1. Régis Borde (regis.borde@eumetsat.int)
Polar Winds
EUMETRAIN
Polar satellite week 2012
Régis Borde

2. After the Wind Workshop
North Cape, NZ
After the Wind Workshop
Feb 2012
Content
Wind extraction from satellite imagery
Polar winds

3. Content
Wind extraction from satellite imagery
Polar winds at EUMETSAT

4. Why do we care about winds from satellites?
For best results, models require information on both the mass field and the wind field.
AMVs are the only observation type to provide good coverage of upper tropospheric wind data over oceans and at high latitudes.
Sondes and wind profilers
Aircraft
For the AMVs each dot represents a single level wind not a wind profile
AMVs

5. Atmospheric Motion Vectors
AMV production centres:
EUMETSAT
NOAA-NESDIS
CIMSS
JMA
CMA
KMA
AMVs from Geo. Sat.:
Meteosat 7, 8 and 9
GOES 13, 15
FY2E
MTSAT-2
AMVs from Polar Sat.:
Metop-A (AVHRR)
Terra, Aqua (MODIS)
NOAA- 15, 16, 18 and 19 (AVHRR)

6. Polar winds Why doing polar winds ?
Assimilation at ECMWF in 2002
12-h sample coverage: used AMVs
at ECMWF in 2012
GOES-13 MET-9 MET-7 MTSAT-2
TERRA AQUA
GOES-15 AMVs under evaluation since December 2011
Lack of observations in
Polar regions

7. Polar winds How doing polar winds ?
Needs:
Low orbit polar satellites: NPP, METOP...
Appropriate instruments: MODIS, AVHRR, VIIRS.
But some new challenges:
Large timeliness (~100 min)
Small areas to track features
Problems of view angles, parallax and varying pixels sizes
No cloudy product (AVHRR)
Polar region specificities like ground colder to air above
...

8. Polar winds Product history
Courtesy of Dave Santek, CIMSS

9. Courtesy of Dave Santek, CIMSS
Unlike geostationary satellites at lower latitudes, it is not be possible to obtain complete polar coverage at a snapshot in time with one or two polar-orbiters. Instead, winds must be derived for areas that are covered by two or three successive orbits, an example of which is shown here. The gray area is the overlap between three orbits.

10. One Day of Arctic Orbits
Courtesy of Dave Santek, CIMSS
MODIS band 31 (11 mm)

11. EUMETSAT Atmospheric Motion Vectors
Algorithm description
Target selection
Tracking
Height assignment
Automatic quality control
Extraction of the motion (speed, direction)
Selection of the pixels
used to set the altitude
Height Assignment QI

12. Two main assumptions and ’limitations’
EUMETSAT Atmospheric Motion Vectors
Physical assumptions and limitations
Two main assumptions and ’limitations’
Feature tracked travels at the exactly same speed and direction than the local wind.
Detected motion represents the ‘cloud top’ motion. Then CTH methods are used to set the altitude

13. Extraction of AMV speed and direction
Target identification and tracking
Initial corrections (image navigation etc.)
Search Area
80 x 80 pixels centred on target box
Target Box / Tracer
24x24 pixels
Pixel – 3 km
New location determined by best match of individual pixel counts of target with all possible locations of target in search area. T
T + 15 min
Infrared Imagery
Need to assign a height to the derived vector

14. AMV Height Assignment Opaque clouds, EBBT B: Energy emitted
T: Blackbody temperature
k: Boltzmann‘s constant
h: Planck‘s constant
c: speed of light
l: wavelenght

15. Pc AMV Height Assignment Best fit
Semi-transparent clouds, CO2 slicing method
Best fit
Rcd : Measured cloudy radiance
Rcf : Measured cloud free radiance
Rbcd : Calculated Planck blackbody radiance for a cloud at level Pc
Rsuf : Calculated clear air radiance
e : Emissivities Pc

