A local patch regression-based generative model for urban flood prediction in data-poor areas

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 127489 - 127489

Published: April 1, 2025

Language: Английский

---

Long-term prediction of Poyang Lake water level by combining multi-scale isometric convolution network with quantile regression

Journal of Hydrology Regional Studies, Journal Year: 2025, Volume and Issue: 59, P. 102365 - 102365

Published: April 17, 2025

Language: Английский

---

Urban Flood Prediction Model Based on Transformer-LSTM-Sparrow Search Algorithm

Water, Journal Year: 2025, Volume and Issue: 17(9), P. 1404 - 1404

Published: May 7, 2025

Global climate change and accelerated urbanization have intensified extreme rainfall events, exacerbating urban flood risks. Although data-driven models shown potential in prediction, the ability of single to capture complex nonlinear relationships their sensitivity hyperparameters still limit prediction accuracy. To address these challenges, this study proposes an model by integrating Transformer, Long Short-Term Memory (LSTM), Sparrow Search Algorithm (SSA), combining Transformer’s global feature extraction with LSTM’s temporal modeling. The SSA was adopted optimize for Transformer-LSTM model. Dropout early stopping techniques were mitigate overfitting. Applied Zhengzhou city Henan province, China, achieves a Nash-Sutcliffe Efficiency (NSE) 0.971, indicating that proposed has high performance flooding. experimental results demonstrate Transformer-LSTM-SSA outperforms standalone LSTM, 12.9%, 10.1%, 2.9% NSE accuracy, respectively, while reducing MAE 62.12%, 56.9%, 34.21%, MAPE 21.69%, 22.2%, 10.89%, respectively. Furthermore, exhibits enhanced stability superior generalization capability. among comparative methods, thereby demonstrating model’s viability providing reliable solution real-time warning.

Language: Английский

---

Incorporating Recursive Feature Elimination and Decomposed Ensemble Modeling for Monthly Runoff Prediction

Water, Journal Year: 2024, Volume and Issue: 16(21), P. 3102 - 3102

Published: Oct. 29, 2024

Monthly runoff prediction is crucial for water resource allocation and flood prevention. Many existing methods use identical deep learning networks to understand monthly patterns, neglecting the importance of predictor selection. To enhance predictive accuracy reliability, this study proposes an RFECV–SSA–LSTM forecasting approach. It iteratively eliminates predictors derived from SSA decomposition PACF using recursive feature elimination cross-validation (RFECV) identify most relevant subset predicting target flow. LSTM modeling then used forecast flows 1–7 months into future. Furthermore, RFECV–SSA framework complements any machine-learning-based method. demonstrate method’s reliability effectiveness, its outputs are compared across three scenarios: direct LSTM, MIR–LSTM, RFECV–LSTM, historical data Yangxian Hanzhong hydrological stations in Hanjiang River Basin, China. The results show that RFECV–LSTM method more robust efficient than MIR–LSTM counterparts, with smallest number outliers NSE, NRMSE, PPTS under all scenarios. approach exhibits worst performance, indicating single-metric-based selection may eliminate valuable information. time–frequency superior, NSE values remaining stably around 0.95 value greater almost scenarios, outperforming other benchmark models. Therefore, effective highly nonlinear series, exhibiting high generalization ability.

Language: Английский

---

Research on a Non-Stationary Groundwater Level Prediction Model Based on VMD-iTransformer and Its Application in Sustainable Water Resource Management of Ecological Reserves

Sustainability, Journal Year: 2024, Volume and Issue: 16(21), P. 9185 - 9185

Published: Oct. 23, 2024

The precise forecasting of groundwater levels significantly influences plant growth and the sustainable management ecosystems. Nonetheless, non-stationary characteristics level data often hinder current deep learning algorithms from precisely capturing variations in levels. We used Variational Mode Decomposition (VMD) an enhanced Transformer model to address this issue. Our objective was develop a called VMD-iTransformer, which aims forecast level. This research nine monitoring stations located Hangjinqi Ecological Reserve Kubuqi Desert, China, as case studies over four months. To enhance predictive performance we introduced novel approach fluctuations Desert region. technique achieve predictions conditions. Compared with classic model, our more effectively captured non-stationarity prediction accuracy by 70% test set. novelty lies its initial decomposition multimodal signals using adaptive approach, followed reconfiguration conventional model’s structure (via self-attention inversion feed-forward neural network (FNN)) challenge multivariate time prediction. Through evaluation results, determined that method had mean absolute error (MAE) 0.0251, root square (RMSE) 0.0262, percentage (MAPE) 1.2811%, coefficient determination (R2) 0.9287. study validated VMD iTransformer offering modeling for predicting context, thereby aiding water resource ecological reserves. VMD-iTransformer enhances projections level, facilitating reasonable distribution resources long-term preservation ecosystems, providing technical assistance ecosystems’ vitality regional development.

Language: Английский

---

High-risk driving factors of rain-induced flooding hazard events on the Loess Plateau and its ecological subregions

Journal of Hydrology, Journal Year: 2024, Volume and Issue: unknown, P. 132475 - 132475

Published: Dec. 1, 2024

Language: Английский

---

Analyzing watershed system state through runoff complexity and driver interactions using multiscale entropy and deep learning

Ecological Indicators, Journal Year: 2024, Volume and Issue: 168, P. 112779 - 112779

Published: Oct. 30, 2024

Language: Английский

---

Innovative Methods Predicting the Remaining Useful Life of Transformer Using Limited Data

Published: Jan. 1, 2024

Language: Английский

---

Research on flood peak prediction in the three gorges region based on similarity search with multisource information fusion

Earth Science Informatics, Journal Year: 2024, Volume and Issue: 18(1)

Published: Dec. 18, 2024

Language: Английский

---

Machine Learning Classification Strategy to Improve Streamflow Estimates in Diverse River Basins in the Colorado River Basin

Earth and Space Science, Journal Year: 2024, Volume and Issue: 11(12)

Published: Dec. 1, 2024

Abstract Streamflow in the Colorado River Basin (CRB) is significantly altered by human activities including land use/cover alterations, reservoir operation, irrigation, and water exports. Climate also highly varied across CRB which contains snowpack‐dominated watersheds arid, precipitation‐dominated basins. Recently, machine learning methods have improved generalizability accuracy of streamflow models. Previous successes with LSTM modeling primarily focused on unimpacted basins, few studies included impacted systems either regional or single‐basin modeling. We demonstrate that diverse hydrological behavior river basins are too difficult to model a single, model. propose method delineate catchments into categories based level predictability, characteristics, influence. Lastly, we each category climate anthropogenic proxy data sets use feature importance assess whether performance improves additional relevant data. Overall, cover at low temporal resolution was not sufficient capture irregular patterns releases, demonstrating having high‐resolution release global scale. On other hand, classification approach reduced complexity has potential improve forecasts human‐altered regions.

Language: Английский

---