Rainfall estimation for food security in Africa, using the Meteosat Second Generation (MSG) satellite. Robin Chadwick
Mar 28, 2015
Rainfall estimation for food security in Africa, using the
Meteosat Second Generation (MSG) satellite.
Robin Chadwick
Contents of presentation
• Motivation for satellite rainfall estimation in Africa• TAMSAT satellite rainfall estimation methodology• Met office NIMROD nowcasting precipitation estimation
product.• Extension of Met office rainfall estimates to Africa• AMMA Sahelian rain-gauge dataset• Comparison of Met office rainfall estimates against
TAMSAT estimates and AMMA gauge data• Current and future work
Motivation for satellite rainfall estimation in Africa
• Accurate near-real time estimates of rainfall are vital for humanitarian applications such as famine prediction and prevention, and flood prediction.
• Very few precipitation radar networks. Rain-gauges sparce and badly maintained.
• Satellite based rainfall estimation algorithms offer one solution to this problem.
• Several algorithms exist, using IR data (from geostationary satellites) , passive microwave data (from polar orbiting satellites) or a combination.
MSG wavelength channels
Visible
IR window channelsWater vapour
Near
IR IR
The TAMSAT rainfall estimation method
• Utilises one infrared channel (10.8 microns) from the MSG
• Simple method based on the concept of Cold Cloud Duration
• Produces operational dekadal rainfall estimates for Africa
• Intercomparisons of various satellite rainfall products over Africa have found that the TAMSAT method is as accurate as more complex algorithms.
• Should be possible to improve on this because of the information on rainfall provided by other channels on the MSG.
TAMSAT methodology
•CCD is the Cold Cloud Duration; the length of time each pixel is below the threshold temperature
•Rainfall, R = a + b(CCD)
•Threshold temperature and coefficients a, b calibrated for each region using historical rain-gauge data
tT
TAMSAT rainfall estimate for 2007 October 1st dekad
Met office NIMROD nowcasting precipitation estimation product
• Rain-rate estimates over Europe produced operationally every 15 minutes
• Radar Satellite Analysis
+ =
Francis et al ‘06
Calibration of NIMROD using only 2 MSG channels
Radar rain-rateMeteosatIR channel
MeteosatVis channel
Nimrod satelliterain-rate
72.3376.8959.8634.364
19.3841.9732.8615.353
2.9013.6620.619.222
0.002.465.995.471
4321% Rain
Infrared (10.8 m) channel
Visible (0.8 m) channel
72.3376.8959.8634.364
19.3841.9732.8615.353
2.9013.6620.619.222
0.002.465.995.471
4321% Rain
Infrared (10.8 m) channel
Visible (0.8 m) channel
10266 /14193
3137 /4080
3877 /6477
706 / 20554
1342 /6925
1340 /3193
1842 /5605
610 /39753
76 /2620
408 /2987
980 /4756
621 /67372
0 /286
59 /2401
279 /4656
1028 /187971
4321Rain /
Total
Infrared (10.8 m) channel
Visible (0.8 m) channel
10266 /14193
3137 /4080
3877 /6477
706 / 20554
1342 /6925
1340 /3193
1842 /5605
610 /39753
76 /2620
408 /2987
980 /4756
621 /67372
0 /286
59 /2401
279 /4656
1028 /187971
4321Rain /
Total
Infrared (10.8 m) channel
Visible (0.8 m) channel
11104
01003
00002
00001
4321Rain /No rain
Infrared (10.8 m) channel
Visible (0.8 m) channel
11104
01003
00002
00001
4321Rain /No rain
Infrared (10.8 m) channel
Visible (0.8 m) channel
BrightDark
Cold
Warm
Extension of NIMROD to multiple channels
SZA Primary correlation method
0o-75o 4-d (0.8/1.6/3.9 refl/10.8)
75o-80o 4-d (0.8/1.6/3.9 refl/10.8) => 3-d (0.8/1.6/10.8)
80o-85o 3-d (0.8/1.6/10.8)
85o-88o 3-d (0.8/1.6/10.8) => 3-d (3.9BT/10.8/12.0)
>88o 3-d (3.9BT/10.8/12.0)
10th/11th October 2006
Extension of Met office algorithm to Africa
Current domain of Met office algorithm extension
The AMMA Sahelian rain-gauge dataset
• O.5 degree resolution gridded rain-gauge dataset for May – September 2004 covering the Sahel.
• Met office estimates processed for this period & region using historical MSG data
• Estimates still use (historical) European radar data for calibration
• Comparison of Met office estimates against AMMA gauge data, for grid cells containing gauges only.
• Comparison of TAMSAT estimates against AMMA gauge data
• Comparison of Met office estimates against TAMSAT estimates
Validation domain
Comparison of Satellite rainfall estimates and gauge data over the Sahel for the July dekad 2 2004
Met office – Raingauge anomaly
TAMSAT – Raingauge anomaly
Met office – TAMSAT anomaly
Met office and TAMSAT vs gauge dekadal estimates for May to September 2004
Met office vs Raingauge TAMSAT vs Raingauge
Met office vs TAMSAT dekadal rainfall totals for all dekads May to September 2004
Statistical summary
Bias RMSE
Met office 14.2 33.9 0.55
TAMSAT 4.3 18.1 0.72
2R
Current and Possible Future work
• Use of historical local radar data (AMMA or TRMM) to calibrate the Met office algorithm
• Use of historical gauge data to constrain or calibrate an MSG based algorithm
• Neural network based algorithm