Comparative study of Discharge Estimation using Non-Parametric and Parametric Methods

Abstract

Abstract Runoff estimation from a watershed is of utmost importance for various hydrologic and hydraulic purposes. The present study aims to evaluate various parametric and non-parametric methods of discharge estimation for the Gambhar watershed in the state of Himachal Pradesh, India. The study employed Artificial Neural Network (ANN) as a non-parametric and statistical modelling and the Green-Ampt (GA) infiltration model as parametric methods. The Meteorological data for the duration of twenty years was procured from Indian Meteorological Department (IMD). By employing the GA infiltration model, the cumulative infiltration and subsequently the runoff was calculated for different intensities while lag time was assessed using ArcGIS. The collected field point data for various soil infiltration parameters was spatially interpolated. The cumulative distribution function (CDF) of discharge were generated with a confidence limit of 95% using statistical modelling. The root mean square error (RMSE) and coefficient of determination (R2) were employed to determine the best ANN architecture. On the basis of the qualitative and quantitative comparative evaluation, the study establishes the efficacy of ANN method over parametric methods i.e., GA and statistical modelling, in discharge estimation with values of average relative error (ARE) & Nash Sutcliffe Efficiency (NSE) coefficient as 0.15 & 0.91 respectively.