2.1 Conceptual model Tundzha

Below the river scheme of the Tundja catchment is shown. This scheme will be modelled and calibrated as a whole because there are not enough gauging stations available to divide the scheme in model sections.

Length is rounded off in km

Additional cross-sections could be used from the JICA project (9) and from gauging stations (4).

3. Model Set Up

3.1 Hydrological Model Set up

The Hydrological Model used in our case is the NAM model provided by Mike11 software.

NAM model is a continuous precipitation-runoff model of the deterministic, lumped, conceptual type. It includes simulation of snowmelt in different altitude zones within the catchment.

NAM represents various components of the rainfall-runoff process by continuously accounting for the water content in four different and mutually interrelated storages.

These storages are Snow, Surface, Root Zone and Ground Water Storages. Using rainfall, potential evaporation and temperature as input, the model simulates Snow accumulation and melting, interception, evapotranspiration, overland flow, groundwater recharge and baseflow.

In case of using snowmelt component, NAM module needs additional snowmelt parameters. The model utilizes the temperature index model for the computation of daily snowmelt. The Snow melt module uses a temperature input time series, usually mean daily temperature. Users may choose the simple snowmelt or extended more sophisticated snowmelt module in the system.

Each hydrological model of the system is first defined conceptually using the watershed separation into subbasins. The models are thus constructed from GIS basin and sub-basins zoning. Their initial definition (ie the first evaluation of parameters) before calibration is based upon the various elements available within the GIS : Surface and shape of the basin, land cover, soil properties, geological information, slopes and elevations, etc.

The calibration of the hydrologic model consists in refining the parameters that make up each single model. For this we made use of historical observed data. By running the model with real observed data as input (rain, snow, temperature, evapotranspiration,..) and comparing its output with the observed discharges we optimise the various parameters in order to minimise the differences between output results and observed data on several measured sequences.

The Mike 11 software provides semi automatic procedures in order to facilitate the calibration process which is rather long and tedious.

Calibration of Luda Yana hydrologic model

3.2 Hydraulic Model Set Up

The used model, Mike 11, is a one dimensional type hydraulic model: flows in reaches and overflows into floodplain where storage and flow are simulated, have a multidirectional though one dimensional formulation.

Hydraulic models are constructed using true channel details, aggregated runoff coefficients, floodplain storage mechanisms, and significant structures such as bridges, weirs, levees, terraces, reclamations, pond bunds…

A detailed topography of the flood plain up to the break of slope to the terraces or hills is thus required in order to build the hydraulic system to be modelled.

Hydraulic modelling is based on nodes linked into a topological network. For nodes linked together in the main channel and the immediate left and right side flood plain, an implicit scheme of resolution of St Venant set of differential equations is used.

For the nodes within the flood plain associated with topographic structures as levees, roads, railway tracks, which delimited retention zones, St Venant equations are also solved, and additionally, overflows can be modelled using spillway discharge equations.

Because of the detailed topographic description used, and because water height is the main controlling variable to model the flows in the flood plain structures, the topographic data have to be accurate enough for the model to be effective in its forecasts. A 10 cm overall accuracy in elevation, for the topographic structure is generally required for these models.

Depending on the channel and embankments topographic variations, the cross section spacing may vary from 2-3 km to as low as 200 m.

When the flood plain and river channel slope is rather flat, as it is the case in the downstream Maritza and Tundja basins, Hydraulic models are very sensitive to topographic inaccuracies. This mean that water height divergences (which may result from a poorly accurate topographic survey) between real data and simulated data will propagate numerically downstream and upstream, leading to an overall poor result.

The main limitation for these hydraulic models is thus the quality of the topographic data used to construct their topology.

The other points (solver methods, implicit or explicit schemes…) hydraulic coefficient values, while important in the calibration process and for computation fiability and performances, are less critical for accurateness issues thanks to the know how gained today on these tools.

The Hydraulic model development contains the following steps:

• Collection and processing of cross-section and topographic data.
• Estimation of the flow conveying and storage (inundation) areas in the cross sections and through the flood plain
• Collection and processing of water level and discharge measurements of gauging stations and stations in first order tributaries.
• Estimation of the hydraulic roughness of the flood plain of the river on basis of vegetation characteristics.
• Collection of data on hydraulic structures and the way these are controlled.
• Model construction by:
  • entering delineation of the river and its relevant tributaries;
  • entering the cross sections;
  • entering the hydraulic roughness of the main channel and the flood plain;
  • entering lateral and boundary inflow;
  • modelling of structures;
  • entering the computational grid;
  • defining computational settings.
• Calibration and validation of the hydraulic models by running the model with historical data and minimizing the differences of model outputs with the observed water levels and discharges at stations.

Note that, although the calibration and validation of the models can only start when all data is collected, analysed and processed, the steps before the calibration do not depend on one another and can be executed separately.

Mike 11 software provides semi automatic procedures in order to facilitate the calibration process of the HD model.