Performance Evaluation of
P
i it ti
R ti
l over
Precipitation
Retrievals
South Eastern South America
South-Eastern
considering
different climatic regions
Paola Salio
DCAO - CIMA – University of Buenos Aires - CONICET. Argentina.
Daniel Vila
DSA – CPTEC. Brazil.
Maria Paula Houbouchian DCAO – University of Buenos Aires
Aires. Argentina
Argentina.
Yanina Garcia Skabar National Weather
Cynthia Matsudo
Service. Argentina.
National Weather Service. Argentina.
October 2010. 5th IPWG. Hamburg. Germany
Overview
 Objective and Motivation
 Methodology
 Results
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Overview
 Objective and Motivation
 Methodology
 Results
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Objecti e
Objective
Evaluate the p
performance of the g
gaugeg
blended precipitation retrievals and satellitebased retrievals on networks with high
spatial resolution over south-eastern South
America.
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Motivation: Why should we care about the performance
of the QPE over SE South America?
Most extreme MCS in TRMM sample over SA – 20 Dec 2003 8:24 Z
2A25 Precipitation mm h-11
85 GHz PCT
IR and LIS
Liu and Zipser 2007
Motivation:
Why should we care about the
performance of the QPE over SE South America?
% Precipitation explainded by
Area < 2000km2
40dBZ top < 6km
Small and shallow systems
Area > 2000km2
40dBZ top >= 6km
Large and deep systems
from Vidal at al 2010, Liu et al 2008, Liu and Zipser 2007 Zipser et 2006,
Nesbitt and Zipser 2003, among others
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Overview
 Objective and Motivation
 Methodology
 Results
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Metodology:
1) Available overlapping periods for QPE
Short Period = July 12, 2009 to Dec 31, 2009
Gauge-Blended retrievals
Satellite Based-retrievals
CoSch (Vila et al 2009)
CMORPH (Joyce et al 2004)
3B42 V6 (H
3B42_V6
(Huffman
ff
ett all 2007)
3B42 RT (H
3B42_RT
(Huffman
ff
ett all 2003)
Long Period = Jan 1, 2003 to Dec 31, 2009
Gauge-Blended retrievals
Satellite based retrievals
3B42_V6
CMORPH
QPEs are accumulated over 24 hours at 12Z
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Metodology:
25º considering more
2) Interpolate precipitation observations to a common grid of 00.25
than 2560 stations not available on GTS (number of stations depends on the day).
Operative Network available on GTS e.g. 2010, Jul 21
122 stations in Argentina and Uruguay
150 Brazil
12 Paraguay
5 Bolivia
1141 stations in Argentina and Uruguay
1253 Brazil
12 Paraguay
5 Bolivia
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Metodology:
3) Evaluation of QPE performance considering different statistics:
RMSE, BIAS, BIAS SCORE, ETS, FAR, PODrain
 over the whole sample
 particular areas associated with dense networks and climatic regions
25 28 75S 64
57 75W)
BA= (39
(39.25-28.75S
64-57.75W)
SE=(30-25S
(
65-61.5W))
SG=(32.25-30S 58.5-55.75W)
SL=(36-31S 68-64W)
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Overview
 Objective and Motivation
 Methodology
 Results
Short Period Analysis
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Precipitation Mean mm/day July 2009 – Dec 2009
Bias
Estimated – Obs
J l 2009 – Dec
July
D 2009
Porcentual
Error
rainy days
July 2009 – Dec 2009
RMSE
July 2009 – Dec 2009
Frequency Precip >= 25 mm/day July 2009 – Dec 2009
BA Region
y to Dec 2009
July
3B42_V6
Probality Density Functions
PDF
18
ETS
0.3
ETS
20
0.4
CMORPH
0.2
CoSch
3B42_RT
16
0.1
14
3B42_V6
12
CMORPH
10
CoSch
0
OBS
8
1
3
5
10
15
20
30
50
3B42_RT
6
4
1
2
FAR
0
0.1
1
3
5
10
15
20
30
0.9
+50
3B42_V6
0.8
CMORPH
FAR
2
CoSch
0.7
1.8
3B42_RT
1.6
0.6
1.2
1
0.8
1
3
5
10
15
20
30
50
3B42_V6
CMORPH
CoSch
3B42_RT
0.5
1
3
5
10
15
20
30
50
0.6
0.8
0.4
0.2
0
PODrain
0.7
Bias Score
0.6
3B42_V6
0.5
POD
Bias Score
1.4
CMORPH
0.4
CoSch
0.3
3B42_RT
02
0.2
0.1
0
1
3
5
10
15
20
30
50
Overview
 Objective and Motivation
 Methodology
 Results
Extreme RMSE Events in the short sample
