Upload
pierluigi-claps
View
464
Download
3
Tags:
Embed Size (px)
DESCRIPTION
Citation preview
Deciphering River Flood Change Vienna, 3-5 September 2012
Extremes of the extremesStill too few to analyse their change?
Pierluigi Claps Politecnico di Torino
www.idrologia.polito.it
www.idrologia.polito.it 2
Hydrologic Monsters
Black Swans: Unexpected phenomena in hydrology
RELATED MEETINGS
www.idrologia.polito.it 3
LITERATURE DISCUSSIONS
4
my meaning:
floods caused by big rainfall outliers Rainfall observations exceedingly different from the rest of the station record
Values almost statistically impossible to observe, in ‘stationary’ terms, in a lifetime (if considering only the station record):
Hovever….if we are observing them relatively frequently in a region, they should not be as highly unlikely!
Definition of REGION is crucial for a correct transfer of information in space
www.idrologia.polito.it
5
Some of the worst Italian Big rainfall Outliers
www.idrologia.polito.it
Cetara 1910It was not an earthquake!
www.idrologia.polito.it 6
Molare disaster (Ortiglieto dam)
www.molare.net
13 August 1935554 mm in 24 h (>30% MAP)111 victims
1954 – Salerno Disaster
25/26 October 1954Worst catastrophe in Italy (318 victims) due to rainfall-induced floods 504 mm in 24 h (40% of the MAP) www.idrologia.polito.it
www.idrologia.polito.it 8
Genova 1970: largest daily rainfall in Italy
7/8 October 1970, 948 mm in 24 h, (90% of the MAP)43 victims
www.idrologia.polito.it 9
The november 2011 “black swan”
Precipitation maps of the two eventsSource: Italian Civil Protection
www.idrologia.polito.it 10
1 Vernazza-Monterosso Oct 25-26 2011
Liguria – Cinque Terre Brugnato station• 143 mm in 1 h• 328 mm in 3 hs• 469 mm in 6 hs (Salerno 1954)
• 511 mm in 12 hs• 538 mm in 24 hs• 542 mm in 30 hs
• Italian Record break
13 victims
11www.idrologia.polito.it
http://www.flickr.com/people/chiara-sibona/
Monterosso
Vernazza
www.idrologia.polito.it 12
Cinque terre event: How extreme?
Pignone: middle-age bridge
www.idrologia.polito.it 13
2 Genova - Nov 4 2011
Radar reflectivity and Sea surface temperature anomalies From Parodi et al., Eos, Vol. 93, N
6 victims
• 181 mm in 1 h (Capoterra 2008)
• 336 mm in 3 hs (Giffone, 1959 )
• 385 mm in 6 hs• 411 mm in 12 hs
Vicomorasso station
www.idrologia.polito.it 14
180 mm/h =50 m3/s/km2
www.idrologia.polito.it 15
12:58:3712:46:41
12:30:00
Via Fereggiano
School exit time
Same synoptyc meteorological configuration for the 2 events
• Pronounced moist air advection from the subtropical Mediterranean areas near Africa.
• Significant sea surface temperature (SST) anomaly (>1°C).
• Strong pressure ridge centered on eastern Europe acting as a block to the motion of the system.
www.idrologia.polito.it 17
Meteorological “bombs”
Quezzi amateur station Genova (11/4/2011)Avg hourly intensity 170 mm/h460 mm / 6 hTotal rainfall (12 h) 556 mm
www.idrologia.polito.it 18
Social relevance of the 2011 Genoa Flash Flood
• Scary event (in an advanced and big city)• Very well forecasted, except exact position, yet
producing victims• Miscommunication between Weather Forecast
Service and Local Civil Protection• Children and schools involved
www.idrologia.polito.it 19
Social relevance of the Big rainfall Outliers and their consequent flash floods
• Cause more and more fatalities, as compared to lowland floods
• Isolated, very fast events, huge unit discharges• Very small basins, interaction with urban
infrastructures (often river reach not even present)• Flood hazard evenly distributed in the space• Frequently associated to heavy debris transport
(coastal hillslopes)
www.idrologia.polito.it 20
Questions arising:
• “Man in the street” question: why are we still calling them exceptional? How frequent are they REALLY?
• Civil Engineer question: How do these ‘new’ observations should impact on design rainfalls?
• Scientist questions: Are STAT toolkits for frequency assessment adequate? Will the frequency increase in the near future?
www.idrologia.polito.it 21
Regional or At-Site estimation of T(h)?
Kysely et al. (2011) and many other conclude with: REGIONAL!
