台灣坐落於西北太平洋颱風的主要路徑上，一年內台灣受到颱風影響數次，颱風除了帶來豐沛的水資源，也可能引發嚴重的災害。台灣的山地連綿起伏，造成洪水在颱風期間移動地更快，對水庫或河川下游地帶的衝擊大。水庫在台灣是非常有效的防洪設施，然而在防災期間，水庫操作面臨最大問題是颱風降雨預報的不準確性。台灣的颱風降雨預報之所以不容易預測準確，是因為預報員對颱風的降雨機制仍理解不足；除了颱風結構複雜，另外颱風也受到地形因素影響，其降雨機制不易被歸納或推測。
了解颱風的降雨情況，有助於水資源與防災系統的管理與規劃。目前已經發展的颱風降雨預報模式有幾種，包括了數值模式 (Numerical Weather Prediction, NWP)，劇烈天氣監測系統 (Quantitative Precipitation Estimation and Segregation Using Multiple Sensors, QPESUMS)，系集模式颱風定量降水預報(Ensemble Typhoon Quantitative Precipitation Forecast, ETQPF)。過去的學者多在研究颱風內部的降雨分布，或者颱風於大面積的降雨情形，像是颱風為整個台灣所帶來的降雨量，卻鮮少有研究在討論集水區的颱風降雨情形。另外，雖然已經有研究利用類神經網路 (Artificial Neural Networks, ANNs) 分析與預測一般的降雨，但是颱風降雨的分析與預報依舊多使用數值方法或統計方法。
本論文利用前饋式、非監督式、競爭式的類神經網路——自組特徵映射類神經網路 (Self-organizing map, SOM) 來分析颱風期間石門水庫集水區的降雨時空分布。研究發現路徑相似的颱風，其降雨的空間變化也會相似。因此，颱風降雨的空間分布與颱風路徑有高度相關性；倘若有兩場路徑不同之颱風中心位於同一個經緯度，其降雨空間分布不一定相同。
由於颱風路徑會影響集水區降雨之時空分布，相似路徑之颱風對同一集水區降雨之時空分布影響相似，所形成雨型也較為相似。因此，本研究使用特徵雨量組體圖 (Feature hydrographs of rainfall, FHR) 來描述路徑類似之颱風所帶來的雨型。每場颱風歷經時間不同、強度不同，造成降雨強度與總雨量亦不相同，故須進行正規化，即將歷經時間換成 0~1 之間的數值，將時雨量轉換成百分比，再將該神經元內所有正規化雨量歷程進行平均，即為特徵雨量組體圖。研究結果顯示，特徵雨量組體圖若呈中央集中型，且峰值較高與較寬，其颱風降雨的破壞性較高。反之，分布相對平均的特徵雨量組體圖，一般表示颱風遠離石門水庫集水區，颱風對水庫的影響較低。一旦中央氣象局發布颱風預報，本研究可根據預報的颱風路徑，利用特徵雨量組體圖來估算該場颱風的雨量組體圖，提供防災機構與水庫操作單位非常有利的資訊。
本研究分析在歷史颱風場次中ETQPF於石門水庫的降雨預報誤差；若颱風距離石門水庫較近，ETQPF的準確度較高。相反的，對於距離石門水庫較遠的颱風，其ETQPF的預報誤差高，這可能是因為颱風降雨機制的不確定性大，發生於石門水庫的降雨主要受到颱風外圍環流的影響，而非颱風主要結構。為了改善颱風降雨預報，本研究篩選出表現優良的系集預報成員，并利用加權平均法重新計算颱風預報降雨量。系集預報成員的權重是比較颱風的預報路徑與實際路徑之決定系數(R2)，倘若某成員的颱風路徑預報越準確，其權重值越高。本研究發現選擇可靠度高之系集預報成員，ETQPF之降雨預報改善率可達最高90%。