16. AMV Quality check Several independent quality test applied
Forecast consistency
Spatial consistency
Temporal consistency (vector, speed, direction)
Temporal pressure consistency
Image correlation (WV 6.2 / IR 10.8)
Final quality index (QI)
Weighted mean of the individual tests

17. AMV Quality check Example of Spatial consistency
red wind vector would not pass the spatial consistency test, it would therefore get a low quality indicator (in a range from 0 to 1)

18. How check the quality of AMVs ?DU
Speed bias
FC Speed at AMV pressure level
AMV pressure DP
Best Fit level pressure
AMV
AMV Speed

19. Collocation AMV / RS RS altitude RadioSonde AMV altitude AMV Pressure
150 km
RadioSonde
RS altitude
AMV
altitude
50 hPa
AMV
Pressure
CGMS Reference
HOR. DIST. < 150 (Km)
VERT. DIST. < 25 (hPa)
QUALITY (Without FC test) >= 80
SPEED DIFF. < 30 (m/s)
DIRECTION DIFF. < 60 (deg)
AMV SPEED > 2.5 (m/s)
Longitude
Latitude

20. EUMETSAT Polar winds Image pairs or triplet ?
increase the coverage
loose temporal quality checks between derived vectors

21. EUMETSAT Polar winds Image pairs or triplet ?
Example for retrieved polar cap winds for the Arctic on 19 January 2010.

22. EUMETSAT Polar winds How to set the AMV heights with AVHRR ?
Main problems:
No WV or CO2 channels on AVHRR
No cloudy product
Then...
Use IR window channel (EBBT method) for opaque clouds
No clean solution for semi-transparent clouds
Possibility to use low-level correction from model fields

23. EUMETSAT Polar winds Opportunity to use IASI heights
24x24
AVHRR pixels
24x24 AVHRR pixels target boxes
IASI foot print sometimes in the box, sometimes not.
Collocation IASI foot print /feature tracked

24. EUMETSAT Polar winds Use of IASI heights, Validation
3 day period Nov 2011 : Z and Z 3 hour forecast data
Use of IASI co-located heights improves a bit wind quality
ARCTIC All Levels
QI > 80
ANTARCTIC All Levels
OLD
NEW
All
IASI
Speed Bias (m/s)
1.14
0.75
0.80
2.57
1.58
1.72
Speed RMS (m/s)
5.62
4.81
5.11
5.58
4.26
4.53
Direction Bias (deg)
0.30
-0.51
-1.91
-4.69
0.04
-0.06
Direction RMS (deg)
14.95
28.20
15.71
19.28
14.55
11.15
NRMS
0.23
0.20
0.16
0.27
0.22

25. from initial target box
EUMETSAT Polar winds Necessity to use the first guess !
Altitude estimated
from initial target box
Main problems:
Timeliness between consecutives images is very long, which makes the tracking difficult and noisy.
This problem is emphasised when using only pair of images
Use of the first guess improves the results, but:
Needs to estimate the altitude first
Refers to the model fields
Source of potential errors if first estimation of altitude is wrong
Speed retrieved in the model fields that is then used to locate the search area

26. FORECAST FIRST GUESS (GS2) FORECAST FIRST GUESS (GS2)
EUMETSAT Polar winds Use of first guess, Validation
PROTOTYPE
FORECAST FIRST GUESS (GS2)
NO FORECAST (GS3)
Speed Bias (m/s)
-1.21
0.87
-1.45
Speed RMS (m/s)
2.67
4.50
8.21
Direction Bias (deg)
1.00
0.45
5.05
Direction RMS (deg)
8.60
15.65
59.40
Mean Speed AMV
19.47
21.52
16.90
Mean Speed Analysis
20.69
20.65
18.35
Sample size
3988
970
1035
Statistics for Arctic winds
Data of 19th January 2010,
12:00 UTC
PROTOTYPE
FORECAST FIRST GUESS (GS2)
NO FORECAST (GS3)
Speed Bias (m/s)
-0.10
1.72
-0.02
Speed RMS (m/s)
2.01
3.97
5.30
Direction Bias (deg)
0.54
2.25
11.39
Direction RMS (deg)
13.33
38.45
66.91
Mean Speed AMV
14.69
12.83
7.65
Mean Speed Analysis
14.79
11.11
7.66
Sample size
1503
393
947
Statistics for Antarctic winds

