River discharge estimation by integrating Sentinel-2 and machine learning： A case study of the Tangnaihai Hydrometry Station

YIN Wenjie; WANG Xuelei; WANG Chen; WANG Hang; HUANG Caisheng; ZHAO Ruixue; MENG Fanle; LIU Jinxiu

doi:10.11834/jrs.20265294

Views : 0 下载量: 0 CSCD: 0

R-PDF
PDF
Export
Share
Collection
Album

River discharge estimation by integrating Sentinel-2 and machine learning： A case study of the Tangnaihai Hydrometry Station
Pages: 1-15(2026)
Received：08 August 2025，

Online First：27 February 2026，
DOI： 10.11834/jrs.20265294
稿件说明：

移动端阅览

尹文杰，王雪蕾，王晨，王航，黄才声，赵瑞雪，孟凡乐，刘锦绣.XXXX.基于Sentinel-2与机器学习的河流流量估算研究——以唐乃亥水文站为例.遥感学报，XX（XX）： 1-15 DOI： 10.11834/jrs.20265294.

YIN Wenjie，WANG Xuelei，WANG Chen，WANG Hang，HUANG Caisheng，ZHAO Ruixue，MENG Fanle，LIU Jinxiu. XXXX. River discharge estimation by integrating Sentinel-2 and machine learning： A case study of the Tangnaihai Hydrometry Station. National Remote Sensing Bulletin， XX（XX）：1-15 DOI： 10.11834/jrs.20265294.

摘要

河流流量是水文循环过程的关键变量，在洪水预警、水资源调度和生态环境管理中具有重要意义。本论文选取黄河源区的唐乃亥水文站为试点，提出一种融合卫星遥感和机器学习方法的河流流量监测技术。首先利用哨兵2号遥感影像提取河流的水面宽度，联合全球陆面数据同化系统模拟的气象水文变量（蒸散发、土壤湿度、温度、陆地水储量和径流量）作为6种驱动因子，分别构建了基于四种统计方法（线性函数、幂函数、指数函数和多项式函数）和四种机器学习方法（XGBoost、Random Forest、LightGBM和CatBoost）的流量监测模型，评估不同模型监测结果间的差异并利用Shapley Additive Explanations （SHAP）方法量化不同驱动因子的重要性。结果表明，四种统计方法中，多项式函数模型在测试期的效果优于其它三种模型；相比于传统统计方法，机器学习方法在精度和稳定性方面表现出显著优势，决定系数（

）提高了46.15%，均方根误差（

RMSE

）和平均绝对误差（

MAE

）分别减少了54.61%和55.65%。Random Forest模型在测试期的模拟效果优于其它三种模型，其

、

NSE、RMSE

和

MAE

分别为0.96，0.89，172.81 m³/s和147.33 m³/s。SHAP方法表明水面宽度在流量监测模型中具有最显著的贡献（189.02），其次为土壤湿度（145.11）和温度（97.41）。本研究证实了联合卫星遥感和机器学习方法在复杂地形与资料匮乏区域开展高精度流量估算的可行性和优越性。

Abstract

(Objective)

River discharge is a pivotal variable within the hydrological cycle

holding significant importance for flood warning

water resource allocation

and eco-environmental management. Traditional ground-based methods are limited by sparse station distribution and high costs of data acquisition

particularly in areas with complex terrains or remote regions

making it difficult to meet the demands of precise w

ater resource management. Satellite remote sensing technology offers extensive coverage and high spatiotemporal resolution

providing new data sources and methodologies for river discharge monitoring. Machine learning (ML) approaches can accurately simulate complex relationships between river discharge and multiple driving factors

offering novel avenues for processing intricate hydrological data and optimizing models. By integrating machine learning algorithms with remote sensing and

in-situ

river discharge

it can provide the innovative measure for the efficient and reliable of river discharge monitoring.

