analisi modellistica di un evento estremo: 3-4 novembre 1966 risultati preliminari dal progetto...

Analisi modellistica di un evento estremo: 3-4

Novembre 1966

Risultati preliminari dal progetto congiunto Flood 66 tra:

ECMWF (R. Buizza, S. Uppala, E. Klinker),

ISAC-CNR (A. Buzzi, P.Malguzzi),

Università di Brescia, Dip. Ing. Civile (G. Grossi, R. Ranzi, B. Bacchi)

FLOOD 66

Major widespread hystorical storm and flood event, which caused disasters mainly in Tuscany (Florence, Grosseto and other towns) and in other parts of central and north-eastern Italy, such as Venice (highest tide on record), Trento and in many valleys situated in Veneto and Friuli, in Eastern Alps, Emilia etc. in the period 3-5 November.

Firenze

Dolomiti bellunesi

Grosseto

Venezia

Objectives:Use the ERA database to analayze the extreme

event and to assess the value of the ERA-40 for a-posteriori studies of the ECMWF forecasting system value for weather risk assessment.

Examine the accuracy of deterministic and probabilistic forecasts during the period of the 1966 flood in Florence and in the Eastern Alps

Investigate the possibility of using forecasts to drive river-basin discharge hydrological models using direct model output from a meteorological model chain, including hydrostatic and non-hydrostatic, cloud-resolving models.

Tuscany (Arno and Ombrone River basins)

extreme event for precipitation intensity, extension and continuity – most of the rain fell on 4 November

200

200

300

200

400

The Eastern Alps (Adige and other river basins)

250

250

400500

700

500400

Most of precip. fell 3-4 November)

Z500 TL511L60 analysis: 12 UTC of 1-4 November

This figure shows the TL511L60 analyses at 12UTC of 1-4 November.

Contour interval is 4dam.

ERA40 and TL511L60 analyses: 12 UTC of 1-2 November

The ERA-40 (TL159L60, left) and the TL511L60 (right) analyses at 12UTC of 1 and 2 November are very similar (ci is 2dam for full fields and 1dam for difference).

MSLP, 4 Nov. 1966, 00 and 12 UTC.

ECMWF ERA-40 special high-resolution (TL511L60), interpolated on the BOLAM grid

950 hPa wind, 4 Nov. 1966, 00 and 12 UTC.

ECMWF ERA-40 special high-resolution (TL511L60), interpolated on the BOLAM grid

500 hPa wind, 4 Nov. 1966, 00 and 12 UTC.

ECMWF ERA-40 special high-resolution (TL511L60), interpolated on the BOLAM grid

24h TP: ECMWF TL511L60 deterministic

forecasts for 3-4 Nov (168h, 120h, 72h)

This figure shows TL511L60 forecasts of 24h-accumulated precipitation started at 12UTC of the 27 (144-168h) and 29 (96-120h) October, and 1 November (48-72h) and valid for 3-4 November. The right-bottom panel shows a proxi for verification defined by the TL511L60 24h forecast started at 12UTC of 3 November.Contour isolines are 2-25-50-75-100-150-300 mm for precipitation.

t+48-72h

t+96-120ht+144-168h

3-4 Nov

24h TP: EPS fc from 30 Oct for 3-4 Nov (96-120h)

This figure shows three EPS (51*TL255L40) probabilistic forecasts started on 30 October (96-120h) and valid for 3-4 November, for 24-h accumulated precipitation in excess of 25, 50 and 75 mm. The right-bottom panel shows a proxi for verification given by the TL511L40 24h forecast started on 3 November.Contour isolines are 2-10-20-40-60-100% for probabilities and 2-25-50-75-100-150-300 mm for precipitation.

P(TP>75mm)

P(TP>25mm

)

P(TP>50mm)

3-4 Nov

Good EPS t+96-120h fc over Friuli for all thresholds, and some signals that 75mm could hit also Tuscany …

Good EPS t+96-120h fc over Friuli for all thresholds, and some signals that 75mm could hit also Tuscany …

24h TP: EPS fc from 31 Oct for 3-4 Nov (72-96h)

This figure shows three EPS (51*TL255L40) probabilistic forecasts started on 31 October (72-96h) and valid for 3-4 November, for 24-h accumulated precipitation in excess of 25, 50 and 75 mm. The right-bottom panel shows a proxi for verification given by the TL511L40 24h forecast started on 3 November.Contour isolines are 2-10-20-40-60-100% for probabilities and 2-25-50-75-100-150-300 mm for precipitation.

