Distributed hydrological modelling in a GIS perspective : an evaluation of the MIKE SHE model

Abstract: The strain on water resources with the increased man-induced environmental changes through water pollution, is creating a growing demand in the understanding and management of complex hydrological systems where both water quality and quantity are concerned. This is beyond the capabilities of lumped models and more advanced tools are needed to predict these effects and hence the need for physically-based distributed hydrological models.
The aim of this study is to integrate a Geographical Information System (GIS) with a physically-based distributed hydrological model, the MIKE SHE model, creating a link between the model and the input data. The further aim is to calibrate and validate the MIKE SHE model and to evaluate the MIKE SHE model against the lumped HBV model.
The main effort was concentrated in retrieving the large amount of input data required by the MIKE SHE model. Most of the data originate from existing digital databases. Additional data were retrieved through remote sensing techniques and from literature. Both models were implemented on a catchment with two suitably placed discharge measurement stations used for evaluation. The catchment is situated in Scania, southern Sweden, and covers an area of 352 km2•
The resolution used in the MIKE SHE model computations is 250x250 metres, and the model has been calibrated for a period of ten years and validated for additional five years. The same time periods were used with the HBV model. The results from the MIKE SHE model calibration indicated problems in seizing the internal variations in the catchment. The total water flow from the entire catchment was however relatively well mapped. The comparison of the two models implies that a distributed hydrological model is redundant when estimating the discharge at a single outflow point. The computational time and effort required by the MIKE SHE model further enhances the recommendation of using the HBV model if possible. The overall calibration and validation results were better for the lumped model, but the limited use of a lumped model does not exclude the need for and use of distributed models. The physically-based modelling approach is not yet fully accomplished due to the lack of adequate data and computation capacity. This will hopefully be achieved with the aid of GIS and remote sensing techniques in the future.
A general conclusion of this study is that GIS are a necessity for handling the large amount of data used in all physically-based distributed hydrological modelling.