臺灣博碩士論文加值系統

台灣位處於太平洋西側間熱帶輻合區，地理位置約為北緯22°至25°、東經120°至122°之間，於每年夏秋兩季之間頻受颱風的侵襲，且受到副熱帶高壓外圍環流影響之下，易導致熱帶氣旋的產生並帶來強風與強雨，其中強雨更是造成淹水、山崩、土石流等災害。本研究目的為發展出可以預測未來時間降雨量模式的方法，掌握未來降雨量資料達到預測功能，讓研究地區能夠及時做好因應措施，降低對生命財產威脅及利於戶外活動規劃等的安排。
在本次研究中開發了一種即時的颱風侵台期間降雨預測模型，用於即時預報未來1、3、6小時內降雨狀況。本研究利用Marshall-Palmer Z-R Relationship和人工智慧法的監督式機器學習演算法，主要包含Linear Regression、Support Vectror Regression(SVR)、Extreme Gradient Boosting(XGB)等。由於雷達回波強度對於颱風降雨雨量估算有高度的相關性，因此本研究模式輸入資料主要有雷達回波因子和地面氣象觀測值作為模型輸入變量之一。本研究設計了兩個演算階段以實現降雨預測，第一階段為降雨反演模式的建立，本階段為分別對各雷達回波仰角進行建模，以便了解逐時段最適合的雷達回波仰角；在第二階段中，本研究進行降雨預測模式建立，本階段再利用第一階段之最適雷達回波仰角資訊作為預測模式建模之基礎，再將之結合地面氣象資訊以建立降雨預測模式。
本研究之研究區域為台灣台東地區及其外島蘭嶼。本研究蒐集了2008-2017年的14場颱風事件，這些颱風事件為中央氣象局有發布颱風警報單中路徑穿過或者對研究區域有影響之颱風事件。本研究模型輸入資料包括中央氣象局成功、台東、大武、蘭嶼四個逐時氣象測站資料和中央氣象局花蓮氣象雷達回波逐時資料。研究結果顯示由地表氣象測站資料、雷達回波資料及其最適仰角所構成資料集，對於未來時刻降雨狀態的預測起到最佳模式輸入因子的作用。為了驗證實驗模型的實用性，實驗過程使用2016年尼伯特、莫蘭蒂和梅姬三場颱風進行了模擬測試。在颱風侵襲台灣期間由實驗模型所產生的預測結果達到了良好的預測性能，有效的預測了1、3、6小時的降雨結果。

Taiwan is located at the tropical convergence zone on the west side of the Pacific Ocean. Its geographical position is about 22° to 25° north latitude and 120° to 122° east longitude. It is frequently affected by typhoons between summer and autumn, and is affected by the circulation of the subtropical high. Underneath, it is easy to cause extreme weather and bring strong winds and strong rains. Rainfall is also causing floods, landslides, earth and rock flows and other disasters. The purpose of this study is to develop a method for predicting the future rainfall pattern, to grasp the future rainfall data to achieve the prediction function, so that the research area can timely respond to the measures, reduce the threat to life and property and facilitate the planning of outdoor activities.
In this study, an instantaneous rainfall prediction model during typhoon was developed to predict the rainfall in the future 1, 3, and 6 hours. This study uses the machine learning algorithm of artificial intelligence method, and the algorithm mainly includes Marshall-Palmer Z-R Relationship, linear regression method, regression support vector machine, XGBoost. Since radar echo intensity has a high correlation with typhoon rainfall estimation, the input data of this research model mainly includes radar echo factor and ground meteorological observation value as one of the model input variables. Two calculation stages are designed to realize rainfall prediction. The first stage is the establishment of rainfall inversion mode. In this stage, the elevation angles of each radar echo are modeled separately to understand the most suitable radar echo elevation angle. In the second stage, this study carried out the establishment of rainfall prediction mode. At this stage, the optimal radar echo elevation information of the first stage was used as the basis of prediction mode modeling, and then combined with ground meteorological information to establish a rainfall prediction mode.
The research area is Taitung, Taiwan and its outer island, Lanyu. The study collected 14 typhoon events from 2008 to 2017. These typhoon events were reported by the Central Meteorological Administration to issue typhoon warnings in the typhoon warning list or impact on the study area. The input data of this research model includes time-of-day data of meteorological stations and hourly data of Hualien meteorological radar echoes. Results show that the data set composed of surface meteorological station data and high-altitude radar intensity data and its optimum elevation angle play the role of the best mode input factor for the prediction of future rainfall state. In order to verify the practicability of the experimental model, the experimental process was simulated using three typhoons, Nibert, Morandi and Meggie in 2016. The prediction results produced by the experimental model during typhoon achieved good prediction performance, effectively predicting the rainfall results of 1, 3, and 6 hours.

