BenefitBenefit of convectiveof convective--scale meteorologicalscale meteorological models models

for for MediterraneanMediterranean flashflash--floods forecastingfloods forecasting

1 GAME/CNRM

(Météo-France, CNRS)

42 Avenue Coriolis31057 Toulouse Cedex 1

FRANCETel: 33 5 61 07 98 55 Fax: 33 5 61 07 96 26

Email: beatrice.vincendon@meteo.fr

The AROME deterministic QPF uncertainties

Ensemble streamflow forecast

Conclusions

Uncertainties in flash-flood forecasting

Larger uncertainties for the flash-flood forecast due to :

Ø catchment topography

Ø non-linearity of hydrological processes

Ø temporal/spatial scales of precipitating systems leading to flash-floods.

The Cévennes-Vivarais region with its three

main watersheds: Ardèche river

(green/2240km²), Ceze river (blue/1110km²)

and Gardons river (red/1090km²).

Hydrometeorological prediction is affected by several uncertainties:

Ø hydrological model error,Ø meteorological uncertainties forecasting.

Ø soil moisture initial conditions,

⇒ Need to develop methods that quantify these uncertainties

=> French Cévennes-Vivarais region is well known to be prone to those hazards.

Mediterranean areas are often affected by heavy rain events and devastating flash-floods

B. Vincendon1, V. Ducrocq1, O. Nuissier1, B. Vié1

References : -Bernard P., 2004 Aladin/AROME dynamical core, status and possible extension to IFS. In ECMWF Seminar Proceeding, Sept. 2004, available from http://www.ecmwf.int/publications/library in NOV.2004.-Bubnovà R., Hello G., Bernard, P, Geleyn JF., 1995: Integration of the fully elastic equations cast in the hydrostatic pressure tarrain-following coordinate in the framework of ARPEGE/ALADIN NWP system. Mon. Wea. Rev., 123 : 515-535-Beven and Kirkby, 1979 : A physically based, variable contributing area model of basin hydrology. Hydrol. Sci. Bull., 24, 43-69. - Bouilloud, L., K. Chancibault, B. Vincendon, V. Ducrocq, F. Habets, G.-M. Saulnier, S. Anquetin, É. Martin, J. Noilhan, 2009: Coupling the ISBA land surface model and the TOPMODEL hydrological model for Mediterranean flash-flood forecasting: Description, calibration and validation. J. Hydrometeor., 11(2), 315-333.,DOI: 10.1175/2009JHM1163.1.-Lafore,J.P., J. Stein, N. Asencio, P. Bougeault, V. Ducrocq, J. Duron, C. Fisher, P. Hereil, P. Mascart, V. Masson, J.P. Pinty, J.L. Redelsperger, E. Richard, J. Vila-Guerau de Arellano, 1998 : The Meso-NH Atmospheric simulation system. Part I: adiabatic formulation and control simulations, Ann. Geophysicae, 16, 90-109-Noilhan J. ans Planton S., 1989 : A simple parametrization of land surface processes for meteorological models. Mon. Wea. Rev.,117,536-549.-Pellarin T., Delrieu G., Saulnier G.M., Andrieu H., Vignal B. and Creutin J.D., 2002 : Hydrological visibility of weather radars operating in mountainous regions: case study for the Ardèche catchment, France, J. of Hydrometeor., vol. 3, No 5, pp 539-555.-Seity Y., P. Brousseau, S. Malardel, G. Hello, P . Bernard, F. Bouttier, C. Lac and V. Masson, 2010 : The AROME –France concvective scale operational model; submitted to Mon. Wea. Review.-Vié, B., Nuissier, O., Ducrocq, V. : Cloud-resolving ensemble simulations of heavy precipitating events : uncertainty on initial conditions and lateral boundary conditions, Mon. Wea. Rev., accepted , doi:10.1175/2010MWR3487.1, 2010.-Wernli, H., Paulat, M., Hagen, M. and Frei, C. : SAL-A novel quality measure for the verification of quantitative precipitation forecasts, 2008, Mon. Wea.. Rev., 136, 4470-4487, 2008.-Vincendon, B., Ducrocq, V., Nuissier, O. and Vié, B.: Introducingperturbation in rainfall fields for an ensemble forecasting of flash-floods, submitted to NHESS.

The AROME/ISBA-TOPMODEL systemTo simulate Mediterranean flash-flood, hydrometeorologicalcoupling (Bouilloud et al., 2010) has been developed between:Ø the land surface model ISBA (Interaction SurfaceBiosphere Atmosphere, Noilhan et Planton, 1989) Ø a version of TOPMODEL (Beven and Kirkby, 1979)adapted to the Mediterranean context (Pellarin et al., 2002).