P(TP>75mm)

P(TP>25mm) P(TP>50mm)

3-4 Nov

EPS t+72-96h fc does not propagate quickly enough, but there is a stronger signal that 75mm could hit also Tuscany …

Compared to previous fc, consistent but delayed signal, with more localized probability values ..

EPS t+72-96h fc does not propagate quickly enough, but there is a stronger signal that 75mm could hit also Tuscany …

Compared to previous fc, consistent but delayed signal, with more localized probability values ..

24h TP: EPS fc from 1 Nov for 3-4 Nov (48-72h)

This figure shows three EPS (51*TL255L40) probabilistic forecasts started on 30 October (96-120h) and valid for 3-4 November, for 24-h accumulated precipitation in excess of 25, 50 and 75 mm.

The right-bottom panel shows a proxi for verification given by the TL511L40 24h forecast started on 3 November.

Contour isolines are 2-10-20-40-60-100% for probabilities and 2-25-50-75-100-150-300 mm for precipitation.

P(TP>75mm)

P(TP>25mm) P(TP>50mm)

3-4 Nov

Good EPS t+48-72h fc for all thresholds …Compared to previous fc, consistent but with higher and correctly localized probability values ..

Good EPS t+48-72h fc for all thresholds …Compared to previous fc, consistent but with higher and correctly localized probability values ..

Rianalisi ECMWF (T511, 45 km)

12 UTC 1 Nov 66

2 nov 4 nov3 novBOLAM 18

BOLAM 6 km

Accumulated precipitation period

MOLOCH 2 km

5 nov

06 UTC 3 Nov

06 UTC 5 Nov

00 UTC 2 Nov

00 UTC 3 Nov

The BOLAM 6km precipitation forecast

The MOLOCH 2 km precipitation forecast

The MOLOCH 2 km precipitation forecast

The MOLOCH 2 km precipitation forecast but starting with the NCEP reanalysis (300 km!)

ECMWF NCEP

Short digression about ECMWF vs NCEP reanalyses …

ECMWF NCEP

ECMWF NCEP

ECMWF NCEP

Some sensitivity experiments: role of orography and sea surface fluxes…

Case with flattened orography : stronger cyclone, slightly weaker LLJ and much weaker precipitation (a 24 hour FC)

Some sensitivity experiments: role of orography and sea surface fluxes…

Reference case: a 24 hour FC

Accumulated precipitation with different sea surface fluxes …

Case with different surface fluxes

Reference

No sens. lat. heat fluxes

SST + 3 K

Isarco a Chiusa A=3059 km²

Adige a Trento A=9763 km²

Noce a S.Giustina A=1050 km²

Avisio a Lavis A=934 km²

Adige a Bronzolo A=6926 km²

Simulazioni idrologiche

Adige River at Bronzolo [6926 km²]

0

500

1000

1500

2000

2500

3 4 5 6 7

days of November 1966

dis

charg

e [

m³/

s]

02468101214

pre

cip

itati

on

[m

m]

observed precipitation prec. MOLOCHprec. BOLAM observed runoffsimulated raingauge runoff simulated Bolam runoffsimulated Moloch runoff

Esempio di risultato della modellistica idrologica (Univ. di Brescia)

Sieve River at Fornacina [831 km²]

0200400600800

1000120014001600

3 4 5 6 7days of November 1966

dis

charg

e [

m³/

s]

0246810121416 p

reci

pit

ati

on [

mm

]

observed precipitation prec. BOLAMprec. MOLOCH observed runoffSimulated raingauge runoff simulated BOLAM runoffsimulated MOLOCH runoff

Conclusioni

L’evento viene ‘ricostruito’ in maniera realistica, nonostante le incertezze nelle analisi e le notevoli differenze tra l’analisi NCEP e quella ECMWF, probabilmente associate alle incertezze sull’Atlantico.

Sorprende la possibilità di ricostruire campi di precipitazione dettagliati a partire dalle analisi NCEP, persino meglio che con le analisi ECMWF.

L’origine appare nelle scale grandi (crescita di un’onda baroclina, associata a jet meridionale e intensa frontogenesi)

Nonostante questo, la predicibilità appare bassa prima di 3 giorni: la natura ‘estrema’ dell’evento non appare ad un orizzonte > 2-3 gg.

Precipitazione orografica sulle Alpi, apparentemente convettiva sull’Italia centrale, quest’ultima predicibile solo con il modello non idrostatico.

L’orografia determina totalmente la distribuzione della precipitazione ‘orografica’ e la ciclogenesi; i flussi superficiali sono relativamente poco importanti.