Hydrological modelling for lowland catchments : a geographical information systems approach
This research applies, evaluates and compares approaches to hydrological modelling and stream flow forecasting within a GIS environment. Three different approaches to modelling stream flow were investigated, namely; TOPMODEL, a regression approach and a GIS-based model, HydroGrid.
TOPMODEL is a parametrically simple, physically-distributed model that allows the topological modelling of catchment processes. Regression modelling is a statistical technique that derives an empirical equation based on the assumption that the values of a dependent variable will depend upon the values of the independent variables. HydroGrid is a purpose-built GIS-based model for catchment modelling using the functionality that GIS offers for modelling the spatial variations of catchment characteristics.
All three approaches were evaluated using readily available data for a lowland catchment, in Staffordshire, U.K. Model validation used six years of data covering the period 1991-92 through to 1996-97 - with years running from March-February. Five performance indicators were used to assess the models enabling both for detailed evaluation of the models and comparisons to be drawn with other research.
The performance of the three models tested showed great similarities, with all approaches tending to over-predict stream flow. Model performance was also evaluated using three different evapotranspiration models - the Penman formula, the Crowe-irrigation method and the sine curve method. All three models performed best during wet years or wetter seasons indicating a common weakness in the accurate modelling of low stream flows. Despite similarities in performance, clear benefits of hydrological modelling within a GIS framework are identified.
Overall, the results show that although the methods used here can help in daily flow modelling, there is a major need to improve methods for catchment modelling with routine data sources. An important development could be to loose-couple hydrological models with a GIS to improve their ability to use available information but also, as shown in this work, to model catchment processes directly within a GIS.