Abstract:
A catchment is a dynamic system whose response changes with time. Given the complexities of real-life catchment, a mathematical model is but a simplified representation of this complex system. Further, the input (precipitation) to this system continuously varies in space and time. An efficient water resources planning and management requires continuous forecast of river discharge over a given period. Such forecast would be helpful in operating the component of water resources systems to avoid or reduce the flooding in flood prone downstream area to the extent possible and to issue warning to evacuate the area along the reach. The flooding is normally due to high intensity of rainfall in the catchment. The development of flood forecasting system is essential to manage natural hazard such as flood. Flood forecasting is an expanding area of application of hydrologic techniques. The goal is to obtain real-time precipitation and stream flow data through a microwave, radio, or satellite communications network, insert the data into rainfall-runoff and stream flow routing programs, and forecast flood flow rates and water levels for periods ranging from a few hours to a days ahead, depending on the size of watershed.
To date, many conceptual and black box models have been developed to compute the forecast at the flood controlling points. Development and calibration of conceptual models require a number of physical parameters, which interact in a complex fashion. In contrast, black box models produce good results from the input-output data in absence of detailed modeling of physical process in the basin. In recent years, Artificial Neural Network (ANN) models have been applied successfully to flood forecasting problems since it can represent any non-linear function given sufficient complexity nature of the underlying process under consideration to be explicitly described in mathematical form.
In this study, it is to proposed to revisit the available flood forecasting techniques particularly the methods based on statistical and Artificial Neural Network Approaches. These are applied to the real-time data of Ciliwung basin, and their result explained. Here, back propagation ANN models are developed and applied.