Typhoons hit Taiwan several times every year, which could cause serious flood disasters. Because mountainous terrains and steep landforms can rapidly accelerate the speed of flood during typhoon events, rivers cannot be a stable source of water supply. Reservoirs become the most effective floodwater storage facilities for alleviating flood damages in Taiwan. Forecasting typhoon rainfall is a long-standing and challenging issue due to the complexity of typhoon.
This study focused on typhoon rainfall including the spatial distribution of rainfall, the temporal distribution of rainfall (hyetographs) and the total rainfall. For the spatial distribution of rainfall, self-organizing map (SOM) is implemented to explore the spatial distribution of typhoon rainfall. The spatial distribution of rainfall has similar change processes for the similar tracks of the typhoons. Rainfall spatial distribution is highly related to the classification of the typhoon track; if two typhoons have different track, their rainfall spatial distributions would be probably different, even they are in the same location. The feature hyetographs of rainfall (FHR) in the same classification of typhoon tracks are constructed. For the classification of typhoon tracks with great impact on the Shihmen watershed, the rainfall hyetographs are central distributed with a higher and wider peak that reveals the destructivity of the typhoon rainfall. In contrast, the relatively flat distribution of rainfall hyetograph means the typhoon is far away from the watershed and less harmful. FHR can be used to estimate the rainfall hyetographs based on the forecast typhoon tracks, this is highly valuable to the emergency response agencies and reservoir operation units. The forecast error of ETQPF in the Shihmen watershed is investigated for improving the total rainfall forecast. If the locations of the typhoon are near the watershed, the ETQPF has higher accuracy. On the other hand, typhoons distancing from the watershed, usually have higher uncertainty in the factors of typhoon rainfall because of the effect of the typhoon outer circulation flow, instead of the typhoon main structure, resulting in lower accurate ETQPF forecast. This study proposes the weighted average method to recalculate the typhoon rainfall by selecting QPFs from well-performed ensemble members. The more reliable the ensemble members are, the higher the values their weights are. The weights of the ensemble members, based on the determination coefficient between the forecast typhoon tracks of ensemble members and the actual tracks, are given to recalculate the total rainfall. When highly reliable ensemble members are chosen, the improvement rate of rainfall prediction could be more than 90% compared to the ETQPF.

謝誌			I
中文摘要		II
Abstract		V
Table of Contents	VII
List of Figures	IX
List of Tables	XI


Chapter 1	Introduction	1
1.1	Background	1
1.2	Study objectives	3

Chapter 2	Literature review	4
2.1	Application of artificial neural network	4
2.2	Estimating and forecasting rainfall using statistical methods	6
2.3	Estimating and forecasting rainfall using satellite images	8
2.4	SOM Clustering of typhoon track	10

Chapter 3	Methodology	12
3.1	Self-organizing map	12
3.1.1	Architecture of SOM	13
3.1.2	Algorithm of SOM	14

Chapter 4	Case study	18
4.1	Study Area	18
4.2	Data Collection	19
4.2.1	Rainfall data from rain gauges	19
4.2.2	QPE data of QPESUMS	20
4.2.3	QPF data of TTFRI’s EPS	21
4.2.4	ETQPF data	22
4.3	Investigating the spatial distribution of typhoon rainfall	27
4.3.1	SOM configuration for the rainfall ratio	27
4.3.2	Exploring the spatial distribution of rainfall along the typhoon path	32
4.3.3	Comparing the rainfall ratios at the intersection of different typhoon paths	32
4.4	Estimating the rainfall patterns during typhoon events	33
4.5	Evaluating the forecast error of ETQPF	34
4.6	Predicting rainfall by selecting well-performed ensemble members	34

Chapter 5	Results and discussions	40
5.1	SOM clustering of typhoon paths	40
5.2	The spatial distribution of typhoon rainfall	42
5.2.1	The changes in rainfall ratio sequence	42
5.2.2	The comparison of the rainfall ratio between typhoons with same location of centers	46
5.3	The rainfall patterns during typhoon events	49
5.4	The forecast error of ETQPF	51
5.5	Predicting rainfall by selecting well performed ensemble members	54

Chapter 6	Conclusions and Suggestions	59
6.1	Conclusions	59
6.2	Suggestions	61

References		62
Appendix A	The results of the pattern changes of rainfall ratios	67
Appendix B	The results of the rainfall forecast by selecting well-performed ensemble members		76



