Forecasting the River Ice Break-Up Date in the Upper Reaches of the Heilongjiang River Based on Machine Learning

Water, Journal Year: 2025, Volume and Issue: 17(3), P. 434 - 434

Published: Feb. 4, 2025

Ice-jam floods (IJFs) are a significant hydrological phenomenon in the upper reaches of Heilongjiang River, posing substantial threats to public safety and property. This study employed various feature selection techniques, including Pearson correlation coefficient (PCC), Grey Relational Analysis (GRA), mutual information (MI), stepwise regression (SR), identify key predictors river ice break-up dates. Based on this, we constructed machine learning models, Extreme Gradient Boosting (XGBoost), Backpropagation Neural Network (BPNN), Random Forest (RF), Support Vector Regression (SVR). The results indicate that reserves Oupu Heihe section have most impact date section. Additionally, accumulated temperature during period average before identified as features closely related river’s opening all four methods. choice method notably impacts performance models predicting Among tested, XGBoost with PCC-based achieved highest accuracy (RMSE = 2.074, MAE 1.571, R2 0.784, NSE 0.756, TSS 0.950). provides more accurate effective for dates, offering scientific basis preventing managing IJF disasters.

Language: Английский

---

Multivariate probabilistic prediction of dam displacement behaviour using extended Seq2Seq learning and adaptive kernel density estimation

Advanced Engineering Informatics, Journal Year: 2025, Volume and Issue: 65, P. 103343 - 103343

Published: April 18, 2025

Language: Английский

---

Spatio-Temporal Deformation Prediction of Large Landslides in the Three Gorges Reservoir Area Based on Time-Series Graph Convolutional Network Model

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(8), P. 4491 - 4491

Published: April 18, 2025

The displacement–time curve of a landslide serves as critical indicator its movement state, with precise deformation prediction being essential for effective disaster early warning. While numerous studies have employed machine learning techniques to predict at individual monitoring points, they often overlook the spatial correlations among points arranged along horizontal and vertical cross-sections. To address this limitation, paper employs Temporal Graph Convolutional Network (T-GCN) model, which integrates strengths Networks (GCNs) Gated Recurrent Units (GRUs). GCN captured while GRU modeled temporal dynamics displacement. T-GCN model was applied spatio-temporal Dawuchang in Three Gorges Reservoir area. Experimental results demonstrated that effectively predicted displacement landslides, offering robust approach warning systems. also incorporated influence external factors, such rainfall reservoir water levels, enhancing accuracy providing valuable insights future research forecasting.

Language: Английский

---

A Combined Landslide Displacement Prediction Model Based on Variational Mode Decomposition and Deep Learning Algorithms

Water, Journal Year: 2024, Volume and Issue: 16(23), P. 3503 - 3503

Published: Dec. 5, 2024

Accurate landslide displacement prediction is an essential prerequisite for early warning systems aimed at mitigating geological hazards. However, the inherent nonlinearity and dynamic complexity of evolution often hinder forecasting performance. Previous studies have frequently combined signal decomposition techniques with individual machine learning methods to enhance reliability. To address limitations uncertainties associated models, this study presents a hybrid framework that combines variational mode (VMD) multiple deep (DL) methods, including long short-term memory neural network (LSTM), gated recurrent unit (GRU), convolutional (CNN), using cloud model-based weighted strategy. Specifically, VMD decomposes cumulative data into trend, periodic, random components, thereby reducing non-stationarity raw data. Separate DL networks are trained predict each component, forecasts subsequently integrated through combination strategy optimally assigned weights. The proposed approach underwent thorough validation utilizing field monitoring from Baishuihe in Three Gorges Reservoir (TGR) region China. Experimental results demonstrate framework’s capacity effectively leverage strengths achieving RMSE, MAPE, R values 12.63 mm, 0.46%, 0.987 site ZG118, 20.50 0.52%, 0.990 XD01, respectively. This substantially enhances accuracy landslides exhibiting step-like behavior.

Language: Английский

---

Prediction of Floor Failure Depth Based on Dividing Deep and Shallow Mining for Risk Assessment of Mine Water Inrush

Water, Journal Year: 2024, Volume and Issue: 16(19), P. 2786 - 2786

Published: Sept. 30, 2024

Understanding and predicting floor failure depth is crucial for both mitigating mine water inrush hazards safeguarding groundwater resources. Mining activities can significantly disturb the geological strata, leading to shifts damage that may result in cracks. These disruptions extend confined aquifers, thereby increasing risk of inrushes. Such events not only pose a threat safety mining operations but also jeopardize sustainability surrounding systems. Therefore, accurately take effective coal seam management measures key reducing impact on Seventy-eight sets data China were collected, main controlling factors considered: (D1), working face inclination length (D2), (D3), thickness (D4). Firstly, distance evaluation function based Euclidean was constructed as clustering effectiveness index, optimal cluster number K = 3 determined. The collected clustered into three categories using K-means algorithm. It found results positively correlated with size D1, indicating D1 played dominant role clustering. dividing points types samples between 407.7~414.9 m 750~900 m. On this basis, grey correlation analysis method used analyze order influence weights depth. For first group, D2 > D3 D4, while, other two, it D4. emerged most influential factor, surpassing D2. 407.7 414.9 could be boundary, group classified shallow mining, second third groups deep mining. Based CatBoost prediction models parts model test set compared calculation empirical formula. exhibited superior accuracy lower mean squared error (MSE) absolute (MAE) higher R-squared (R2) This study helps enhance understanding behavior, guide management, protect resources by defining predict

Language: Английский

---