| For a long time,typhoon storm surges have been one of the main marine disasters faced by China’s southeastern coastal areas.Against the background of accelerating global climate change trends,extreme weather events are becoming increasingly frequent,the risk of storm surge disasters caused by typhoons is expanding in the influence of natural-social-ecological system,which posing enormous natural and social ecological security and development pressures to coastal areas.Therefore,conducting research on storm surge disaster risk in coastal areas and carrying out risk assessment work is one of the important ways to reduce the impact of storm surge disasters.Based on an in-depth analysis of the current research status and results of domestic and foreign storm surge disaster risk assessments,this article constructs a method for storm surge disaster risk assessment based on machine learning algorithms.In light of the existing shortcomings in current storm surge disaster assessment work,such as high numerical simulation costs,difficulties in data collection,lack of generality in index analysis and integration between hazard and vulnerability assessments,this research provides a complete rapid,comprehensive,and scientific potential storm surge risk area identification method in areas with relatively scarce hydro-meteorological data,effectively improving the evaluation efficiency,expanding the methods and means of storm surge disaster risk assessment.The main research achievements of this study are as follows:(1)Based on the ADCIRC model,this article obtained the maximum value of storm surge simulation results under nine typhoon scenarios,and taking typhoon "Mangkhut" as an example,the accuracy of the maximum value of storm surge simulation results was verified.In the Verification station,the maximum value of simulated results caused by Typhoon "Mangkhut" was 121 cm,and the maximum relative error between the simulated and measured value was 8.03%.The simulated results are reliable.Based on the maximum value of storm surge simulation results under different typhoon scenarios and the DEM elevation data,the theoretical inundation range of Dongshan Bay was delineated under nine typhoon scenarios.Secondly,based on the In VEST model,this article conducted a vulnerability assessment of the coastal zone of Dongshan Bay under storm surge disaster scenarios,according to the results of vulnerability assessment,the vulnerability index ranged from 1.92 to 4.73.Using the natural breakpoint method,areas with vulnerability index greater than 3.73 were identified as potential fragile areas.Finally,based on geographic information technology,this article spatially overlaid the theoretical storm surge disaster inundation range with the potential vulnerable areas of the coastal zone to obtain the potential storm surge disaster risk area of the coastal zone of Dongshan Bay under different typhoon scenarios.(2)Based on the identified potential storm surge disaster risk areas of the coastal zone of Dongshan Bay under nine typhoon scenarios,this study established a dataset which includes 45,000 samples for identifying potential storm surge disaster risk areas.Based on the random forest algorithm,the decision rules for determining the risk areas of storm surge disasters were learned and classified through model training.A random forest generalization model was constructed to automatically classify and identify potential disaster risk areas of storm surge in the bays and coastlines.The accuracy of the classification model was 98.2%,with a precision of84.1% and a recall rate of 85.9%.(3)This study applied the constructed random forest generalized model to evaluate the storm surge disaster risk of the coastal zone in the Luoyuan Bay of Fujian Province,identified the potential storm surge disaster risk area and completed the construction of a machine learning-based storm surge disaster risk assessment model.Overall,this article broadens the risk assessment methods for storm surge disasters and constructs a storm surge disaster risk assessment model based on machine learning algorithms,provides important references for coastal areas to carry out storm surge disaster assessment and theoretical support,for scientific utilization and ecological protection-restoration of coastal zone. |
