A New Custom Deep Learning Model Coupled with a Flood Index for Multi-Step-Ahead Flood Forecasting
Accurate and prompt flood forecasting is essential for effective decision making in flood control to help minimize or prevent flood damage. We propose a new custom deep learning model, IF-CNN-GRU, for multi-step-ahead flood forecasting that incorporates the flood index (<inline-formula><mat...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Hydrology |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2306-5338/12/5/104 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate and prompt flood forecasting is essential for effective decision making in flood control to help minimize or prevent flood damage. We propose a new custom deep learning model, IF-CNN-GRU, for multi-step-ahead flood forecasting that incorporates the flood index (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mrow><mi>I</mi></mrow><mrow><mi>F</mi></mrow></msub></mrow></semantics></math></inline-formula>) to improve the prediction accuracy. The model integrates convolutional neural networks (CNNs) and gated recurrent neural networks (GRUs) to analyze the spatiotemporal characteristics of hydrological data, while using a custom recursive neural network that adjusts the neural unit output at each moment based on the flood index. The IF-CNN-GRU model was applied to forecast floods with a lead time of 1–5 d at the Baihe hydrological station in the middle reaches of the Han River, China, accompanied by an in-depth investigation of model uncertainty. The results showed that incorporating the flood index <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><msub><mrow><mi>I</mi></mrow><mrow><mi>F</mi></mrow></msub></mrow></semantics></math></inline-formula> improved the forecast precision by up to 20%. The analysis of uncertainty revealed that the contributions of modeling factors, such as the datasets, model structures, and their interactions, varied across the forecast periods. The interaction factors contributed 17–36% of the uncertainty, while the contribution of the datasets increased with the forecast period (32–53%) and that of the model structure decreased (32–28%). The experiment also demonstrated that data samples play a critical role in improving the flood forecasting accuracy, offering actionable insights to reduce the predictive uncertainty and providing a scientific basis for flood early warning systems and water resource management. |
|---|---|
| ISSN: | 2306-5338 |