4B.4 IMPROVING RADAR RAINFALL MEASUREMENT STABILITY USING MOUNTAIN RETURNS IN
REAL TIME
Daniel Sempe e-To es* (1), Ra ael Sanchez-Diezma (1), Ma c Be engue (1), Ramon Pascual (2), Isz a
Zawadzki (3).
(1) G up de Rece ca Aplicada en Hid ome eo ologia. Uni e si a Poli ecnica de Ca alunya, Ba celona (Spain).
(2) Cen e Me eo ologic Te i o ial de Ca alunya, Ins i u o Nacional de Me eo ologia, Ba celona (Spain).
(3) Ma shall Rada Obse a o y, McGill Uni e si y, Mon eal (Quebec, Canada).
1. INTRODUCTION
A enua ion is one o he mos se e e p oblems
a ec ing he ain all es ima es by me eo ological
ada s a sho wa eleng hs (5 cm and less).
The e o e, mos o he ope a ional ne wo ks,
(essen ially C-band ada s ou side USA), a e se iously
a ec ed by a enua ion, pa icula ly in Eu ope. In
gene al ada a enua ion is ela ed o he e lec i i y
o he a ge s along a gi en beam pa h, and i is a ield
wi h alues a ying all o e ada measu emen s.
Addi ionally, in some cases, a enua ion can a ec
he whole ada e lec i i y ield due o he occu ence
o hea y ain o e he adome. In his case, he we
adome su ace can a enua e he signal in all
di ec ions (Manz e al., 1999), p oducing an e ec o
gene al subde ec ion on he measu ed e lec i i y
ields. A simila e ec could also be obse ed due o
ins abili ies on he ada signal ela ed o elec onic
luc ua ions, which could lead o gene al unde o
o e es ima ions o he en i e ield (Della-B una e al.,
1997). This kind o o e /subde ec ion could become
c i ical in some cases, and i also may lead o
imp ope ly apply o he usual co ec ion p ocedu es i
signal s abili y is no con olled and co ec ed be o e.
Du ing he e en o June he 10 h, 2000 in
Ca alunya, he ada o he Spanish Ins i u o Nacional
de Me eo ología in Ba celona (INM) su e ed an
imp essi e case o gene al subde ec ion (see Figu e
1). This se e e ain all e en yielded mo e han 200
mm in less han 3 hou s, p oducing impo an
damages (a b idge in he mo o way be ween Mad id
and Ba celona collapsed).
When he squall line app oached he ada , he
educ ion o he signal be ween wo consecu i e
images was so impo an ha i led us o ca e ully
s udy he e en , and o de elop an algo i hm able o
de ec signal ins abili ies o he ada measu emen s
and wa ning in eal- ime. A i s algo i hm was
de eloped and implemen ed a he INM cen e in
Ba celona in 2000. This algo i hm was based on he
Moun ain Re e ence Technique (Del ieu and C eu in,
1995) using he moun ain e u ns as e e ences o
con ol s abili y (Sempe e-To es e al., 2001). A e
wo yea s o ope a ional expe ience he algo i hm has
been imp o ed and e o mula ed.
2. ASCMORE: A enua ion and S abili y Con ol
using he MOun ain RE u ns.
The basis o he p oposed me hodology is o
con ol ada signal s abili y h ough he analysis o
empo al a ia ions o g ound clu e e u ns. The wo
essen ial hypo hesis a e ha (a) luc ua ions a ec all
pixels in he same way, (b) he moun ains’ e u n
* Co esponding au ho add ess: Daniel Sempe e-To es,
GRAHI-UPC, Jo di Gi ona, 1-3 D1, E-08034 Ba celona e-
mail: [email p o ec ed].
dis ibu ion shape is no a ec ed, bu jus biased by a
cons an ac o . So he compa ison o he obse ed
clu e agains he mean clu e map can p o ide a
measu emen o he s abili y o he signal.
