Predicting Chlorophyll-a Concentrations in the World’s Largest Lakes Using Kolmogorov-Arnold Networks DOI
Mohammad Javad Saravani, Roohollah Noori, Changhyun Jun

et al.

Environmental Science & Technology, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 16, 2025

Accurate prediction of chlorophyll-a (Chl-a) concentrations, a key indicator eutrophication, is essential for the sustainable management lake ecosystems. This study evaluated performance Kolmogorov-Arnold Networks (KANs) along with three neural network models (MLP-NN, LSTM, and GRU) traditional machine learning tools (RF, SVR, GPR) predicting time-series Chl-a concentrations in large lakes. Monthly remote-sensed data derived from Aqua-MODIS spanning September 2002 to April 2024 were used. The based on their forecasting capabilities March August 2024. KAN consistently outperformed others both test forecast (unseen data) phases demonstrated superior accuracy capturing trends, dynamic fluctuations, peak concentrations. Statistical evaluation using ranking metrics critical difference diagrams confirmed KAN's robust across diverse sites, further emphasizing its predictive power. Our findings suggest that KAN, which leverages KA representation theorem, offers improved handling nonlinearity long-term dependencies data, outperforming grounded universal approximation theorem algorithms.

Language: Английский

Deep learning in hydrology and water resources disciplines: concepts, methods, applications, and research directions DOI Creative Commons
Kumar Puran Tripathy, Ashok K. Mishra

Journal of Hydrology, Journal Year: 2023, Volume and Issue: 628, P. 130458 - 130458

Published: Nov. 15, 2023

Language: Английский

Citations

95
Assessing the Physical Realism of Deep Learning Hydrologic Model Projections Under Climate Change DOI
Sungwook Wi, Scott Steinschneider

Water Resources Research, Journal Year: 2022, Volume and Issue: 58(9)

Published: Aug. 30, 2022

Abstract This study examines whether deep learning models can produce reliable future projections of streamflow under warming. We train a regional long short‐term memory network (LSTM) to daily in 15 watersheds California and develop three process (HYMOD, SAC‐SMA, VIC) as benchmarks. force all with scenarios warming assess their hydrologic response, including shifts the hydrograph total runoff ratio. All show shift more winter runoff, reduced summer decline ratio due increased evapotranspiration. The LSTM predicts similar but some an unrealistic increase then test two alternative versions which model outputs are used either additional training targets (i.e., multi‐output LSTM) or input features. Results indicate that does not correct hybrid using estimates evapotranspiration from SAC‐SMA feature produces realistic projections, this hold for VIC HYMOD. suggests method depends on fidelity model. Finally, we climate change responses trained over 500 across United States find Ultimately, work modeling may support use LSTMs change, so large, diverse set watersheds.

Language: Английский

Citations

86
A Multiscale Deep Learning Model for Soil Moisture Integrating Satellite and In Situ Data DOI
Jiangtao Liu, Farshid Rahmani, Kathryn Lawson

et al.

Geophysical Research Letters, Journal Year: 2022, Volume and Issue: 49(7)

Published: March 15, 2022

Abstract Deep learning (DL) models trained on hydrologic observations can perform extraordinarily well, but they inherit deficiencies of the training data, such as limited coverage in situ data or low resolution/accuracy satellite data. Here we propose a novel multiscale DL scheme simultaneously from and to predict 9 km daily soil moisture (5 cm depth). Based spatial cross‐validation over sites conterminous United States, obtained median correlation 0.901 root‐mean‐square error 0.034 m 3 /m . It outperformed Soil Moisture Active Passive mission's product, alone, land surface models. Our product showed better accuracy than previous 1 downscaling products, highlighting impacts improving resolution. Not only is our useful for planning against floods, droughts, pests, generically applicable geoscientific domains with multiple scales, breaking confines individual sets.

Language: Английский

Citations

73
Temperature outweighs light and flow as the predominant driver of dissolved oxygen in US rivers DOI
Wei Zhi, Wenyu Ouyang, Chaopeng Shen

et al.

Nature Water, Journal Year: 2023, Volume and Issue: 1(3), P. 249 - 260

Published: March 9, 2023

Language: Английский

Citations

72
Deep learning based data-driven model for detecting time-delay water quality indicators of wastewater treatment plant influent DOI

Yituo Zhang,

Chaolin Li,

Hengpan Duan

et al.

Chemical Engineering Journal, Journal Year: 2023, Volume and Issue: 467, P. 143483 - 143483

Published: May 15, 2023

Language: Английский

Citations

65
Deep learning for water quality DOI
Wei Zhi, Alison P. Appling, Heather E. Golden

et al.

Nature Water, Journal Year: 2024, Volume and Issue: 2(3), P. 228 - 241

Published: March 12, 2024

Language: Английский

Citations

64
Widespread deoxygenation in warming rivers DOI
Wei Zhi, Christoph Klingler, Jiangtao Liu

et al.

