| 摘要: | 近年全球氣候快速變遷，伴隨著全球暖化，海平面的變化對於人類經濟活動以及身家財產的衝擊相當於劇烈天氣所帶來的災害，自二十世紀至今，研究指出全球海平面在時空上的特性具有短週期大尺度空間的變化，儘管上升速率與季節變化幅度以每年數個釐（毫）米間遊走，但在換算成總量變化則成為地表最大尺度的水氣（物質）重新分布。隨著時變重力場資料處理技術的增長以及時間序列的處理，本研究針對近年來（2003年至2010年）全球及區域海平面變化作綜合性的探討與研究，利用太空測地學的資料，計算季節週期、速率以及非季節變化的時空變化與相關地球物理訊號（如：全球海洋洋底壓力數據、Argo量測海溫剖面的混和層深度變化...等等）之間的關連性。 海平面變化主要由熱膨脹（SLVsteric） 以及陸地水體（SLVmass） 進入海洋所控制， 利用海洋雷達測高儀所得的全球海平面變化（1993−2010）， 上升速率約為+3.2mm/yr ；GRACE（2003−2010）所測得的上升速率約為+0.81mm/yr；Argo（2004−2010）則約為+0.25mm/yr。在年週期上，SLVmass 以及SLVtotal 顯示同一相位（最高振幅期間發生於九、十月）， 而SLVsteric 則相位相反（振幅最高發生在三月左右）， 此現象在其他文獻中都有提及。海平面變化在非季節項中與聖嬰現象的事件有明顯的相關，但是GRACE的資料在ENSO區域的訊號有較大誤差，導致SLVmass 在經驗正交經驗函數（EOF）的結果中空間分布情況沒有提取的相當完整。除了主要兩種自然的（熱膨脹與水體質量）貢獻於海平面變化之外，人類對於全球海平面的影響也相當大，藉由統計人類建造水庫將水體留置陸地上（自1950−2008年）推估出約有3 cm 等量（−0.55mm/yr）的海平面高度應該增長卻因為人類活動而未增高。這不但加劇了全球海平面變化貢獻量無法自洽的差距，也增加了對於目前已知兩項自然貢獻海平面變化的來源量的未確性。 本研究更利用赤道太平洋、印度洋的ENSO地區作為主要探討區域，擷取其海平面變化中非熱膨脹變化所扮演的比例。使用GRACE衛星計畫所得的時變重力場求得海洋中質量的變化，搭配多種最新的去相關以及濾波方法，利用一系列經驗正交函數（EOF）等工具，求得非季節性的聖嬰現象在海洋中的時空分布，但在赤道地區的較大的誤差雜訊的干擾下還是無法完整求得聖嬰東西向震盪分布訊號；此外，由於陸地上水文訊號遠大於海洋訊號，對於海陸交界有明顯的訊號洩漏，利用消除陸地洩漏訊號的工具輔助下，海洋內質量對應聖嬰現象的訊號提取仍效果有限，且訊號將隨著不同種經驗參數的濾波設置，而求得不同型態特徵分布。在另一方面，搭配海洋衛星測高儀以及Argo實地測量數據之比較後，明顯表示儘管有東西向震盪訊號，但是聖嬰現象中海平面變化質量貢獻量比熱膨脹效應小了一個數量級。 為求得聖嬰現象相關訊號與其他地球物理訊號作比對，除一般經驗正交函數EOF之外，CEOF、CCA、MSSA、Wavelet...等時間序列分析方法， 也一併使用擷取多變量、時空分布特徵訊號。CEOF為複數經驗正交函數，能擷取波動傳播訊號，針對聖嬰現象空間傳遞特性，能提供強度與相位的特徵，此方法可能作為另一個聖嬰指數的參考；CCA為典型相關分析，利用正交函數的經驗特性，求得最大時序相關性的特徵分布，呈現出聖嬰現象海洋內質量分布之特性，結果顯示印尼附近海域與其他赤道太平洋地區有震盪現象；MSSA為多變量特徵頻譜分析，用於多變量（不同地球物理量）中不同選定窗段下的相關特徵，找出聖嬰現象空間、時間特徵序列，其顯示與SOI、Nino3.4序列相符合；Wavelet小波分析顯示出時−頻分布狀態，顯示出最近這幾年中，此種兩年震盪普遍存在於各種資料變量中。 As recent global climate fast shifting, the impact of sea-level change to human socioeconomics is almost equivalent to the disaster which is stricken by severe weather accompanying with global warming. The variability of global sea-level change is characterized by small magnitude in short time span but large scale distribution from the last century. Even though the secular trend and seasonal strength are merely a few mm to cm, the quantity is an immense amount of hydrological redistribution which happens near Earth surface. By means of enriched post processes and longer time series of global time-variable gravity measurements, this study is dedicated to analyzing contemporary sea-level variation from global to regional scale in an integrative view. Using space geodetic satellites data, the secular trend, seasonal and non-seasonal terms are examined and intercompared to various geophysical signals, e.g. ocean bottom pressure from a global ocean model, temperature profiles (mixed layer depth) from Argo floats, etc. Sea-level variation (SLV) is dominated by steric (SLVsteric) and water mass discharge from land (SLVmass). Multi-oceanic altimeters reveal the global SLV rises at a rate of +3.2mm/yr (1993−2010); SLVmass, GRACE, is about +0.81mm/yr; SLVsteric, Argo, is about +0.25mm/yr (2004−2010). For the seasonal variations, the periodic time series peak of both SLVmass and SLVtotal are located around September to October, but SLVmass shows out of phase