Enhanced water quality prediction model using advanced hybridized resampling alternating tree-based and deep learning algorithms

Environmental Science and Pollution Research, Год журнала: 2025, Номер unknown

Опубликована: Фев. 24, 2025

Язык: Английский

---

Enhancing flow rate prediction of the Chao Phraya River Basin using SWAT–LSTM model coupling

Journal of Hydrology Regional Studies, Год журнала: 2024, Номер 53, С. 101820 - 101820

Опубликована: Май 24, 2024

Chao Phraya River Basin—a major river with unique characteristics located in Thailand. This study sought to simulate the flow rates Basin, which is a tidal that poses challenges traditional modeling approaches. The soil and water assessment tool (SWAT) hydrological model extensively employed for simulating rates. However, limitations arise applying SWAT Basin due its nature, resulting an unsatisfactory performance. To address this, long short-term memory (LSTM) model, i.e., SWAT–LSTM was introduced complement model. collaborative coupling of information derived from LSTM notably enhanced improvement assessed using Nash-Sutcliffe efficiency (NSE), demonstrating increase 0.13 0.72. incorporation topographic static data also investigated provide basic basin results yielded NSE exceeding 0.79. shoreline level identified as crucial input feature indicating patterns. findings highlight effectiveness predicting rates, implying their applicability similar scenarios across different basins.

Язык: Английский

---

Deep Learning Empowered Water Quality Assessment: Leveraging IoT Sensor Data with LSTM Models and Interpretability Techniques

International Journal of Computational and Experimental Science and Engineering, Год журнала: 2024, Номер 10(4)

Опубликована: Окт. 20, 2024

Addressing the imperative demand for accurate water quality assessment, this paper delves into application of deep learning techniques, specifically leveraging IoT sensor datasets classification and prediction parameters. The utilization LSTM (Long Short-Term Memory) models navigates intricacies inherent in environmental data, emphasizing balance between model accuracy interpretability. This equilibrium is achieved through deployment interpretability methods such as LIME, SHAP, Anchor, LORE. Additionally, incorporation advanced parameter optimization techniques focuses on fine-tuning essential parameters like rates, batch sizes, epochs to optimize performance. comprehensive approach ensures not only precise predictions but also enhances transparency model, addressing critical need actionable information management. research significantly contributes convergence learning, IoT, science, offering valuable tools informed decision-making while highlighting importance optimal performance

Язык: Английский

---

Modeling freshwater plankton community dynamics with static and dynamic interactions using graph convolution embedded long short-term memory

Water Research, Год журнала: 2024, Номер 266, С. 122401 - 122401

Опубликована: Сен. 6, 2024

Язык: Английский

---

A state-of-the-art review of long short-term memory models with applications in hydrology and water resources

Applied Soft Computing, Год журнала: 2024, Номер unknown, С. 112352 - 112352

Опубликована: Окт. 1, 2024

Язык: Английский

---

HydroEcoLSTM: A Python package with graphical user interface for hydro-ecological modeling with long short-term memory neural network

Ecological Informatics, Год журнала: 2025, Номер 85, С. 102994 - 102994

Опубликована: Янв. 5, 2025

Язык: Английский

---

Spatiotemporal estimation of groundwater and surface water conditions by integrating deep learning and physics-based watershed models

Water Research X, Год журнала: 2024, Номер 23, С. 100228 - 100228

Опубликована: Май 1, 2024

The impacts of climate change on hydrology underscore the urgency understanding watershed hydrological patterns for sustainable water resource management. conventional physics-based fully distributed models are limited due to computational demands, particularly in case large-scale watersheds. Deep learning (DL) offers a promising solution handling large datasets and extracting intricate data relationships. Here, we propose DL modeling framework, incorporating convolutional neural networks (CNNs) efficiently replicate model outputs at high spatial resolution. goal was estimate groundwater head surface depth Sabgyo Stream Watershed, South Korea. consisted input variables, including elevation, land cover, soil type, evapotranspiration, rainfall, initial conditions. conditions target were obtained from HydroGeoSphere (HGS), whereas other inputs actual measurements field. By optimizing training sample size, design, CNN structure, hyperparameters, found that CNNs with residual architectures (ResNets) yielded superior performance. optimal reduces computation time by 45 times compared HGS monthly estimations over five years (RMSE 2.35 0.29 m water, respectively). In addition, our framework explored predictive capabilities responses future scenarios. Although proposed is cost-effective simulations, further enhancements needed improve accuracy long-term predictions. Ultimately, has potential facilitate decision-making, complex

Язык: Английский

---

Dissolved Oxygen Forecasting for Lake Erie’s Central Basin Using Hybrid Long Short-Term Memory and Gated Recurrent Unit Networks

Water, Год журнала: 2024, Номер 16(5), С. 707 - 707

Опубликована: Фев. 28, 2024

Dissolved oxygen (DO) concentration is a pivotal determinant of water quality in freshwater lake ecosystems. However, rapid population growth and discharge polluted wastewater, urban stormwater runoff, agricultural non-point source pollution runoff have triggered significant decline DO levels Lake Erie other lakes located populated temperate regions the globe. Over eleven million people rely on Erie, which has been adversely impacted by anthropogenic stressors resulting deficient concentrations near bottom Erie’s Central Basin for extended periods. In past, hybrid long short-term memory (LSTM) models successfully used time-series forecasting rivers ponds. prediction errors tend to grow significantly with period. Therefore, this research aimed improve accuracy taking advantage real-time (water temperature concentration) monitoring network establish temporal spatial links between adjacent stations. We developed LSTM that combine LSTM, convolutional neuron (CNN-LSTM), CNN gated recurrent unit (CNN-GRU) models, (ConvLSTM) forecast near-bottom Basin. These their capacity handle complicated datasets variability. can serve as accurate reliable tools help environmental protection agencies better access manage health these vital Following analysis 21-site dataset 2020 2021, ConvLSTM model emerged most reliable, boasting an MSE 0.51 mg/L, MAE 0.42 R-squared 0.95 over 12 h range. The foresees future hypoxia Erie. Notably, site 713 holds significance indicated outcomes derived from Shapley additive explanations (SHAP).

Язык: Английский

---

Determining water and solute permeability of reverse osmosis membrane using a data-driven machine learning pipeline

Journal of Water Process Engineering, Год журнала: 2024, Номер 64, С. 105634 - 105634

Опубликована: Июнь 15, 2024

Язык: Английский

---

Common issues of data science on the eco-environmental risks of emerging contaminants

Environment International, Год журнала: 2025, Номер 196, С. 109301 - 109301

Опубликована: Янв. 27, 2025

Язык: Английский

---