Runoff prediction model using unit-step and transfer functions approach for a small watershed

In this study, two parameters conceptual model based on unit step and transfer functions approach was developed for runoff prediction on storm basis from Shenda Park watershed, treating the watershed as lumped, linear and timeinvariant system. In this model, rainfall data in blocks of finite duration is represented by the unit-step function. Generally the ordinates of direct runoff are obtained directly as the inverse Laplace transform of the product of Laplace transform of the input and the transfer function of the system. The model parameters were estimated by the methods of cumulants and moments. The autocorrelation function of residuals is confined within the lines of confidence that is ±2 SE (r_k) lines, confirming adequacy of the model for study watershed. The goodness of fit of developed model was evaluated. The average values of correlation coefficient, special correlation coefficient, coefficient of efficiency, root mean square error, percentage absolute deviations in peak flow rates and direct runoff volumes of the model were 0.96, 0.92, 0.92, 0.002, 19.85 and 1.33 respectively for calibration events and 0.96, 0.92, 0.88, 0.001, 20.06 and 0.10 respectively for prediction events. Based on performance evaluation criteria, model can be easily applied to predict direct runoff hydrographs from small watersheds for design of various soil and water conservation structures.