To con ol he co ec pe o mance o he ada , a
mean clu e map has been gene a ed o he
olume ic scanning C-band ada o he INM loca ed
a Ba celona Main ada cha ac e is ics a e 0.9° 3-dB
beamwid h, λ=5.6 cm, 20 ele a ion angles. A se ies o
700 ada maps (i.e app oxima ely 5 days o ada
da a) we e selec ed o calcula e he mean map, in
which only alues o e 23 dBZ a e conside ed as
alid pixels ( o compa e only clea in e cep ions o he
main lobe). These da a we e ca e ully o e iewed o
ensu e non-p ecipi a ing condi ions and absence o
anomalous p opaga ion.
A each ime s ep ada s abili y is analyzed by
compa ing he g ound clu e measu ed a he i s
ele a ion o he p esen ada scan agains he
a e age clu e map, which is aken as he co ec
e e ence. The algo i hm is applied in wo s eps.
2.1. Selec ion o he alid pixels.
The obse ed echoes loca ed a he posi ion o he
labeled e e ence g oundclu e a e selec ed, and a
double c i e ia o alidi y is applied in o de o accep
hem as alid pixels. The wo selec ed es s a e:
2.1.1. Tes o ejec ion by ain
To a oid he use o g oundclu e pixels in which
he obse ed g ound echoes a e a ec ed by ain, he
egion o 5 km a ound he labeled pixel is analyzed
and only i he mean alue o his neighbo hood is
unde 20 dBZ he pixel is conside ed as alid o he
compa ison.
2.1.2. Tes o ejec ion by a enua ion
To a oid he use o pixels which alues could be
a ec ed by in- ain a enua ion, he Pa h In eg a ed
A enua ion (PIA) be ween he ada and he
g oundclu e loca ion is calcula ed using he di ec ly
compu ed PIA om he obse ed da a ( he ‘appa en ’
PIA sugges ed by Hildeb and, (1978) and Meneghini
(1978).
2.2. Compu a ion o he co ec ion ac o .
In his new e sion o he algo i hm, he
Cumula i e Dis ibu ion Func ion (CDF) o he
obse ed alues a he alid labeled pixels a e
compa ed wi h he CDF o he a e aged ones. The
co ec ion ac o is compu ed by he di e ence
be ween means, and a quali y c i e ia is calcula ed
using he Nash-e iciency be ween bo h CDFs.
The analysis o di e en case s udies ha e shown
ha his new me hodology o calcula e he co ec ion
ac o is be e han he one used in he i s e sion o
he algo i hm (using linea eg ession be ween
obse ed and e e ence alues). In pa icula his
me hod u ns ou o be less in luenced by ou laye s,
mo e s able unde anap op condi ions and i is able o
p opose ai es ima es e en wi h a educed numbe o
alid poin s (a less 10 poin s a e equi ed).
In he case o insu icien numbe o alid poin s o
i he simila i y be ween bo h CDFs is no high
(E iciency < 0.7), he co ec ion ac o canno be
compu ed as p oposed. The e o e, o a oid sudden
changes on he ada ields ( lashes be ween wo
consecu i e images due o he applica ion/non
applica ion o he co ec ion) he con inui y o he
ain all ield mean is imposed. So he CDF o he
p e iously obse ed ada scan and he p esen one
(only he pixels conside ed as ain all) a e compa ed
and he co ec ion ac o is calcula ed as he
di e ence be ween he means.
Some case s udies a e p esen ed in Figu e 2.
3. CONCLUSIONS
The expe ience o he ope a ional use o he
algo i hm and he analysis o di e en case s udies
ha e allowed us o poin ou he ollowing conclusions:
• E en du ing ligh ain e en s i exis s a enua ion
due o ain o e he adome, ha in C-band ada s
a e usually a ound 4 dB.
• When hea y ain is alling o e he ada , his
gene al a enua ion can p oduce se e e
unde es ima ion (se e al cases o e 10 dB ha e
been obse ed).
• Elec onic luc ua ions no ela ed o p ecipi a ion
e ec s can some imes appea . In he s udied
ada hey can oscilla e be ween +2 and –2 dB.
• The o iginal algo i hm was p ope ly wo king in mos
o cases, bu i was e y sensi i e o he numbe o
alid e e ence pixels, si ua ion ha occu s when
in ense cells a e jus o e he ada . The new
e sion o he algo i hm is able o wo k be e in
hese cases ( educed numbe o alid pixels) and
i is able o use he ain all ield o ensu e
con inui y. This solu ion can be accep able i i is
applied o ill sho pe iods (10 o 20 minu es).
