATOVS radiances at ECMWF: Bias correction and impact in NWP
Enza Di Tomaso* and Niels Bormann
ECMWF
*EUMETSAT fellow

3. Assimilated ATOVS radiances
HIRS: channel 4-7, 11, 14, 15 over sea; 12 over sea + low orography only
AMSU-A: channels 5,6 over sea + low orography; 7-14 land+sea
AMSU-B/MHS: channel 5 over sea only; 3,4 sea+low orography
HIRS
( 3 used)
AMSU-A
(5 used)
AMSU-B/MHS
(3 used)
NOAA-15
no: unstable
yes
(not ch 6, 11, 14)
no: quality
NOAA-17
Instrument failed
no (since Dec 09)
NOAA-18
NOAA-19
yes (not ch 8)
yes (not ch 3)
AQUA
n/a
(not ch 5 & 7;
6 over sea only)
METOP-A
(not ch 7)

4. Assimilated ATOVS radiances
HIRS: channel 4-7, 11, 14, 15 over sea; 12 over sea + low orography only
AMSU-A: channels 5,6 over sea + low orography; 7-14 land+sea
AMSU-B/MHS: channel 5 over sea only; 3,4 sea+low orography
HIRS
( 3 used)
AMSU-A
(5 used)
AMSU-B/MHS
(3 used)
NOAA-15
no: unstable
yes
(not ch 6, 11, 14)
no: quality
NOAA-17
Instrument failed
no (since Dec 09)
NOAA-18
NOAA-19
yes (not ch 8)
yes (not ch 3)
AQUA
n/a
(not ch 5 & 7;
6 over sea only)
METOP-A
(not ch 7)

5. Assimilated ATOVS radiances
HIRS: channel 4-7, 11, 14, 15 over sea; 12 over sea + low orography only
AMSU-A: channels 5,6 over sea + low orography; 7-14 land+sea
AMSU-B/MHS: channel 5 over sea only; 3,4 sea+low orography
Part 1
HIRS
( 3 used)
AMSU-A
(5 used)
AMSU-B/MHS
(3 used)
NOAA-15
no: unstable
yes
(not ch 6, 11, 14)
no: quality
NOAA-17
Instrument failed
no (since Dec 09)
NOAA-18
NOAA-19
yes (not ch 8)
yes (not ch 3)
AQUA
n/a
(not ch 5 & 7;
6 over sea only)
METOP-A
(not ch 7)

6. Assimilated ATOVS radiances
HIRS: channel 4-7, 11, 14, 15 over sea; 12 over sea + low orography only
AMSU-A: channels 5,6 over sea + low orography; 7-14 land+sea
AMSU-B/MHS: channel 5 over sea only; 3,4 sea+low orography
Part 1
HIRS
( 3 used)
AMSU-A
(5 used)
AMSU-B/MHS
(3 used)
NOAA-15
no: unstable
yes
(not ch 6, 11, 14)
no: quality
NOAA-17
Instrument failed
no (since Dec 09)
NOAA-18
NOAA-19
yes (not ch 8)
yes (not ch 3)
AQUA
n/a
(not ch 5 & 7;
6 over sea only)
METOP-A
(not ch 7)
Part 2

7. Assimilated ATOVS radiances
HIRS: channel 4-7, 11, 14, 15 over sea; 12 over sea + low orography only
AMSU-A: channels 5,6 over sea + low orography; 7-14 land+sea
AMSU-B/MHS: channel 5 over sea only; 3,4 sea+low orography
Part 1
HIRS
( 3 used)
AMSU-A
(5 used)
AMSU-B/MHS
(3 used)
NOAA-15
no: unstable
yes
(not ch 6, 11, 14)
no: quality
NOAA-17
Instrument failed
no (since Dec 09)
NOAA-18
NOAA-19
yes (not ch 8)
yes (not ch 3)
AQUA
n/a
(not ch 5 & 7;
6 over sea only)
METOP-A
(not ch 7)
Part 2

