Coupling the remote sensing data-enhanced SWAT model with the bidirectional long short-term memory model to improve daily streamflow simulations

Journal of Hydrology, Год журнала: 2024, Номер 634, С. 131117 - 131117

Опубликована: Март 24, 2024

Global climate change has led to an increase in the frequency and scale of extreme weather events worldwide, there is urgent need develop better-performing hydrological models improve accuracy streamflow simulations facilitate water resource planning management. The Soil Water Assessment Tool (SWAT) a notable physical foundation widely used research. However, it uses simplified vegetation growth model, introducing uncertainty into simulation results. This study focused on improving model based remotely sensed phenological leaf area index (LAI) data. Phenological data were define dormancy, LAI replaced corresponding simulated by original model. approach improved describing dynamics. Then, enhanced SWAT was coupled with bidirectional long short-term memory (BiLSTM) validate processes upstream Hei River. During validation, performance simulating (R2 = 0.835, NSE 0.819) better than that 0.821, 0.805). In terms evapotranspiration, demonstrated even greater advantages. verification period, compared those R2 values for daily-scale increased from 0.196 −0.269 0.777 0.732, respectively. monthly-scale 0.782 0.678 0.906 0.851, Simultaneously, levels two coupling approaches prediction compared, i.e., direct BiLSTM (SWAT-BiLSTM) (enhanced SWAT-BiLSTM). results showed SWAT-BiLSTM always performed during entire especially which could more accurately predict peak changes. deep learning accuracy.

Язык: Английский

---

Training deep learning models with a multi-station approach and static aquifer attributes for groundwater level simulation: what is the best way to leverage regionalised information?

Hydrology and earth system sciences, Год журнала: 2025, Номер 29(4), С. 841 - 861

Опубликована: Фев. 18, 2025

Abstract. In this study, we use deep learning models with advanced variants of recurrent neural networks, specifically long short-term memory (LSTM), gated unit (GRU), and bidirectional LSTM (BiLSTM), to simulate large-scale groundwater level (GWL) fluctuations in northern France. We develop multi-station collective training for GWL simulations, using dynamic variables (i.e. climatic) static basin characteristics. This approach can incorporate features cover more reservoir heterogeneities the study area. Further, investigated performance relevant feature extraction techniques such as clustering wavelet transform decomposition simplify network regionalised information. Several modelling tests were conducted. Models trained on different types GWL, clustered based spectral properties, performed significantly better than whole dataset. Clustering-based reduces complexity data targets information efficiently. Applying without prior lead preferentially learn dominant behaviour, ignoring unique local variations. respect, pre-processing was found partially compensate clustering, bringing out common temporal characteristics shared by all available time series even when these are “hidden” (e.g. if their amplitude is too small). When employed along a technique improves model performances, particularly GWLs dominated low-frequency interannual decadal advances our understanding simulation learning, highlighting importance approaches, potential pre-processing, value incorporating attributes.

Язык: Английский

---

Assessing atmospheric influences for improving time-varying data-driven decadal predictions of Mediterranean spring discharge

Hydrological Sciences Journal, Год журнала: 2025, Номер unknown

Опубликована: Март 17, 2025

Язык: Английский

---

Large-scale seasonal forecasts of river discharge by coupling local and global datasets with a stacked neural network: Case for the Loire River system

The Science of The Total Environment, Год журнала: 2023, Номер 897, С. 165494 - 165494

Опубликована: Июль 13, 2023

Язык: Английский

---

Groundwater level reconstruction using long-term climate reanalysis data and deep neural networks

Journal of Hydrology Regional Studies, Год журнала: 2023, Номер 51, С. 101632 - 101632

Опубликована: Дек. 20, 2023

Northern Metropolitan France. Assessing long-term changes in groundwater is crucial for understanding the impacts of climate change on aquifers and managing water resources.However, level (GWL) records are often scarce, limiting historical trends variability. In this paper, we present a deep learning approach to reconstruct GWLs up several decades back time using recurrent-based neural networks with wavelet pre-processing reanalysis data as inputs. reconstructed two different datasets distinct spatial resolutions (ERA5: 0.25° x & ERA20C: 1° 1°) monthly resolution, performance simulations were evaluated. Long term GWL timeseries now available northern France, corresponding extended versions observational early 20th century. All three types piezometric behaviours could be reliably consistently capture multi-decadal variability even at coarser resolutions, which hydroclimatic cycles. GWLs'multidecadal was consistent Atlantic multidecadal oscillation. From synthetic experiment involving modified series, highlighted need longer training some low-frequency signals. Nevertheless, our study demonstrated potential DL models together extend observations improve interactions.

