T4-A2：捕捉丛林火灾蔓延预测中的不确定性.pdf

1、T4-A2:Capturing uncertainty in bushfire spread prediction Natural Hazards Research Australia 2025Michael StoreyAssociate ResearcherUniversity of WollongongGoals A Bayesian probabilistic bushfire rate of spread model for operational use.Tools and approaches to generate and communicate probabilistic p

2、redictions.A rate of spread database and processing packageCompletion Early 2027Team University of Wollongong:Michael Storey,Michael Bedward and Owen PriceContact mstoreyuow.edu.auCapturing uncertainty in bushfire spread predictionWhy a Bayesian model for rate of spread?Deterministic PredictionWe us

3、ually predict spread based on a few variables.But bushfire spread complex:Topography variationWind gustsExtreme fire behavioursSpotting interactions Atmospheric interactionWe need a way to represent this uncertainty in ROS predictionsBased on all the examples of fire spread we have from line scans,a

4、nd given these conditions,there is a 49%chance that fire ROS will be faster 3 kmh.After one hour,the fire will most likely end up somewhere between the black lines.If the fire has extreme behaviour,there is a small chance it could spread as far as the black dotted line.Bayesian prediction examples B

5、ayesianBayesian prediction exampleFire progression dataWork to dateCollating data from agencies and previous work(Thanks RFS,CFA,DEECA,DFES)Collating data from previous work at UOWMapping fire progressionsBuilding prototype processing packageBuilding prototype databaseApproach to mapping progression

6、s and measuring ROSFire progression dataWork to dateCollating data from agencies and previous work(Thanks CFA,DEECA,RFS,DFES)Collating data from previous work at UOW.Mapping fire progressionsBuilding prototype processing packageBuilding prototype database20 min 5 hour timesteps.Processing packagePro