Figure 2.1	CWB typhoon tracks categorisation between 1911 and 2015.  (Central Weather Bureau, 2016)	10
Figure 2.2 	SOM topological map of typhoon track.  (Tamkang University Information Center for Water Environment, 2018)	11
Figure 3.1	Architecture of Self-Organizing Map (SOM).	13
Figure 3.2	Schematic diagram of a SOM topological neighbourhood,  showing a monotonic decrease in the neighbourhood.	15
Figure 3.3	Flow chart of SOM algorithm.	17
Figure 4.1	Location of the Shihmen watershed and 10 rainfall gauging stations.	18
Figure 4.2	The grid for vectorising typhoon paths for Shihmen watershed.  (Tamkang University Information Center for Water Environment, 2018)	19
Figure 4.3	The grid points of QPESUMS and ETQPF of Shihmen watershed.	20
Figure 4.4	Three nested domains used in TTFRI-EPS. (Wu et al., 2018)	23
Figure 4.5	ETQPF Generation by selecting the ensemble cases, which are the numerical weather predictions (black dots) along a chosen typhoon path (dashed red line) that fall within a distance (pink circle). (Central Weather Bureau, 2018)	24
Figure 4.6	SOM topological map of the hourly rainfall ratio (10 × 10).  Black arrows indicate the changes in rainfall ratio along the track of Typhoon Soudelor.			29
Figure 4.7	SOM topological map of the hourly rainfall ratio (10 × 10).  Black arrows indicate the changes in rainfall ratio along the track of Typhoon Dujuan.			30
Figure 4.8	The selected typhoon paths with filtered ensemble cases in ETQPF (A) and  the 48-hour accumulated rainfall forecasts (B) of the Typhoon Matmo  from 00 UTC on July 22nd through 00 UTC on July 24th in 2014. (1), (2) and (3) are the forecasts of the left track, official track, and right track, respectively. (Central Weather Bureau, 2018)	35
Figure 4.9	The best track of Typhoon Matmo (Central Weather Bureau)	36
Figure 4.10	The predicted and observed rainfall amounts for Typhoon Matmo.  (Central Weather Bureau)	36
Figure 5.1	SOM topological map of the hourly rainfall ratio of Shihmen watershed.			42
Figure 5.2  Changes in rainfall ratio along typhoon tracks of 3rd neuron’s typhoon track.	43
Figure 5.3	The path of Typhoon Soudelor (ID: 201513) with labelled rainfall ratio.	44
Figure 5.4	The path of Typhoon Dujuan (ID: 201521) with labelled rainfall ratio.	45
Figure 5.5	The path of Typhoon Megi (ID: 201617) with labelled rainfall ratio.	45
Figure 5.6	The rainfall ratio at the intersection of Typhoons Fung-Wong and Trami			47
Figure 5.7 	The rainfall ratios at the intersection of Typhoons Morakot and Megi	47
Figure 5.8 	The rainfall ratios at the intersection of Typhoons Sepat and Soudelor	48
Figure 5.9	The feature hyetograph of rainfall of every neuron in  SOM clustering of typhoon path.	49
Figure 5.10	Scatter diagram of observed rainfall and ETQPF forecast rainfall.	53
Figure 5.11	Plot of the absolute forecast error against the ETQPF forecast rainfall.	53



Table 4.1	Selected typhoon events and collected data.	25
Table 4.2	QPE data of QPESUMS separation for SOM model’s training and testing.			31
Table 4.3	Similar typhoons whose pattern changes of the rainfall ratio are compared.		32
Table 5.1	SOM Clustering of typhoon paths.  (Tamkang University Information Center for Water Environment, 2018)	40
Table 5.2	Comparison of observed rainfalls and ETQPF forecast rainfalls.	52
Table 5.3	Comparison of the performance of ensemble members’ typhoon track forecast for training events. (Neuron 3 of SOM clustering of typhoon path)		56
Table 5.4	QPFs forecasted by the ensemble members and corresponding weights calculated from R2 in Table 5.3. (Neuron 3 of SOM clustering of typhoon path)	57
Table 5.5	Comparison of three groups of total QPFs forecasted by the ensemble members, total observed rainfall and ETQPF’s total rainfall prediction. (Neuron 3 of SOM clustering of typhoon path)	58

