1. EUMETSAT Precipitation Week
Nowcasting SAF: Precipitating Clouds (PC) and Convective Rainfall Rate (CRR) products
EUMETSAT Precipitation Week
4 - 8 February 2013
Cecilia Marcos
Antonio Rodríguez

2. Where are you from?

3. Overview I. Introduction to EUMETSAT Nowcasting SAF
II. Precipitating Clouds Product
Algorithm description
Applications, limitations and visual examples
III. Convective Rainfall Rate Product
Algorithm description (Matrices)
Algorithm description (Analytical functions)
IV. Precipitation products through microphysical cloud top information (in development)
Precipitating Clouds Product
Convective Rainfall Rate Product
Limitations, improvements and visual examples
V. Comparison of precipitation products: visual examples

4. Introduction to EUMETSAT NWCSAF
The Nowcasting Satellite Application Facility was established in 1996 between Eumetsat and former INM (Instituto Nacional de Meteorología).
Consortium:
Objectives:
Development of Nowcasting products derived from both GEO and PPS satellite systems
To be delivered to users as SW Packages
Responsible for
Development and maintenance of the NWC products
Development and maintenance of the SW Packages
User's support tasks made through dedicated Help Desk

5. Introduction to EUMETSAT NWCSAF
Products are generated in the users’ premises
Features of the products:
Near Real Time (NRT)
Full resolution
Frequency to be selected by the user (default every repeat cycle)
Region to be selected by the user
More information on the project is available at Nowcasting SAF Web site:

6. Introduction to EUMETSAT NWCSAF

7. Introduction to EUMETSAT NWCSAF
MSG Precipitation
Products

8. Precipitating Clouds Product for MSG v2012
INTRODUCTION:
The objetive is the delineation of non-precipitating and precipitating clouds.
PC serves as a geneal tool for Nowcasting of precipitation, especially for areas where no surface data is available.

9. Precipitating Clouds Product for MSG v2012
SEVIRI

10. Precipitating Clouds Product for MSG v2012
ALGORITHM DESCRIPTION:
A Precipitation Index (PI) has been built through a linear combination of those spectral features which have the highest correlation with precipitation:
Two different calibrations:
European Synops
Precipitation rates from rain gauge over France (using the Cloud Type product)
Different algorithms for day and night situations
The choice of the algorithm to be used is done through the Configuration File
PI=a0 +a1*TSurf +a2*T10.8+a3*(T10.8-T12.0)+a4*abs(a5-R0.6/R1.6)+
a6*R0.6+a7*R1.6+a8*T6.2+a9*T7.3+a10*T3.9)

11. Precipitating Clouds Product for MSG v2012
OUTPUTS:
The PC product shall consist of a numerical value for the likelihood
The following probability classes will be used:
      0% (= 0-5%)      10% (= 5-15%)      20% (= 15-25%)      30% (= 25-35%)      40% (= 35-45%)      50% (= 45-55%)      60% (= 55-65%)      70% (= 65-75%)      80% (= 75-85%)      90% (= 85-95%)      100% (= %)
FLAG: information about the processing conditions

12. Precipitating Clouds Product for MSG v2012
Applications: Provide information about precipitation ocurrence likelihood over extensive areas, out of the Radar coverage or as a Radar complement.
RADAR PC
Example: 9th September 2012 at 15:00 UTC

13. Precipitating Clouds Product for MSG v2012
Limitations:
16th January 2013 at 10:30 UTC
RADAR PC
IR 10,8
- Similar to the cloud tops
- Too big estimated precipitation area
Different result for day and night
12th April 2012 at 06:30 UTC PC

14. Precipitating Clouds Product for MSG v2012
Precipitating Clouds Loop:

15. Convective Rainfall Rate Product v2012
INTRODUCTION:
The CRR goal is to estimate rainfall rates from convective systems, using IR, WV and VIS MSG SEVIRI channels and lightning information (as optional input).

16. Convective Rainfall Rate Product v2012

17. Convective Rainfall Rate Product v2012
CALIBRATION MATRICES:
Calibration matrices have been built through a statistic method using:
SEVIRI data
Composite radar data: (2Km spatial resolution)
Reflectivities
Echotop (only to select convective areas)
Two different calibrations:
R = f (IR, IR-WV, VIS), for 3-D calibration (day time)
R = f (IR, IR-WV), for 2-D calibration (night time)
For each type of calibration there are two regional matrices:
Spanish matrices
Nordic matrices
The CRR retrieval can be latitude dependant (difference matrices)
PROCESSING:
Basic CRR mm/h value for each pixel obtained from the calibration matrices.
Filtering process
Corrections: Moisture Correction, Evolution Correction (Cloud-top Temperature Gradient Correction), Parallax correction and Orographic correction
Lightning algorithm