*AROME (Seity et al., 2010)

o2.5km resolution => resolve deep convection processes.

o Same non-hydrostatic dynamic as Aladin (Bubnovà et al., 1995; Bernard,

2004) model

o Physical parametrization from Meso-NH (Lafore et al, 1998).

o Own data assimilation cycle running at 2.5km with analyses performed

each 3hours.

To increase the anticipation delay, ISBA-TOPMODEL can use as input convective-scale meteorological model (as AROME*) forecasts.

ISB

AIS

BA

-- TO

PM

OD

EL

TO

PM

OD

EL

Surface scheme

ISBA

Routing

Hydrological model

TOPMODEL

Soil waterRunoffDrainage

Total Discharge

Atmospheric forcing

AROME

AROME deterministic operational quantitative precipitation forecasts (QPF) uncertainties were assessed through a comparison with radar based quantitative precipitation estimate (QPE) :

Component L

Com

pon

en

t A

Composante S

SAL diagram for rainy objects# SAL components (Wernli et al., 2008) :

o A is a relative measure of the bias of

the model over the whole domain.

o L permits to evaluate the global shift

of the simulated field compared to

observations. It also gives information

about the spatial precipitation

distribution.

o S informs about the gradient within

the rainy objects.

§Positive values of S ⇔ simulated

objects too large and/or flat objects

compared to observation.

§Negative values ⇔ simulated

objects too small and/or peaked

Ø focus on hourly precipitation,

Ø on a sample of significant rainy days (>70 mm / day) within theperiod of AROME forecasts achieve (from sept. 2008),

Ø using an object-based approach (to avoid double penalty problem).

Objects identified with connex pixels where hourly rain exeeds :

Ø 2 mm / hour (rainy objects),

Ø 9 mm / hour (convective objects).

Use of SAL# diagnostic:

Ø median for A component close to 0 => no bias.

Ø L component values generally weak

ØA and S signs mostly identical

Determination of probability density functions (pdf) of QPF errors for amplitude and location and both types of objects :

Ø in 70% of the cases, the location errors do not exceed 50km

Results of the comparison :

Assessment of the high resolution meteorological simulations => no biases of the AROME forecats. General drawback of heavy precipitation underestimation of coarser NWP models not found for convecting permitting AROME model. Location errors can however highly affect the hydrological response at the catchment scale.

Design of an ensemble discharge forecasting method based on perturbations generation=> results as goog as the ones obtained using a state-of-art research convective scale NWP ensemble. Cheap computer time cost with respect to convective-scale NWP ensemble

⇒ see Vincendon et al, 2010 submitted to NHESS

Future workFurther verification on a larger sample on flash-flood cases needed to confirm the promising

results => Hymex SOP et EOP good framework

Method can be applied to each members of the convective scale NWP ensemble to increase its size.

Focus on the other sources of uncertainty ( hydrological modelling, initial soil water contents).

Tuning of the final method and choice of the assement criteria together with end-users.

Development of a perturbation method to produce short-term precipitation ensemble :

Ø to take advantage of the relevant information included into the deterministic AROME forecast,

Ø to sample the uncertainty that affect this forecast,

Ø using the PDFs obtained in the object-based AROME QPF verification.

Principle of the perturbation method

Verification on two flash flood cases that occurred during the fall 2008 :

Ø forecasting ensemble performs better that the deterministic forecast,

Ø Performance as good as the one obtained with a state-of-art research convection permitting NWP ensemble (called AROME-PEARP, Vié et al., 2010)

Ensemble median

Discharges simulated from :

50 members of the ensemble

Observed discharge

Ø Location perturbation step has the strongest impact on the results but best ensemble streamflowsimulation obtained with the three perturbation steps.

Ardèche river

N=10Ardèche river

N=50

Streamflow ensembles for the case of november 2008 :

Exemple of sensitivity to N value

Streamflow ensembles for the case of november 2008 :

discharges observed and simulated from 01-nov-2008

12UTC to 02-nov-2008 22UTC

Parameters of the method :ØGeographical shift

XY = 50 kmØNumber of members

N = 50

Precipitation scenarios obtained ingested into ISBA-TOPMODEL => ensemble streamflowforecasts.

operational AROME

Interquartiles range

Sensitivity of the method to its degrees of freedom :

Ø 50 (N parameter) members ensemble => better median and ensemble spread than ensemble with fewer members.