September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Precipitation Validation
Network in West Africa set up for AMMA
Perspectives for Megha-Tropiques and GPM
Michel Desbois et al. (LMD), Thierry Lebel et al. (IRD/LTHE)
Abdou Ali et Abou Amani (AGRHYMET)
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Quick reminder : what is AMMA ?
Radiovent / Pibal (Wind)RadioSounding
Theorical networkRadiosounding - Pibal
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
What is available as pluviometric data over Sahel ?
Cilss network : 650 stations available several weeks after the rainy seasonCRA network : average of 280 stations available every 10 daysSYN : average of 85 stations available each day
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Summary on the kriging method used to create the areal rainfall for validation of satellite rainfall estimates
dlongclatblatalatlongdrift +++= ***),( 2
1) Regression kriging method is used to estimate the mean areal rainfall for the three grids (0.5°x0.5°, 1°x1° and 2.5°x2.5°) from AGRHYMET rainfall data base (around 650 gauges on average)
2. A drift model with quadratic component for the latitude and linear one for the longitude is used:
the values of the coefficients a, b, c and d depend on the period (on the 10-day under consideration). There are estimated using the generalized least squares (GLS) in order to account for the spatial correlation of residuals.
3. Also in order to account for the non-homogeneity of the drift coefficients in the Sahelian region, the studied
area is divided in five regions. The central region between longitude 13°W and 10°E and latitude 12°N and 15°N. The different sides of this central region are considered separately.
4. A nested variogram using two anisotropic exponential functions with an anisotropy coefficient of for the first
structure (decorrelation distance = 0.2°) and 2/3 for the second structure nested (decorrelation distance = 3) is used.
5. For details on the kriging method and raingauge data description used to compute the validation areal
rainfall, see Ali et al., J. Appl. Meteo, 2005 (accepted).
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Regional « supersites » are available forhigher resolution studies
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Niger River
-40 -20 0 20 40 60 80 100 120
Local coordinates in km; Origin at 2°E;13°N
0
20
40
60
80
100
Say
NIAMEY
Baleyara
Banizoumbou
1.5 2 2.5
9
9.5
10
KPAYEROUN
INA2
BASSILA
BETEROU
BIRNI
DJOUGOU
OKPARA
PARAKOU
PARTAGO
PENESSOULOU
KOKOUBOU
TCHETOU
Observateurs RivièrePluie Aquifère
Réseaux AMMA-CATCH
Enregistreurs IDAF-Photon
Pluie Aquifère Rivière Météo
Végétation
-2 0 2 44
6
8
10
12
14
16
18
Hombori
Bani
Djougou
Supersites of Niamey and Djougou
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Raingauge network – Mesoscale sites Available data sets for 2004
Niamey square degree
33 recording rain gauges
Raw data 5 min, ~ from May to September (depends on the station)
Validated data:
• 10-day periods - station and grid (krigging 5 km x5 km)
• rainfall events (i.e. >30% rainy stations) - station and grid (krigging 5 km x5 km)
North Benin
35 recording rain gauges
Raw data 5 min, all over the year (with some missing data)
Validated data: daily rainfall (stations)
Gridded estimates are not available (yet) by lack of climatological knowledge
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Example of large scale validation of a particular algorithm : SRA
SRA means Sliding Rescaling Algorithm (Franck Chopin and Jean-Claude Bergès)
It is based on the estimation of rainfall probability on MSG images (15minutes intervals, 3 km pixels), calibrated on coincidences with TRMM radar.
A second step is to establish maps of « precipitation efficiencies » based on comparisons of monthly rainfall accumulation with rainfall probabilities
Then, precipitation estimates can be provided at all time scales below one month (but errors need to be estimated)
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Data and Experimentation areaInput dataset:
Full flow of MSG/HRI TRMM/PR from 3G68Land GPCP 1° daily rainfall
Validation dataset :
Dense raingauges network (IRD & AGRHYMET supplied)
Experimentation Area :
A window on W-Africa from June to Sep. 2004 (25W to 25E and 5S to 20N) This is the AMMA West Africa Region
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Example of Rainfall Probability Image
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Collocation between Pr image and 2A25 TRMM data
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Examples of Potential Intensity Image
Mm/h
Potential rainfall intensity 15th August 2004
Mm/h
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Example of rainfall estimation image
mm
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Results for 3rd decade of August 2004
mm
Number of stations Krigged data set
SRA result SRA on area and space resolution of krigged data set
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Duration of rain (hours)
raingauges
GPCP 1DD
raingauges
SRA
Third decade of August, 2004
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Third decade of August 2004
Absolute value difference between raingauges and SRA
Number of raingauges
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Results for all the decades of 2004 are better for most of the indicators for SRA than for GPCP. But better validation fields are needed at smaller scales.