2011
although “there are no means of validating the estimates”
All regional models (ROI) agree on return periods in the order of several hundreds of years;
www.idrologia.polito.it 22
However…. Traditional regional methods must face the effects of variable record lengths
www.idrologia.polito.it 23
Another problem (specially in Italy):station birth/death
ARPA (1988-2006)
SIMN (1928-1996)
active rain gauges over Piemonte - Valle d’Aosta 1928-2006
www.idrologia.polito.it 24
Spatial correlation of eventsROI or and other regional methods cannot account if similar events occur in different years or in the same year. (If similar events occur in different years the individual hazard is higher)
De Luca et al, 2010
www.idrologia.polito.it 25
Why not At-Site estimation of T(h)?
At site estimation depends whether or not the last observation is included in the sample(*) (T=283 vs. T=45110)
(*) Answer: the outlier can be included in the sample if the Maximim-Value test is passed
www.idrologia.polito.it 26
Maximum-Value TestGrubbs (1969) Rossi et al. (1984)
Test question: Can the maximum value X(n) (of a sample of lenght n) considered as extracted from the proposed parent distribution FX(x) , with a significance level a ?
0 500 1000 1500 2000 2500 3000
0.01 0.02 0.05 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80
0.90
0.95
0.99
0.995
0.999
Cu
mu
lativ
e P
rob
ab
ility
2
5
10
20
100
[T]
Annual maximum flood peaks [m3/s]
EV1200
1000
Test Answer
www.idrologia.polito.it 27
where X(n) is the sample maximum value, Q is the estimated parameter set and a is the significance level of the test.
YES, if
However, Q should not depend on X(n) itself!
Modified maximum value Test Laio, Allamano, Claps, HESS, 2010
Suggested solution: observed X(n) can be substituted by the median of the distribution of the n-year maxima.
In distributions characterized by position, scale and shape parameters:
z(u,θ3) is the quantile function of the standardized variable z=(x−θ1)/θ2, which only depends on the probability level u and on the shape parameter θ3
THE USE OF L-MOMENTS ESTIMATED MAKES THE TEST EXPLICIT
www.idrologia.polito.it 29
To get the best from local information, despite data patchiness….
Year-after-Year KrigingSpatial variability of extreme precipitation in each year, using the available data
1930 … 1950 … 1970 … 1990 … 2004…
At-a-point Time series of the annual maximum precipitation for d=1h
Allamano et al., (2011). GEOPHYSICAL RESEARCH ABSTRACTS,
Weighted by the annual
variance of prediction
L-moments computed on reconstructed data“local” T(h) estimation and verification possible
Average h1
a)
L-CV (h1)
L-CA (h1)
Respective of the spatial correlation of events
www.idrologia.polito.it 32
What about h(T) in ungauged sites?
• “Man in the street” question: why are we still calling them exceptional? How frequent are they REALLY?
• Civil Engineer question: How do these ‘new’ observations should impact on design rainfalls?
• Scientist questions: Are STAT toolkits for frequency assessment adequate? Will the frequency increase in the near future?
www.idrologia.polito.it 33
h(T) as a “man in the street” question:
Limited response with the individual-year data reconstruction
Given a new ‘VERY UNUSUAL’ event in a station, shouldn’t it affect the rainfall hazard in the surroundings?
Piemonte, h12, 1935
Probability distributions for h(T): mixed?
www.idrologia.polito.it 35
About the Physically-based grounds of the TCEV
Conte, M., P
iervitali, E., a
nd Colacino, M.: T
he
meteorological “bomb” in the Mediterranean,
in: INM/W
MO International symposiu
m on
cyclones and hazardous w
eather in the
Mediterranean, M
MA/UIB, pp. 283–287, 1997.
Meteorologists and Hydrologists collaborations
www.idrologia.polito.it 36
Meteorological basis for classification of events
Towards a physical-meteorological identification of more hazardous areas
www.idrologia.polito.it 37average cyclogenesis of all cyclones in the Mediterranean
Kouroutzoglou et al., Int. J. Climatol. (2011)
Why Italy in general …
… and Mar Ligure in particular
www.idrologia.polito.it 38
How to identify areas with “similar” Big Rainstorm distribution?
• Accurate selection of event occurrences
• More effort to select and verify very old events
• Interaction with meteorologists as regards statistical analysis
www.idrologia.polito.it 39
European-wide collaboration
www.idrologia.polito.it 40
ARCHIVAL SOURCES
Archivio S.I.C.I.
Annuario dati Ambientali
Archivio S.C.I.A.
Progetto Annali
Rapporti evento A.R.P.A.
Pubblicazioni locali
Siti Web amatoriali
Archivio “La Stampa”
>300 mm in 24 h
Macchia, 2011
Issues related to change
www.idrologia.polito.it 41
Number of observing stations
19231926
19291932
19351938
19411944
19471950
19531956
19591962
19651968
19711974
19771980
19831986
19891992
19951998
20012004
20072010
0
1
2
3
4
5
6
7
8
stations (th's)Events in Italy
www.idrologia.polito.it 42
Explosive cyclones in the Mediterranean (1962-2001)
NO TREND SO FAR!