Multiple-Day-Ahead Flood Prediction in the South Asian Tropical Zone Using Deep Learning

AbstractA reliable and accurate flood forecasting procedure is crucial due to the hazardous nature of such disasters. Despite the growing interest in machine learning models over traditional methods for enhanced accuracy, there is a notable gap in guidelines for selecting the best-suited learning al...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of hydrologic engineering 2025-02, Vol.30 (1)
Hauptverfasser: Madhushanka, G. W. T. I., Jayasinghe, M. T. R., Rajapakse, R. A.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:AbstractA reliable and accurate flood forecasting procedure is crucial due to the hazardous nature of such disasters. Despite the growing interest in machine learning models over traditional methods for enhanced accuracy, there is a notable gap in guidelines for selecting the best-suited learning algorithm for flood forecasting. Furthermore, there is a lack of deep learning–based flood simulation studies in the South Asian Tropical Zone. This research addresses these gaps by investigating the viability of artificial neural network (ANN), long short-term memory (LSTM), bidirectional LSTM (BLSTM), two-dimensional (2D) convolutional LSTM (ConvLSTM2D), and transformer models for multiple-day-ahead flood simulation. A forecasting window of 3 days was selected for the task, focusing on the lower reaches of the Mahaweli catchment in Sri Lanka, which is heavily affected by the Northeast Monsoon. Observed rainfall data from three nearby rain gauges and historical discharges from the target river gauge serve as input features for the models. Unlike previous studies that used limited data sets, this study utilizes daily data spanning 28 years in order to examine the behavior of the transformer handling an extensive hydrological data set. The study finds that the ANN model performed the worst, with a mean Nash-Sutcliffe efficiency (NSE) of 0.67, while the transformer model showed superior performance, especially in multiday forecasts, with a mean NSE of 0.72 and a mean root-mean square error of 32.52, showcasing the effectiveness of handling this extensive data set. Practical ApplicationsArtificial intelligence (AI) is the most important invention of this era. Deep learning (DL) is a subset of AI, which tries to imitate the way the human brain works. Using these powerful DL techniques, Machines are trained to identify patterns in data and use the learned patterns to forecast future values. In this paper, we have investigated the performances of five DL techniques in flood forecasting. Predicting future flood conditions is an important subject area that helps to save lives and property in flood-prone regions. Accurate and timely flood forecasts allow disaster management agencies, and other relevant authorities for early evacuation of people, ultimately lowering the loss of life. Our results indicate that transformer, the latest DL architecture, has the highest accuracy among the models. Transformer architecture was first proposed in 2019 and started a revolution in AI.
ISSN:1084-0699
1943-5584
DOI:10.1061/JHYEFF.HEENG-6296