全球暖化與氣候變遷不斷加劇，導致極端降雨之強度與頻率日益嚴重，更造成淹水事件頻傳，為儘早得知淹水風險的嚴重程度，降雨監視、水情即時監控系統及淹水預報扮演著至關重要的角色，目前既有的淹水資訊預測主要可分為兩種方法，包含水文模型及數據驅動模型，但預測之資訊仍不夠充足，如政府所裝設之淹水感測器只可監控即時淹水資訊，而無法掌握其未來淹水之變化量，若能獲得該資訊，則可提供災害防救及應變之參考。
有鑑於此，本研究旨在運用深度學習結合注意力機制建立預測未來時刻之淹水感測器的深度。由兩階段預測模型組成，第一部分以分類型之隨機森林（Random forest）建構淹水事件偵測模型；第二部分以時空注意力機制結合具轉換閥之長短期記憶（Transformation-gated LSTM）建置未來三小時淹水預測模型。與機器學習基線模型相比，本研究於預測未來三小時的RMSE損失為最低值，而預測第一小時（T+1）準確的淹水感測器達76%，相較於傳統水文模型，本研究使用數據驅動模型具有快速運算的優勢，在數秒內即可得知未來淹水深度之資訊。透過視覺化時空注意力機制中的權重值，以解釋各淹水感測器與各變量之間的對應關係，並可針對不同地理特性之淹水感測器進行深入分析，給予不同層面之解釋能力及參考依據。本研究特點在於不僅考慮模型績效，還可掌握影響各感測器之時空資訊，提供相應的防災訊息。

With global warming and climate change, the intensity and occurrence rate of heavy rainfall events are significantly increasing in many areas, resulting in the potential for flood hazards. Therefore, it is crucial to forecast the future trends of floods to understand the timely risk and control flood effectively. However, current information is not sufficient, e.g., the flood sensors installed by the government can only observe real-time information, it is hard to grasp the sudden change in the amount of flood depth values. Nonetheless, given the abovementioned information, the future flood can be known and provided to the related decision maker to enhance decision quality.
Focusing on the realistic needs, this study constructs a method to forecast the flood depth of flood sensors. The description of each stage is illustrated as follows: First, based on the information provided by the precipitation data, a classifier model based on Random Forest algorithm is built to detect whether the flood sensor will be flooded or not. Second, a regression prediction model based on Transformation-gated LSTM and attention mechanism is deployed, including using real-time and forecasting rainfall information to forecast the flood depth values for the flood sensors. Compare with traditional hydrological models, this study applies data-driven models which is less time-consuming. By visualizing the attention weights, we can gain insight into the relationship between the input features and the model. This information can be interpreted and derived to a decision strategy, the potential risks of floods can be mitigated, leading to a reduction in the extent of flooding damage.

摘要 I
Abstract II
目錄 III
圖目錄 V
表目錄 VI
第一章 緒論 1
1.1研究背景與動機 1
1.2研究目的 4
1.3論文架構 5
第二章 文獻回顧 6
2.1淹水預報 6
2.2淹水預報模型 8
第三章 研究方法 13
3.1資料前處理 14
3.2特徵工程 14
3.3模型建構 17
3.3.1 淹水事件偵測模型 17
3.3.2 淹水深度預測模型 19
3.4模型評估 25
3.5模型解釋力 27
第四章 個案研究 29
4.1資料準備與處理 30
4.2預測結果與模型評估 35
4.3注意力機制解釋分析 40
第五章 結論與未來研究 44
參考文獻 45

國家災害防救科技中心. 民生示警公開資料平台 https://alerts.ncdr.nat.gov.tw/calamityAlertSearch_history.aspx

淡江大學. (2020). 整合物聯網監測資料與機器學習技術建置智慧城市淹水預報系統. 台灣: 經濟部水利署 Retrieved from https://www.itdr.tw/dispPageBox/getFile/GetView.aspx?FileLocation=PJ-SITEVC%5CFiles%5CPrjFiles%5C68%5C&FileFullName=%E5%85%A8%E6%96%87%E5%A0%B1%E5%91%8A.pdf&FileName=FR5739108054Ga1jm.PDF

Adnan, R. M., Liang, Z., Trajkovic, S., Zounemat-Kermani, M., Li, B., & Kisi, O. (2019). Daily streamflow prediction using optimally pruned extreme learning machine. Journal of Hydrology, 577, 123981.

