1. Second Land Surface Analysis SAF Workshop 8-10 March 2006, Lisbon
Short-Range NWP Programme
Use of Satellite Information for Surface Variable Analysis in Short-Range NWP: A Small Survey
Compiled by Jean Quiby
EUMETNET Short-Range NWP Programme Manager
Second Land Surface Analysis SAF Workshop
8-10 March 2006, Lisbon

2. The Questionnaire (1) Surface variable Operational / Development
Sensor:
Carrying satellite:
Channels used:
Are you using products from the SAF Land Surface Analysis?
If yes, which ones?
References:
Scientist to contact:
Remarks:

3. The Questionnaire (2) List of the Surface Variables:
Sea Surface Temperature
Lake Surface Temperature
Soil Surface Temperature
Snow cover
Vegetation (NDVI)
Albedo
Emissivity
Other Surface Parameter

4. The Questionnaire (3) The Answers:
Questionnaire sent to the 25 NWS Members of the EUMETNET SRNWP Programme
14 NWS have replied
NWS not working with satellite data had a strong tendency not to reply
From the 14 which have replied, 11 are actively using satellite data for development works

5. O D (D) Belgium Finland France Portugal Austria Denmark Germany Italy
SST
LkST
LST
Snow
NDVI
Albed
Emis.
Solar
Moist.
E.Tr.
Belgium O
(D) D
Finland
France
O Ar
D OSI
Pl.
LSA
Pl.NESD
Pl. LSA
ECOCL
(SPOT)
Verif.
Portugal
Austria
Denmark
Germany
NESDIS
clim
Italy
Spain
Switzerland UK
O &
OSI
OSTIA
D LAI
MODIS
D-BDRF
SEVIRI
ASCAT

6. Some Comments (1)
There is a reasonable amount of work done in Europe in surface parameter analysis with the use of satellites (thanks to the LSA SAF)
Operational SST everywhere derived from satellites (mostly NOAA/NESDIS radiation)
Outside L-SAF, largest development work done for the determination of the snow cover
LSA SAF products not yet operationally utilized in SR-NWP. But plans to do it exist at the ZAMG (next to L-SAF participants).

7. Some Comments (2)
The NWP community is only one user among many of the SAF products
Land SAF does not seem to be known from all the European NWS
Land SAF thinks “full disk”, “global”
But SEVIRI has the right temporal and spatial resolution for meso-scale LAM’s
Land SAF should improve its visibility among the NWS
Land SAF should have as slogan: “We are global and regional”.

8. Next steps
Put the full information of this short survey in the SRNWP Programme web site (
Ask for information when indications not clear
Try to get answers of missing NWS
Possibly: Prepare (with specialists) a better questionnaire that will be sent to the NWS active in surface data analysis
To snow

9. Soil Surface Temperature
Operational / Development: Applications under development
Sensor: SEVIRI
Carrying satellite: Meteosat-8
Channels used: IR 10.8 and IR 12.0
Are you using products from the SAF Land Surface Analysis? No.
If yes, which ones?
References:
Scientist to contact: Alexander Jann
Remarks: Use of the LST product of the LSA SAF is a goal for the near future.
ZAMG back

10. Soil Surface Temperature
Development in the framework of the LSA-SAF
Sensor: SEVIRI
Carrying satellite: MSG
Channels used: -
Are you using products from the SAF Land Surface Analysis? Yes
If yes, which ones? Albedo, DSSF, DSLF and LST from LSA-SAF
References: RMI takes part to the LSA-SAF consortium for the development of the ET (evapotranspiration) product
Scientist to contact: Françoise Meulenberghs
Remarks:
RMI back

11. Sea Surface Temperature
Operational : YES
Sensor: AVHRR; Carrying satellite: NOAA 17, NOAA 18
Channels used: 3b, 4 and 5
Sensor: Seviri; Carrying satellite: METEOSAT-8
Channels used: 4, 7, 9 and 10
Sensor: -; Carrying satellite: GOES
Channels used: 2, 4 and 5
Are you using products from the SAF Land Surface Analysis? NO
If yes, which ones? -
References: -
Scientist to contact: Bjarne Amstrup
Remarks:
All the SAF data are received via Eumetcast. The channels referred to above all apply to the infrared and near-infrared spectral regime.
DMI back

12. Sea Surface Temperature
Operational : operational SST analysis from NESDIS, which assimilates satellite obs, is used as a relaxation field in the global ARPEGE SST analysis done with in situ observations only (buoys, ships)
Development : SAF-OSI SST product is currently tested for ALADIN model
Sensor: for SAF-OSI (AVHRR/NOAA and SEVIRI/MSG)
Carrying satellite:
Channels used: SAF-OSI web site
Are you using products from the SAF Land Surface Analysis?
If yes, which ones?
References:
Scientist to contact: Francoise Taillefer (CNRM/GMAP), Hervé Roquet (CMS/Lannion)
Remarks:
M-F back

13. Lake Surface Temperature
Operational
Sensor:AVHRR
Carrying satellite:tiros
Channels used:10.8m, 12m, 3.7m
Are you using products from the SAF Land Surface Analysis? no
If yes, which ones?
References:
Scientist to contact: Francesco Zauli
Remarks:
UGM back