(Method)

This study selected the Tangnaihai Hydrometry Station as the study area

and proposed a river discharge monitoring method by integrating satellite remote sensing and ML methods. Firslty

the Sentinel-2 imagery was utilized to extract river water surface width based on the Google Earth Engine cloud platform. The GLDAS v2.2 model-simulated five variables were served as predictor variables

namely evapotranspiration

soil moisture

temperature

terrestrial water storage and runoff. Discharge monitoring models were subsequently developed based on four statistical methods (linear function

power function

exponential function

and polynomial function) and four ML algorithms (XGBoost

Random Forest

LightGBM

and CatBoost). The discrepancies among different models were assessed

and the Shapley Additive Explanation (SHAP) method was employed to quantify the importance of different input variables.

(Result)

The results demonstrate that the polynomial function model demonstrates superior performance over other three statistical models during the testing period

with an

²of 0.67

and its error metrics (MAE: 319.01 m³/s

RMSE: 393.14 m³/s) were lower than those of the other three statistical models. Compare to traditional statistical approaches

ML models exhibit significant improvements overall in both simulation a

ccuracy and stability

and the coefficient of determination (

) increased by 46.15%

while the root mean square error (

RMSE

) and mean absolute error (

MAE

) decreased by 54.61% and 55.65

respectively. Notably

the Random Forest model achieved the optimal performance in the testing phase

with the

of 0.96

Nash-Sutcliffe efficiency coefficient (

NSE

) of 0.89

RMSE

of 172.81 m

and

MAE

of 147.33 m

reflecting robust generalization capability and stability. SHAP analysis revealed that water surface width contributed most significantly to the discharge monitoring model (189.02)

followed by soil moisture (145.11) and temperature (97.41). The runoff variable exhibited the minimal degree of influence on the river discharge monitoring model with the value of 14.14%.

(Conclusion)

This study confirms the feasibility and superiority of integrating satellite remote sensing and ML approaches for high-accuracy discharge estimation in regions characterized by complex topography and data scarcity. Future work could be optimized by integration of higher-resolution satellite imagery and mechanistic models with physical processes.

关键词

Keywords

references

Bai X J , Wang Z Y , Liu T X , Li X D and Wan T Z . 2024 . Response to climate changes and prediction mod-el of runoff in the upper Yellow River Basin . Science Technology and Engineering , 24 ( 18 ): 7502 - 7509

白小晶 , 王中玉 , 刘泰兴 , 李晓丹 , 万铁庄 . 2024 . 黄河上游径流量对气候变化的响应和预测模型 . 科学技术与工程 , 24 ( 18 ): 7502 - 7509 [ DOI: 10.12404/j.issn.1671-1815.2305557 http://dx.doi.org/10.12404/j.issn.1671-1815.2305557 ]

Crochemore L , Isberg K , Pimentel R , Pineda L , Hasan A and Arheimer B . 2019 . Lessons learnt from checking the quality of openly accessible river flow data worldwide . Hydrological Sciences Journal , 65 ( 5 ), 699 - 711 [ DOI: 10.1080/02626667.2019.1659509 http://dx.doi.org/10.1080/02626667.2019.1659509 ]

Dai J J , Liu T Y , Zhao Y Y , Tian S F , Ye C Y and Nie Z . 2023 . Remote sensing inversion of the Zabuye S-alt Lake in Tibet, China using LightGBM algorithm . Frontiers in Earth Science , 10 : 1022280 [ DOI: 10.3389/feart.2022.1022280 http://dx.doi.org/10.3389/feart.2022.1022280 ]

Fang S , Liang F B and Liu Y J . 2021 . A survey on statistical regression models and optimization algorithms . Journal of Fuzhou University(Natural Science Edition) , 49 ( 05 ): 638 - 654

方升 , 梁飞豹 , 刘勇进 . 2021 . 统计回归模型及其优化算法综述 . 福州大学学报(自然科学版) , 49 ( 05 ): 638 - 654 [ DOI: 10.7631/issn.1000-2243.21283 http://dx.doi.org/10.7631/issn.1000-2243.21283 ]

Filippucci P , Brocca L , Bonafoni S , Saltalippi C , Wagner W and Tarpanelli A . 2022 . Sentinel-2 high-resolution data for river discharge monitoring . Remote Sensing of Environment , 281 [ DOI: 10.1016/j.rse.2022.113255 http://dx.doi.org/10.1016/j.rse.2022.113255 ]