27. 5 day period May 2011 : 060000Z and 120000Z 6 hour forecast
EUMETSAT Polar winds Validation against ECMWF analysis - Arctic QI > 80
Low Level
(>=700 hPa)
Mid Level
( hPa)
High Level
(<=400 hPa)
All Levels
OLD
NEW
Speed Bias (m/s)
1.81
1.21
1.91
1.18
0.57
1.14
1.82
Speed RMS (m/s)
4.06
3.57
4.54
3.98
5.45
4.28
4.43
3.91
Direction Bias (deg)
0.44
2.51
0.66
0.49
2.26
-0.86
0.63
0.86
Direction RMS (deg)
16.00
18.41
12.77
13.75
10.96
10.04
13.83
14.70
Mean Speed AMV
16.64
15.43
21.62
20.21
29.23
26.51
20.33
19.53
Mean Speed Analysis
14.83
14.22
19.71
19.10
28.66
25.38
18.51
18.39
NRMS
0.27
0.25
0.23
0.21
0.19
0.17
0.24
Sample size
3002
2111
5909
5754
393
685
9268
8508
% of Winds 32 25 64 67 4 8
100
5 day period May 2011 : Z and Z 6 hour forecast

28. 5 day period May 2011 : 060000Z and 120000Z 6 hour forecast
EUMETSAT Polar winds Validation against ECMWF analysis - Antarctic QI > 80
Low Level
(>=700 hPa)
Mid Level
( hPa)
High Level
(<=400 hPa)
All Levels
OLD
NEW
Speed Bias (m/s)
1.93
1.32
1.43
0.97
-0.19
0.40
1.09
0.75
Speed RMS (m/s)
4.24
3.74
5.09
4.62
5.25
4.78
5.05
4.63
Direction Bias (deg)
-0.71
-1.82
0.18
0.01
-0.15
1.00
0.32
Direction RMS (deg)
12.81
11.31
13.96
15.50
14.55
17.12
14.00
15.98
Mean Speed AMV
18.42
17.82
24.72
23.68
27.81
28.53
24.82
25.37
Mean Speed Analysis
16.49
16.50
23.29
22.71
28.09
28.13
23.73
24.61
NRMS
0.26
0.23
0.22
0.20
0.19
0.17
0.21
Sample size
953
398
6554
2869
2237
2468
9714
5718
% of Winds 10 7 67 50 23 43
100
5 day period May 2011 : Z and Z 6 hour forecast

29. Polar winds EUMETSAT product from AVHRR, summary
Current status
AVHRR L1B IR-Window images from channel 4 (11µm)
For METOP-AVHRR full spatial resolution Local Area Coverage (LAC) data are available globally
Use image pairs
Use IASI heights when possible
Use first guess to locate search area for tracking
Future
Use triplet of images for METOP A and METOP B
Derive dual METOP winds (global coverage)

30. Polar winds Example of EUMETSAT AVHRR winds

31. AMV Impact on FC
24 h ‘Forecast Error Contribution’ sliced by specific AMVs observations, June 2011 (C. Cardinali, ECMWF)
Impact of different AMV products from
MODIS,
Meteosat-7, -8, -9,
GOES-11, -12

32. AMV Impact on FC
24 h ‘Forecast Error Contribution’ of the AMVs by instrument and for altitude intervals, June 2011 (C. Cardinali, ECMWF)
Impact of AMVs from
MODIS,
Meteosat-7, -9,
GOES-11, -12
sliced by
pressure intervals

33. Thank You