MeteoCAST: A Neural Ensemble
Nowcasting Model based on
Geostationary Multispectral Imagery for
Hydro-Meteorological Applic...
Outline









Introduction
The goal
The starting point
The model
The case studies
The rainfall estimation
The p...
Introduction






A relevant part of environmental risk can be
ascribed to meteorological severe events with
high prec...
The goal






Develop a model based on the MSG frames to
nowcast (from 30 Mins to 60 Mins) the rain field.
The model s...
The starting point


The NeuCAST (Marzano et al.)






Meteosat 7's images application
IR channel (10.8 μm) nowcast ...
The model: the multi-channels approach


MeteoCAST: Meteorological Combined
Algorithm for Storm Tracking







Appl...
The model: the multi-channels model tools








Cao’s method to find the optimal temporal
window
PCA (Principal Comp...
The model: the multi channels approach
layout

2013/09/16

Eumetsat Conference
2013 – Vienna
The case-studies




The area of interest ranges from longitude 7° E
to 18° E and from latitude 36.5° N to 48° N
Trainin...
The case studies: Ensemble setup






3 GLMs for each case-study: one GLM for the
lower correlation frame, one for th...
The case studies: the benchmarks


The Persistence

Ft + ∆t = Ft


The Steady State Displacement (SSD)

Ft + ∆t

2013/09...
The case studies: the performance indexes


BIAS
mε ( t k ) =



1
N points

RMSE
 1
sε ( t k ) = 
N
 points



∑ [...
The case studies: training set 60 mins
ahead mean performance
BIAS
2
1.5
1
K

0.5
0

2013/09/16

MeteoCAST

SSD

Persisten...
The case studies: training set 60 mins
ahead mean performance
RMSE
20
15
10
K
5
0

2013/09/16

MeteoCAST

SSD

Persistence...
The case studies: training set 60 mins
ahead mean performance
Correlation
100
80
60

% 40
20
0

2013/09/16

MeteoCAST

SSD...
The case studies: 2007/03/20 13:30 UTC
60 mins ahead
BIAS

2
1.5
1

K 0.5
0
-0.5

2013/09/16

MeteoCAST

SSD

Persistence
...
The case studies: 2007/03/20 13:30 UTC
60 mins ahead
RMSE
16
14
12
10
K 8
6
4
2
0

2013/09/16

MeteoCAST

SSD

Persistence...
The case studies: 2007/03/20 13:30 UTC
60 mins ahead
Correlation
100
80
60

%

40
20
0

2013/09/16

MeteoCAST

SSD

Persis...
The case studies: tornado over Modena

2013/09/16

Eumetsat Conference
2013 – Vienna
The case studies: tornado over Modena from
MSG

2013/09/16

Eumetsat Conference
2013 – Vienna
The case studies: 2013/05/03 14:00 UTC
60 mins ahead
BIAS

2.5
2
1.5

K

1
0.5
0

2013/09/16

MeteoCAST

SSD

Persistence
...
The case studies: 2013/05/03 14:00 UTC
60 mins ahead
RMSE

16
14
12
10
8
K 6
4
2
0

2013/09/16

MeteoCAST

SSD

Persistenc...
The case studies: 2013/05/03 14:00 UTC
60 mins ahead
Correlation

80
70
60
50
40
% 30
20
10
0

2013/09/16

MeteoCAST

SSD
...
The rainfall estimation




Use the produced synthetic images in a waterfall
manner
Some intermediate products are gener...
The rainfall estimation: the model layout
MSG
BTs

First level

Second level

Third level
2013/09/16

DEM

LST
Estimator

...
The rainfall estimation : tornado over Modena

Performance Indexes 60 Min
BIAS

mm/h

RMSE

10.49

mm/h

Correlation

2013...
The rainfall estimation: a static case
2010/01/26 10:15 UTC - 60 Mins ahead.

Performance Indexes 60 Min
BIAS

1.33

mm/h
...
The present







The www.mondometeo.org website publishes
the near real time outputs of the MeteoCAST
model
The KMZ...
The future











CellTrack integration (attend the talk of Davide
Melfi tomorrow morning)
RSS integration in or...
Acknowledgements

Thanks to the Italian Air Force
Meteorological Office

for the support

2013/09/16

Eumetsat Conference
...
Thank you for your attention
michele.derosa@geok.it
mic_der@yahoo.it

2013/09/16

Eumetsat Conference
2013 – Vienna
Upcoming SlideShare
Loading in...5
×

MeteoCAST: a nowcasting model to predict extreme meteorological events

123

Published on

Presentation about the MeteoCAST very short term forecast model at the Eumetsat Conference 2013.

