摘要
WhenlinearregressivemodelssuchasARorARMAmodelareusedforfittingandpredictingclimatictimeseries,resultsareoftennotsufficientlygoodbecausenonlinearvariationsinthetimeseries.Inthispaper,anonlinearself-excitingthresholdautoregressive(SETAR)modelisappliedtomodelingandpredictingthetimeseriesofflood/droughtrunsinBeijing,whichwerederivedfromthegradedhistoricalflood/droughtrecordsinthelast511years(1470—1980).TheresultsshowthatthemodelingandpredictingwiththeSETARmodelaremuchbetterthanthatoftheARmodel.Thelattercanpredicttheflood/droughtrunswithalengthonlylessthantwoyears,whiletheformalcanpredictmorethanthree-yearlengthruns.ThismaybeduetothefactthattheSETARmodelcanrenewthemodelaccordingtotherun-turningpointsintheprocessofpredic-tion,thoughthetimeseriesisnonstationary.
出版日期
1990年04月14日(中国期刊网平台首次上网日期,不代表论文的发表时间)