Язык: Английский

---

Identification of Groundwater Contamination Sources Based on a Deep Belief Neural Network

Water, Год журнала: 2024, Номер 16(17), С. 2449 - 2449

Опубликована: Авг. 29, 2024

Groundwater Contamination Source Identification (GCSI) is a crucial prerequisite for conducting comprehensive pollution risk assessments, formulating effective groundwater contamination control strategies, and devising remediation plans. In previous GCSI studies, various boundary conditions were typically assumed to be known variables. However, in many practical scenarios, these are exceedingly complex difficult accurately pre-determine. This practice of presuming as may significantly deviate from reality, leading errors identification results. Moreover, the outcomes influenced by multiple factors or conditions, including fundamental information about source polluted area. study primarily focuses on unknown conditions. Innovatively, three deep learning surrogate models, Deep Belief Neural Network (DBNN), Bidirectional Long Short-Term Memory Networks (BiLSTM), Residual (DRNN), employed validation simulate highly no-linear simulation model directly establish mapping relationship between outputs inputs model. approach enables direct acquisition inverse results variables based actual monitoring data, thereby facilitating rapid identification. Furthermore, account uncertainty noise inversion accuracy methods compared, method with higher selected analysis. Multiple experiments conducted, such tests, robustness cross-comparative ablation studies. The demonstrate that all models effectively complete research tasks, DBNN showing most exceptional performance experiments. achieved an R2 value 0.982, RMSE 3.77, MAE 7.56%. Subsequent analysis, robustness, further affirm adaptability tasks.

Язык: Английский

---

Training deep learning models with a multi-station approach and static aquifer attributes for groundwater level simulation: what’s the best way to leverage regionalised information?

Опубликована: Май 13, 2024

Abstract. In this study, we used deep learning models with recurrent structure neural networks to simulate large-scale groundwater level (GWL) fluctuations in northern France. We developed a multi-station collective training for GWL simulations, using both “dynamic” variables (i.e. climatic) and static aquifer characteristics. This approach offers the possibility of incorporating dynamic features cover more reservoir heterogeneities study area. Further, investigated performance relevant feature extraction techniques such as clustering wavelet transform decomposition, intending simplify network regionalised information. Several modelling tests were conducted. Models specifically trained on different types GWL, clustered based spectral properties data, performed significantly better than whole dataset. Clustering-based reduces complexity data targets information efficiently. Applying without prior can lead learn dominant station behavior preferentially, ignoring unique local variations. respect, pre-processing was found partially compensate clustering, bringing out common temporal characteristics shared by all available time series even when these are “hidden” because too small amplitude. When employed along thanks its capability capturing essential across scales (high low), decomposition technique provided significant improvement model performance, particularly GWLs dominated low-frequency advances our understanding simulation learning, highlighting importance approaches, potential preprocessing, value attributes.

Язык: Английский

---

Characteristics and Mechanism of Karst Groundwater Cycle Evolution Under Large-Scale Exploitation: A Case Study of the Yanxi Karst Groundwater System

Опубликована: Янв. 1, 2024

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

---

Characteristics and Mechanism of Karst Groundwater Cycle Evolution Under Large-Scale Exploitation: A Case Study of the Yanxi Karst Groundwater System

Опубликована: Янв. 1, 2024

Study region: The Yanxi karst groundwater system in northern China.Study focus: By analyzing long timeseries monitoring data of the level, quality, and withdrawal over past 30 years, this paper aims to evaluate regime characteristics fault block guide rational exploitation utilization groundwater. Using analysis, hydrogeochemical isotope evolution under large-scale conditions is analyzed.New hydrological insights for results reveal that before after exploitation, cycle changed fundamentally, groundwaterlevel continued fall below sea table a time, main discharged from lateral runoff centralized source field. spatial distribution quality closely related surface water coal measure strata. hydrochemical components are mainly controlled by dissolution minerals Ordovician limestone and, certain extent, silicate-rock minerals. mineral precipitation concentration caused evaporation relatively weak. It urgent take series management protection measures resources curb trend environment.

Язык: Английский

---

Comment on egusphere-2024-794

Опубликована: Июнь 11, 2024

Abstract. In this study, we used deep learning models with recurrent structure neural networks to simulate large-scale groundwater level (GWL) fluctuations in northern France. We developed a multi-station collective training for GWL simulations, using both “dynamic” variables (i.e. climatic) and static aquifer characteristics. This approach offers the possibility of incorporating dynamic features cover more reservoir heterogeneities study area. Further, investigated performance relevant feature extraction techniques such as clustering wavelet transform decomposition, intending simplify network regionalised information. Several modelling tests were conducted. Models specifically trained on different types GWL, clustered based spectral properties data, performed significantly better than whole dataset. Clustering-based reduces complexity data targets information efficiently. Applying without prior can lead learn dominant station behavior preferentially, ignoring unique local variations. respect, pre-processing was found partially compensate clustering, bringing out common temporal characteristics shared by all available time series even when these are “hidden” because too small amplitude. When employed along thanks its capability capturing essential across scales (high low), decomposition technique provided significant improvement model performance, particularly GWLs dominated low-frequency advances our understanding simulation learning, highlighting importance approaches, potential preprocessing, value attributes.

Язык: Английский

---