high-resolution non-hydrostatic numerical weather prediction of mediterranean torrential rain events...

High-resolution Non-hydrostatic Numerical High-resolution Non-hydrostatic Numerical Weather Prediction of Mediterranean torrential Weather Prediction of Mediterranean torrential

rain eventsrain events

Véronique DUCROCQ, Cindy LEBEAUPIN, Olivier NUISSIER, Didier RICARD,

Hervé GIORDANI

CNRM/GAMEMétéo-France & CNRS

World Weather Research Program Symposium on Nowcasting and Very Short Range Forecasting, Toulouse, 5-9 Sept. 2005

Algiers, 10 Nov. 2001~260 mm in less than 24 hours>700 deaths, 4 billion € damages(Hamadache et al, 2002)

Gandia, 3 Nov. 1987 ~800 mm in 24 hours(Fernandez et al, 1995)

Vaison La Romaine, 22 Nov. 1992~300 mm in 24 hours>50 deaths, 1 billion € damages(Sénési et al, 1996)

Piedmont, 4-5 Nov. 1994~300 mm in 36 hours> 60 deaths, 12 billion € damages(Buzzi et al, 1998)

Spain

Italy

France

Mediterranean

Sea

Western Mediterranean region is prone to torrential rain events © Midi-Libre

Algiers, 10 Nov. 2001~260 mm in less than 24 hours>700 deaths, 4 billion € damages(Hamadache et al, 2002)

Gandia, 3 Nov. 1987 ~800 mm in 24 hours(Fernandez et al, 1995)

Vaison La Romaine, 22 Nov. 1992~300 mm in 24 hours>50 deaths, 1 billion € damages(Sénési et al, 1996)

Piedmont, 4-5 Nov. 1994~300 mm in 36 hours> 60 deaths, 12 billion € damages(Buzzi et al, 1998)

Spain

Italy

France

Mediterranean

Sea

Western Mediterranean region is prone to torrential rain events © Midi-Libre

Piedmont, 4-5 Nov. 1994~300 mm in 36 hours> 60 deaths, 12 billion €(Buzzi et al, 1998)

Spain

Italy

France

Mediterranean

Sea

For France, the southeastern France (Massif Central) is a threatened regionNumber of days withdaily precipitation > 200 mm from 1958 to 2000 for Southern France

Mas

sif

Cent

ral

Alps

Pyrenees

Toulouse

Gardon d’Anduze Watershed (545 km2)

Discharges

hou

rly r

aig

au

ge a

nd

rad

ar

rain

fall d

ep

ths

over

the w

ate

rsh

ed

660 700 740 780

Lam bert II é tendu (km )

1840

1880

1920

1960

2000

Lam

bert

II é

tend

u (k

m)

M t Aigoual

M ontpellier

Privas

M ende

M illau

Le Puy

Valence

Alès

M t M ézenc

M t Lozère

0

200

400

600

800

1000

1200

1400

1600

m ètres

HéraultVidourle

Gard

GardonsCèze

ArdècheChassezac

Eyrieux

Rhô

ne

A lot of rivers have their sources in the Massif Central Steep slopes of the rivers and small catchments increasing the speed of the streamflow

*AnduzeAnduze

Rainfall peak

LTHE

Flow peak

~6 h

The heavy precipitation touch areas that have a high-level flash-flood risk due to orography : small catchments (500km2-2000 km2) with fast responses to rain showers

Need for very short range model forecasts to provide longer lead times to prepare for flash-flooding and secure people.

High Conditional Convective instability

A very moist Low-Level Jet

Steep orography to

trigger/enhance the

convection

Slow progressing synoptic patterns

Quasi-stationary high-geopotentials over Central Europe

Mesoscale PV

anomalies

Synoptic forcing and topography forcing (reliefs + Mediterranean sea) increase the predictability of flash-flood driven thunderstorms.

Synoptic conditions propitious to torrential rain events

In most cases, the forecast of propitious conditions at synoptic scale allows to forecast the occurrence of an heavy precipitation event over the region, however it is still difficult to forecast : - the magnitude (normal or extreme heavy precipitation event) - the precise location (at a scale of less than 100 km).

100km

?

Vigilance map

High Conditional Convective instability

A very moist Low-Level Jet

Steep orography to

trigger/enhance the

convection

Slow progressing synoptic patterns

Quasi-stationary high-geopotentials over Central Europe

Mesoscale PV

anomalies

Synoptic forcing and topography forcing (reliefs + Mediterranean sea) increase the predictability of flash-flood driven thunderstorms.

Synoptic conditions propitious to torrential rain events

In most cases, the forecast of propitious conditions at synoptic scale allows to forecast the occurrence of an heavy precipitation event over the region, however it is still difficult to forecast : - the magnitude (normal or extreme heavy precipitation event) - the precise location (at a scale of less than 100 km).