Hao R N and Bai Z X . 2023 . Comparative study for daily streamflow simulation with different machine learni-ng methods . Water , 15 ( 6 ): 1179 - 1179 [ DOI: 10.3390/W15061179 http://dx.doi.org/10.3390/W15061179 ]

Ji P and Yuan X . 2020 . Underestimation of the warming trend over the tibetan plateau during 1998 – 2013 by g-lobal land data assimilation systems and atmospheric reanalyses. Journal of Meteorological Research , 34 ( 1 ): 88 - 100 [ DOI: 10.1007/s13351-020-9100-3 http://dx.doi.org/10.1007/s13351-020-9100-3 ]

Lan X J , He Y L and Wu S W . 2024 . River water quality prediction based on RF-BiLSTM model . Journal of Changjiang River Scientific Research Institute , 41 ( 7 ): 57 - 63

兰小机 , 贺永兰 , 武帅文 . 2024 . 基于RF-BiLSTM模型的河流水质预测 . 长江科学院院报 , 41 ( 7 ): 57 - 63 [ DOI: 10.11988/ckyyb.20230244 http://dx.doi.org/10.11988/ckyyb.20230244 ]

Le Y , Liu J T and Wen H . 2024 . Extraction and spatiotemporal variation of poyang lake water body based on multi-source and multi-phase images . Journal of Changjiang River Scientific Research Institute , 41 ( 8 ): 164 - 171

乐颖 , 刘聚涛 , 文慧 . 2024 . 基于多源多时相影像的鄱阳湖水体提取及时空变化分析 . 长江科学院院报 , 41 ( 8 ): 164 - 171 [ DOI: 10.11988/ckyyb.20230268 http://dx.doi.org/10.11988/ckyyb.20230268 ]

Li X T , Li Z L and Han R C . 2023 . Evaluations of different bias correction methods on the GCM precipitati-on data . Journal of China Hydrology , 43 ( 03 ): 93 - 100

李昕潼 , 李占玲 , 韩孺村 . 2023 . 不同偏差校正法对GCM降水数据的应用效果分析 . 水文 , 43 ( 03 ): 93 - 100 [ DOI: 10.19797/j.cnki.1000-0852.20220060 http://dx.doi.org/10.19797/j.cnki.1000-0852.20220060 ]

Li Z Q . 2022 . Extracting spatial effects from machine learning model using local interpretation method: An ex-ample of SHAP and XGBoost . Computers, Environment and Urban Systems , 96 : 101845 [ DOI: 10.1016/j.co-mpenvurbsys.2022.101845 http://dx.doi.org/10.1016/j.co-mpenvurbsys.2022.101845 ]

Liu D Y , Dong W Z , Gou R and Su W C . 2025 . Spatial and temporal evolution of ecosystem service supply and demand and analysis of driving factors in chongqing based on InVEST model and XGBoost-SHAP . E-nvironmental Science , 1 - 17

刘东岳 , 董文卓 , 勾容 , 苏维词 . 2025 . 基于InVEST模型和XGBoost-SHAP的重庆市生态系统服务供需时空演变及驱动因素分析 . 环境科学 , 1 - 17 [ DOI: 10.13227/j.hjkx.202504257 http://dx.doi.org/10.13227/j.hjkx.202504257 ]

Lundberg S M and Lee S I . 2017 . A unified approach to interpreting model predictions . Advances in Neural Information Processing Systems , 30 : 4768 - 4777 [ DOI: 10.48550/arXiv.1705.07874 http://dx.doi.org/10.48550/arXiv.1705.07874 ]

Madhushani C , Dananjaya K , Ekanayake I.U , Meddage D . P .P, Kantamaneni K and Rathnayake U. 2024 . Modeli-ng streamflow in non-gauged watersheds with sparse data considering physiographic, dynamic climate, and anthropogenic factors using explainable soft computing techniques. Journal of Hydrology , 631 : 130846 [ DOI: 10.1016/J.JHYDROL.2024.130846 http://dx.doi.org/10.1016/J.JHYDROL.2024.130846 ]

