1. EVALUATION OF A NON-HYDROSTATIC MESOSCALE MODEL FOR DISCHARGE FORECASTING
K. Chancibault, V. Ducrocq, F. Habets
CNRM/GAME, Météo-France
S. Anquetin, G-M. Saulnier
CNRS, LTHE
WSN05, Toulouse, France, 5-9 September 2005

2. Meso-NH or AROME TOPMODEL
INTRODUCTION
Flash-floods in Southeastern France :
Quasi-stationnary MCS => high cumulative rainfall amounts (>200 mm/day)
Small mountainous basins (200 – 2500 km²) => short lag-time (<12 hrs)
Next NWP models : non-hydrostatic high-resolution
model (AROME)
Meso-NH or
AROME
TOPMODEL
Meteorological model
Hydrological model
1-way
2-way
Southeastern France is often affected by flash-floods. They are due in most cases to quasi-stationnary MCS that lead to high cumulative rainfall amounts and to small mountainous basins with short lag-time.
With the next generation of NWP models that will be non-hydrostatic high-resolution models, like AROME in France discharge forecasting from forecasted rainfall becomes conceivable.
We are currently developping a hydrological forecasting chain using Meso-NH (a non-hydrostatic research model) or AROME for the meteorological model and a hydrological model from the TOPMODEL family. This chain can be used in two configurations : with one-way or2 way coupling
So,first, I will present some results with the one-way coupling for an extreme French flash-flood case and in a second part I will describe how we are developping the hydrometeorological 2-way coupling.
One-way coupling : 8-9 September 2002 case
Two-way coupling : configuration
WSN05, Toulouse, France, 5-9 September 2005

3. The hydrological model : TOPMODEL (Beven and Kirkby, 1979)
ONE-WAY COUPLING
The hydrological model : TOPMODEL (Beven and Kirkby, 1979)
DTM : 50 m
1 model / basin
9 basins (200 km² km²)
Calibration (4 parameters)
Time step : 1 hour
Areal rainfall depths
ARDECHE
CEZE
GARD
VIDOURLE *
Vogüé
St Martin *
Remoulins
Alès
Anduze
Boucoiran *
La Roque *
Quissac
Sommières
The hydrological model used to simulate this flash-flood case is one of the TOPMODEL family which simulate the surface run-off, a predominant phenomena in the region. Here is a cross section of a basin, with the river and the water level indicated with this blue line. During the rainfall event, the water level goes up and the soil water can’t flow anymore. The surface run-off appears on the saturated soil. This model show its efficiency for numerous flash-flood cases over the region.
It uses a digital terrain model with a resolution of 50 m. There is one model per basin and we studied 9 subbasins of these four main basins with surface ranging from 200 to 2200 km².
TOPMODEL needs to be calibrated for each basin thanks to four parameters : for the saturation soil conductivity, the vertical profil of the soil transmissivity, and two to parameterize the water interception and the evapotranspiration.
It needs, at each time step of one hour the mean areal rainfall depths calculated for each simulated catchment.
WSN05, Toulouse, France, 5-9 September 2005

4. Rainfall data ONE-WAY COUPLING Simulated rainfall :
Meso-NH (Lafore et al, 1998), non-hydrostatic mesoscale model at 2.4 km
Same physical package and resolution as AROME
Initial conditions : mesoscale initialization
(surf obs + radar + IR sat data) from Ducrocq et al ( 2000)
Interpolated (kriging method) rain-gauge data
These rainfall depth were calculated from simulated rainfall fields with the non-hydrostatic mesoscale model Meso-NH. It ran with the same physical phackage and the same resolution as AROME.
The initial conditions were provided by a mesoscale initialization procedure using surface data, satellite IR data and radar data, at 12 UTC, on the 8th of September.
For the control run, we used also as input to TOPMDEL rain gauge data that were previously interpolated following a kriging method.
WSN05, Toulouse, France, 5-9 September 2005

5. + + + ONE-WAY COUPLING Case study : 8-9 September 2002 Phase I
24 casualties
Damage : 1.2 billion euros mm
12-22 UTC +
12-22 UTC, 08/09/02
Phase I
Nîmes
22-04 UTC +
22-04 UTC, 08-09/09/02
Phase II
Nîmes
04-12 UTC +
04-12 UTC, 09/09/02
Phase III
Delrieu et al (2005)
Nîmes
For the 1-way coupling, the case studied is the 8th and 9th of September 2002 case which was responsible for one of the most important floods ever recorded in the Southeastern region. It caused 24 casualties and an economic damage of 1.2 billion euros.
This event can be described in three phases. During the first phase, a quasi-stationnary MCS formed and stayed over plains about ten hours and lead to rainfall amounts larger than 300 mm.
During the second phase, the MCS moved northward to the foothills of the Cevennes ridge. The maximal rainfall amounts were greater than 200 mm.
In the morning of September the 9th, a surface cold front with embedded convection, passed over the region and afterward, there was no more convection.
Cumulated rainfall (Nîmes radar)
WSN05, Toulouse, France, 5-9 September 2005

