A Novel Strategy for Automatic Selection of Cross-Basin Data to Improve Local Machine Learning-Based Runoff Models
文献类型：期刊论文
| 作者 | Nai, Congyi3,4; Liu, Xingcai3,4; Tang, Qiuhong3,4; Liu, Liu2; Sun, Siao1; Gaffney, Paul P. J.4 |
| 刊名 | WATER RESOURCES RESEARCH
 |
| 出版日期 | 2024-05-01 |
| 卷号 | 60期号:5页码:e2023WR035051 |
| 关键词 | LSTM deep learning basin-affinity data selection |
| DOI | 10.1029/2023WR035051 |
| 产权排序 | 1 |
| 文献子类 | Article |
| 英文摘要 | Previous studies have shown that regional deep learning (DL) models can improve runoff prediction by leveraging large hydrological datasets. However, training a DL regional model using all data without screening may degrade local performance. This study focuses on constructing enhanced local models through the utilization of cross-basin data. To this end, we propose an approach that employs a novel training strategy to optimize DL model training for specific basins. The approach measures the impact of any one basin's gradient on the loss of the basin of interest, providing insights into the relationships between different basins. The approach was validated using 531 basins from the CAMELS dataset. Results suggest that local performance degradation is a common occurrence in regional models, and imbalanced data are likely to result in a specific pattern dominating the entire regional model. In comparison to a regional model simply trained with all basins, the median Nash-Sutcliffe efficiency (NSE) for our models is 0.031 higher. In particular, the increase in NSE can exceed 0.2 for some dry basins. Our findings indicate that this novel DL strategy can significantly improve model performance in specific basins using large hydrological datasets, while mitigating local performance loss. In the realm of deep learning, incorporating more data into the training process typically results in a more potent model. Conventionally, large datasets have been employed to train regional models with the intention of predicting rainfall-runoff processes across all basins. However, such regional models often encounter performance degradation when applied to local basins. This degradation can be attributed to the extraction of overly general features, leading to a loss of specificity. In this paper, our objective is to harness information from a large dataset to establish a more robust local model. We have introduced a method that autonomously learns the similarities in rainfall-runoff behavior among basins. Subsequently, we utilize this learned similarity to selectively choose data that proves advantageous for training the local model. Our results demonstrate that this strategy can significantly enhance runoff prediction, particularly in arid basins. Training a DL regional model using all data without screening may degrade local performance Proper selection of training data is crucial for enhancing DL model training for individual basins, especially arid basins Data from different basins might act as mutual noise during the training process |
| WOS研究方向 | Environmental Sciences & Ecology ; Marine & Freshwater Biology ; Water Resources |
| WOS记录号 | WOS:001230335500001 |
| 源URL | [http://ir.igsnrr.ac.cn/handle/311030/205167]  |
| 专题 | 陆地水循环及地表过程院重点实验室_外文论文 |
| 通讯作者 | Liu, Xingcai |
| 作者单位 | 1.Chinese Acad Sci, Key Lab Reg Sustainable Dev Modeling, Inst Geog Sci & Nat Resources Res, Beijing, Peoples R China 2.China Agr Univ, Ctr Agr Water Res China, Beijing, Peoples R China 3.Univ Chinese Acad Sci, Beijing, Peoples R China 4.Chinese Acad Sci, Key Lab Water Cycle & Related Land Surface Proc, Inst Geog Sci & Nat Resources Res, Beijing, Peoples R China |
| 推荐引用方式 GB/T 7714 | Nai, Congyi,Liu, Xingcai,Tang, Qiuhong,et al. A Novel Strategy for Automatic Selection of Cross-Basin Data to Improve Local Machine Learning-Based Runoff Models[J]. WATER RESOURCES RESEARCH,2024,60(5):e2023WR035051. |
| APA | Nai, Congyi,Liu, Xingcai,Tang, Qiuhong,Liu, Liu,Sun, Siao,&Gaffney, Paul P. J..(2024).A Novel Strategy for Automatic Selection of Cross-Basin Data to Improve Local Machine Learning-Based Runoff Models.WATER RESOURCES RESEARCH,60(5),e2023WR035051. |
| MLA | Nai, Congyi,et al."A Novel Strategy for Automatic Selection of Cross-Basin Data to Improve Local Machine Learning-Based Runoff Models".WATER RESOURCES RESEARCH 60.5(2024):e2023WR035051. |
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。