18. Convective Rainfall Rate Product v2012
Automatic station data provided by IM (Portuguese Meteorological Service)
Madeira 20th February 2010 – 12:00 UTC
Moisture Correction : (PWRH)
Depends on total precipitable water (surface- 500 hPa) and relative humidity
No corrections applied
Only Moisture correction applied

19. Convective Rainfall Rate Product v2012
Evolution Correction / Gradient Correction :
No Evolution Correction
Radar
Evolution Correction:
Rain rate decreases if
the analysed pixel becomes warmer in the second image
Cloud-top Temperature Gradient Correction:
Rain rate decreases if the analysed pixel has a temperature maximum, which indicates that this pixel is warmer than its surroundings
Evolution Correction
Gradient Correction

20. Convective Rainfall Rate Product v2012
Parallax correction:
A spatial shift is applied to every pixel with precipitation according the basic CRR value
Only Moisture correction applied
Moisture + Evolution
corrections applied
Moisture + Evolution + Parallax
corrections applied

21. Convective Rainfall Rate Product v2012
Orographic correction:
This correction uses the interaction between the wind vector (taken from the 850 hPa. numerical model) and the local terrain height gradient in the wind direction to create a multiplier that enhances of diminishes the previous rainfall estimate, as appropriate.
Moisture + Evolution + Parallax
corrections applied
Moisture + Evolution + Parallax + Orographic
corrections applied

22. Convective Rainfall Rate Product v2012
LIGHTNING ALGORITHM:
The Lightning algorithm assumes that the higher are the spatial and temporal density of lightning, the higher are the probability and the intensity of convective precipitation.
The rain rates assigned to every lightning takes into account:
- the time distance between the lightning event and scanning time of the processing region centre.
- the location of the lightning.
- the spatial density of lightning in a time interval.
Only Cloud-to-Ground lightning flashes are used by this algorithm.
When the lightning precipitation pattern has been computed, it is compared to the CRR precipitation one in order to obtain the final product, that will contain the highest rain rates of the two.
Lightning algorithm
CRR without lightning
CRR with lightning

23. Convective Rainfall Rate Product v2012
OUTPUTS:
CRR rainfall rates expressed in classes
CRR rainfall rates expressed in mm/h (only for developing purposes)
CRR Hourly Accumulations
CRR-QUALITY
CRR-DATAFLAG
Where:
Ai: hourly accumulation, in mm, corresponding to the time i.
T: time interval between scenes in hours (T= 0.25)
Ф: part of T that corresponds to the time that takes the satellite to reach the centre of the region.
Ii: Instantaneous rainfall rate for each scene in mm/h

24. Convective Rainfall Rate Product v2012
Applications: Estimation of rain rates for convective events over extensive areas, out of the radar coverage or as a radar complement.
RDT
CRR
RADAR +
Lightning info

25. Courtesy of Jean-Marc Moisselin
Links between SAF/NWC products Use of CRR in RDT (Rapid Development Thunderstorm)–Implemented in v2012
RDT takes advantage of CRR data with different options:
CRR can be used to identify maximum convective rain rate under cloud cell
High CRR values can help to qualify a cloud cell as significant and thus to encode this cell in BUFR output (only for the last BUFR version)
Very high CRR values can be used to set diagnosis of convection to « YES » (only for the last BUFR version)
RDT main outlines
Attributes
PGE11 can take advantage of CRR data (PGE05) as optional input or for additional
attribute :
o CRR data are used to identify maximum convective rain rate over cloud cell
extension
o High CRR data may be used additionally to qualify a cloud cell as significant and to
encode this cell in BUFR output (for the last BUFR version only)
The default threshold for that purpose is 10mm/h. It can
be customized through parameter “–crr” of the model configuration file
o Very High CRR data may be used on request to force convective diagnostic in last
BUFR output version.
The default threshold is 50mm/h.
This option is activated only when the parameter “–crrdiscri” is set to 1 in the model
configuration file.
For a given detected cloud system, the corresponding maximum convective rain rate
is defined as the 99th percentile of processed CRR intensity over the cell’s horizontal
Attributes coming from other NWC SAF products : Cloud Top Pressure, Cloud Type, Cloud Phase, Convective Rain Rate

26. Convective Rainfall Rate Product v2012
Problems:
26th July :00 UTC
CRR
IR 10.8
RADAR
- Similar to the cloud tops
- Too big estimated precipitation area and lower rain rates than Radar
Different result for day and night
CRR
31th August :00 UTC

