Integration of Lag Time Analysis and Antecedent Rainfall Scenarios in Artificial Neural Network Models for Streamflow Prediction

Abstract

Artificial Neural Networks (ANN) play a vital role in hydrological modeling, adept at capturing intricate, non-linear relationships in dynamic environmental systems. This study investigates the interplay between lag time in a basin and the predictive accuracy of ANN models for streamflow forecasting. Three distinct ANN models, employing varied antecedent rainfall scenarios as inputs, were evaluated. Results indicate a robust correlation between the lag time of the basin and the efficacy of the ANN model. Notably, the model incorporating a 3-day antecedent rainfall scenario demonstrated superior predictive performance, aligning precisely with the calculated lag time. This synchronicity underscores the relevance of tailoring input features to the inherent characteristics of the watershed. Beyond model optimization, the study's significance lies in its contribution to flood risk management, offering advanced tools for more accurate and efficient predictions in the face of increasing climate-induced extreme weather events.