監測資料補遺技術開發與應用

本研究提出利用迴歸方程結合統計分析的方式進行監測資料補遺與除錯，針對連續的監測資料進行多元迴歸補遺，可以合理描述不同監測井間的時空關係，同時加入除錯的程序，進行統計值分析，剔除超過95%發生機率的離群值資料，讓補遺成果更為正確。本研究取礁溪溫泉2012～2017年的監測資料進行補遺，完整的資料量共計有683,904筆小時水位資料，但有129,490筆資料缺漏。結果顯示確實可針對缺漏的資料進行合理補遺，不僅符合長時間的趨勢，也呈現了短時間的擾動。進一步將補遺資料應用在礁溪溫泉區抽水型態的辨識，成功改善了辨識的結果，發現礁溪溫泉水位呈現一天一次的起伏震盪，主要用水行為發生在5:00～9:00、13:00～22:00。結果也顯示了如果直接使用監測資料，而沒有適當的補遺，分析所得可能失真。

關鍵字

資料補遺；多元迴歸；礁溪溫泉；抽水型態

並列摘要

This study proposes the use of regression analysis combined with statistical analysis for missing data interpolation and debug of monitoring data. Using continuous monitoring data in multiple regression can reasonably describe the temporal and spatial relationship between different monitoring wells, while applying the debug procedure can eliminate the outlier data with value out off the interval of 95% probability of occurrence. The proposed addendum technique makes the results more correct. In this study, the monitoring data of Jiaoxi Hot Spring from 2012 to 2017 is supplemented. The total amount of data is 683,904 counts of hourly groundwater level data, but there are 129,490 missing data. The addendum results show that it is possible to make reasonable additions to the missing data, which not only meets the long-term trend, but also presents short-term disturbances. The addendum data is used to the identification of the pumping type in Jiaoxi Hot Spring Area and successfully improve the identification results. It is found that the main hot spring use behavior of Jiaoxi occurs during 5:00~9:00 and 13:00~22:00, thus inducing a once-a-day fluctuation of water level. The results also show that if the monitoring data is used directly without an appropriate supplement, the analysis may be distorted.

並列關鍵字

Missing data interpolation ； Multiple regression ； Jiaoxi hot spring ； Pumping type

國際替代計量

監測資料補遺技術開發與應用

全文下載

主題瀏覽