| 32 | 0 | 2 |
| Downloads | Citas | Reads |
[Objective] Runoff prediction in river basins plays a crucial role in improving hydrological prediction accuracy and water resources management at the basin scale. However， existing runoff prediction methods still have certain limitations in characterizing hydrological interactions and spatial dependencies among stations. To improve the accuracy of multi-site daily runoff prediction， a hybrid model coupling recurrent neural network (RNN) with a graph neural network (GNN) is established to conduct daily-scale runoff prediction in river basins. [Methods] Three hydrological stations in the Ganjiang River Basin were selected as the study sites. Two RNN models， namely long short-term memory (LSTM) and gated recurrent unit (GRU)， were constructed and separately coupled with GNN to compare and analyze the runoff prediction performance of different hybrid models. Furthermore， different input-feature combinations and graph-structure construction methods were set to evaluate the effects of differences in input features and graph-structure construction on the prediction accuracy and robustness of hybrid models. [Results] The results showed that introducing GNN could effectively reduce the errors between observed and predicted runoff of single RNN-based models， with the maximum error reduction reaching 22.94%. In terms of input features， using a single high-quality remote sensing product to replace the corresponding ground observation data as model inputs could achieve relatively good prediction performance， while the simultaneous use of precipitation and potential evapotranspiration remote sensing data could further significantly improve model performance， reducing the relative error (RE) by 7.1%–97.4%. In addition， different construction methods of graph structure directly determined the variation patterns and trends of their monthly mean RE. [Conclusion] The findings indicate that incorporating GNN can enhance the ability of RNN-based models to predict multi-site runoff processes in the river basin， helping reduce prediction errors and improve model robustness. A reasonable graph structure should reflect the runoff relationships between upstream and downstream stations， as well as the driving effects of precipitation on the runoff processes at each station， which has an important impact on model prediction results. The findings can provide a reference for the construction of multi-site runoff prediction models， graph-structure design， and the integrated application of multi-source hydrometeorological factors.
[1] Allan R P， Barlow M， PYRNE M P， et al. Advances in understanding large-scale responses of the water cycle to climate change[J]. Annals of the New York Academy of Sciences， 2020， 1472(1): 49-75.
[2] WANG H， QIN H， LIU G J， et al. Hierarchical attention network for short-term runoff forecasting[J]. Journal of Hydrology， 2024， 638: 131549.
[3] 张挺，马睿佳，程泳铭，等. 基于 VMD-LSTM 模型的流域径流预报技术[J]. 南水北调与水利科技（中英文），2025，23（5）：1196-1203.
[4] WU M M， WEI X， GE W， et al. Analyzing the spatial scale effects of urban elements on urban flooding based on multiscale geographically weighted regression[J]. Journal of Hydrology， 2024， 645: 132178.
[5] 黄一凡， 张翔， 邓梁堃， 等. 基于机器学习的汉江流域径流模拟与时滞变化分析[J]. 水资源保护， 2024， 40(6): 173-180.
[6] WANG S， ZHANG H J， WANG T T， et al. Simulating runoff changes and evaluating under climate change using CMIP6 data and the optimal SWAT model: a case study[J]. Scientific Reports， 2024， 14(1): 23228.
[7] 师鹏飞， 赵酉键， 徐辉荣， 等. 融合相空间重构和深度学习的径流模拟预测[J]. 水科学进展， 2023， 34(3): 388-397.
[8] BEVEN K， FREER J. Equifinality， data assimilation， and uncertainty estimation in mechanistic modelling of complex environmental systems using the GLUE methodology[J]. Journal of Hydrology， 2001， 249(1-4): 11-29.
[9] REFSGAARD J C， STISEN S， KOCH J. Hydrological process knowledge in catchment modelling: Lessons and perspectives from 60 years development[J]. Hydrological Processes， 2022， 36(1): e14463.
[10] SANG Y F， SINGH V P， WEN J， et al. Gradation of complexity and predictability of hydrological processes[J]. Journal of Geophysical Research: Atmospheres， 2015， 120(11): 5334-5343.
[11] PANCHANATHAN A， AHRARI A， GHAG K S， et al. An overview of approaches for reducing uncertainties in hydrological forecasting: Progress and challenges[J]. Earth-Science Reviews， 2024， 258: 104956.
[12] PAPACHARALAMPOUS G， KOUTSOYIANNIS D， MONTANARI A. Quantification of predictive uncertainty in hydrological modelling by harnessing the wisdom of the crowd: Methodology development and investigation using toy models[J]. Advances in Water Resources， 2020， 136: 103471.
[13] FENG D P， LIU J T， LAWSON K， et al. Differentiable， learnable， regionalized process-based models with multiphysical outputs can approach state-of-the-art hydrologic prediction accuracy[J]. Water Resources Research， 2022， 58(10): e2022WR032404.