14. Snow cover
Operational / Development: Development
Sensor: SSM/I and MODIS
Carrying satellite: DMSP and Terra
Channels used: Derived products
Are you using products from the SAF Land Surface Analysis? No
If yes, which ones?
References:
Scientist to contact: Alberto Cansado:
Remarks:
INM back

15. Other surface parameter
Which one? Soil moisture
Operational / Development Development
Sensor: ASCAT
Carrying satellite: METOP
Channels used:
Are you using products from the SAF Land Surface Analysis? No
If yes, which ones?
References:
Scientist to contact:
Remarks:
Planned EUMETSAT CGS product.
We have been involved in development activities through NWP SAF Associate Scientist activity with Technical University of Vienna.
Met Office back

16. Determination of the Snow Cover with METEOSAT-8
Martijn de Ruyter de Wildt
EUMETSAT Research Fellowship
Institute of Geodesy and Photogrammetry of the ETH Zurich
and
Swiss Federal Office of Meteorology and Climatology

17. Main Problem: to separate snow from cloudiness
ice cloud
BT10.8
:12 UTC

18. The Spectral Differentiation (1)
channel used for the spectral differentiation

19. The Spectral Differentiation (2)no
(r0.64 > 0.25) AND (r1.6> 0.30) OR
BT3.9 – BT10.8 > 10cosθsun OR
BT10.8 < 253 – z OR
BT10.8 – BT12.0 > 1.5
NDSI > AND
R064 > AND
R0.81 > AND
BT10.8 < K
cloud
snow
yes
Snow free surface

20. Normalized Difference Snow Index (NDSI)
As snow has a larger reflectance at 0.64 μm and a lower reflectance at 1.64 μm than snow-free land, we can similarly to vegetation create a Normalized Difference Snow Index for a differentiation between snow and snow-free land

21. The Spectral Differentiation (3)
snow
ice cloud
white: snow
dark grey: clouds
light grey: snow-free land
black: sea
12:12 UTC

22. MSG versus NOAA Low Earth Orbit
Small differences between the spectral channels
But very large differences between the temporal resolutions
Every 15 minutes versus 2 twice a day (for one satellite)
We should try to use this tremendous superiority of MSG in time frequency

23. Use of the time frequency of MSG (1)
How the high temporal frequency of MSG (15 minutes) can help to improve separation of clouds and snow?
Fact:
In general, in a short period of time, say one hour, clouds show more variability than the Earth surface.
Consequence:
For most of the channels, the temporal differences in the values of the pixels are generally larger for cloudy pixels than for clear pixels.

24. Use of the time frequency of MSG (2)
For each scene, for each pixel, the differences dm,i,j,t are computed for the following channels or difference of channels Im,i,j,t = r0.64, r0.81, r1.6, r0.64-r1.6, BT3.9, BT3.9-BT10.8 and r0.75 (h. res.) according to the following formula:
I = reflectance or brightness temperature
m = channel number or combination of channels
i,j

25. Use of the time frequency of MSG (3)
For the each channel or difference of channels
r0.64, r0.81, r1.6, r0.64-r1.6, BT3.9, BT3.9-BT10.8 and r0.75
the dm,i,j,t values are averaged separately for the cloudy
pixels and for clear pixels.
Thus we have 7 values <dm,cloudy> and 7 values <dm,clear>
To know whether a pixel is cloudy or clear, we use the spectral differentiation.

26. Use of the time frequency of MSG (4)
Let us go back to the spectral differentiation: no
(r0.64 > 0.25) AND (r1.6> 0.30) OR
BT3.9 – BT10.8 > 10cosθsun OR
BT10.8 < 253 – z OR
BT10.8 – BT12.0 > 1.5
NDSI > AND
R064 > AND
R0.81 > AND
BT10.8 < K
cloud
snow
yes
Let us consider only the pixels of the class “snow”
Snow free surface

27. Use of the time frequency of MSG (5)
A snow pixel is categorized as “cloudy” (i.e. no longer as “snow”) if for one or more of the 7 channels or differences of channels
r0.64, r0.81, r1.6, r0.64-r1.6, BT3.9, BT3.9-BT10.8 and r0.75
its normalized distance to the class “cloudy” is smaller than its normalized distance to the class “clear”, i.e. if
σ = standard deviation

28. Use of the time frequency of MSG (6)
Temporal
test
cloud
(r0.64 > 0.25) AND (r1.6> 0.30) OR
BT3.9 – BT10.8 > 10cosθsun OR
BT10.8 < 253 – z OR
BT10.8 – BT12.0 > 1.5
NDSI > AND
R064 > AND
R0.81 > AND
BT10.8 < K
yes no
Snow free surface
snow

29. Supplementary checks Spatial check
Any snow pixel surrounded by 6, 7 or 8 cloudy pixels is re-classified as cloudy
Temporal check
If a pixel is classified as snow in an image at time t but as cloud in the images at times (t-1) and (t+1), it is re-classified as cloud

30. Effect of the time differentiation
Spectral differentiation only
With temporal differentiation
white: snow
dark grey: clouds
light grey: snow-free land
:12 UTC

31. Thanks to Martijn de Ruyter de Wildt
who has given his permission to show his work at the
Second Land Surface Analysis SAF Workshop