[Objective]Infrasound is an effective approach for debris flow warning.Traditional threshold-based warning methods focus solely on individual infrasound characteristics,which can lead to false alarms or missed de-tections.Thus,incorporating multiple time-frequency characteristics is essential to improve warning accuracy.[Methods]Infrasound data from Caodianshui Village,Beijing,were analyzed to differentiate the infrasound char-acteristics of debris flows from environmental background.A random forest algorithm was employed to establish an infrasound warning model for debris flows.[Results]The effective pressure of debris flow infrasound ranges from 0.4 to 1.0 Pa,while environmental infrasound typical remains below 0.1 Pa,though noise can raise it above 0.4 Pa.Noise infrasound energy is primarily concentrated below 6 Hz,whereas debris flow infrasound exhibits signifi-cantly higher energy in the 6-15 Hz.Therefore,comprehensive time-frequency characteristics,especially the ener-gy in the range of 6-15 Hz,should be considered when identifying debris flow infrasound.Using effective infra-sound pressure,infrasound pressure within 6-15 Hz,zero crossing rate,dominant frequency,and its amplitude as characteristic variables,a debris flow warning model was constructed based on a random forest algorithm.The mod-el achieved an AUC of 0.99,with a 90％recognition accuracy for test data,a 15％improvement over threshold methods.[Conclusion]The random forest-based infrasound warning model substantially improves warning accura-cy for debris flows and is applicable to typical basins in the Beijing mountainous areas.This approach offers a valu-able reference for infrasound-based debris flow warning research in other areas.
关键词
泥石流/次声/时频特征/预警模型/随机森林/北京山区
Key words
debris flow/infrasound/time-frequency characteristic/early warning model/random forest/Beijing mountainous area
维普
万方数据