Mo J Y , Tian Y C , Wang J L , Du J Z , Zhang Q , Zhang Y L , Tao J and Lin J L . 2025 . Remote sensing inv-ersion of COD in Maowei Sea and nearshore aquaculture ponds based on machine learning . Haiyang Xuebao , 47 ( 5 ): 128 - 140

莫锦英 , 田义超 , 王家乐 , 杜金泽 , 张强 , 张亚丽 , 陶进 , 林俊良 . 2025 . 基于机器学习的茅尾海及近岸养殖池水体COD遥感反演 . 海洋学报 , 47 ( 5 ): 128 - 140 [ DOI: 10.12284/hyxb2025046 http://dx.doi.org/10.12284/hyxb2025046 ]

Papa F , Crétaux J F , Grippa M , Robert E , Trigg M , Tshimanga R M , Kitambo B , Paris A , Carr A , Fleischmann A S , de Fleury M , Gbetkom P G , Calmettes B and Calmant S . 2022 . Water Resources in Africa under Global Change: Monitoring Surface Waters from Space . Surveys in geophysics , 44 ( 1 ): 51 - 51 [ DOI: 10.1007/s10712-022-09700-9 http://dx.doi.org/10.1007/s10712-022-09700-9 ]

Prasad S D , Bhabagrahi B , Kumar T M and Kumar B G . 2022 . Integrated remote sensing and machine learnin-g tools for estimating ecological flow regimes in tropical river reaches . Journal of Environmental Manage-ment , 322 : 116121 [ DOI: 10.1016/J.JENVMAN.2022.116121 http://dx.doi.org/10.1016/J.JENVMAN.2022.116121 ]

Qian Y P , Lin Y P , Jin S Y , Song R P and Jiang X H . 2004 . Analysis of water resources changes in the sou-rce region of the Yellow River . Water Resources and Hydropower Engineering , ( 05 ): 8 - 10

钱云平 , 林银平 , 金双彦 , 宋瑞鹏 , 蒋秀华 . 2004 . 黄河河源区水资源变化分析 . 水利水电技术(中英文) , ( 05 ): 8 - 10 [ DOI: 10.13928/j.cnki.wrahe.2004.05.003 http://dx.doi.org/10.13928/j.cnki.wrahe.2004.05.003 ]

Ren J , Wang J K , Chen R , Li H , Xu D L , Yan L H and Song J Y . 2024 . Remote sensing identification of s-hallow landslide based on improved otsu algorithm and multi feature threshold . Frontiers in Earth Science , 12 : 1473904 [ DOI: 10.3389/FEART.2024.1473904 http://dx.doi.org/10.3389/FEART.2024.1473904 ]

Samat A , Li E Z , Du P J , Liu S C , Miao Z L and Zhang W . 2020 . CatBoost for RS image classification with pseudo label support from neighbor patches-based clustering . IEEE Geoscience and Remote Sensing Letters , 19 : 1 - 5 [ DOI: 10.1109/LGRS.2020.3038771 http://dx.doi.org/10.1109/LGRS.2020.3038771 ]

Shi J , Zhou J G , Wang H , Gan L , Shen J L , Li X , Li Y Y and Wu J S . 2019 . Analyzing the uncertainty in-duced by methods used to calculate the missing data in time series: A case study based on meteorological and hydrological data in small watershed . Journal of Irrigation and Drainage . 38 ( 02 ): 84 - 92

石锦 , 周脚根 , 王辉 , 甘蕾 , 沈健林 , 李希 , 李裕元 , 吴金水 . 2019 . 点源时间序列数据缺失值的估值不确定性分析—以小流域气象和水文数据为例 . 灌溉排水学报 , 38 ( 02 ): 84 - 92 [ DOI: 10.13522/j.cnki.ggps.2017.0421 http://dx.doi.org/10.13522/j.cnki.ggps.2017.0421 ]

Shi Y M , Liu X S , Zhu W B and Song H L . 2022 . Research on inversion of river discharge in high mountai-n region based on GEE platform . Remote Sensing Technology and Application , 37 ( 1 ): 186 - 195

