Comparative study of artificial neural network and hydrological models towards runoff estimation
Author(s): Anusmita Bhowmik, Ankana Moulik and Tanmoy Majhi
Abstract: Runoff is a vital hydrological factor, governing water flow into systems and returning excess precipitation to the seas. Water resource managers use model-derived runoff data to comprehend, regulate, and monitor water resources. The process of obtaining this data is arduous. The study provides a framework for grasping the model-specific building blocks, their influence on closure effects, and parameter calibration's simplicity and physical meaning. It surpasses mere flow rate replication, delving into deeper model capabilities in the process of integrating model outcomes, three distinct approaches are considered: the simple average technique (SAM), the weighted average method (WAM), and the neural network method (NNM). In forecasting stream flow utilizing seven parameters, the SIMHYD model examines daily rainfall and areal potential evapotranspiration data. The foremost and pivotal component of the ARNO model entails delineating the soil moisture equilibrium, while the subsequent aspect involves depicting the course of runoff transfer towards basin outflow. The degree of calibration achieved in a conceptual rainfall-runoff (CRR) model governs its potential efficacy in practical implementation. In spite of the prevalent application of conceptual rainfall-runoff (CRR) models, research findings indicate that achieving distinct optimal parameter values through automatic calibration techniques is often challenging. The adoption of Artificial Neural Networks (ANNs) has become increasingly common in the analysis of hydrological and water resource complexities. Furthermore, the Probability Distributed Model (PDM), a compendium of model functions, was developed as a lumped rainfall-runoff model capable of elucidating a spectrum of hydrological behaviours at the catchment scale. As a result, this paper offers insights into both the merits and limitations of various rainfall-runoff models for the readership's enrichment.