27. Convective Rainfall Rate Product v2013
From matrices to analitical functions
Matrices
Analitical functions + recalibration

28. Convective Rainfall Rate Product v2013
Visual examples
22th August 2008
14:00 UTC
RADAR (PPI)
CRR 2D MATRIX
CRR 3D MATRIX
CRR 2D FUNCTION
CRR 3D FUNCTION

29. Precipitation products through microphysical cloud top information
This algorithm is based on the one developed by Roebeling and Holleman*
Cloud Top Microphysical Properties used by this algorithm:
Phase (Ph) Effective radius (Reff) Cloud optical thickness (COT)
Rain rate
(*) Roebeling, R. A. and I. Holleman, 2009: SEVIRI rainfall retrieval and validation using weather radar observations. J. Geophys. Res., VOL. 114, D21202.

30. Precipitation products through microphysical cloud top information
Precipitating clouds: likelihood of precipitation
Comparison reprojected Radar with CWP
CWP Thresholds Precipitation Probability

31. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
Some dependance on illumination conditions – better VIS normalization needed
Improvements:
More confidence on the assignment of the precipitation likelihood
PC (Currently available)
PC (CTMP)
Radar
8th September :30 UTC

32. Precipitation products through microphysical cloud top information
Datasets:
Spain: 40 storms, May-September 2009, 12:00 UTC
Hungary: 18 storms, May-September 2009, 10:00-12:00 UTC
CRR Calibration: Precipitation area
Radar (PPI)
Precipitation area:
Ph = water:
Reff  14 μm
CWP  200 gm-2
Ph = ice:
CRR Calibration: Rain Rates
Cloud Top Microphysical Properties Function

33. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
High dependance on illumination conditions – better VIS normalization needed
9th September 2012
12:00 UTC
Illumination Quality flag
Radar
CRR CTMP
Illumination Quality Flag

34. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
High dependance on illumination conditions – better VIS normalization needed
9th September 2012
13:00 UTC
Illumination Quality flag
Radar
CRR CTMP
Illumination Quality Flag

35. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
High dependance on illumination conditions – better VIS normalization needed
9th September 2012
14:00 UTC
Illumination Quality flag
Radar
CRR CTMP
Illumination Quality Flag

36. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
High dependance on illumination conditions – better VIS normalization needed
9th September 2012
15:00 UTC
Illumination Quality flag
Radar
CRR CTMP
Illumination Quality Flag

37. Precipitation products through microphysical cloud top information
Limitations:
Only day time
Only for estimated phase
High dependance on illumination conditions – better VIS normalization needed
9th September 2012
16:00 UTC
Illumination Quality flag
Radar
CRR CTMP
Illumination Quality Flag

38. Precipitation products through microphysical cloud top information
CRR matrices
Improvements:
Precipitation patterns closer to the radar ones
28 September 2008
12:00 UTC
CRR CTMP
Radar

39. Precipitation products through microphysical cloud top information
CRR matrices
Improvements:
Precipitation intensities closer to the radar ones
5th August 2008
13:00 UTC
CRR CTMP
Radar

40. Precipitation products through microphysical cloud top information
CRR matrices
IR 10.8
Improvements:
No Cold Rings and detection of smaller precipitation nuclei
CRR CTMP
Radar
9th September 2008
13:00 UTC

41. Precipitation products through microphysical cloud top information
CRR matrices
IR 10.8
Improvements:
Detection of precipitation for warm top clouds
CRR CTMP
Radar
11th August 2012
14:00 UTC

42. Comparison of precipitation products
PC (Currently available)
Visual example
12th July 2008 at 13:00 UTC
PC (CTMP)
Radar

43. 12th July 2008
13:00 UTC
CRR 2D MATRIX
Version 2012
CRR 2D FUNCTION
Version 2013
CRR 3D MATRIX
Version 2012
CRR 3D FUNCTION
Version 2013
CRR CTMP
Version 2013
RADAR (PPI)

44. Quiz
Which CRR correction allows a better location of the raining clouds?
Orographic correction
Parallax correction
No CRR correction tries to adjust the position of the raining clouds
Why the estimation of precipitation with GEO satellite data gives better results during daytime?
There is no difference on the results between day and night time
Because during night time only IR SEVIRI channels give information and during day time only solar SEVIRI channels give information
Because solar channels, which provide microphysical information on the cloud tops, are useful only during day time.

45. This image shows the CRR rain rates over Europe at dawn or at dusk, can you indicate where is the transition between 2D and 3D algorithms and put an star at the day time side?

46. Solution CRR Rain rates CRR-QUALITY DAY TIME NIGHT TIME DAY TIME

47. Thank you for your attention!
Question time