Modeling of Rainfall- Runoff relationship using Artificial Neural Networks

Abstract

Abstract— Relationship between rainfall and runoff plays a significant role in generation of stramflows. The objective of the study is to model the rainfall – runoff using daily, weekly and monthly data for a catchment in the coastal Karnataka region using Artificial Neural Networks. The study uses data from two rainguage stations and a riverguage station located within the catchment. Fifteen models were developed using different input combinations which included 11 daily, 2 weekly and 2 monthly models. The efficiency of the models were compared using the statistical parameters - Coefficient of Correlation (r), Index of Agreement (d), Nash-Sutcliffe Efficiency (NSE) and Root Mean Square Error (RMSE). The results indicate that the daily model with daily past one day rainfall, past 2 day rainfall, past one day maximum temperature and past one day runoff as inputs was the best. The results can be used for any future studies of the catchment.