[14] 晁丽君， 张珂， 陈新宇，等. 基于多源降水融合驱动的WRF-Hydro模型在中小河流洪水预报中的适用性[J]. 河海大学学报(自然科学版)， 2022， 50(3): 55-64.
[15] THYER M， GUPTA H， WESTRA S， et al. Virtual hydrological laboratories: Developing the next generation of conceptual models to support decision making under change[J]. Water Resources Research， 2024， 60(4): e2022WR034234.
[16] DENG C， LIU P， WANG W G， et al. Modelling time-variant parameters of a two-parameter monthly water balance model[J]. Journal of Hydrology， 2019， 573: 918-936.
[17] PATHIRAJA S， ANGHILERI D， BURLANDO P， et al. Insights on the impact of systematic model errors on data assimilation performance in changing catchments[J]. Advances in Water Resources， 2018， 113: 202-222.
[18] 王磊，牛庚，桑学锋，等. 基于径流组分约束的机器学习融水径流模型[J]. 南水北调与水利科技（中英文），2025，23（5）：1173-1184.
[19] TRIPATHY K P， MISHRA A K. Deep learning in hydrology and water resources disciplines: Concepts， methods， applications， and research directions[J]. Journal of Hydrology， 2024， 628: 130458.
[20] FRAME J M， KRATZERT F， GUPTA H V， et al. On strictly enforced mass conservation constraints for modelling the Rainfall-Runoff process[J]. Hydrological Processes， 2023， 37(3): e14847.
[21] DENG C， SUN P Y， YIN X， et al. Assessment of monthly runoff simulations based on a physics-informed machine learning framework: The effect of intermediate variables in its construction[J]. Journal of Environmental Management， 2024， 362: 121299.
[22] KHATUN A， CHATTERJEE C， SAHU G， et al. A novel smoothing-based long short-term memory framework for short-to medium-range flood forecasting[J]. Hydrological Sciences Journal， 2023， 68(3): 488-506.
[23] YANG S Y， JHONG Y D， JHONG B C， et al. Enhancing flooding depth forecasting accuracy in an urban area using a novel trend forecasting method[J]. Water Resources Management， 2024， 38(4): 1359-1380.
[24] MOLINA-NAVARRO E， ANDERSEN H E， NIELSEN A， et al. The impact of the objective function in multi-site and multi-variable calibration of the SWAT model[J]. Environmental Modelling & Software， 2017， 93: 255-267.
[25] CHANG L C， LIOU J Y， CHANG F J. Spatial-temporal flood inundation nowcasts by fusing machine learning methods and principal component analysis[J]. Journal of Hydrology， 2022， 612: 128086.
[26] 张子凡， 刘时银， 马凯， 等. 基于LSTM和气候要素分带的金沙江上游流域径流模拟研究[J]. 地理科学进展， 2023， 42(6): 1139-1152.
[27] MOHAMADIAZAR N， EBRAHIMIAN A， HOSSEINY H. Integrating deep learning， satellite image processing， and spatial-temporal analysis for urban flood prediction[J]. Journal of Hydrology， 2024， 639: 131508.
[28] SUN A Y， JIANG P S， YANG Z L， et al. A graph neural network (GNN) approach to basin-scale river network learning: The role of physics-based connectivity and data fusion[J]. Hydrology and Earth System Sciences， 2022， 26(19): 5163-5184.
[29] LUO Y X， ZHOU Y L， CHEN H， et al. Exploring a spatiotemporal hetero graph-based long short-term memory model for multi-step-ahead flood forecasting[J]. Journal of Hydrology， 2024， 633: 130937.
[30] LI X H， YUAN C Y， SUN T， et al. Identifying the spatiotemporal patterns of drought-flood alternation based on IMERG product in the humid subtropical Poyang Lake basin， China[J]. Journal of Hydrology: Regional Studies， 2024， 54:101912.
[31] DENG C， YIN X， ZOU J C， et al. Assessment of the impact of climate change on streamflow of Ganjiang River catchment via LSTM-based models[J]. Journal of Hydrology: Regional Studies， 2024， 52: 101716.
[32] 余小波， 黄领梅， 申曼华， 等. 基于CN05.1数据集驱动SWAT模型的玉龙喀什河流域径流模拟[J]. 人民珠江， 2024， 45(9): 19-26.
[33] LIU X， ZHANG L P， SHE D X， et al. Postprocessing of hydrometeorological ensemble forecasts based on multisource precipitation in Ganjiang River basin， China[J]. Journal of Hydrology， 2022， 605: 127323.
[34] WANG W， TANG S N， ZOU J C， et al. Runoff prediction in different forecast periods via a hybrid machine learning model for Ganjiang River Basin， China[J]. Water， 2024， 16(11): 1589.
[35] FARFÁN-DURÁN J F， CEA L. Streamflow forecasting with deep learning models: A side-by-side comparison in Northwest Spain[J]. Earth Science Informatics， 2024， 17(6): 5289-5315.
