CS(AR)-2/2004 : Hydrological modeling of streamflows using artificial neural networks for Sindh basin

Abstract

The problem of transformation of rainfall to runoff has been a very active area of research throughout the evolution of the subject of hydrology. The relationship of rainfall-runoff is known to be highly non-linear, complex, time varying and spatially distributed. It involves many highly complex components such as interception, depression storage, infiltration, overland flow, interflow, percolation, evaporation and transpiration. Transformation of rainfall to runoff is to be understood in order to forecast the stream flows for water supply, flood control, irrigation, drainage, water quality, power generation and wild life propagation. Every model is an attempt to capture the essence of the complex hydrologic system in a meaningful and manageable way, but it is important that the conceptualization involves considerable degree of simplification. Conceptual rainfall runoff models are designed to approximate within their structures the general internal sub-processes and physical mechanisms, which govern the hydrologic cycle. Conceptual models provide daily, monthly or seasonal estimates of the stream flow for short-term and long-term forecasting by mathematically formulating the entire physical process in the hydrologic cycle. A6-parameter conceptual model of simple structure has been developed to represent the rainfall-runoff relationship. The efficiency of the model varies between 0.67 and 0.83 during calibration and between 0.76 and 0.82 during validation. The percentage difference in volume between the observed and computed annual flows vary between — 5.84 % and 25.65 %. The correlation coefficient between the observed and computed flow series varies between 0.90 and 0.96