基于歷史排位降雨閥值的粵港澳大灣區(qū)滑坡危險(xiǎn)性預(yù)警

打開文本圖片集

（1.SchoolofCivilEngeing，SnYatsenUniversityZuhai9o82，Cna;2.CenterfrWaterResourceandEnvirotu Yat-senUniversity，Guangzhou51O275，China）

Abstract：ThisstudyfocusedontheGuangdong-HongKong-MacaoGreaterBayAreaandconstructedagrid-basedlandslidehazard assessment model toenhanceregionaldisasterpreventionandmitigationcapabilities.Asemi-supervisedlearing methodwasused to optimizetheproporioalselectionoflandsldepointsandnn-landslidepoints toreducetheuncertaintyofsusceptibilitymodelingA historicalrankingainfallthreshold-basedmethodwasproposedtoclasifydailyrainfall，3-daycumulativerainfall，and7-day cumulativerainfalldata.Thespatialsusceptibilityoflandslidesandrainfall-inducedprobabilitywerequantitativelycoupledto establishadynamiclandslidehazardwarningsystem.Theresultsindicatethatwhena1.5-meterevaluationunitscaleisusedwithin theGuangdong-HongKong-MacaoGreaterBayArea，theoptialratiooflandslidepoints tonon-landslidepointsis1：4.Furthore， theareaundercurve（AUC）valueof tesusceptibiltymodelreachesashighasO.973.InpracticalapplicationduringJune2O18，the warning system accurately predicted 25 rainfall-induced landslide events，with 72% occurring in extremely high-risk warning zones and 28% inhigh-rsk warning zones，validatingthemodel'seffectiveness.Thissystemachieves fine-scalelandslidehazard warnings in the Greater Bay Area，providing scientific support for regional landslide risk management.

Keywords：semi-supervised machinelearning;non-lndslidesample;rainfallthreshold;landslidehazard;Guangdong-Hong Kong Macao Greater Bay Area

滑坡是斜坡巖土體沿著一定的軟弱面或軟弱帶順坡向下滑移的自然現(xiàn)象，不僅發(fā)生頻率高，造成的損失也極為嚴(yán)重。（剩余10481字）