around March when the initial time set on January 1st which has been fully documented in recent studies. A significant and prevailing correlation to interannual ENSO events exists in global nonseasonal SLV including SLVsteric & SLVmass from conventional EOF analysis. Unfortunately, SLVmass from GRACE is highly contaminated by north-south stripes in low latitude regions, hence the SLVmass shows blurred results. In addition to two dominate contributors to SLV, anthropogenic contribution to GSL also suggest that the impoundments of reservoirs on land may reduce global sea-level change as large as −0.55mm/yr since 1950-2008. The equivalent −3 cm sea-level height has been stored on land. The latter not only disagree the global SLV budget but enlarges the discrepancy within the contributions from two natural causes we have known. This study therefore selects Tropical Pacific and Indian Ocean as the major study region. Extracting SLVmass accompanied with ENSO is to quantify the portions of non-steric induced SLV from GRACE, and collaborates with updated post processing techniques, e.g. de-correlations, various smoothing filters, etc., to retrieve time-variable gravity over oceans. Owing to a suite of empirical orthogonal functions (EOF) the ENSO-like interannual expansion coefficients are extracted in Tropical Pacific. However, the SLVmass from GRACE is less characterized as seesaw oscillating patterns due to large error in low latitude. For tropical regions, the hydrological effects from land is enormously larger than oceanic response, thus the leakage reduction is applied to eliminate coastal leakages. Unfortunately, either little ENSO signals in Tropical Pacific or strong error in low latitude, the compensated outputs remain obscure. Moreover, the eigen vectors, EOF patterns, shows filter-depend reactions when the empirical parameterized filters are chosen. On the other hand, after comparisons between oceanic altimeter data and in-situ Argo measurements, the SLVmass shows see-saw oscillation despite of an order less than SLVsteric which is correlated to ENSO events. As a sequence of extraction and comparison with relative ENSO signals, Complex EOF (CEOF), Canonical Correlation Analysis (CCA), Multivariate Signular Spectral Analysis (MSSA), Wavele, etc. time series analyses are employed. CEOF can fully reveal spatio-temporal propagating signals with magnitude and phase shifts. ENSO is a non-stationary wave in Tropical Pacific, and the pairs of strength and phase terms could be another ENSO index; CCA is dedicated to maximizing the correlation based on empirical orthogonal functions which adequately reveal the SLVmass patterns as breathing over oceans but oscillating against to SLVmass around Indonesia; MSSA shows multivariates of singular decomposition within empirical window length is selected where the index matches SOI, Nino3.4 well; wavelet provides time-frequency distribution which suggests a quasi-biennial component prevailing in various geophysical responses in oceans and indirect connection to QBO. |
