GPS 船載海水面高測量

Abstract

精密單點定位(precise point positioning, PPP)技術由蒐集高精度軌道及時錶資訊進行動態定位解算，具可獨立彈性作業、不受基線長度影響、不需基準站資料等優點。船載測高無近岸雷達波波形不佳之問題，為一近岸海面高(sea surface heights, SSH)測量以及海洋重力測量之潛力技術。本研究嘗試應用PPP於船載測高動態定位作業，與動態差分定位(Kinematic relative positioning, KRP)成果比較，並以(1)靜態站定位測試、(2)海水面高交叉點差異值分析 (3) SSH與DNSC08海水面高程值比較，共三種方法進行PPP及KRP定位方法精度評估。研究資料以2007年五月至八月期間，「國家基本測量發展計畫」–台灣鄰近海域船載重力測量作業蒐集的GPS相位觀測資料(取樣間隔為1 秒)為實驗數據，研究區域分為基隆、小琉球以及綠島三區。研究工具於PPP部分，使用瑞士伯恩大學天文研究所研發之Bernese 5.0 GPS處理軟體及加拿大Waypoint研發之GRAFNAV GPS後處理軟體；於KRP部分，使用Bernese 5.0、GRAFNAV、以及MIT研發之GAMIT動態定位模組TRACK。

研究成果顯示，(1)於靜態站定位測試，PPP與KRP解算成果在超過百公里的長基線部分，精度量級接近，RMS為 6~8公分，PPP有替代長基線之潛力。(2)海水面高程交叉點差異值分析的均方根值，PPP部分，平差後Bernese PPP為2~16公分，GRAFNAV PPP為4~16公分，二者精度接近；KRP部分，Bernese KRP為11~37公分、GRAFNAV KRP為5~24公分；TRACK KRP為6~11公分；本研究半數地區的定位成果，反應出一般狀況下KRP較PPP解算精度為優，但在岸邊站資料品質不佳的地區，以PPP不受基站資料影響，解算較有優勢。(3)與DNSC08海水面高程值比較得出之標準偏差值，GRAFNAV 軟體於PPP與KRP二者標準偏差最小且一致，約10~60公分；Bernese PPP為38~62公分，Bernese KRP 為13~58公分，以交叉點差異值分析解算成果，顯示Bernese軟體經參數吸收可達一定的內部精度，其標準偏差與其他軟體無明顯差異；與DNSC08海水面高程值比較得其外部精度，其標準偏差偏高，流程尚有需要改善的地方。TRACK KRP為18~32公分，使用30秒一筆穩定接收的國家基準站解算為實驗對照組。

Sea surface heights (SSHs) from shipborne GPS measurements can potentially fill data gaps in coastal areas with bad altimeter observations caused by bad altimeter radar waveforms and poor geophysical corrections, and also be used for vertical datum connection. In this study, such ship-based SSHs are made possible using the 1-Hz GPS data collected in the campaigns of shipborne gravity from May to August, 2007. Precise point positioning (PPP), which uses zero differenced phase observables, is experimented for GPS positioning to determine SSHs, and its performance is compared with that of kinematic relative positioning (KRP). The accuracy of such GPS-derived SSHs is assessed by (1) static tests, (2) crossover difference analysis, (3) comparison with DNSC08 model. The computer packages Bernese and GRAFNAV are used for PPP, while Bernese, GRAFNAV , and TRACK are used for KRP.

The static tests at a long baseline case (>100km) shows that the RMS height variations from PPP and KRP are close, and are about 6-8cm. The RMS crossover differences of SSHs are 2-16 cm and 4-16 cm for Bernese PPP and GRAFNAV PPP, respectively. Such RMS values are 11-37cm and 5-24 cm for Bernese KRP and GRAFNAV KRP, respectively. A special case of TRACK KRP with 30-s GPS data yields a RMS value of differenced SSHs at 6-11 cm. About half of the research data sets agree with the positioning accuracy of KRP is better than PPP methods. However, PPP outperforms KRP especially in areas suffered from bad data quality of base stations, displaying the advantage of baseline independency. Comparisons between GPS-derived SSHs with DNSC08 SSHs lead to the following standard deviations (std.): 10-60 cm for both GRAFNAV PPP and KRP, which are the most consistent result of all softwares. With regard to Bernese, both Bernese PPP and KRP achieve good internal accuracy in crossover analysis by parameters estimation, but the external accuracy derived from the comparison with DNSC08 is worse than other softwares. 38-62cm for Bernese PPP and 13-58 cm for Bernese KRP demonstrate the feasibility of improving the process procedure. As a control group accomplished by using national base stations in 30 seconds data interval, the std. of TRACK KRP is 18-32 cm.