8. Enza Di Tomaso* and Niels Bormann ECMWF *EUMETSAT fellow
Assimilation of ATOVS radiances at ECMWF: Bias correction and impact in NWP (Part 1)
Enza Di Tomaso* and Niels Bormann
ECMWF
*EUMETSAT fellow

9. Part 1: revision of AMSU-A bias correction
Bias correction of ch12 & ch14 (Part 1a)
AMSU/A
(from

10. Part 1: revision of AMSU-A bias correction
Bias correction of ch12 & ch14 (Part 1a)
Bias correction of ch5 to 8 (Part 1b, ongoing work)
AMSU/A
(from

11. Part 1: revision of AMSU-A bias correction
Bias correction of ch12 & ch14 (Part 1a)
Bias correction of ch5 to 8 (Part 1b, ongoing work)
Assimilation of surface-sensitive channels (future work)
AMSU/A
(from
(by Tom Greenwald)

12. Bias correction of ch 12 & 14: interaction between forecast model error and bias correction
T511 experiment (black) versus T255 experiment(red)
Radiosonde T
N.Hemis
Issues with high spatial model resolution: radiosondes show resolution-dependent temperature biases in the stratosphere
T1279 experiment (black) versus T255 experiment(red)
Radiosonde T
N.Hemis

13. Experiment description
Revision of the bias correction of AMSU-A stratospheric channels peaking where the forecast model error is particularly significant
“noBC experiment”: no bias correction applied to AMSU-A ch12 and ch14
“sBC experiment”: scan bias correction (polynomial in the scan angle and with no constant) applied to AMSU-A ch12 and ch14
“N19 anchor experiment”:
scan bias correction (with no constant) applied to AMSU-A ch12 and ch14 on NOAA-19
scan bias and offset correction applied to AMSU-A ch12 and ch14 on other satellites
Experiments were run over ‘summer’ (20 Jul – 31 Oct 2009) and ‘winter’ (6 Dec – 31 Mar 2010) at T511 resolution

14. Experiment description
Revision of the bias correction of AMSU-A stratospheric channels peaking where the forecast model error is particularly significant
“noBC experiment”: no bias correction applied to AMSU-A ch12 and ch14
“sBC experiment”: scan bias correction (polynomial in the scan angle and with no constant) applied to AMSU-A ch12 and ch14
“N19 anchor experiment”:
scan bias correction (with no constant) applied to AMSU-A ch12 and ch14 on NOAA-19
scan bias and offset correction applied to AMSU-A ch12 and ch14 on other satellites
Experiments were run over ‘summer’ (20 Jul – 31 Oct 2009) and ‘winter’ (6 Dec – 31 Mar 2010) at T511 resolution

15. Departure statistics of the first guess and analysis
“noBC experiment” (black) versus control (red)
“noBC experiment” BC (pink) versus control BC (green)
Radiosonde T
N.Hemis
No bias correction of AMSU-A ch12 ad ch14 improves the fit to temperature observations
MetOp AMSU-A TB

16. Comparison with the SPARC climatology
“noBC experiment” minus control
control minus climate

17. Forecast impact
“noBC experiment” versus control (verified against observations), summer
The impact for the forecast of the 50hPa geopotential of the “noBC experiment” is positive in the extra-Tropics
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

18. Forecast impact
“noBC experiment” versus control (verified against observations), winter
The impact for the forecast of the 50hPa geopotential of the “noBC experiment” is positive in the extra-Tropics
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

19. Forecast impact
“noBC experiment” versus control (verified against own-analysis), summer
The impact for the forecast of the 500hPa geopotential of the “noBC experiment” is slightly negative in the Southern Hemisphere
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

20. Forecast impact
“noBC experiment” versus control (verified against own-analysis), winter
The impact for the forecast of the 500hPa geopotential of the “noBC experiment” is slightly negative in the Northern Hemisphere
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

