海洋通报
海洋通報
해양통보
MARINE SCIENCE BULLETIN
2014年
4期
383-390
,共8页
连丽珍%王解先%沈云中%王卫安
連麗珍%王解先%瀋雲中%王衛安
련려진%왕해선%침운중%왕위안
相对海平面变化%平均海平面%对比分析
相對海平麵變化%平均海平麵%對比分析
상대해평면변화%평균해평면%대비분석
relative sea-level change%mean sea level%comparison analysis
采用多种方法和3种模型分析了东方、海口、北海、闸坡、香港、汕尾、厦门、坎门、吕四、那霸、名濑、连云港、石臼所、大连共14个验潮站22年逐时海平面序列的相对海平面变化,主要包括趋势与周期项提取和未来月均海平面预测两大方面,比较了各种方法分析所得结果的相似性与差异性。结果表明:各验潮站的海平面均在逐渐上升,上升速率在1~3 mm/a之间,平均值为2.3 mm/a;各站的周期项也不一样,但基本都包含周年和半年项。发现一些方法组合时可能会出现异常情况,例如模型趋势项选择不同的形式,可能会导致所求速率出现正负截然相反的情形,不同的周期寻找方法所确定的周期也存在一定差异。各种方法的预测效果相差不大,预测残差基本都在依2 dm以内,基于奇异谱分析的均生函数的长期预测效果要明显优于带周期项的灰色模型和传统模型,但短期预测效果相差不大。
採用多種方法和3種模型分析瞭東方、海口、北海、閘坡、香港、汕尾、廈門、坎門、呂四、那霸、名瀨、連雲港、石臼所、大連共14箇驗潮站22年逐時海平麵序列的相對海平麵變化,主要包括趨勢與週期項提取和未來月均海平麵預測兩大方麵,比較瞭各種方法分析所得結果的相似性與差異性。結果錶明:各驗潮站的海平麵均在逐漸上升,上升速率在1~3 mm/a之間,平均值為2.3 mm/a;各站的週期項也不一樣,但基本都包含週年和半年項。髮現一些方法組閤時可能會齣現異常情況,例如模型趨勢項選擇不同的形式,可能會導緻所求速率齣現正負截然相反的情形,不同的週期尋找方法所確定的週期也存在一定差異。各種方法的預測效果相差不大,預測殘差基本都在依2 dm以內,基于奇異譜分析的均生函數的長期預測效果要明顯優于帶週期項的灰色模型和傳統模型,但短期預測效果相差不大。
채용다충방법화3충모형분석료동방、해구、북해、갑파、향항、산미、하문、감문、려사、나패、명뢰、련운항、석구소、대련공14개험조참22년축시해평면서렬적상대해평면변화,주요포괄추세여주기항제취화미래월균해평면예측량대방면,비교료각충방법분석소득결과적상사성여차이성。결과표명:각험조참적해평면균재축점상승,상승속솔재1~3 mm/a지간,평균치위2.3 mm/a;각참적주기항야불일양,단기본도포함주년화반년항。발현일사방법조합시가능회출현이상정황,례여모형추세항선택불동적형식,가능회도치소구속솔출현정부절연상반적정형,불동적주기심조방법소학정적주기야존재일정차이。각충방법적예측효과상차불대,예측잔차기본도재의2 dm이내,기우기이보분석적균생함수적장기예측효과요명현우우대주기항적회색모형화전통모형,단단기예측효과상차불대。
Using several methods and three models to analyze the relative sea-level change with 22 years of hourly water level sequence of 14 tide gauge stations that are Dongfang, Haikou, Beihai, Zhapo, Hong Kong, Shanwei, Xiamen, Kanmen, Lusi, Kaohsiung, Keelung, Lianyungang, Shijiusuo and Dalian, trend and cycle extraction and future sea level forecast are major aspects in this paper. In addition, the similarity and difference of analysis results by various methods is also one of the main research contents. The results showed that:sea level at each tide gauge station gradually increased, and the rates of rise were between 1 mm/a and 3 mm/a, and the average was about 2.3 mm/a;cycle of each station was not the same, but it basically contained annual and semiannual data. Through the analysis, it was found that some abnormal cases occurred by the combination of some methods, and for example, choosing different forms of the trend model, might induce the positive or negative calculated rates that are poles apart;moreover, there were some differences of the cycles determined by different periodicity analysis methods. Prediction effect by various methods differed little, and the prediction residuals were generally within ± 2 dm. Long-term prediction effect by mean generating function model based on singular spectrum analysis was significantly better than those by the gray model and the traditional model, but their short-term prediction effect was similar.