### MikeSHE

Report Lesson 2.1 - MIKE SHE

Team 1 - HydroEurope

Benoît BESSEAS, Océane CALMELS, Laura DAUL, Guillaume HAZEMANN, Quentin MOLIERES, Blazej SMOLINSKI, Arianna VARRANI, Robin WARNER, Taline ZGHEIB, Zhengmin LEI

Supervisors : Philippe GOUBERSVILLE, Olivier DELESTRE

# I.                   Introduction

MIKESHE is one of the MIKE package’s software developed by DHI. MIKESHE enables to produce integrated modelling of groundwater, surface water, recharge and evapotranspiration fluxes. In this course, we will learn how to build a hydrological model for flood event simulations with the help of MIKESHE software. The flood event which we are using is the flood event of 1994 in the Var region. The simulation will be run using two different grid data resolutions (300m and 75m DEM). A comparative analysis of the results will be done to see the influence of the grid resolution.

# II.                Methodology

When building our model, we took into consideration the following hypothesis and assumptions:

-          the simulation is the for a short period, so losses due to the evapotranspiration will be ignored in comparison to the rainfall depth ;

-          the soil was highly saturated before the flood, so the infiltration module will be neglected

The main steps used to build the hydrological model in MIKE SHE are the following (Figure 1):

Figure 1: Main steps of the hydrological modeling process in MIKESHE.

The simulation specifications require a simulation period and a control time step. We consider the beginning of the flood event of 1994 on November 4th at 6pm, and the end on November 6th at 12am. We specify a variable time step (from 0.1 to 1 hour) and a maximum precipitation rate of 10mm/time_step. The time step will be reduced if we get more than the limit of 10mm/time_step to maintain stability of the model.

There is also the possibility to change the number of iterations used in the simulation. The higher the number is, the higher the simulation accuracy and the computation time.

For the geometrical data, we import the topography configuration with two DEM files (300m and 75m grid resolution). Then, we use the Thiessen_polygons file which gives us the repartition of our sub-basins according to Thiessen’s method. Finally, we import our rainfall data which will give us the repartition of the rainfall over our sub-basins in function of the area of our Thiessen’s polygons.

The net rainfall fraction is considered uniform and equals to 90% of the total rainfall. Friction coefficient specifications: the Strickler coefficient is set by using a landcover file. Finally, we specify a 1-hour time step for the overland flow and a 24 hours’ time step for the precipitations. Then we specify 4 observation points to get the accumulated flow at these points, which represent our observation location. To get the total amount of flow in the area of interest we have to sum these 4 observation points.

# III.             Results and discussion

### III.1. Result analysis of Exercise 1

In this first part, we draw the hydrograph for the 300m coupling with Mike11. We observe the simulated hydrograph in comparison to the observed hydrograph of the 1994 flood (figure 2).

Before coupling, we notice a severe underestimation of both flow volume and flow peak. This is why we couple the model, because it enables a better connection between streams to the extent that the coupling enables to take into account smaller streams (water flows) which were ignored by the initial model. Figure 2: Discharge and coupled model results with a 300m resolution map.

The first observation is that the simulated hydrograph underestimates the peak flow and volumes at the outlet. We also notice that the peak time occur earlier in the simulated model than in the observed hydrograph.

### III.2. Result analysis of Exercise 2

For the second exercise, we used a 75m DEM. As you can see in (figure 3), the simulated peak flow is slightly lower than the peak flow simulated by the 300 m DEM. Figure 3 : Discharge and coupled model results with a 300 m and a 75m resolution map.

In order to reduce uncertainties of data, we need to do model calibration which can be carried out by adjusting certain input data to make the simulated result match the observed ones. In this case, we will adjust the Strickler values. This coefficient for the forest can vary from 12 to 20 according to different vegetation cover. In this part, we changed it from 12 to 20. The grey line and the dark blue line in figure 4 represent respectively the different hydrographs.

Figure 4 : Discharge and calibrated coupled model results for a 75m resolution map.

As the figure 4 shows, the simulated volumes are larger with the smaller grid cell.

In this part, we changed the Strickler coefficient from 12 (grey line) to 20 (blue line), meaning we decreased the bed resistance and we observed that we produced a hydrograph with a higher peak intensity and a time to peak occurring shortly sooner.

These results are consistent because if we decrease the bed resistance, then the water is flowing with less resistance, meaning less infiltration, and then more water volume (hence a higher peak occurring sooner). On the other hand, an increased resistance produces a lower peak because the water is flowing slower and infiltration can occur in a more significant way.

# IV.              Conclusion

This study enables to model the river Var behavior thanks to geometry, rainfall and land use data. We also see the limits between simulation and reality as the simulation tends to underestimate the event in our case. It highlights the fact that it is essential to have observed data to calibrate our model. Calibration appears as a key step in the elaboration of an operational model. For example, calibrating the manning coefficient affect both peak and time to peak. But this not the only parameter. Many other parameters such as the NRF (net rainfall fraction) and the retention storage parameter.

Ċ
laura DAUL,
2 Dec 2016, 11:51