Advancements in rainfall-runoff prediction: Exploring state-of-the-art neural computing modeling approaches

Abstract

Rainfall-runoff (RR) is a vital process as it is a key component of the Earth’s water cycle, which is required for the survival of life on our planet. It is responsible for water resource management as it will alter the water quality and availability for living things and environmental requirements. Most of the previous research, in this domain, focused on short-term modelling using data from a specific region. However, fewer studies have been conducted to predict water availability for longer periods. There is an urgent need to explore a model that can predict RR in diverse locations for varied periods and climate circumstances. In this context, predictive models for RR prediction in literature are reviewed in this study. The findings are highlighted, and the discussion of the results are condensed. The review has been carried out for 80 articles that were published within last 21 years (2003–2023) on the competency of the predictive models used in RR prediction in the analysis of the input variables and the data size of the time series. The publications include relevant information such as the model limitation and the suggestions for further research that will be useful to researchers who intend to perform similar studies in RR predictions in the future. In addition, researchers from previous studies found that the hybrid deep learning (DL) models are greater than the hybrid machine learning (ML) models, DL models and standalone ML models. In this study, four new models are suggested to forecast the RR.