Generalization of an Encoder-Decoder LSTM model for flood prediction in ungauged catchments

被引:21
|
作者
Zhang, Yikui [1 ,5 ]
Ragettli, Silvan [2 ]
Molnar, Peter [1 ]
Fink, Olga [3 ]
Peleg, Nadav [4 ]
机构
[1] Swiss Fed Inst Technol, Inst Environm Engn, Zurich, Switzerland
[2] Hydrosolut Ltd, Zurich, Switzerland
[3] Ecole Polytech Fed Lausanne, Lab Intelligent Maintenance & Operat Syst, Lausanne, Switzerland
[4] Univ Lausanne, Inst Earth Surface Dynam, Lausanne, Switzerland
[5] Res Ctr Julich, Inst Bio & Geosci, Agrosphere IBG 3, Julich, Germany
基金
瑞士国家科学基金会;
关键词
Flood prediction; Ungauged catchments; Deep learning; LSTM; BASINS; REGIONALIZATION;
D O I
10.1016/j.jhydrol.2022.128577
中图分类号
TU [建筑科学];
学科分类号
0813 ;
摘要
Flood prediction in ungauged catchments is usually conducted by hydrological models that are parameterized based on nearby and similar gauged catchments. As an alternative to this process-based modelling, deep learning (DL) models have demonstrated their ability for prediction in ungauged catchments (PUB) with high efficiency. Catchment characteristics, the number of gauged catchments, and their level of hydroclimatic heterogeneity in the training dataset used for model regionalization can directly affect the model's performance. Here, we study the generalization ability of a DL model to these factors by applying an Encoder-Decoder Long Short-Term Memory neural network for a 6-hour lead-time runoff prediction in 35 mountainous catchments in China. By varying the available number of catchments and model settings with different training datasets, namely local, regional, and PUB models, we evaluated the generalization ability of our model. We found that both quantity (i.e. number of gauged catchments available) and heterogeneity of the training dataset used for the DL model are important for improving model performance in the PUB context, due to a data synergy effect. The assessment of the sensitivity to catchment characteristics showed that the model performance is mainly correlated to the local hydro-climatic conditions; the more arid the region, the more likely it is to have a poor model performance for prediction in ungauged catchments. The results suggest that the regional ED-LSTM model is a promising method to predict streamflow from rainfall inputs in PUB, and outline the need for preparing a representative training dataset.
引用
收藏
页数:11
相关论文
共 50 条
  • [1] Pedestrian behavior prediction model with a convolutional LSTM encoder-decoder
    Chen, Kai
    Song, Xiao
    Han, Daolin
    Sun, Jinghan
    Cui, Yong
    Ren, Xiaoxiang
    [J]. PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS, 2020, 560
  • [2] Using LSTM encoder-decoder for rhetorical structure prediction
    de Moura, Gustavo Bennemann
    Feltrim, Valeria Delisandra
    [J]. 2018 7TH BRAZILIAN CONFERENCE ON INTELLIGENT SYSTEMS (BRACIS), 2018, : 278 - 283
  • [3] An LSTM based Encoder-Decoder Model for Multi-Step Traffic Flow Prediction
    Du, Shengdong
    Li, Tianrui
    Yang, Yan
    Gong, Xun
    Homg, Shi-Jinn
    [J]. 2019 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2019,
  • [4] A novel Encoder-Decoder model based on read-first LSTM for air pollutant prediction
    Zhang, Bo
    Zou, Guojian
    Qin, Dongming
    Lu, Yunjie
    Jin, Yupeng
    Wang, Hui
    [J]. SCIENCE OF THE TOTAL ENVIRONMENT, 2021, 765
  • [5] Unsupervised Encoder-Decoder Model for Anomaly Prediction Task
    Wu, Jinmeng
    Shu, Pengcheng
    Hong, Hanyu
    Li, Xingxun
    Ma, Lei
    Zhang, Yaozong
    Zhu, Ying
    Wang, Lei
    [J]. MULTIMEDIA MODELING, MMM 2023, PT II, 2023, 13834 : 549 - 561
  • [6] Crossing-Road Pedestrian Trajectory Prediction via Encoder-Decoder LSTM
    Xue, Peixin
    Liu, Jianyi
    Chen, Shitao
    Zhou, Zhuoli
    Huo, Yongbo
    Zheng, Nanning
    [J]. 2019 IEEE INTELLIGENT TRANSPORTATION SYSTEMS CONFERENCE (ITSC), 2019, : 2027 - 2033
  • [7] Long-Term Traffic Prediction Based on LSTM Encoder-Decoder Architecture
    Wang, Zhumei
    Su, Xing
    Ding, Zhiming
    [J]. IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, 2021, 22 (10) : 6561 - 6571
  • [8] Short-term Wind Power Prediction Model Based on Encoder-Decoder LSTM<bold> </bold>
    Lu, Kuan
    Sun, Wen Xue
    Wang, Xin
    Meng, Xiang Rong
    Zhai, Yong
    Li, Hong Hai
    Zhang, Rong Gui
    [J]. 2018 INTERNATIONAL CONFERENCE OF GREEN BUILDINGS AND ENVIRONMENTAL MANAGEMENT (GBEM 2018), 2018, 186
  • [9] Regional ship collision risk prediction: An approach based on encoder-decoder LSTM neural network model
    Lin, Chenyan
    Zhen, Rong
    Tong, Yanting
    Yang, Shenhua
    Chen, Shengkai
    [J]. OCEAN ENGINEERING, 2024, 296
  • [10] A multitask encoder-decoder model for quality prediction in injection moulding
    Muaz, Muhammad
    Yu, Hanxin
    Sung, Wai Lam
    Liu, Chang
    Drescher, Benny
    [J]. JOURNAL OF MANUFACTURING PROCESSES, 2023, 103 : 238 - 247