Published in: Education, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
123
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

MeteoCAST: a nowcasting model to predict extreme meteorological events

  1. 1. MeteoCAST: A Neural Ensemble Nowcasting Model based on Geostationary Multispectral Imagery for Hydro-Meteorological Applications Dr. Michele de Rosa1,2, Prof. Frank S. Marzano1 1. “Sapienza” University of Rome, via Eudossiana, 18 - 00184 Rome – Italy 2. GEO-K srl, via del Politecnico, 1 – 00133 Rome - Italy 2013/09/16 Eumetsat Conference 2013 – Vienna
  2. 2. Outline         Introduction The goal The starting point The model The case studies The rainfall estimation The present The future 2013/09/16 Eumetsat Conference 2013 – Vienna
  3. 3. Introduction    A relevant part of environmental risk can be ascribed to meteorological severe events with high precipitation rate. Heavy precipitation associated to severe weather may cause serious damages in terms of economic losses and, in extreme cases, of human life losses. Managing the environmental risk due to precipitation is strictly linked to monitoring and understanding the storms that produces hazards such as flash floods. 2013/09/16 Eumetsat Conference 2013 – Vienna
  4. 4. The goal    Develop a model based on the MSG frames to nowcast (from 30 Mins to 60 Mins) the rain field. The model should predict the MSG IR channels in order to predict the rain field. The model should be flexible, accurate and quick. 2013/09/16 Eumetsat Conference 2013 – Vienna
  5. 5. The starting point  The NeuCAST (Marzano et al.)     Meteosat 7's images application IR channel (10.8 μm) nowcast (30 mins) Rain estimation from MW and IR sources, using the nowcasting of the IR channels Model for IR-RR mapping (Neural net) 2013/09/16 Eumetsat Conference 2013 – Vienna
  6. 6. The model: the multi-channels approach  MeteoCAST: Meteorological Combined Algorithm for Storm Tracking      Application on MSG images IR channels (4,5,6,7,8,9,10,11) nowcasting from 30 mins to 60 mins Bayesian approach to train the model GLM nowcasting model Model for IR-to-Rain Rate mapping 2013/09/16 Eumetsat Conference 2013 – Vienna
  7. 7. The model: the multi-channels model tools     Cao’s method to find the optimal temporal window PCA (Principal Component Analysis) to reduce the number of information sources: the 8 IR channels are replaced by a linear combination of them. Bayesian model to make nowcasting about the next MSG image The Dynamically Averaging Network (DAN) Ensemble 2013/09/16 Eumetsat Conference 2013 – Vienna
  8. 8. The model: the multi channels approach layout 2013/09/16 Eumetsat Conference 2013 – Vienna
  9. 9. The case-studies   The area of interest ranges from longitude 7° E to 18° E and from latitude 36.5° N to 48° N Training   Validation   2007-03-20 Test   2006-07-24, 2006-08-13, 2006-09-14 2013-05-03 Each frame consists of 275x344 pixels 2013/09/16 Eumetsat Conference 2013 – Vienna
  10. 10. The case studies: Ensemble setup     3 GLMs for each case-study: one GLM for the lower correlation frame, one for the higher correlation frame and one for the median correlation frame (like the worst, best and mean case in computer science). 3 PCA channels 9 components and 27 GLMs Each bayesian GLM consists of 726 inputs Pixel to project ahead (i,j) (nc=5, embed=6), 1 output. 2013/09/16 Eumetsat Conference 2013 – Vienna
  11. 11. The case studies: the benchmarks  The Persistence Ft + ∆t = Ft  The Steady State Displacement (SSD) Ft + ∆t 2013/09/16  = Ft + v Eumetsat Conference 2013 – Vienna
