Thesis colloquium MARK WILDE

Spatial parameterization of the wflow_SBM model using literature-based pedotransfer functions in the Alsace-Lorraine region. Regionalization of these functions aims to aid streamflow prediction in gauged and ungauged basins.

Organised by Hydrology and Quantitative Water Management

Fri 7 July 2017 13:30 to 14:00

Venue Lumen, gebouwnummer 100
L.E 112

Spatial Parameter Regionalization for the SBM model in the Frech part of the Rhine Catchment

Parametrization of a spatialized gridded hydrological model is not straightforward. Calibrating each grid cell individually is not possible due to the excessive computational task involved, and because it would lead to unavoidable equifinality. One method to parametrize a spatial model is by implementing pedo-transfer functions, which relate physiographic basin descriptors to parameters. These pedo-transfer functions are parameterized empirically, and can be taken from literature or derived through calibration. If the correct hydrological processes are captured by the model structure, and if the parameters are correctly inferred by the transfer functions, the functions should, in theory, also be applicable to other catchments.

In order to model spatially, first the subcatchments were extracted based on different resolutions of a digital elevation model. Subcatchments in the French part of the Rhine catchment appeared to be derived best from a digital elevation model with a resolution of 300m. Finer resolutions appeared to be too much influenced by human disturbances in elevation, whereas a courser resolution was not able to capture significant detail in elevation.

Pedo-transfer functions were tested based on a nested catchment consisting of four subcatchments. The functions were tested by first implementing the function and calibrating all other parameters. Next, all pedo-transfer functions that were already accepted were implemented too, and the remaining parameters were calibrated. Thirdly, the model runs were validated based on the best performing parameter set. Implementing the Height Above Nearest Drainage descriptor in transfer functions for the M parameter did not lead to increased performance, because of the limited model sensitivity to the specific parameter.

Regionalization with the transfer functions as used in this research improved model performance in pilot catchments, but did not yield adequate results in adopting catchments. However, calibration of the not regionalized parameters in the adopting catchments with the accepted pedo-transfer functions implemented, indicated a possibility to improve regionalization performance by implementing more functions for different parameters.