21. Experiment description
Revision of the bias correction of AMSU-A stratospheric channels peaking where the forecast model error is particularly significant
“noBC experiment”: no bias correction applied to AMSU-A ch12 and ch14
“sBC experiment”: scan bias correction (polynomial in the scan angle and with no constant) applied to AMSU-A ch12 and ch14
“N19 anchor experiment”:
scan bias correction (with no constant) applied to AMSU-A ch12 and ch14 on NOAA-19
scan bias and offset correction applied to AMSU-A ch12 and ch14 on other satellites
Experiments were run over ‘summer’ (20 Jul – 31 Oct 2009) and ‘winter’ (6 Dec – 31 Mar 2010) at T511 resolution

22. Departure statistics of the first guess and analysis
“sBC experiment” (black) versus “noBC experiment” (red)
“sBC experiment” BC (pink) versus “noBC experiment” BC (green)
MetOp-A AMSU-A TB
S.Hemis
The bias correction of AMSU-A ch12 (and ch14) onboard NOAA-18 is not adequately correcting the scan bias, as it tries to correct for
inter-satellite biases
NOAA-18 AMSU-A TB
S.Hemis

23. Experiment description
Revision of the bias correction of AMSU-A stratospheric channels peaking where the forecast model error is particularly significant
“noBC experiment”: no bias correction applied to AMSU-A ch12 and ch14
“sBC experiment”: scan bias correction (polynomial in the scan angle and with no constant) applied to AMSU-A ch12 and ch14
“N19 anchor experiment”:
scan bias correction (with no constant) applied to AMSU-A ch12 and ch14 on NOAA-19
scan bias and offset correction applied to AMSU-A ch12 and ch14 on other satellites
Experiments were run over ‘summer’ (20 Jul – 31 Oct 2009) and ‘winter’ (6 Dec – 31 Mar 2010) at T511 resolution

24. Forecast impact
“N19 anchor experiment” versus control (verified against own-analysis), summer
The impact for the forecast of the 500hPa geopotential of the “noBC experiment” is neutral also in the Southern Hemisphere
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

25. Forecast impact
“N19 anchor experiment” versus control (verified against own-analysis), winter
The impact for the forecast of the 500hPa geopotential of the “noBC experiment” is neutral also in the Northern Hemisphere
control
GOOD
“noBC experiment”
“noBC experiment” RMSE – control RMSE

26. Departure statistics of the first guess and analysis
“N19 anchor experiment” (black) versus “noBC experiment” (red)
“N19 anchor exp.” BC (pink) versus “noBC experiment” BC (green)
MetOp-A AMSU-A TB
S.Hemis
The bias correction of AMSU-A ch12 (and ch14) onboard NOAA-18 is now adequately correcting the scan bias
NOAA-18 AMSU-A TB
S.Hemis

27. Conclusions of part 1a
We considered a revision of the bias correction of high stratospheric channels because of the interaction between the variational bias correction scheme (VarBC) and large forecast model biases in the upper atmosphere
no bias correction of channels 12 and 14 has some negative forecast impact
scan bias correction alone is affected by inter-satellite biases
using one satellite as anchor for the others offers improvements to the previous solutions

28. Bias correction of ch5 to 8: gamma-delta correction
The observed bias is modelled with a constant fractional error in the optical depth (gamma) and a global constant (delta):
Bias = offset + bias due to errors in the channel transmittance
Gamma coefficients are currently used in the radiative transfer up to NOAA-18 (not for NOAA-19 and MetOp-A), (work by P. Watts & A. McNally)
Sources of error in the channel transmittance (not necessarily constant):
errors in the assumed gas concentration
errors in the absorption coefficient
inaccurate channel spectral response function

29. Mean first guess departures with different gamma values
control experiment (gamma = 1)
“gamma experiment” (gamma = 1.05)

30. Conclusions of part 1b
Values of gamma have been estimated for AMSU-A channels 5 to 8
Experiments are running to show
the impact of the updated gamma values for all AMSU-A
whether gamma can correct air-mass dependent biases without the need of specific predictors in VarBC for channels 5 to 8