1.	Burpee, R.W., and Black, M.L. (1989). Temporal and Spatial Variations of Rainfall Near the Centers of Two Tropical Cyclones. Monthly Weather Review 117, 2204-2218.
2.	Central Weather Bureau (2016). Typhoon Knowledge. Retrieved from https://www.cwb.gov.tw/V7e/knowledge/encyclopedia/ty017.htm
3.	Central Weather Bureau (2018). Meteorological Education and Training. Retrieved from https://metet.cwb.gov.tw/cwb/
4.	Chang, C.P., Yeh, T.C., and Chen, J.M. (1993). Effects of Terrain on the Surface Structure of Typhoons over Taiwan. Monthly Weather Review 121, 734-752.
5.	Chang, F.-J., Chang, K.-Y., and Chang, L.-C. (2008). Counterpropagation fuzzy-neural network for city flood control system. Journal of Hydrology 358, 24-34.
6.	Chang, F.-J., Chiang, Y.-M., Tsai, M.-J., Shieh, M.-C., Hsu, K.-L., and Sorooshian, S. (2014a). Watershed rainfall forecasting using neuro-fuzzy networks with the assimilation of multi-sensor information. Journal of Hydrology 508, 374-384.
7.	Chang, L.C., Shen, H.-Y., and Chang, F.-J. (2014b). Regional flood inundation nowcast using hybrid SOM and dynamic neural networks. Journal of Hydrology 519, 476-489.
8.	Cheung, K.K.W., Huang, L.R., and Lee, C.S. (2008). Characteristics of rainfall during tropical cyclone periods in Taiwan. Nat Hazards Earth Syst Sci 8, 1463-1474.
9.	Elsner, J.B., Kossin, J.P., and Jagger, T.H. (2008). The increasing intensity of the strongest tropical cyclones. Nature 455, 92.
10.	French, M.N., Krajewski, W.F., and Cuykendall, R.R. (1992). Rainfall forecasting in space and time using a neural network. Journal of hydrology 137, 1-31.
11.	Glahn, H.R., and Lowry, D.A. (1972). The Use of Model Output Statistics (MOS) in Objective Weather Forecasting. Journal of Applied Meteorology 11, 1203-1211.
12.	Griffith, C.G., Woodley, W.L., Grube, P.G., Martin, D.W., Stout, J., and Sikdar, D.N. (1978). Rain Estimation from Geosynchronous Satellite Imagery—Visible and Infrared Studies. Monthly Weather Review 106, 1153-1171.
13.	Habib, E., Henschke, A., and Adler, R.F. (2009). Evaluation of TMPA satellite-based research and real-time rainfall estimates during six tropical-related heavy rainfall events over Louisiana, USA. Atmospheric Research 94, 373-388.
14.	Hong, J.-S., Fong, C.-T., Hsiao, L.-F., Yu, Y.-C., and Tzeng, C.-Y. (2014). Ensemble Typhoon Quantitative Precipitation Forecasts Model in Taiwan. Weather and Forecasting 30, 217-237.
15.	Hsiao, M.-C. (2012). Investigation on Watershed Inflow Forecast during Typhoon Period Using Artificial Neural Network. Master Thesis in Department of Water Resources and Environmental Engineering (Tamkang University)
16.	Hung, N.Q., Babel, M.S., Weesakul, S., and Tripathi, N.K. (2009). An artificial neural network model for rainfall forecasting in Bangkok, Thailand. Hydrol Earth Syst Sci 13, 1413-1425.
17.	Kang, N.-Y., and Elsner, J.B. (2015). Climate Mechanism for Stronger Typhoons in a Warmer World. Journal of Climate 29, 1051-1057.
18.	Klein, W.H. (1971). Computer Prediction of Precipitation Probability in the United States. Journal of Applied Meteorology 10, 903-915.
19.	Klein, W.H., and Glahn, H.R. (1974). forecasting local weather by means of model output statistics. Bulletin of the American Meteorological Society 55, 1217-1227.
20.	Kuligowski, R.J., and Barros, A.P. (1998). Localized Precipitation Forecasts from a Numerical Weather Prediction Model Using Artificial Neural Networks. Weather and Forecasting 13, 1194-1204.
21.	Lee, C.S., Huang, L.-r., Shen, H.-s., and Wang, S.-t. (2006). A Climatology Model for Forecasting Typhoon Rainfall in Taiwan. Natural Hazards 37, 87-105.
22.	Lin, G.-F., and Chen, L.-H. (2005). Application of an artificial neural network to typhoon rainfall forecasting. Hydrological Processes 19, 1825-1837.
23.	Lin, G.F., and Wu, M.-C. (2009). A hybrid neural network model for typhoon-rainfall forecasting. Journal of Hydrology 375, 450-458.