As gene al conclusion i can be no ed ha he
moun ains echoes a e no jus an annoying p oblem
(some hing o elimina e) bu ha hey can be an use ul
e e ence, so i can be sugges ed o sa e and use
hem o imp o e ada da a quali y.
Acknowledgemen s: This s udy has been ca ied ou
in he amewo k o he EC p ojec VOLTAIRE (EVK2-CT-
2002-00155) and o he Spanish CICYT p ojec HIDRADMET
(REN2000-1755-C03-01). Thanks a e due o he Spanish
INM o p o iding ada da a, and o he ETSECCPB o
pa ially g an he pa icipa ion a he con e ence.
4. REFERENCES
Della-B una, G., J. Joss, and R. Lee, 1997: Au oma ic calib a ion and
e i ica ion o ada accu acy o p ecipi a ion es ima ion. 27 h
Con e ence On Rada Me eo ology, Vail, Co, USA.
Del ieu, G. and J. D. C eu in, 1995: Simula ion o ada moun ain
e u ns using a digi ized e ain model. J. A mos. Oceanic
Technol., 12, 1038-1049.
Hildeb and, P. H., 1978: I e a i e co ec ion o a enua ion o 5 cm
ada in ain. J. Appl . Me eo ., 17, 508-514.
Manz, A., J. Handwe ke , M. Lö le -Mang, R. Hannesen, and H. Gysi,
1999: Radome in luence on wea he ada sys ems wi h
emphasis o ain e ec s. 29 h Con e ence On Rada
Me eo ology, Mon eal, Quebec, Canada, 918-921.
Meneghini, R., 1978: Rain- a e es ima es o an a enua ing ada .
Radio Sci., 13, 459-470.
Sempe e-To es, D., R. Sánchez-Diezma, M. A. Có doba, R. Pascual,
and I. Zawadzki, 2001: An ope a ional me hodology o con ol
ada measu emen s s abili y om moun ain e u ns. 30 h
In e na ional Con e ence on Rada Me eo ology, Munich,
Ge many, 264-266.
Raw Da a June 10 h 2000 @ 0220
-100 -50 0 50 100
Dis ance om he ada (km)
-100
-50
0
50
100
Dis ance om he ada (km)
5
10
15
20
25
30
35
40
45
50
55
60
65
Z (dBZ)
Raw Da a June 10 h 2000 @ 0230
-100 -50 0 50 100
Dis ance om he ada (km)
-100
-50
0
50
100
Dis ance om he ada (km)
G oundclu e
Rada
Rada
Figu e 1: Consecu i e aw images om he INM ada a Ba celona du ing he e en o June 10 h 2000 a Ca alunya. In en minu es (be ween
02:20 and 02:30) he e is a gene al change due o a enua ion by ain o e he adome o mo e han 10 dB. The e ec is clea when
compa ing he g ound clu e echoes (ci cled) which should be o app oxima ely cons an alue.
e iciencia= 1.00
Means` di e ence= 0.42
e iciency= 1.00
20 30 40 50 60
dBZ
0.0
0.2
0.4
0.6
0.8
1.0
Cumula i e Dis ibu ion Func ion
e iciencia= 1.00
Means` di e ence= 4.2
e iciency= 1.00
20 30 40 50 60
dBZ
0.0
0.2
0.4
0.6
0.8
1.0
Cumula i e Dis ibu ion Func ion
e iciencia= 0.94
Means` di e ence= 9.8
e iciency= 0.94
20 30 40 50 60
dBZ
0.0
0.2
0.4
0.6
0.8
1.0
Cumula i e Dis ibu ion Func ion
Figu e 2: Compa ison o Moun ain e u ns CDFs o h ee cases: a) Clea ai si ua ion on 21/10/00 @ 07:00, 862 alid poin s; b) Ligh ain o e
he adome on 3/4/02 @ 15:30, 128 alid poin s and c) Hea y ain o e he adome on 15/07/01 @ 11:50, 8 alid poin s. In black he
e e ence mean CDF and in g ey he obse ed CDF co ec ed by he di e ence o means (should be iden ical).