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Root Mean Square over BA box
45
Dec 20
20, 2009
40
35
30
25
20
15
10
5
0
12/07/2009
12/08/2009
3B42_V6
_
12/09/2009
12/10/2009
CMORPH
12/11/2009
CoSch
Jul 21, 2009
12/12/2009
3B42_RT
_
Precipitation average over BA box
45
40
Dec 20, 2009
35
30
25
20
15
10
5
0
12/07/09
12/08/09
Observations
12/09/09
3B42_V6
12/10/09
CoSch
12/11/09
3B42_RT
12/12/09
CMORPH
Extreme RMSE Cases
J l 21
July
21, 2009 12UTC Wi
Winter
t C
Case - associated
i t d with
ith a cold
ld front
f
t and
d a MCSs
MCS
development
30
25
Probability density funtions
20
3B42_V6
CMORPH
15
CoSch
OBS
10
3B42_RT
5
0
0.1
1
3
5
10
15
20
30
+50
Extreme
RMSE
Cases
Dec 20,, 2009 Squall
q
line event
30
25
Probability density funtions
20
3B42_V6
CMORPH
15
CoSch
OBS
10
3B42_RT
5
0
0.1
1
3
5
10
15
20
30
+50
Overview
 Objective and Motivation
 Methodology
 Results
Long Period Analysis
 Conclusions and Future Steps
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Bias
Jan 2003 – Dec 2009
RMSE
Porcentual
Error
Jan 2003 – Dec 2009
BA Region – Long Period
close 500 stations
20
Probality Density FunctionsJJA
18
16
14
20
Probality Density Functions
18
16
12
3B42_V6
10
CMORPH
Obs CMORPH
8
6
14
4
12
3B42 V6
3B42_V6
2
10
CMORPH
0
Obs CMORPH
8
0.1
1
3
5
10
15
20
30
50
6
4
2
20
0
DJF
18
0.1
1
3
5
10
15
20
30
50
16
14
2
3B42 V6
3B42_V6
10
CMORPH
6
1.6
4
2
3B42_V6
1.2
1
0.8
1
3
5
10
15
20
30
50
CMORPH
06
0.6
0.4
0.2
0
Obs CMORPH
8
1.8
1.4
Bias Score
12
Bias Score
0
0.1
1
3
5
10
15
20
30
50
Santiago del Estero –
Long Period
Dry region
40 stations
Salto Grande –
Long Period
140 stations
20
20
Probality Density Functions
18
Probality Density Functions
18
16
16
14
14
12
3B42_V6
12
3B42_V6
10
CMORPH
10
CMORPH
Obs CMORPH
8
6
Obs CMORPH
8
6
4
4
2
2
0
0.1
1
3
5
10
15
20
30
0
50
0.1
1
3
5
10
15
20
30
50
2
3
1.8
2.5
1.6
Bias Score
3B42_V6
1.5
CMORPH
1
1
0.5
0
3
5
10
15
20
30
50
Bias Score
1.4
2
3B42_V6
1.2
1
0.8
1
3
5
10
15
20
30
50
CMORPH
06
0.6
0.4
Bias Score
0.2
0
Bias Score
San Luis
Long Period
Dry region
40 stations
20
Probality Density Functions
18
16
14
12
3B42_V6
10
CMORPH
Obs CMORPH
8
6
4
2
0
0.1
1
3
5
10
15
20
30
50
3
2.5
Bias Score
2
3B42_V6
1.5
CMORPH
1
1
3
5
10
15
20
30
50
0.5
0
Bias Score
Summary

Overestimation is observed by all QPE techniques.
g
error values can be observed over central
Larger
section of SESA, associated with the area of
maximum extent of MCSs.
 Gauge-blended
g
retrievals show a better p
performance
than pure satellite estimates, showing different
behavior during stratiform and convective cases.
 Gauge-blended
g
retrievals over a network with large
g
number of observations and a long period of time,
affected principally by mature stage period of MCSs,
shows
a perfect
precipitation
h
f t fit exceptt for
f large
l
i it ti rates.
t
But over MCSs initiation area and dry regions,
problems are observed over light precipitation events
events.
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
MCS Mature stage centroid
position from IR data BT
lower -55ºC
Spring
p g
-10
-15
-20
-25
-30
-35
-40
-10-75
-70
-65
-60
-70
-65
-60
Summer
-55
-50
-45
-40
-15
-20
2
-25
-30
-35
-40
-75
Salio et al 2007
-55
-50
-45
-40
Future Steps
 Include
I l d
more available
il bl networks
k over the
h area.
 Evaluate
E l t
th
the performance
f
against
i t hourly
h l observations
b
ti
from manual and automatic stations principally over
d and
dry
d mountain
t i areas.
 Run
CoSch for a longer period to evaluate the
performance.
 Include
level 2 data (2A25 and 2A42) in the
evaluation
evaluation.
 Deploy
and participate on CHUVA experiment in Foz
do Iguazú, considering this experiment is over the
area with really strong MCSs.
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
CHUVA Project
Lead: Luiz Agusto Toledo Machado
Experiments
New experiment at Foz de Iguazu
From 10-2012 to 1-2013
Joint effort with LPB field activities
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany
Thanks for y
your attention
Paola Salio. October 2010. 5th IPWG. Hamburg. Germany