史宜梦 , 刘希胜 , 朱文彬 , 宋宏利 . 2022 . 基于GEE云平台的黄河源区河流径流量遥感反演研究 . 遥感技术与应用 , 37 ( 1 ): 186 - 195 [ DOI: 10.11873/j.issn.1004-0323.2022.1.0186 http://dx.doi.org/10.11873/j.issn.1004-0323.2022.1.0186 ]

Su L P , Zhou X and Zhang S P . 2024 . Research on small and medium-sized watershed flow prediction based on GRU and XGBoost algorithm . Jilin Water Resources , ( 10 ): 72 - 78

苏林鹏 , 周祥 , 张守平 . 2024 . 基于GRU和XGBoost算法的中小流域流量预测研究 . 吉林水利 , ( 10 ): 72 - 78 [ DOI: 10.15920/j.cnki.22-1179/tv.2024.10.014 http://dx.doi.org/10.15920/j.cnki.22-1179/tv.2024.10.014 ]

Sun W C , Wang X C and Xu Z X . 2024 . Estimating streamflow using remote sensing: Progress and prospects . ACTA GEOGRAPHICA SINICA , 79 ( 03 ): 565 - 583

孙文超 , 王星灿 , 徐宗学 . 2024 . 河道流量遥感估算研究进展与展望 . 地理学报 , 79 ( 03 ): 565 - 583 [ DOI: 10.11821/dlxb202403002 http://dx.doi.org/10.11821/dlxb202403002 ]

Van den Broeck G , Lykov A , Schleich , M and Suciu D . 2022 . On the tractability of SHAP explanations . Jour-nal of Artificial Intelligence Research , 74 : 851 - 886 [ DOI: 10.1613/jair.1.13283 http://dx.doi.org/10.1613/jair.1.13283 ]

Wang J P , Wu X D , Ma D J , Wen J G and Xiao Q . 2023 . Remote sensing retrieval based on machine learni-ng algorithm: Uncertainty analysis . National Remote Sensing Bulletin , 27 ( 3 ): 790 - 801

汪静平 , 吴小丹 , 马杜娟 , 闻建光 , 肖青 . 2023 . 基于机器学习的遥感反演:不确定性因素分析 . 遥感学报 , 27 ( 3 ): 790 - 801 [ DOI: 10.11834/jrs.20221172 http://dx.doi.org/10.11834/jrs.20221172 ]

Wang X C , Sun W C , Lu F and Zuo R . 2023 . Combining satellite optical and radar image data for streamflo-w estimation using a machine learning method . Remote Sensing , 15 ( 21 ): 5184 [ DOI: 10.3390/RS15215184 http://dx.doi.org/10.3390/RS15215184 ]

Xu H Q . 2005 . A study on information extraction of water body with the modified normalized difference water index(MNDWI) . National Remote Sensing Bulletin , ( 05 ): 589 - 595

徐涵秋 . 2005 . 利用改进的归一化差异水体指数(MNDWI)提取水体信息的研究 . 遥感学报 , ( 05 ): 589 - 595 [ DOI: 10.11834/jrs.20050586 http://dx.doi.org/10.11834/jrs.20050586 ]

Yang S T , Wang P F , Wang J , Lou H Z and Gong T L . 2021 . River flow estimation method based on UAV aerial photogrammetry . National Remote Sensing Bulletin , 25 ( 6 ): 1284 - 1293

杨胜天 , 王鹏飞 , 王娟 , 娄和震 , 巩同梁 . 2021 . 结合无人机航空摄影测量的河道流量估算 . 遥感学报 , 25 ( 6 ): 1284 - 1293 [ DOI: 10.11834/jrs.20209082 http://dx.doi.org/10.11834/jrs.20209082 ]

Zhang S T , Lu X , Lu Y , Cheng L , Li M C and Yang K . 2021 . Tracking dynamic river networks in the Tibetan Plateau with high-resolution CubeSat imagery . National Remote Sensing Bulletin , 25 ( 10 ): 2142 - 2152

章斯腾 , 陆欣 , 陆瑶 , 程亮 , 李满春 , 杨康 . 2021 . 青藏高原河流网络高分CubeSat遥感监测 . 遥感学报 , 25 ( 10 ): 2142 - 2152 [ DOI: 10.11834/jrs.20219268 http://dx.doi.org/10.11834/jrs.20219268 ]