本研究應用深度學習理論發展颱風降雨即時預報模式，針對臺北、花蓮、臺南及阿里山四個中央氣象局氣象站，收集1994年至2021年共155場中央氣象局有發布警報單之颱風場次，其中前102場做為模式率定使用，後53場做為驗證使用。本研究運用Pearson、Kendall's tau、Spearman相關係數將各颱風及氣象因子與累積雨量進行相關性分析，據以決定輸入變量。本研究以門閥遞迴單元（gated recurrent unit, GRU）方法建立颱風雨量即時預報模式，來預報未來1至6小時的颱風累積雨量，並以是否加入颱風移動速度做為輸入變量來建置兩類模式，再以誤差指標探討累積雨量及小時雨量的預報效能。預報結果顯示，應用GRU發展的颱風雨量預報模式在未來1至6小時累積雨量預報值與觀測值的相關係數皆有0.96以上。而將累積降雨量預報值轉換為小時雨量的結果顯示未來1小時的相關係數約為0.65。最後本研究探討各測站間預報模式之率定與驗證的差異，發現本研究建置的颱風雨量預報模式有過度擬合率定資料的情形。

SUMMARY
This study applied the deep learning theory to develop the real-time forecasting models for typhoon rainfall. Meteorological data from four weather stations (Taipei, Hualien, Tainan, and Alishan) were collected, and a total of 155 typhoon events for which the Central Weather Bureau, Taiwan issued typhoon warnings from 1994 to 2021 were compiled. The first 102 events were used for model calibration, and the last 53 events were used for validation. In this study, the gated recurrent unit method was used to develop real-time typhoon rainfall forecasting models to predict the cumulative rainfall for lead times of 1 to 6 hours. Forecasting results showed that the forecasting performance in terms of correlation coefficients were all above 0.96 with respect to cumulative rainfall for different lead times and the forecasting performance in terms of correlation coefficients is 0.65 with respect to hourly rainfall for 1-hour lead time.
Keywords：typhoon rainfall, real-time forecasting, gated recurrent unit

Chang, L. C., Chang, F. J., Yang, S. N., Tsai, F. H., Chang, T. H., and Herricks, E. E. (2020). Self-organizing maps of typhoon tracks allow for flood forecasts up to two days in advance. Nature Communications, 11(1), 1-13.
Chen, S. T. (2021). Probabilistic typhoon rainfall forecasting using a modified fuzzy inference model. Terrestrial, Atmospheric and Oceanic Sciences, 32(4), 583-596.
Chen, S. T. (2013). Multiclass support vector classification to estimate typhoon rainfall distribution. Disaster Advances, 6(10), 110-121.
Cho, K., Merrienboer, B. V., Gulcehre, C., Bahdanau, D., Bougares, F., Schwenk, H., and Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation.
Teng, H. F., Done, J. M., Lee, C. S., and Kuo, Y. H. (2019). Dependence of probabilistic quantitative precipitation forecast performance on typhoon characteristics and forecast track error in taiwan. Weather and Forecasting, 35(2), 585-607.
Hsieh, P. C., Tong, W. A., and Wang, Y. C. (2019). A hybrid approach of artificial neural network and multiple regression to forecast typhoon rainfall and groundwater-level change. Hydrological Sciences Journal, 64(14), 1793-1802.
Hu, C., Wu, Q., Li, H., Jian, S., Li, N., and Lou, Z. (2018). Deep learning with a long short-term memory networks approach for rainfall-runoff simulation. Water, 10(11), 1543.
Hsu, L. H., Kuo, H. C., and Fovell, R. G. (2013). On the geographic asymmetry of typhoon translation speed across the mountainous island of taiwan. Journal of the Atmospheric Sciences, 70(4), 1006-1022.
Lin, G. F., Jhong, B. C., and Chang, C. C. (2013). Development of an effective data-driven model for hourly typhoon rainfall forecasting. Journal of Hydrology, 495, 52-63.
Lin, G. F. and Wu, M. C. (2009). A hybrid neural network model for typhoon-rainfall forecasting. Journal of Hydrology, 375(3-4), 450-458.
Wei, C. C., and Hsu, C. C. (2021). Real-time rainfall forecasts based on radar reflectivity during typhoons: case study in southeastern taiwan. Sensors, 21(4), 1421.
Wei, C. C., and Huang, T. H. (2021). Modular neural networks with fully convolutional networks for typhoon-induced short-term rainfall predictions. Sensors, 21(12), 4200.
Wei, C. C., and Hsieh, P. Y. (2020). Estimation of hourly rainfall during typhoons using radar mosaic-based convolutional neural networks. Remote Sensing, 12(5), 896.
Yen, M. H., Liu, D. W., Hsin, Y. C., Lin, C. E., and Chen, C. C. (2019). Application of the deep learning for the prediction of rainfall in Southern Taiwan. Scientific Reports, 9(1), 12774.
Yu, P. S., Chen, S. T., Chen, C. J., and Yang, T. C. (2005) The potential of fuzzy multi-objective model for rainfall forecasting from typhoons. Nat Hazards, 34, 131-150.
郭鴻基（2010）地形影響颱風異常變化之探討，交通部中央氣象局委託研究計畫期末成果報告。
陳政安（2011）侵台颱風路徑變化對台灣降雨影響，中國文化大學理學院地學研究所大氣科學組碩士論文。
陳清田、陳儒賢、陳奕任（2012）以支撐向量分類與倒傳遞神經網路為基礎的颱風降雨預報模式，水保技術，7卷3期，第138頁至第151頁。