Reply on RC2

Опубликована: Июль 25, 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

---

A Systematic Review of Neural Network Applications for Groundwater Level Prediction

EarthArXiv (California Digital Library), Год журнала: 2024, Номер unknown

Опубликована: Июль 30, 2024

This systematic review investigates the application of neural networks (NNs) for groundwater level (GWL) prediction. The study employs Preferred Reporting Items Systematic Reviews and Meta-Analysis (PRISMA) technique to screen synthesize relevant data, focusing on input variables, data size, performance metrics. results indicate a growing preference hybrid models, which are effective in capturing hidden relationships between GWL environmental factors. root mean square error (RMSE) emerges as predominant metric, highlighting its significance evaluating NNs. incorporation lagged values is identified crucial enhancing predictive accuracy. In conclusion, this provides concise overview NN applications prediction, emphasizing efficacy models importance RMSE metric. findings contribute understanding trends research, addressing both technical nuances broader challenges.

Язык: Английский

---

A Systematic Review of Machine Learning Algorithms in Groundwater Level Simulations and Forecasting

EarthArXiv (California Digital Library), Год журнала: 2023, Номер unknown

Опубликована: Дек. 29, 2023

Over two billion individuals worldwide rely on subterranean water as their primary reservoir of clean water. Ensuring the sustainable management this heavily burdened resource necessitates a comprehensive quantitative evaluation groundwater reserves. This becomes even more critical resources face escalating demands resulting from socioeconomic growth, population expansion, and impacts climate change. research paper undertakes an extensive investigation in context special issue dedicated to utilization machine learning (ML) algorithms for modeling predicting levels (GWL). It offers concise overview prevalent Machine Learning(ML) techniques, encompassing general architecture, key hyper-parameters, methods fine-tuning, strategies optimal feature selection. Drawing insights scrutiny 170 papers across three prominent onlinedatabases, our findings indicate that well-constructed machine-learning models exhibit commendable capacity accurately levels. Based review we realized model GWLs is quite common. Typically, past are used input data, artificial neural networks (ANN) popular choice purpose. Our existing provides useful guide researchers interested applying algorithmsfor level forecasting. We also suggest new improve quality highlight areas future field.

Язык: Английский

---

Reply on RC1

Опубликована: Июнь 20, 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

---

Large-Scale Seasonal Forecasts of River Discharge by Coupling Local and Global Datasets with a Stacked Neural Network: Case for the Loire River System

Опубликована: Янв. 1, 2023

Accurate prediction of river discharge is critical for a wide range sectors, from human activities to environmental hazard management, especially in the face increasing demand water resources and climate change. To address this need, multivariate model that incorporates both local global data sources, including piezometer gauges, sea level, parameters. By employing phase shift analysis, optimizes correlations between target 12 parameters related hydrologic climatic systems, all sampled daily. In addition, stacked LSTM - more complex neural network architecture used improve information extraction ability.Exploring dynamics Loire-Bretagne basin its surroundings, investigation delves into predictions daily time steps one, three, six months ahead. The resulting forecast features high accuracy efficiency predicting fluctuations, showcasing superior performance forecasting drought periods over flood peaks. A detailed examination on highlights significance datasets discharge, where former dictates short-term predictions, while latter drives long-range forecasts. Seasonally extended confirms strong connection leading correlation, with lower correlation at lag 3 due seasonal changes affecting quality, compensated by higher longer 6 months. Such mutual effect multi-time-step improves predictive quality six-month horizon, thus encourages progress long-term scale. research establishes practical foundation effectively utilizing big leverage dynamics.

Язык: Английский

---