  12. 12. The case studies: the performance indexes  BIAS mε ( t k ) =  1 N points RMSE  1 sε ( t k ) =  N  points  ∑ [T ( P ,t ) − T ( P ,t ) ] est b i k b i k 1 2 2 est ∑ Tb ( Pi ,tk ) − Tb ( Pi ,tk )    [ ] Correlation index rε (t k ) = [ ∑T ( P ,t ) − T (t ) ][T ( P ,t ) − T (t ) ] est b [  ∑T  2013/09/16 est b i est b k ( Pi ,t k ) − T (t k ) est b k b i k ] ∑T ( P ,t [ 2 Eumetsat Conference 2013 – Vienna b i b k ) − Tb (t k ) ] 2 k 1 2  
  13. 13. The case studies: training set 60 mins ahead mean performance BIAS 2 1.5 1 K 0.5 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  14. 14. The case studies: training set 60 mins ahead mean performance RMSE 20 15 10 K 5 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  15. 15. The case studies: training set 60 mins ahead mean performance Correlation 100 80 60 % 40 20 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  16. 16. The case studies: 2007/03/20 13:30 UTC 60 mins ahead BIAS 2 1.5 1 K 0.5 0 -0.5 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  17. 17. The case studies: 2007/03/20 13:30 UTC 60 mins ahead RMSE 16 14 12 10 K 8 6 4 2 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  18. 18. The case studies: 2007/03/20 13:30 UTC 60 mins ahead Correlation 100 80 60 % 40 20 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  19. 19. The case studies: tornado over Modena 2013/09/16 Eumetsat Conference 2013 – Vienna
  20. 20. The case studies: tornado over Modena from MSG 2013/09/16 Eumetsat Conference 2013 – Vienna
  21. 21. The case studies: 2013/05/03 14:00 UTC 60 mins ahead BIAS 2.5 2 1.5 K 1 0.5 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  22. 22. The case studies: 2013/05/03 14:00 UTC 60 mins ahead RMSE 16 14 12 10 8 K 6 4 2 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  23. 23. The case studies: 2013/05/03 14:00 UTC 60 mins ahead Correlation 80 70 60 50 40 % 30 20 10 0 2013/09/16 MeteoCAST SSD Persistence Eumetsat Conference 2013 – Vienna Ch. Ch. Ch. Ch. Ch. Ch. Ch. Ch. 10 11 4 5 6 7 8 9
  24. 24. The rainfall estimation   Use the produced synthetic images in a waterfall manner Some intermediate products are generated:     CM LST RR Integration with the PGE01 and PGE05 products of the NWCSAF (for calibration purposes) 2013/09/16 Eumetsat Conference 2013 – Vienna
  25. 25. The rainfall estimation: the model layout MSG BTs First level Second level Third level 2013/09/16 DEM LST Estimator GLM Cloud Mask RR Classifier RR Estimator Eumetsat Conference 2013 – Vienna
  26. 26. The rainfall estimation : tornado over Modena Performance Indexes 60 Min BIAS mm/h RMSE 10.49 mm/h Correlation 2013/09/16 2.17 53.00 % Eumetsat Conference 2013 – Vienna
  27. 27. The rainfall estimation: a static case 2010/01/26 10:15 UTC - 60 Mins ahead. Performance Indexes 60 Min BIAS 1.33 mm/h RMSE 9.05 mm/h Correlation 2013/09/16 Eumetsat Conference 2013 – Vienna 68.47 %
  28. 28. The present      The www.mondometeo.org website publishes the near real time outputs of the MeteoCAST model The KMZ service The Augmented Reality service The Twitter service Covered countries: Italy, Swiss, Austria (almost all covered) and Brazil (Sao Paulo region). 2013/09/16 Eumetsat Conference 2013 – Vienna
  29. 29. The future       CellTrack integration (attend the talk of Davide Melfi tomorrow morning) RSS integration in order to improve the performance on heavy dynamic events Integration with the NWCSAF v.2013 Synthetic images (extended to VIS) as input to the NWCSAF Coverage of other countries: Africa and South America Extension to other satellites: GOES and MTSAT 2013/09/16 Eumetsat Conference 2013 – Vienna
  30. 30. Acknowledgements Thanks to the Italian Air Force Meteorological Office for the support 2013/09/16 Eumetsat Conference 2013 – Vienna
  31. 31. Thank you for your attention michele.derosa@geok.it mic_der@yahoo.it 2013/09/16 Eumetsat Conference 2013 – Vienna
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×