Abstract:
A large number of methods have been proposed in the past to predict the flood discharges resulting due to a storm event in a catchment. Unit Hydrograph method described by the response functions is the most popular and widely used practical tool, among the various available methods. But, one of the basic assumptions followed in this method is that the effective rainfall is occurring simultaneously over the whole catchment uniformly, which is seldom true, in case of large catchments. The spatial variation generally becomes more and more pronounced as the size of the catchment increases. The unit hydrograph hypothesis is particularly adequate in the range of floods experienced on small catchments (size smaller about 500 km2). For large catchments, it may even happen that some part of the catchment would not have any rain at all during the storm period. Also, these methods are derived on the concept of lumped linear response functions.
In the present thesis, a multiple-input single-output nonlinear model based on systems approach proposed by Muftuoglu (1984 & 1991) has been formulated for forecasting flows during flood events. The total rainfalls have been considered as the inputs and the total runoff as the output as against to the use of effective rainfall and direct runoff as input and output respectively used in conventional unit hydrograph method. The whole catchment area was divided into a number of sub-catchments which receive uniform rainfall, approximately. Rainfall of each sub-catchment was treated as separate lumped-inputs to the model to incorporate spatial variations of rainfall as well as catchment heterogeneities. To derive composite response functions of the catchment, various data sets in calibration period were stacked together.
Rainfall and runoff data of the Wardha catchment upto Ghugus gauging site have been used for calibration and verification of the model. Results have been obtained for one, two and three rainfall inputs for the Wardha catchment with its discharge at Ghugus being the output. Both linear and nonlinear approaches have been studied. Multiple-input approach by linear and nonlinear methods was also compared with the rest of the results. Better results have been obtained when number of inputs in the nonlinear model were increased from one to three.