31. Assimilation of ATOVS radiances at ECMWF: Bias correction and impact in NWP (Part 2)
Enza Di Tomaso* and Niels Bormann
ECMWF
*EUMETSAT fellow
Thanks to Alan Geer for the IVER package

32. Part 2: orbit constellation OSEs
characterise the benefit for NWP of having ATOVS data from three evenly-spaced orbits versus data from a less optimal coverage
assess the benefit for NWP of assimilating ATOVS data from more than three satellites
Satellite equatorial crossing times (local)
MetOp-A
NOAA-17 T i m e
NOAA-16
NOAA-15
NOAA-18
NOAA-19
Aqua

33. Data coverage Sample coverage from a 6-hour period around 0Z
“NOAA-15 experiment”
* MetOp-A * NOAA * NOAA-15
“two-satellite experiment”
* MetOp-A * NOAA-18
“NOAA-19 experiment”
* MetOp-A * NOAA * NOAA-19

34. Experiment description
“no-MW sounder experiment”: no AMSU-A and AMSU-B/MHS were assimilated
“two-satellite experiment”: AMSU-A and AMSU-B/MHS on MetOp-A and NOAA-18 were assimilated
“three-satellite experiments”:
“NOAA-15 experiment”: AMSU-A data were added from a third satellite NOAA-15
“NOAA-19 experiment”: AMSU-A data were added from a third satellite NOAA-19
“all-satellite experiment”: all available ATOVS observations were assimilated
The above set of experiments was run also in the case in which the advanced sounder instruments IASI and AIRS were denied
Experiments were run over more than three months (14 April 2009 to 4 August 2009)
at T255 resolution

35. Departure statistics of the first guess and analysis
“three-satellite experiment” (black) versus “two-satellite experiment” (red)
Radiosonde T
Tropics
Both NOAA-15 and NOAA-19 bring some small improvement to the fit to temperature observations
“NOAA-15 experiment” (black) versus “NOAA-19 experiment” (red)
MetOp AMSU-A TB
Departure statistics for MetOp-A AMSU-A show some benefits from assimilating observations from NOAA-15 rather than NOAA-19

36. Forecast impact
“NOAA-19 experiment”
GOOD
“NOAA-15 experiment”
When averaged over the extra-Tropics the impact for the forecast of the geopotential of “NOAA-15 experiment” versus “NOAA-19 experiment” is neutral to slightly positive
“NOAA-15 exp” RMSE – “NOAA-19 exp” RMSE

37. Forecast impact
“no-MW sounder experiment”
GOOD
“two-”, “three-”, “all-satellite experiment”
Both the assimilations of NOAA-15 and NOAA-19 data have a clearly positive forecast impact in the Southern Hemisphere compared to the use of two satellites only
Having ATOVS-like data from more than three satellites adds further benefit in terms of the forecast impact
“two-satellite” RMSE – “no-Mw sounder” RMSE
“three-satellite” RMSE – “no-Mw sounder” RMSE
“all-satellite” RMSE – “no-Mw sounder” RMSE

38. Forecast impact
“no-MW sounder experiment”
GOOD
“two-”, “three-”, “all-satellite experiment”
When IASI and AIRS are denied, the results show in general a stronger positive impact when additional ATOVS data are assimilated into the NWP system
“two-satellite” RMSE – “no-Mw sounder” RMSE
“three-satellite” RMSE – “no-Mw sounder” RMSE
“all-satellite” RMSE – “no-Mw sounder” RMSE

39. Less thinning of data
Comparing “three-satellite experiments” with a new “two-satellite experiment” where less data are removed
less thinning of AMSU-A data
additional field of view on each side of the scan

40. “two-satellite experiment
Forecast impact
“three-satellite experiment” versus “two-satellite experiment (less thinning)” (verified against operational analysis)
“NOAA-15 experiment”
GOOD
“two-satellite experiment
(less thinning)”
“NOAA-15 exp” RMSE – “two-satellite (less thinning)” RMSE
There is still some advantage in using three AMSU-A rather than two