6. Meso-NH simulations results
ONE-WAY COUPLING
Meso-NH simulations results mm
OBS
OBS
OBS
Phase I: UTC
Phase II: UTC
Phase III: UTC
Here are the results of the Meso-NH simulation : On these figures are presented the cumulated rainfall for the three phases. Observation from radar are on the top panels, and simulation on the bottom ones. We can see that during the first ten hours of simulation, Meso-NH simulates a MCS, with a good location. However, during the two following phases, the model does not succeed in simulating a MCS with a good location as the simulated MCS moves to the noth of the domain
SIMU
SIMU
SIMU
Cumulated rainfall
WSN05, Toulouse, France, 5-9 September 2005

7. Cumulated rainfall during
ONE-WAY COUPLING
Meso-NH simulations results
Vidourle
Gard
Cèze
Ardèche mm
Simulated rainfall depths
errors on the phase I
Quissac
Sommières
Alès
Anduze
Boucoiran
Rémoulins
La Roque
Vogüé
St Martin
OBS
In order to evaluate this simulation from a hydrological point of view, we computed the mean areal rainfall depths over the 9 basins for the simulation and for the rain gauge data. These mean areal rainfall depths will then be used as input to TOPMODEL.
errors on these rainfall depths, calculated for the 10 first hours of the simulation, are drawn on this graph for both subbasins of the Vidourle, for the subasins in the Gard catchment, for the Ceze basin and two on the Ardeche basin.
the rainfall depths are underestimated at all the catchments but the Ales one. It is on the southern catchment that the underestimation is the largest one. On the other catchments, the rainfall depths are less underestimated.
In the following slide, I will present the simulated discharges for the Anduze catchment and the Vogue catchment.
SIMU
Cumulated rainfall during
the phase I
WSN05, Toulouse, France, 5-9 September 2005

8. Hydrological simulations results
ONE-WAY COUPLING
Hydrological simulations results
Anduze (540 km²)
Vogüé (620 km²)
observed discharges
Simulated discharges from :
observed rainfall
simulated rainfall
For the Anduze basin, here are the measured discharges in black dotted line, the simulated discharges from observed rainfall for the control run in blue and the simulated discharges from the simulated rainfall are in red.
Clearly, these latter are largely underestimated compared to the control run and the observation.
The underestimation calculated on the rainfall depth is largely due to the last fourteen hours of the meteorological simulation for which the MCS is not over the anduze catchment anymore.
On the contrary, for the Vogue basin, during the two next phases, the rainfall depth are overestimated and despite the underestimation of rainfall for the 10 first hours of simulation, the discharges are overestimated.
These results highlight the need, for such a convective case, to useupdating high-resolution forecasted rainfall for discharge forecasting.
WSN05, Toulouse, France, 5-9 September 2005

9. TWO-WAY COUPLING Configuration ISBA t = 5 min x = 2-3 km L = 1000 km
Meso-NH or Arome
TOPMODEL
Wmob As
flux
ISBA
t = 1h
x = 50 m
L = 1 km
t = 5 min
x = 2-3 km
L = 1000 km
The other configuration, for the forecasting hydrological chain, is the 2-way coupling.
This 2-way coupling already exists between the research mesoscale model Meso-NH (or in the future this will be AROME) and the SVAT scheme called ISBA. What we have done is to implement the horizontal transfert of soil water, following the TOPMODEL concept, in the SVAT scheme. These two models exchange different variables: the distribution soil water that can be transfered along the slope and the portion of saturated area of each ISBA grid mesh, to allow ISBA to calculate a surface run-off.
The routing module will be developped soon, in order to calculate the discharges at the outlet.
Notice, that these different models run at different time steps, and different spatial scales,
Routing
WSN05, Toulouse, France, 5-9 September 2005

10. TWO-WAY COUPLING Idealized simulation + Soil water horizontal
distribution
TOPMODEL 50m
Vogüé
Anduze
m3/m3
0.47
0.45
0.43
0.41
0.39
0.37
0.35
0.33
0.31
0.29
Vogüé +
ISBA 2.5 km
Anduze
Soil water
increment
Now, let’s see a preliminary result : an idealized simulation has been performed, with the ISBA-TOPMODEL system.
The ISBA grid has a soil with a uniforme composition, with a soil water content corresponding to a wet but not saturated soil.
At each iteration, the soil water content is increased uniformly on the grid with a value corresponding to a flash-flood event.
Here are the results, on the ISBA grid, after ten iterations. We can see the impact of TOPMODEL for both simulated basins, where the soil moisture ranges from dry (in red) to saturated (in dark blue).
coupleur
WSN05, Toulouse, France, 5-9 September 2005

11. CONCLUSIONS AND OUTLOOKS
One-way coupling :
Discharge forecasting needs updated high-resolution forecasted rainfall
Ensemble forecasting rainfalls to perform ensemble discharges forecasting
Two-way coupling :
Soil water horizontal distribution was implemented in the SVAT scheme
Soon, routing will be added  discharge simulation
Sensitivity studies (e.g. time step, soil depth, …)
Validation for numerous basins of the Southeastern France
Applications
Validation of high-resolution simulated rainfall
Study of the feedback of hydrology to the atmosphere
Evaluation of the 2-way coupling vs the 1-way coupling
WSN05, Toulouse, France, 5-9 September 2005