[36] LIU M Y， ZHANG P P， CAI Y P， et al. Spatial-temporal heterogeneity analysis of blue and green water resources for Poyang Lake basin， China[J]. Journal of Hydrology， 2023， 617: 128983.
[37] DENG C， ZHANG Y C， MA M M， et al. Compound temporal-spatial extreme precipitation events in the Poyang Lake Basin of China[J]. Journal of Hydrology: Regional Studies， 2025， 58: 102270.
[38] ADLER R F， WANG J J， SAPIANO M， et al. Global Precipitation Climatology Project (GPCP) Climate Data Record (CDR)， Version 1.3 (Daily)[DB/OL]. NOAA National Centers for Environmental Information， 2017. DOI:10.7289/V5RX998Z.
[39] HUFFMAN G J， STOCKER E F， BOLVIN D T， et al. GPM IMERG Final Precipitation L3 1 day 0.1 degree x 0.1 degree V07[DB/OL]. NASA Goddard Earth Sciences Data and Information Services Center， 2023. DOI:10.5067/GPM/IMERGDF/DAY/07.
[40] LI B L， RODELL M， SHEFFIELD J， et al. Long-term， non-anthropogenic groundwater storage changes simulated by three global-scale hydrological models[J]. Scientific Reports， 2019， 9(1): 10746.
[41] MIRALLES D G， BONTE O， KOPPA A， et al. GLEAM4: Global land evaporation and soil moisture dataset at 0.1° resolution from 1980 to near present[J]. Scientific Data， 2025， 12(1): 416.
[42] HOCHREITER S， SCHMIDHUBER J. Long short-term memory[J]. Neural Computation， 1997， 9(8): 1735-1780.
[43] VAN HOUDT G， MOSQUERA C， NÁPOLES G. A review on the long short-term memory model[J]. Artificial Intelligence Review， 2020， 53(8): 5929-5955.
[44] DEY R， SALEM F M. Gate-variants of Gated Recurrent Unit (GRU) neural networks[C]//IEEE. 2017 IEEE 60th International Midwest Symposium on Circuits and Systems (MWSCAS). Boston， MA， USA: Institute of Electrical and Electronics Engineers， 2017:1597-1600.
[45] WEERAKODY P B， WONG K W， WANG G J， et al. A review of irregular time series data handling with gated recurrent neural networks[J]. Neurocomputing， 2021， 441: 161-178.
[46] WU Z H， PAN S R， CHEN F W， et al. A comprehensive survey on graph neural networks[J]. IEEE Transactions on Neural Networks and Learning Systems， 2021， 32(1): 4-24.
[47] NASH J E， SUTCLIFFE J V. River flow forecasting through conceptual models part I: A discussion of principles[J]. Journal of Hydrology， 1970， 10(3): 282-290.
[48] SAMANTARAY S， SAHOO A， YASEEN Z M， et al. River discharge prediction based multivariate climatological variables using hybridized long short-term memory with nature inspired algorithm[J]. Journal of Hydrology， 2025， 649: 132453.
[49] XU J R， CHEN J R， YOU S Q， et al. Robustness of deep learning models on graphs: A survey[J]. AI Open， 2021， 2: 69-78.
[50] WANG W， LIN H， CHEN N C， et al. Evaluation of multi-source precipitation products over the Yangtze River Basin[J]. Atmospheric Research， 2021， 249: 105287.
[51] CHAROENSUK T， LUCHNER J， BALBARINI N， et al. Enhancing the capabilities of the Chao Phraya forecasting system through the integration of pre-processed numerical weather forecasts[J]. Journal of Hydrology: Regional Studies， 2024， 52: 101737.
[52] WANG J， ZHUO L， RICO-RAMIREZ M A， et al. Interacting effects of precipitation and potential evapotranspiration biases on hydrological modeling[J]. Water Resources Research， 2023， 59(3): e2022WR033323.
[53] PAN X X， HOU J M， WANG T， et al. Study on the influence of temporal and spatial resolution of rainfall data on watershed flood simulation performance[J]. Water Resources Management， 2024， 38(8): 2647-2668.
[54] ZHANG Y F， WU C H， YEH P J， et al. Evaluation and comparison of precipitation estimates and hydrologic utility of CHIRPS， TRMM 3B42 V7 and PERSIANN-CDR products in various climate regimes[J]. Atmospheric Research， 2022， 265: 105881.
Basic Information:
China Classification Code:TP79
Citation Information:
[1]FANG Yuchen,DENG Chao,ZOU Jiacheng ,et al.Research on multi-site daily runoff prediction based on RNN-GNN under multi-source remote sensing data and graph-structure constraints[J].Water Resources and Hydropower Engineering().
Fund Information:
国家重点研发计划项目(2022YFC3202802)
2026-09-16
2026-09-16
2026-09-16