Prediction of peak discharge from earth–rock dam failures based on Bayesian optimization XGBoost regression algorithm

Accurately predicting the peak discharge at the breach of an earth–rock dam is crucial for formulating downstream flood control measures and risk assessments. This paper established a database of 162 earth–rock dam failure cases. Based on the analysis of earth–rock dam failure processes and paramete...

Full description

Saved in:
Bibliographic Details
Published in:Natural hazards (Dordrecht) Vol. 121; no. 12; pp. 13961 - 13984
Main Authors: Wang, Ting, Gao, Panrui, Tian, Zhiwen, Li, Yanlong, Qiu, Wen, Shi, Ning, Xu, Zengguang
Format: Journal Article
Language:English
Published: Dordrecht Springer Netherlands 01.07.2025
Springer Nature B.V
Subjects:
Accuracy
Algorithms
Bayesian analysis
Bayesian theory
Case studies
Civil Engineering
Correlation analysis
Dam failure
Dams
Datasets
Decision trees
Discharge
Earth
Earth and Environmental Science
Earth Sciences
Empirical models
Environmental Management
Failure
Failure modes
Failures
Flood control
Flood peak
Geophysics/Geodesy
Geotechnical Engineering & Applied Earth Sciences
Goodness of fit
Height
Hydrogeology
Impact analysis
Machine learning
Mathematical models
Natural Hazards
Neural networks
Optimization
Original Paper
Prediction models
Probability theory
Regression analysis
Risk assessment
Rock
Rocks
Database
SHAP
XGBoost
Peak discharge
Earth–rock dam
ISSN:0921-030X, 1573-0840
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurately predicting the peak discharge at the breach of an earth–rock dam is crucial for formulating downstream flood control measures and risk assessments. This paper established a database of 162 earth–rock dam failure cases. Based on the analysis of earth–rock dam failure processes and parameter correlation analysis, we selected dam type, failure mode, the height of water above breach invert, and volume of water stored above breach invert at failure as the control variables. A prediction model of the peak discharge from the earth–rock dam failures is established using the Bayesian optimization (BO) XGBoost algorithm. To show the superiority of the BO-XGBoost regression prediction model, we also selected three additional algorithms including Decision Tree, Random Forest, and Gradient Boosting Decision Tree to establish the prediction models. These models were compared against the BO-XGBoost model and empirical models. The results indicate that the BO-XGBoost model has a maximum goodness of fit R 2 of 0.98, indicating that the proposed model possesses higher predictive accuracy. To analyze the impact of each control variable on the prediction values, the BO-XGBoost model is combined with the interpretable machine learning Shapley Additive exPlanations theory. The global analysis indicates that the volume of water stored above the breach invert at failure has the most significant impact on the peak discharge, followed by the height of water above the breach invert at failure. Finally, the feasibility and application of the BO-XGBoost model are further verified by analyzing the typical cases of the Shimantan dam and the Teton dam.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0921-030X
1573-0840
DOI:10.1007/s11069-025-07338-5