Interpretable deep learning hybrid streamflow prediction modeling based on multi-source data fusion
文献类型：期刊论文
| 作者 | Wang, Zhaocai2; Ding, Cheng2; Xu, Nannan2; Wang, Weilong1; Zhang, Xingxing3 |
| 刊名 | ENVIRONMENTAL MODELLING & SOFTWARE
 |
| 出版日期 | 2026-01-30 |
| 卷号 | 196页码:106796 |
| 关键词 | Streamflow prediction Deep learning Feature extraction Mode decomposition Nonlinear integration Interpretability |
| ISSN号 | 1364-8152 |
| DOI | 10.1016/j.envsoft.2025.106796 |
| 产权排序 | 3 |
| 文献子类 | Article |
| 英文摘要 | Accurate streamflow forecasts are critically important for monitoring flood disasters and managing water resources. The factors influencing streamflow are complex, characterized by significant non-linearity and intricacy. Developing a data-driven hybrid deep learning model for streamflow prediction represents an effective strategy. Consequently, this study introduces an enhanced deep learning model, named CEEMDAN-ISMA-CNN-LSTM-AMRF (CICLAR), for predicting both streamflow and extreme flood events. This study integrates multi-source heterogeneous data, including remote sensing, meteorological, hydrological, and streamflow data. The Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is utilized to reduce the complexity, and then multi-source data are input into the CNN-LSTM-AM model. Additionally, the Improved Slime Mould Algorithm (ISMA) is employed to optimize the neural network's hyperparameters. Finally, Random Forest (RF) is used to perform non-linear summation. The study conducted daily streamflow predictions at 11 stations located in the upstream, midstream, and downstream sections of the Jialing River in China, demonstrating that the CICLAR model significantly outperforms other benchmark models. Taking the Beibei Hydrological Station as an example, compared to the conventional Long Short-Term Memory (LSTM) model, the Nash-Sutcliffe Efficiency Coefficient (NSE) of the CICLAR model's prediction results increased by 30 %, and the Mean Absolute Error (MAE) decreased by 75 %. For extreme flood forecasting, compared to the LSTM, the CICLAR model reduced the Mean Relative Error (MRE) by 0.86 and improved the Qualification Rate (QR) by 150 %. The results of this study show that the CICLAR model has significant application value in extreme flood forecasting and water resources management. |
| URL标识 | 查看原文 |
| WOS关键词 | NEURAL-NETWORKS ; FORECAST |
| WOS研究方向 | Computer Science ; Engineering ; Environmental Sciences & Ecology ; Water Resources |
| 语种 | 英语 |
| WOS记录号 | WOS:001631956200001 |
| 出版者 | ELSEVIER SCI LTD |
| 源URL | [http://ir.igsnrr.ac.cn/handle/311030/219693]  |
| 专题 | 资源与环境信息系统国家重点实验室_外文论文 |
| 通讯作者 | Wang, Weilong; Zhang, Xingxing |
| 作者单位 | 1.Shanghai Ocean Univ, Ctr Res Environm Ecol & Fish Nutr, Minist Agr, Shanghai 201306, Peoples R China; 2.Shanghai Ocean Univ, Coll Informat, Shanghai 201306, Peoples R China; 3.Chinese Acad Sci, Inst Geog Sci & Nat Resources Res, Beijing 100101, Peoples R China |
| 推荐引用方式 GB/T 7714 | Wang, Zhaocai,Ding, Cheng,Xu, Nannan,et al. Interpretable deep learning hybrid streamflow prediction modeling based on multi-source data fusion[J]. ENVIRONMENTAL MODELLING & SOFTWARE,2026,196:106796. |
| APA | Wang, Zhaocai,Ding, Cheng,Xu, Nannan,Wang, Weilong,&Zhang, Xingxing.(2026).Interpretable deep learning hybrid streamflow prediction modeling based on multi-source data fusion.ENVIRONMENTAL MODELLING & SOFTWARE,196,106796. |
| MLA | Wang, Zhaocai,et al."Interpretable deep learning hybrid streamflow prediction modeling based on multi-source data fusion".ENVIRONMENTAL MODELLING & SOFTWARE 196(2026):106796. |
入库方式： OAI收割
来源：地理科学与资源研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。