The aim of the study is to examine the capability of a high-resolution non-hydrostatic model to forecast torrential rain events at very short range (in view of the next generation operational model AROME).

emphasis on the sensitivity to initial conditions (atmosphere and surface)

Date Type Maximum of 24-h rainfall

13-14/10/1995 quasi-stationary MCS ~ 250 mm

3-4 /10/1995 quasi-stationary MCS ~ 200 mm

12-13/11/1999The Aude extreme flash-flood event ~30deaths, 3 millions € damages

quasi-stationary MCS ~ 550 mm

6-7/10/2001 quasi-stationary MCS ~350 mm

8-9/09/2002 The Gard extreme flash-flood event~20 deaths, 1.2 billion € damages

quasi-stationary MCS and front ~ 700 mm

3/12/2003The 3rd day of a major Rhône flooding

quasi-stationary frontal system with embedded convection

~ 180 mm

including cases of

- quasi-stationary back-building MCS

- quasi-stationary frontal systems

The case studies

Gard flash-flood, 2002

Rhône flooding 3 Dec.2003

Torrential rain events over Southeastern France

Performed with the non-hydrostatic MESO-NH model (Lafore et al, 1998)

Two-way grid-nesting (Stein et al, 2000) :

•Domain 1 ~ 10 km•Domain 2 ~ 2.5 km (centred on the convective event)

Characteristics of the high-resolution simulations

Almost the same physical package as in AROME, including a bulk microphysic parameterization (Pinty et Jabouille, 1998, Caniaux,

1994) with 6 prognostic water variables : water vapour, cloud water rainwater, primary ice, graupel and snowThe convection is parameterized following the Kain and Fritsch parameterization for the 10-km domain whereas convection is explicitly resolved for the 2.5 km domain (no deep convection scheme).

10-km domain

2.5-km domain

~500-600 km

Initial conditions provided either by a large scale analysis (ARPEGE/IFS) or by a mesoscale analysis following Ducrocq et al, 2000 (objective analysis of mesonet surface observations, qv, qr and qs bogussing based on radar and satellite data)

>100 mm

>75 mm>50 mm>30 mm>20 mm

>10 mm

>5 mm>2 mm>1 mm

Rhône flood event (3 December 2003)

9-h accumulated precipitation from 03 UTC to 12 UTC, 3 Dec. 2003

Initial conditions : large scale ARPEGE analysis at 00 UTC, 3 Dec. 2003

Raingauge data 2.4 km MESO-NH forecast

100 km

+Nîmes

+Nîmes

Observations

Nîmes radar

Raingauges

Same Initial Conditions (ARPEGE analysis12UTC, 08/09/02) and same boundary conditions (3-hourly ARPEGE forecast)

+

12-h accumulated precipitation from 12 UTC, 8 Sept to 00 UTC, 9 Sept 2002

Gard flash-flood (8-9 Sept.2002)Aladin (9.4 km)operational forecast

MESO-NH (2.5km)

100 km

+Nîmes

+Nîmes

Observations

Nîmes radar

Raingauges

Initial Conditions : ARPEGE analysis12UTC, 08/09/02

+

MESO-NH (2.5km)

12-h accumulated precipitation from 12 UTC, 8 Sept to 00 UTC, 9 Sept 2002

Initial Conditions : Ducrocq et al (2000) Initialisation12UTC, 08/09/02

100 km

+

Gard flash-flood (8-9 Sept.2002)

SST for the Aude flash-flood case :

SST=17.8°C SST=17.4°C

- In-situ buoys and ships SST data analysis (ARPEGE)

- High-resolution satellite SST (AVHRR/NOAA )

Sensitivity to the Sea Surface Temperature of the Mediterranean Sea

- Warming [Cooling] of the ARPEGE analysis by 1.5°C (SST analysis error range)

Several SST fields tested :

Initial cond.:ARPEGE analysis, 12 UTC, 12 Nov. 1999

SST ARPEGE SST AVHRR

SST ARPEGE -1.5°CSST ARPEGE +1.5°C

18-hour accumulated precipitation from 12 UTC, 12 Nov. 1999 to 06 UTC, 13 Nov. 1999

Max:303mm Max:296mm

Max obs.:

485mm

Max:368mm Max:274mm

100 km

(mm)

Several cases of flash-flood have been simulated with the high-resolution non-hydrostatic MESO-NH model

get an insight into the capabilities of the next-generation operational high-resolution NWP systems, including the future Météo-France AROME model, to forecast Mediterranean torrential rain events.

In all the cases, the 2.5 km simulations improve the very short range QPF with respect to the current operational models. Forecast rainfall amounts are close to the observed ones for normal flash-flood event, but still underestimated for extreme events.

The improvement in location and in intensity is however in most cases a tribute to the mesoscale initial conditions. Dynamical adaptation from large scale analyses in most cases is not sufficient : mesoscale data assimilation improves rainfall forecast even in cases with strong synoptic forcing.

Primary importance of the assimilation of observations in the boundary layer (mesoscale surface observations) : cold pool, low-level convergence line, low-level moisture flow,

The SST acts to modulate the magnitude and location of the convective system : A significant dependency on the mean value of the SST over the Western Mediterranean basin, but a weak one to the mesoscale patterns of the SST field. also a role of the parameterization of the sea-atmosphere fluxes (not shown here).

Errors in location of the same order than the size of the small/medium watersheds plead in favor of a probabilistic post-processing of the high-resolution forecasts for hydrological applications.

Conclusion and outlook

Thank you for your attention

ARPEGE+1.5° ARPEGE

ARPEGE+3° ARPEGE-1.5° ARPEGE-3°

AVHRR

Aude flash-flood (12-13 Nov. 1999)

Synthetic infrared METEOSAT radiances at 00 UTC, 13 Nov. 1999 from 12-h MESO-NH forecasts using various initial SST fields

Initial conditions : large scale ARPEGE analysis at 12 UTC, 12 Nov. 1999