Integrating principal component analysis, fuzzy inference systems, and advanced neural networks for enhanced estuarine water quality assessment

Journal of Hydrology Regional Studies, Journal Year: 2025, Volume and Issue: 57, P. 102182 - 102182

Published: Jan. 15, 2025

Language: Английский

---

Harnessing artificial neural networks for coastal erosion prediction: A systematic review

Marine Policy, Journal Year: 2025, Volume and Issue: 178, P. 106704 - 106704

Published: April 11, 2025

Language: Английский

---

Research on Method for Intelligent Recognition of Deep-Sea Biological Images Based on PSVG-YOLOv8n

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(4), P. 810 - 810

Published: April 18, 2025

Deep-sea biological detection is a pivotal technology for the exploration and conservation of marine resources. Nonetheless, inherent complexities deep-sea environment, scarcity available organism samples, significant refraction scattering effects underwater light collectively impose formidable challenges on current algorithms. To address these issues, we propose an advanced biometric identification framework based enhanced YOLOv8n architecture, termed PSVG-YOLOv8n. Specifically, our model integrates highly efficient Partial Spatial Attention module immediately preceding SPPF layer in backbone, thereby facilitating refined, localized feature extraction organisms. In neck network, Slim-Neck (GSconv + VoVGSCSP) incorporated to reduce parameter count size while simultaneously augmenting performance. Moreover, introduction squeeze–excitation residual (C2f_SENetV2), which leverages multi-branch fully connected layer, further bolsters network’s global representational capacity. Finally, improved head synergistically fuses all modules, yielding substantial enhancements overall accuracy. Experiments conducted dataset images acquired by Jiaolong manned submersible indicate that proposed PSVG-YOLOv8n achieved precision 79.9%, mAP50 67.2%, mAP50-95 50.9%. These performance metrics represent improvements 1.2%, 2.3%, 1.1%, respectively, over baseline model. The observed underscore effectiveness modifications addressing associated with detection, providing robust accurate identification.

Language: Английский

---

Interpretable machine learning for coastal wind prediction: Integrating SHAP analysis and seasonal trends

Journal of Coastal Conservation, Journal Year: 2025, Volume and Issue: 29(3)

Published: May 6, 2025

Language: Английский

---

Simulating Future Exposure to Coastal Urban Flooding Using a Neural Network–Markov Model

Journal of Marine Science and Engineering, Journal Year: 2024, Volume and Issue: 12(5), P. 800 - 800

Published: May 11, 2024

Urbanization and climate change are two major challenges of the 21st century, effects change, combined with urbanization coastal areas, increase frequency flooding area exposed to it, resulting in increased risk larger numbers people properties being vulnerable. An urban growth modeling system was used simulate future scenarios along coast Vendée region western France, potential exposure each scenario evaluated. The model an Artificial Neural Network a Markov Chain, using data obtained by remote sensing geographic information techniques predict three scenarios: business as usual, environmental protection, strategic planning. High-risk flood areas sea level projections from Sixth Assessment Report Intergovernmental Panel on Climate Change were then assess under study area. According results, different associated development patterns, planning significantly reduces compared other scenarios. However, rise considerably expands vulnerable flooding. Finally, methodology adopted can be prepare for impact develop strategies mitigate future.

Language: Английский

---

Early warning system for floods at estuarine areas: combining artificial intelligence with process-based models

Natural Hazards, Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 28, 2024

Abstract Floods are among the most common natural disasters, causing countless losses every year worldwide and demanding urgent measures to mitigate their impacts. This study proposes a novel combination of artificial intelligence process-based models construct flood early warning system (FEWS) for estuarine regions. Using streamflow rainfall data, deep learning model with long short-term memory layers was used forecast river discharge at fluvial boundary an estuary. Afterwards, hydrodynamic simulate water levels in The predictors were trained using different forecasting windows varying from 3 h 36 assess relationship between time window accuracy. insertion attention into network architecture evaluated enhance capacity. FEWS implemented Douro River Estuary, densely urbanised flood-prone area northern Portugal. results demonstrated that Estuary is reliable discharges up 5000 m /s, predictions made advance. For values higher than this, uncertainties increased; however, they still capable detecting occurrences.

Language: Английский

---

A Significant Wave Height Prediction Method Based on Improved Temporal Convolutional Network and Attention Mechanism

Electronics, Journal Year: 2024, Volume and Issue: 13(24), P. 4879 - 4879

Published: Dec. 11, 2024

Wave prediction is crucial for ensuring the safety and disaster mitigation of coastal areas, helping to support marine economic activities. Currently, many deep learning models, such as temporal convolutional network (TCN), have been applied wave prediction. In this study, a model based on improved TCN-Attention (ITCN-A) proposed. This incorporates improvements in two aspects. Firstly, address difficulty calibrating hyperparameters traditional TCN whale optimization algorithm (WOA) has introduced achieve global hyperparameters. Secondly, we integrate dynamic ReLU implement an adaptive activation function. The then combined with attention mechanism further enhance extraction long-term features height. We conducted experiments using data from three buoy stations varying water depths geographical locations, covering lead times ranging 1 h 24 h. results demonstrate that proposed integrated reduces RMSE by 12.1% MAE 18.6% compared long short-term memory (LSTM) model. Consequently, effectively improves accuracy height predictions at different stations, verifying its effectiveness general applicability.

Language: Английский

---

Estimation of the Input Height of Irregular Waves Generated at a Wave Maker by Using Gaussian Process Regression and Artificial Neural Networks

Journal of Korean Society of Coastal and Ocean Engineers, Journal Year: 2024, Volume and Issue: 36(6), P. 225 - 232

Published: Dec. 31, 2024

A neural network (NN) and Gaussian process regression (GPR) model for predicting the input wave height of generator were established, their performance was evaluated compared using irregular data acquired in a two-dimensional flume. Both models able to predict that can produce target waves with very high accuracy. Among two models, GPR showed better than NN model. Based on results this study, it is expected reduce time required experiments by shortening trial error setting when conducting physical

Language: Английский

---