臺灣博碩士論文加值系統

臺灣於1885年開始進行有系統之氣象觀測，隨著時代更迭以及科技的進步，雨量站的精度也越來越高。雨量站資料不僅在網路上開放供民眾參考，國內也有許多防災研究係利用雨量站資料作為雨量來源，但是雨量計、風速計等等實地觀測儀器的設置，容易受到當地的電力、通訊、成本等因素受到限制，因此站與站之間普遍有相當的距離。而利用遙測技術，藉由全台雷達測站之雷達回波，進行氣象水文整合，提供即時性天氣監測資訊的QPESUMS，即可讓沒有雨量站的地區做累積雨量的參考。
本研究以台灣東北角之台二線為研究區域，而鄰近本路段之雨量站僅有瑞芳及鼻頭角雨量站，若僅以此二雨量站所蒐集之雨量資料對應整個區域，則略有稍嫌不足，比起雨量站的一對大範圍區域，QPESUMS在空間上一對一的優勢更能針對災害區域反饋較為準確的雨量資料。
本研究蒐集108筆災害事件紀錄，針對災害發生地點進行分群，最終分得WEST、MIDDLE、EAST三個區域，後找出致災區域之中心點，並以中心點向外取7 x 7個網格範圍作為與量控制區塊，針對這49個網格之雨量資料分析雨量對災害發生之影響，觀察災害發生前後降雨情形可以發現：(1)多數QPESUMS網格於海面上多數雨量觀測值為0，故後續進行分析時僅取陸地上網格雨量觀測值為主，並且因雨量值為多筆，故選擇三種雨量計算指標進行後續研究，(2)所有災害事件發生前都有降雨的情形，判斷災害發生原因與長時間的降雨有關。
接著以三天累積降雨量以及一週累積降雨設立降雨門檻值，並計算雨量紀錄年份內超過降雨門檻值的降雨事件次數，再使用Poisson分布推估未來再次發生同樣規模降雨事件的機率，並計算發生超過門檻值降雨事件時可能發生災害事件的機率，將兩個機率相乘即可得到未來發生超過降雨門檻值降雨事件同時也發生災害事件的機率。
成果顯示，在相同降雨門檻值下，WEST區域以及MIDDLE區域發生超過降雨門檻值的降雨事件次數明顯高於EAST區域，但是EAST區域發生災害事件的機率卻遠高於兩區，故應於EAST區域加強災前防治，另外在參考災害事件發生機率的同時加入平均超過降雨門檻雨量事件的次數，利用計算災害風險值找出在門檻值改變時風險最高的雨量值，並以其作為雨量警戒值的設定依據，最後，根據降雨型態的不同，可以使用不同降雨延時作為雨量因子進行災害發生機率的推估。

Taiwan began to carry out systematic meteorological observations in 1885. With the changes of the ages and the advancement of science and technology, the accuracy of rainfall stations has become better and better. The data from rainfall stations are not only available for public reference on the internet, but are also used by many disaster prevention studies in Taiwan. However, the settings of on-site observation instruments such as rain gauges and anemometers are easily restricted by factors such as local power, communication, and cost. Therefore, there is usually a considerable distance between stations. QPESUMS, which uses radar echoes from all radar stations in Taiwan, integrates meteorology and hydrology, and provides real-time weather monitoring information, can be used as a reference for accumulated rainfall in areas without rainfall stations.
This study takes the second line of Taiwan in the northeast corner of Taiwan as the research area, and the adjacent rainfall stations are only Ruifang and Bitoujiao rainfall stations. If only the rainfall data collected by these two rainfall stations correspond to the whole area, it is slightly insufficient. Compared with the large area of one-to-one rainfall stations, the one-to-one advantage of QPESUMS in space can provide more accurate rainfall data for disaster areas.
This study collected 108 disaster event records, found the center point of the disaster area, and selected 7 x 7 grid areas from the center point as the quantity control blocks. The rainfall data from 49 grids were used to analyze the impact of rainfall on the disaster occurrence. By observing the pre-disaster and post-disaster rainfall, it can be found that there is rainfall before all disaster events, and it can be judged that the cause of the disaster is related to long-term rainfall.
Then, set the rainfall threshold according to the cumulative rainfall in 72 hours and the cumulative rainfall in one week, and calculate the number of rainfall events exceeding the rainfall threshold in the rainfall record year. Use the Poisson distribution to estimate the probability of another rainfall event of the same size in the future, and calculate the probability of occurrence. The probability that a disaster event may occur when the rainfall event exceeds the threshold. And multiply these two probabilities to get the probability that the rainfall event exceeds the rainfall threshold and the probability of future disaster events.
The results showed that under the same rainfall threshold, the number of rainfall events exceeding the rainfall threshold in the WEST and MIDDLE regions was significantly higher than in the EAST region. However, the probability of disaster events in the EAST region was much higher than in the two regions, so the EAST region should strengthen pre-disaster prevention and control.