BIAS RMSD NRMSD R² SKILLMIN RAIN
(mm)
MAX RAIN (mm)
MEAN RAIN (mm)
GPCP 1dd june dec 1 10,33 1,15 2,02 0,72 -0,440 58,6 18,4
SRA 1-1 june dec 1 4,29 0,76 1,33 0,64 0,37
GPCP 1dd june dec 2 7,52 1,04 2,37 0,58 -0,50 60,2 13,3
SRA 1-1 june dec 2 4,72 0,74 1,68 0,61 0,24
GPCP 1dd june dec 3 4,22 0,87 1,46 0,62 0,410,3 73,2 28,3
SRA 1-1 june dec 3 2,02 0,67 1,12 0,77 0,65
GPCP 1dd july dec 1 14,75 1,98 2,71 0,66 -0,925 121,5 37,8
SRA 1-1 july dec 1 9,23 1,17 1,6 0,74 0,33
GPCP 1dd july dec 2 12,24 1,53 1,59 0,65 0,610,4 204,6 44,9
SRA 1-1 july dec 2 1,91 1,34 1,39 0,63 0,63
GPCP 1dd july dec 3 -9,23 2,56 1,84 0,41 0,384,5 334,9 70,5
SRA 1-1 july dec 3 -6,09 2,21 1,59 0,58 0,54
GPCP 1dd august dec 1 11,94 2 2,34 0,31 -0,453,6 159 46,7
SRA 1-1 august dec 1 1,35 1,53 1,79 0,32 0,15
GPCP 1dd august dec 2 3,55 1,73 1,61 0,42 -0,038,4 157,8 60,8
SRA 1-1 august dec 2 1,66 1,38 1,29 0,49 0,34
GPCP 1dd august dec 3 10,1 1,48 1,4 0,75 0,623,6 199,2 61,1
SRA 1-1 august dec 3 -1,2 1,11 1,05 0,79 0,78
GPCP 1dd sept dec 1 0,66 0,95 1,11 0,68 0,651,3 128,6 37,6
SRA 1-1 sept dec 1 -0,04 0,7 0,83 0,81 0,8
GPCP 1dd sept dec 2 -2,67 1,18 1,2 0,65 0,640 146,3 32,4
SRA 1-1 sept dec 2 -7,48 1,32 1,34 0,62 0,56
GPCP 1dd sept dec 3 5,63 0,96 1,53 0,67 0,520 104,7 18,5
SRA 1-1 sept dec 3 1,74 0,75 1,2 0,72 0,71
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Present work for Amma is concentrated on giving results at smaller scales, for needs of soil moisture modelling and hydrology. Further validation will use the supersites of AMMA, which were already operated in 2004.
Year 2005 will also be processed, as well as 2006 (the special observing period of AMMA, and 2007).
After 2007, it is foreseen to maintain these supersites, in particular for the needs of Megha-Tropiques and GPM ground validation.
It seems very difficult in Africa to establish and maintain validation sites in the definition of GPM, in particular with well calibrated rainfall radars. However, considering the crucial and specific problems of rainfall estimation in these regions of the world, the Megha-Tropiques science team is considering AMMA-type supersites as necessary for ground validation over these regions. (does GPM support this statement ?)
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Perspectives of Megha-Tropiques (launch 2008-2009) nominal life time 3 years
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
•ScaRaB : a wide band radiometer for radiative fluxes measurement (transversal scan, 40 km)
•Saphir : a microwave water vapour sounder. 6 channels around 183.31 GHz. (transversal scan, 10 km)
•Madras : a microwave imager5 channels 18, 23, 37, 89 and 157 GHz(conical scan, <10 km to 40 km)
Marfeq is the radiofrequency part of this instrument
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
QuickTime™ et undécompresseur TIFF (LZW)
sont requis pour visionner cette image.
30 N
30 S
15 N
15 S
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
2005 : new start of the scientific studies
•2004 : French mission team constituted : Michel Desbois (P.I.); Laurence Eymard (SAPHIR); Rémy Roca (mission); Nicolas Viltard (MADRAS); Michel Viollier (ScaRaB); Sophie Cloché (data bases)
•January 2005 : 1st meeting of a preliminary French « science team »
•April 2005 : Indo-French meeting of the mission teams in Bangalore (Indian team chaired by R. Narasimha, PI’s J. Srinivasan, M.S. Narayanan, participants P.C. Joshi, V.K. Agarwal, Kusuma Rao, members of SAC Ahmedabad, of ISRO HQ)
•October 2005 : enlarged Indo-French meeting in Ahmedabad aimed to start common projects around missions Megha-Tropiques and Oceansat III (French altimeter Altika)
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
•Actions devoted to algorithm developments
-Satellite data bases for testing algorithms and processing procedures of the Megha-Tropiques instrumental package.
-Microwave rain algorithm developments : adaptation and evaluation of TRMM-based algorithms; specific effort on the ice phase.
-Combined geostationary - microwave algorithms.
-Radiative fluxes retrievals at different space-time scales.
-Humidity retrievals in the perspective of SAPHIR.
Preliminary list of actions
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
•Actions devoted to validation of satellite products
-Precipitation retrieval validation (raingauges and radars) : sampling and scaling effects.
-Validation for other quantities (ice microphysics, others…)
-Intercomparisons of methods for water vapour determination (IR and MW sounders, GPS, … )
-General use of the AMMA campaigns and data bases for satellite validation studies
- Promoting specific campaigns for water vapour, clouds and radiative budget .
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
•Actions devoted to MT sampling effects
-Use of geostationary observations to simulate the MT sampling of convective systems and their life cycle. To be performed over Africa, India and Brazil.
-Development of a simulator for MT from analyses / satellite observations to analyze the impact of MT orbit on the various fields retrieved (radiative budget, water vapour …)
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
•Actions devoted to MT data scientific exploitation
-Assimilation of microwave data including MT in forecast models.
-Impact of dry subsidence areas in the subtropics on the global climate .
-Upper tropospheric drying influences on monsoon cloud systems and hence on monsoon variability. -Construction of cloud climatology over the Indian ocean area.
-Structure of mesoscale convective complexes/systems over Indian Monsoon Region.
-Monsoon systems in India and Africa
-Use of microwave emissivity in complement of VIS-IR data for land surface properties.
-Cloud resolving models for Indian region using Microwave and radar data.
September 29, 2005 Michel Desbois, Taipei, GPM GV meeting
Conclusion
•Numerous research subjects related to the satellite mission, but also to general tropical climate issues
•Presently, not enough scientists between France and India to perform all this research
•Mission teams in France and India look for a widening of the science community working on Megha-Tropiques related themes
•Concerning ground validation, same tropical sites should be used for GPM and MEGHA-TROPIQUES