Forpredictionoftheextremesignificantwaveheightintheoceanareaswherelongtermwavedataarenotavailable,theempiricalmethodofextrapolatingshorttermdata(1～3years)isusedindesignpractice.Inthispapertwomethodsareproposedtopredictextremesignificantwaveheightbasedonshort-termdailymaxima.AccordingtothedaarecordedbytheOceanographicStationofLiaodongBayattheBohaiSea,itissupposedthatdailymaximumwaveheightsarestatisticallyindependent.Thedatashowthatdailymaximumwaveheightsobeylog-normaldistribution,andthatthenumbersofdailymaximavaryfromyeartoyear,obeyingbinomialdistribution.Basedonthesestatisticalcharacteristics,thebinomial-log-normalcompoundextremumdistributionisderivedforpredictionofextremesignificantwaveheights(50～100years).Forexaminationofitsaccuracyandvalidity,thepredictionofextremewaveheightsisbasedon12years′dataatthisstation,andbasedoneach3years′datarespectively.Theresultsshowthatwithconsiderationofconfidenceintervals,thepredictedwaveheightsbasedon3years′dataareveryclosetothosebasedon12years′data.TheobserveddatainsomeoceanareasintheAtlanticOceanandtheNorthSeashowitisnotcorrecttoassumethatdailymaximumwaveheightsarestatisticallyindependent;theyaresubjecttoMarkovchaincondition,obeyinglog-normaldistribution.Inthispaperananalyticalmethodisderivedtopredictextremewaveheightsinthesecases.AcomparisonofthecomputationsshowsthatthedifferencebetweentheextremewaveheightsbasedontheassumptionthatdailymaximaarestatisticallyindependentandthattheyaresubjecttoMarkovChainconditionissmallerthan10%.