Nature Climate Change, Journal Year: 2023, Volume and Issue: 13(10), P. 1105 - 1113

Published: Sept. 14, 2023

Language: Английский

Citations

61
Prediction modelling framework comparative analysis of dissolved oxygen concentration variations using support vector regression coupled with multiple feature engineering and optimization methods: A case study in China DOI Creative Commons
Xizhi Nong, Laifei Cheng, Lihua Chen

et al.

Ecological Indicators, Journal Year: 2023, Volume and Issue: 146, P. 109845 - 109845

Published: Jan. 2, 2023

Dissolved oxygen (DO) is an essential indicator for assessing water quality and managing aquatic environments, but it still a challenging topic to accurately understand predict the spatiotemporal variation of DO concentrations under complex effects different environmental factors. In this study, practical prediction framework was proposed based on support vector regression (SVR) model coupling multiple intelligence techniques (i.e., four data denoising techniques, three feature selection rules, hyperparameter optimization methods). The holistic tested using matrix (17,532 observation in total) 12 indicators from vital monitoring stations longest inter-basin diversion project world Middle-Route South-to-North Water Diversion Project China), during year 2017 2020 period. results showed that we advocated could successfully concentration variations geographical locations. used "wavelet analysis–LASSO regression–random search–SVR" combination Waihuanhe station has best performance, with Root Mean Square Error (RMSE), (MSE), Absolute (MAE), coefficient determination (R2) values 0.251, 0.063, 0.190, 0.911, respectively. combined methods can significantly promote robustness accuracy provide new universal way investigating understanding drivers variations. For management department, comprehensive also identify reveal key parameters should be concerned monitored factors change. More studies terms potential integrated risk multi-indicators mega projects and/or similar bodies are required future.

Language: Английский

Citations

60
The suitability of differentiable, physics-informed machine learning hydrologic models for ungauged regions and climate change impact assessment DOI Creative Commons
Dapeng Feng, Hylke E. Beck, Kathryn Lawson

et al.

Hydrology and earth system sciences, Journal Year: 2023, Volume and Issue: 27(12), P. 2357 - 2373

Published: June 30, 2023

Abstract. As a genre of physics-informed machine learning, differentiable process-based hydrologic models (abbreviated as δ or delta models) with regionalized deep-network-based parameterization pipelines were recently shown to provide daily streamflow prediction performance closely approaching that state-of-the-art long short-term memory (LSTM) deep networks. Meanwhile, full suite diagnostic physical variables and guaranteed mass conservation. Here, we ran experiments test (1) their ability extrapolate regions far from gauges (2) make credible predictions long-term (decadal-scale) change trends. We evaluated the based on hydrograph metrics (Nash–Sutcliffe model efficiency coefficient, etc.) predicted decadal For in ungauged basins (PUB; randomly sampled representing spatial interpolation), either approached surpassed LSTM metrics, depending meteorological forcing data used. They presented comparable trend for annual mean flow high but worse trends low flow. (PUR; regional holdout extrapolation highly data-sparse scenario), advantages became prominent. In addition, an untrained variable, evapotranspiration, retained good seasonality even extrapolated cases. The models' pipeline produced parameter fields maintain remarkably stable patterns data-scarce scenarios, which explains robustness. Combined interpretability assimilate multi-source observations, are strong candidates global-scale simulations climate impact assessment.

Language: Английский

Citations

57
Improving River Routing Using a Differentiable Muskingum‐Cunge Model and Physics‐Informed Machine Learning DOI Creative Commons
Tadd Bindas, Wen‐Ping Tsai, Jiangtao Liu

et al.

Water Resources Research, Journal Year: 2024, Volume and Issue: 60(1)

Published: Jan. 1, 2024

Abstract Recently, rainfall‐runoff simulations in small headwater basins have been improved by methodological advances such as deep neural networks (NNs) and hybrid physics‐NN models—particularly, a genre called differentiable modeling that intermingles NNs with physics to learn relationships between variables. However, hydrologic routing simulations, necessary for simulating floods stem rivers downstream of large heterogeneous basins, had not yet benefited from these it was unclear if the process could be via coupled NNs. We present novel method ( δ MC‐Juniata‐hydroDL2) mimics classical Muskingum‐Cunge model over river network but embeds an NN infer parameterizations Manning's roughness n ) channel geometries raw reach‐scale attributes like catchment areas sinuosity. The trained solely on hydrographs. Synthetic experiments show while geometry parameter unidentifiable, can identified moderate precision. With real‐world data, produced more accurate long‐term results both training gage untrained inner gages larger subbasins (>2,000 km 2 than either machine learning assuming homogeneity, or simply using sum runoff subbasins. parameterization short periods gave high performance other periods, despite significant errors inputs. learned pattern consistent literature expectations, demonstrating framework's potential knowledge discovery, absolute values vary depending periods. traditional models improve national‐scale flood simulations.

Language: Английский

Citations

27