目錄
摘要 I
Abstract II
目錄 III
圖目錄 V
表目錄 VIII
第一章 前言 1
1.1 研究動機 1
1.2 研究目的 1
1.3 論文架構 2
第二章 文獻回顧 3
第三章 研究方法 6
3.1 研究流程 6
3.2 研究模型 9
3.2.1 Marshall-Palmer Z-R Relationship 9
3.2.2線性回歸(Linear Regression) 9
3.2.3回歸支援向量機 (SVR) 9
3.2.4 Extreme Gradient Boosting (XGBoost) 9
3.3 研究工具 10
第四章 資料來源與處理 11
4.1 颱風路徑與場次篩選 11
4.2 中央氣象局局屬測站逐時觀測資料處理 13
4.3 花蓮雷達回波資料處理 16
4.4 資料集彙整 18
第五章 降雨反演模式建立與分析 19
5-1 案例設計 19
5-2 參數檢定 20
5-3 成功站模式結果 22
5-4 台東站模式結果 24
5-5 大武站模式結果 26
5-6 蘭嶼站模式結果 28
5-7 綜合評估 30
第六章 降雨預測模式建立與分析 32
6-1 最佳延遲時間決定 32
6-2 案例設計 33
6-3 案例結果 33
第七章 模擬與分析 39
7.1 NEPARTAK尼伯特颱風 40
7.1.1 預測結果 40
7.1.2 評估指標 44
7.2 MERANTI莫蘭蒂颱風 46
7.2.1 預測結果 46
7.2.2 評估指標 50
7.3 MEGI梅姬颱風 52
7.3.1 預測結果 52
7.3.2 評估指標 56
7.4 綜合評估 58
7.4.1 所有颱風的指標 58
7.4.2 雨量分級評估 62
7.4.3 比較中央氣象局預報結果 76
第八章 結論與建議 79
8-1 結論 79
8-2 建議 80
參考文獻 81

 

參考文獻
1. Asklany, S. A., Elhelow, I. K., and El-wahab, M. A. (2011). Rainfall events prediction using rule-based fuzzy inference system. Atmos.Res, 101, 228–236.
2. Borga, M., Anagnostou, E. N., and Frank, E. (2000). On the use of real-time radar rainfall estimates for flood prediction in mountainous basins. J. Geophys. Res, 105, 2269–2280, doi:10.1029/1999JD900270.
3. Chen, K. S., Wang, J. T., and Mitnik, L. M. (2001). Satellite and ground observations of the evolution of Typhoon Herb near Taiwan. Remote Sens. Environ, 75, 397–411.
4. Chiang, Y. M., Chang, F. J., Jou, B. J. D., and Lin, P. F. (2007). Dynamic ANN for precipitation estimation and forecasting from radar observations. J. Hydrol, 334, 250–261.
5. Chen, T., and Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System, KDD 16, August 13-17, 2016, San Francisco, CA, USA.
6. Diop, C. A., Sauvageot, H., and Mesnard, F. (2013). Partitioning the distribution function of radar reflectivity in convective storms using maximum likelihood method. Atmos. Res, 124, 123–136.
7. Gabella, M., Joss, J., Perona, G., and Galli, G. (2001). Accuracy of rainfall estimates by two radars in the same Alpine environment using gage adjustment. J. Geophys. Res, 106(D6), 5139–5150, doi:10.1029/2000JD900487.
8. Georgakakos, K. P. (2000). Covariance propagation and updating in the context of real-time radar data assimilation by quantitative precipitation forecast models. J. Hydrol, 239, 115–129.
9. Gerstner, E. M., and Heinemann, G. (2008). Real-time areal precipitation determination from radar by means of statistical objective analysis. J. Hydrol, 352, 296–308.
10. Ho, C. H., and Lin, C. J. (2012). Large-scale Linear Support Vector Regression. Department of Computer Science National Taiwan University.
11. Kidd, C., Kniveton, D. R., Todd, M. C., and Bellerby, T. J. (2003). Satellite rainfall estimation using combined passive microwave and infrared algorithms. J. Hydrometeorol, 4, 1088–1104.
12. Li, P. W., and Lai, E. S. T. (2004). Short-range quantitative precipitation forecasting in Hong Kong. J. Hydrol, 288, 189–209.
13. Marshall, J. S., Langille, R. C., and Palmer, W. M. K. (1947). Measurement of rainfall by radar. J. Meteor, 5, P165.
14. Marshall, J. S., and Palmer, W. M. (1948). The distribution of raindrops with size. Journal of Meteorology, 5, 165–166.
15. Seo, D. J., and Breidenbach, J. P. (2002). Real-time correction of spatially nonuniform bias in radar rainfall data using rain gauge measurements. Journal of Hydrometeorology, National Weather Service, 1325 East–West Highway, Silver Spring, MD 20910-3283.
16. Sauvageot, H. (1992). Radar Meteorology, 366 pp., Artech House, Inc., Boston, Ma.
17. Taylor, R. (1990). Interpretation of the correlation coefficient: A basic review. J. Diagn, Med, Sonography, 1, 35–39.
18. Wei, C. C. (2014). Simulation of operational typhoon rainfall nowcasting using radar reflectivity combined with meteorological data. J. Geophys. Res. Atmos, 119, 6578–6595.
19. 黃威雄(2000)，應用類神經網路於颱風期間雷達降雨模擬之研究。國立臺灣大學土木工程，碩士論文
20. 楊政潭(2003)，雷達回波應用於颱風降雨空間分佈與總量之研究—以納莉颱風為例。國立中央大學，碩士論文。
21. 謝章廷(2007)，應用雷達降雨於分布型水文模式與不確定性分析。國立成功大學，碩士論文。
22. 林國峰，吳爵廷和廖信華等人(2010)，以新型人工智慧技術建立具整合颱風資訊能力之雨量預報模式。科技部專題研究計畫。
23. 尤心瑜(2011)，使用都卜勒氣象雷達資料改善模式定量降雨預報之可行性研究－以模擬資料測試之實驗結果。大氣科學期刊，第三十九期第一號。
24. 陳莉(2014)，人工智慧模式改進雷達回波預測颱風降雨量準確度之研究。中華大學土木工程學系，研究計畫。
25. 吳家佑(2015)，雷達回波圖做短期降雨預測。國立屏東大學，碩士論文。
26. 中央氣象局氣象衛星中心(2015)，https://www.cwb.gov.tw/V7/service/notice/dow
nload/publish_20150608145912.pdf