目 錄

摘 要................i
Abstract................iii
誌 謝................v
目 錄................vi
表目錄................viii
圖目錄................x
第一章 緒論................1
1.1研究動機與目的................1
1.2 文獻回顧................3
1.2.1 降雨門檻之界定................3
1.2.2 邊坡災害模型................4
1.3 研究架構與流程................6
第二章 暴雨頻率分析................8
2.1 雷達降雨網格資料................8
2.2 研究區域概況................9
2.3 雷達降雨資料與雨量站之差異.........11
2.3.1 雨量站分析結果................11
2.3.2 雷達降雨資料分析結果與差異......14
第三章 災害機率模型................18
3.1 災害區域規劃................18
3.2 災害事件紀錄................24
3.3 災害機率分析................34
3.3.1 Poisson分布................34
3.3.2 條件機率................35
第四章 雨量警戒值探討................36
4.1 雨量事件挑選................36
4.2 致災機率預測................45
4.3 雨量警戒值探討................57
第五章 結論與建議................83
5.1 結論...............................83
5.2 後續研究方向及建議................84
參考文獻................................86
附 錄................................87
附錄1　災害事件災前降雨情形................87
附錄2　災害事件災前三天累積降雨紀錄................102
附錄3　災害事件災前七天累積降雨紀錄................114

[1]魏倫瑋、李璟芳、黃韋凱、紀宗吉、林錫宏、陳宏宇、林銘朗，2013。台2線68K八斗子落石災害致災成因初探，社團法人中華水土保持學會102年度年會。
[2]葉彥駒，2018。結合多年期崩塌目錄及降雨因子建立崩塌機率模型，國立中興大學水土保持學系碩士學位論文。
[3]盧瑞興，張炎銘，林秀娟，2003。「台灣地區雨量測站降雨強度-延時Horner公式分析」，經濟部水利署。
[4]聯合大地工程顧問股份有限公司，2020。「台2線地質敏感路段(67K~89K)易致災邊坡測繪、調查、評估工作暨落石告警系統及防護工法」，交通部公路總局第一區養護工程處。
[5]陳振宇、劉維則、許家祥，2017。使用QPESUMS雨量資料建立崩塌災害預警模式，中華水土保持學報，48(1):44-55。
[6]陳振宇、陳均維、陳國威、林詠喬，2019。坡地降雨致災熱區警戒模式，中華水土保持學報，50(1):1-10。
[7]吳罕夏，2020。山區道路因降雨及地震致災崩塌因素之探討-以中橫公路(台8線)為例，國立宜蘭大學土木工程學系碩士學位論文。
[8]Robert A. Crovelli，2000。Probability Models for Estimation of Number and Costs of Landslides，Open-File Report。
[9]D. Hantz, J. M. Vengeon, C. Dussauge-peisser，2003，An historical, geomechanical and probabilistic approach to rock-fall hazard assessment，Natural Hazards and Earth System Sciences，(2003)693-701。
[10]M. Krautblatter and M. Moser，2009，A nonlinear model coupling rockfall and rainfall intensity basedon a four year measurement in a high Alpine rock wall (Reintal,German Alps)，Natural Hazards and Earth System Sciences，(2009)1425-1432。