41. Conclusions of part 2
ATOVS data in a more evenly-spaced orbit configuration give slightly better results in terms of forecast impact in the Southern Hemisphere than data from a less optimal coverage
Both the assimilations of NOAA-15 and NOAA-19 observations have a positive forecast impact in the Southern Hemisphere in comparison to the use of just two satellites, and there is a clear advantage in assimilating all available ATOVS data
The benefit of evenly-spaced orbits is expected to be stronger in limited area systems where the coverage plays a more crucial role

42. Danke und Frohe Weihnachten!

43. Additional slides: gamma-delta correction
Watts and McNally

44. Modelling absorption coefficient errors

45. Estimating gamma

46. Additional slides: variational bias correction (VarBC)
Dick Dee and Niels Bormann

47. Variational analysis and bias correction: A brief review of variational data assimilation
Minimise
background constraint (Jb)
observational constraint (Jo)
The input xb represents past information propagated by the forecast model
(the model background)
The input [y – h(xb)] represents the new information entering the system
(the background departures - sometimes called the innovation)
The function h(x) represents a model for simulating observations
(the observation operator)
Minimising the cost function J(x) produces an adjustment to the model background based on all used observations
(the analysis)

48. Variational analysis and bias correction: Error sources in the input data
Minimise
background constraint (Jb)
observational constraint (Jo)
Errors in the input [y – h(xb)] arise from:
errors in the actual observations
errors in the model background
errors in the observation operator
There is no general method for separating these different error sources
we only have data about differences
there is no true reference in the real world
The analysis does not respond well to contradictory input information
A lot of work is done to remove biases prior to assimilation:
ideally by removing the cause
in practise by careful comparison against other data

49. Past* scheme for radiance bias correction at ECMWF
Scan bias and air-mass dependent bias for each sensor/channel were estimated off-line from
background departures, and stored on files (Harris and Kelly 2001)
Error model for brightness temperature data:
where
Periodically estimate scan bias and predictor coefficients:
typically 2 weeks of background departures
2-step regression procedure
careful masking and data selection
Average the background departures:
Predictors, for instance:
hPa thickness
hPa thickness
surface skin temperature
total precipitable water
*Replaced in operations September 2006 by VarBC (Variational Bias Correction)

50. The need for an adaptive bias correction system
The observing system is increasingly complex and constantly changing
It is dominated by satellite radiance data:
biases are flow-dependent, and may change with time
they are different for different sensors
they are different for different channels
How can we manage the bias corrections for all these different components?
This requires a consistent approach and a flexible, automated system

51. Variational bias correction: The general idea
The bias in a given instrument/channel is described by (a few) bias parameters:
typically, these are functions of air-mass and scan-position (the predictors)
These parameters can be estimated in a variational analysis along with the model state
(Derber and Wu, 1998 at NCEP, USA)
The standard variational analysis minimizes
Modify the observation operator to account for bias:
Include the bias parameters in the control vector:
Minimize instead
What is needed to implement this:
The modified operator and its TL + adjoint
A cycling scheme for updating the bias parameter estimates
An effective preconditioner for the joint minimization problem

52. Variational bias correction: The modified analysis problem
The original problem:
Jb: background constraint
Jo: observation constraint
Jb: background constraint for x
J: background constraint for 
Jo: bias-corrected observation constraint
The modified problem:
Parameter estimate
from previous analysis

53. Limitations of VarBC: Interaction with model bias
VarBC introduces some extra degrees of freedom in the analysis, to help improve the fit to the (bias-corrected) observations:
This works well where
the analysis is well-constrained by observations, and
“anchoring” observations are available (e.g.,
radiosondes, GPSRO data).
VarBC will correct any biased observations and produce a consistent consensus analysis.
model
abundant observations
This may lead to undesired effects where
model bias is present, and
few observations are available, or
only observations with VarBC are present.
VarBC will, over time, force agreement with the model background.
model
observations
VarBC may wrongly attribute model bias to the observations