简介：Mesoscaleensembleisanencouragingtechnologyforimprovingtheaccuracyofheavyrainfallpredictions.Occurrencesofheavyrainfallarecloselyrelatedtoconvectiveinstabilityandtopography.Inmid-latitudes,perturbedinitialfieldsformedium-rangeweatherforecastsareoftenconfiguredtofocusonthebaroclinicinstabilityratherthantheconvectiveinstability.Thus,alternativeapproachestogenerateinitialperturbationsneedtobedevelopedtoaccommodatetheuncertaintyoftheconvectiveinstability.Inthispaper,aninitialconditionperturbationapproachtomesoscaleheavyrainfallensembleprediction,namedasDifferentPhysicsModeMethod(DPMM),ispresentedindetail.BasedonthePSU/NCARmesoscalemodelMM5,anensemblepredictionexperimentonatypicalheavyrainfalleventinSouthChinaiscarriedoutbyusingtheDPMM,andthestructureoftheinitialconditionperturbationisanalyzed.Further,theDPMMensemblepredictioniscomparedwithamulti-physicsensembleprediction,andtheresultsshowthattheinitialperturbationfieldsfromtheDPMMhaveareasonablemesoscaiecirculationstructureandcouldreflectthepredictionuncertaintyinthesensitiveregionsofconvectiveinstability.AnevaluationoftheDPMMinitialconditionperturbationindicatesthattheDPMMmethodproducesbetterensemblemembersthanthemulti-physicsperturbationmethod,andcansignificantlyimprovetheprecipitationforecastthanthecontrolnon-ensemblerun.