Modelling of pollutant distribution in surface runoff in ungauged catchments using geographical information systems

Abstract

In developing countries the scarcity of environmental information poses a difficult problem to the planner who is interested in assessing different control strategies for water quality problems. Although sophisticated models exist, the data available to run such models is usually limited. Furthermore, few of the models are easy to use, can operate over a variety of conditions, and can integrate the wide range of data that planners and managers need to compare water quality control strategies. The aim of this thesis has been to develop a water quality model, suitable for Brazilian conditions and that is capable of estimating the effects of future land use scenarios on water quality. The model is designed to be consistent with the availability of data in less developed countries and one of its major advantages is that the parameters involved are easily defined in terms of physical characteristics alone. A methodology is presented which has been designed specifically to support the determination of water quality changes resulting from point and non-point sources in a large river basin with varied land use, taking into account both the magnitude and spatial distribution of the loads produced. The methodology is capable of implementation on land surfaces having heterogeneous distribution of water pollution sources. The model is based on a cellular configuration where information on land use, topography, soil type and rainfall is manipulated using geographical information systems (GIS). Information on land use in each cell is obtained by classifying remotely sensed satellite data. Topographic parameters for each cell are derived from digital elevation models (DEM). Rainfall values for each cell are obtained from the interpolation of point data derived from meteorological stations. Other more conventional data are acquired by digiti sing maps. Using this cellular structure, runoff, and chemical outputs from the individual cell, are routed through the catchment using a physically based mixing model to provide input to the drainage network. The network, derived automatically from a digital elevation model (DEM), defines the river system in the model. The link between the catchment and the river network defines a river 'buffer' zone where point and non-point sources are stored. The model enables spatial relationships between point and non-point sources to be investigated and the consequences to the river system can then be modelled using the river network topology. The model is demonstrated for the Sinos catchment in southern Brazil. Analysis of the case study results indicate the model is capable of generating reasonable trends in water quality which reflect the impact of management activities. However, only limited water quality data availability precludes exhaustive testing of the system, and further work in this area is needed. The use of the model to investigate future water quality scenarios is also illustrated. Finally, the propagation of errors (due to spatial variation) associated with input variables is investigated. The problem is examined by first order theory and stochastic modelling, combined with the physical models, to show the possible magnitude of error within the model predictions.