24.	Liu, C.-C., Liu, G.-R., Lin, T.-H., and Chao, C.-C. (2010). Accumulated Rainfall Forecast of Typhoon Morakot (2009) in Taiwan Using Satellite Data. Journal of the Meteorological Society of Japan Ser II 88, 785-798.
25.	Lonfat, M., Marks, F.D., and Chen, S.S. (2004). Precipitation Distribution in Tropical Cyclones Using the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager: A Global Perspective. Monthly Weather Review 132, 1645-1660.
26.	Luk, K.C., Ball, J.E., and Sharma, A. (2001). An application of artificial neural networks for rainfall forecasting. Math Comput Model 33, 683-693.
27.	Marks, F.D., Kappler, G., and DeMaria, M. (2002). Development of a Tropical Cyclone Rainfall Climatology and Persistence (R-CLIPER) Model. 25th Conference Hurricane and Tropical Meteorology, 327-328.
28.	Mase, H., Yasuda, T., and Mori, N. (2011). Real-Time Prediction of Tsunami Magnitudes in Osaka Bay, Japan, Using an Artificial Neural Network. Journal of Waterway, Port, Coastal, and Ocean Engineering 137, 263-268.
29.	Parrish, J.R., Burpee, R.W., Marks, F.D., and Grebe, R. (1982). Rainfall Patterns Observed by Digitized Radar During the Landfall of Hurricane Frederic (1979). Monthly Weather Review 110, 1933-1944.
30.	Raftery, A.E., Gneiting, T., Balabdaoui, F., and Polakowski, M. (2005). Using Bayesian Model Averaging to Calibrate Forecast Ensembles. Monthly Weather Review 133, 1155-1174.
31.	Rodgers, E.B., Chang, S.W., and Pierce, H.F. (1994). A Satellite Observational and Numerical Study of Precipitation Characteristics in Western North Atlantic Tropical Cyclones. Journal of Applied Meteorology 33, 129-139.
32.	Scofield, R.A. (1987). The NESDIS Operational Convective Precipitation- Estimation Technique. Monthly Weather Review 115, 1773-1793.
33.	Sloughter, J.M.L., Raftery, A.E., Gneiting, T., and Fraley, C. (2007). Probabilistic Quantitative Precipitation Forecasting Using Bayesian Model Averaging. Monthly Weather Review 135, 3209-3220.
34.	Tamkang University Department of Water Resources and Environmental Engineering (2013). Developing an Operation System of Flood Control ,Sediment Removal and Water Supply for Shihmen Reservoir and Providing Operation Consulting Service. M.o.E.A. Water Resources Agency
35.	Tamkang University Department of Water Resources and Environmental Engineering (2015). Maintaining an Operation System of Flood Control, and Anti-Sediment and Water Supply for Shihmen Reservoir and Providing Operation Consulting Service for 2015. M.o.E.A. Water Resources Agency
36.	Tamkang University Information Center for Water Environment (2018). 2018 Shihmen Reservoir Flood Control, Anti-silt and Water Supply System Maintenance and Operational Consulting. M.o.E.A. Water Resources Agency
37.	Tsai, F.-H. (2015). Investigate typhoon paths by self-organizing maps and predict long-term flow during typhoon periods. Master Thesis in Department of Bioenvironmental Systems Engineering (National Taiwan University)
38.	Wei, C., Hung, W.-C., and Cheng, K.-S. (2006). A multi-spectral spatial convolution approach of rainfall forecasting using weather satellite imagery. Advances in Space Research 37, 747-753.
39.	Wu, L., and Zhao, H. (2011). Dynamically Derived Tropical Cyclone Intensity Changes over the Western North Pacific. Journal of Climate 25, 89-98.
40.	Wu, M.-C., Hong, J.-S., Hsiao, L.-F., Hsu, L.-H., and Wang, C.-J. (2017). Effective Use of Ensemble Numerical Weather Predictions in Taiwan by Means of a SOM-Based Cluster Analysis Technique. Water 9.
41.	Wu, M.-C., Yang, S.-C., Yang, T.-H., and Kao, H.-M. (2018). Typhoon Rainfall Forecasting by Means of Ensemble Numerical Weather Predictions with a GA-Based Integration Strategy. Atmosphere 9, 425.
42.	Yang, Y., Lin, H., Guo, Z., and Jiang, J. (2007). A data mining approach for heavy rainfall forecasting based on satellite image sequence analysis. Computers & Geosciences 33, 20-30.