Alfieri, L., Salamon, P., Pappenberger, F., Wetterhall, F., & Thielen, J. (2012). Operational early warning systems for water-related hazards in Europe. Environmental Science & Policy, 21, 35-49.

Chen, C.-Y., Lin, L.-Y., Yu, F.-C., Lee, C.-S., Tseng, C.-C., Wang, A.-H., & Cheung, K.-W. (2007). Improving debris flow monitoring in Taiwan by using high-resolution rainfall products from QPESUMS. Natural hazards, 40(2), 447-461.

Chen, C., Jiang, J., Liao, Z., Zhou, Y., Wang, H., & Pei, Q. (2022). A short-term flood prediction based on spatial deep learning network: A case study for Xi County, China. Journal of Hydrology, 607, 127535.

Dembélé, M., Hrachowitz, M., Savenije, H. H., Mariéthoz, G., & Schaefli, B. (2020). Improving the predictive skill of a distributed hydrological model by calibration on spatial patterns with multiple satellite data sets. Water Resources Research, 56(1), e2019WR026085.

Ding, Y., Zhu, Y., Feng, J., Zhang, P., & Cheng, Z. (2020). Interpretable spatio-temporal attention LSTM model for flood forecasting. Neurocomputing, 403, 348-359.

Ghaeini, R., Fern, X. Z., & Tadepalli, P. (2018). Interpreting recurrent and attention-based neural models: a case study on natural language inference. arXiv preprint arXiv:1808.03894.

Haile, A. T., Tefera, F. T., & Rientjes, T. (2016). Flood forecasting in Niger-Benue basin using satellite and quantitative precipitation forecast data. International journal of applied earth observation and geoinformation, 52, 475-484.

Hu, J., & Zheng, W. (2020). Multistage attention network for multivariate time series prediction. Neurocomputing, 383, 122-137.

Kao, I.-F., Zhou, Y., Chang, L.-C., & Chang, F.-J. (2020). Exploring a Long Short-Term Memory based Encoder-Decoder framework for multi-step-ahead flood forecasting. Journal of Hydrology, 583, 124631.

Liao, K.-H., Le, T. A., & Van Nguyen, K. (2016). Urban design principles for flood resilience: Learning from the ecological wisdom of living with floods in the Vietnamese Mekong Delta. Landscape and Urban Planning, 155, 69-78.

Liu, J., Xu, L., & Chen, N. (2022a). A spatiotemporal deep learning model ST-LSTM-SA for hourly rainfall forecasting using radar echo images. Journal of Hydrology, 609, 127748.

Liu, J., Yuan, X., Zeng, J., Jiao, Y., Li, Y., Zhong, L., & Yao, L. (2022b). Ensemble streamflow forecasting over a cascade reservoir catchment with integrated hydrometeorological modeling and machine learning. Hydrology and Earth System Sciences, 26(2), 265-278.

Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in neural information processing systems, 30.

Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S. L., Péan, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., & Gomis, M. (2021). Climate change 2021: the physical science basis. Contribution of working group I to the sixth assessment report of the intergovernmental panel on climate change, 2.

Merkuryeva, G., Merkuryev, Y., Sokolov, B. V., Potryasaev, S., Zelentsov, V. A., & Lektauers, A. (2015). Advanced river flood monitoring, modelling and forecasting. Journal of computational science, 10, 77-85.

Miau, S., & Hung, W.-H. (2020). River flooding forecasting and anomaly detection based on deep learning. IEEE Access, 8, 198384-198402.

Mishra, K., & Sinha, R. (2020). Flood risk assessment in the Kosi megafan using multi-criteria decision analysis: A hydro-geomorphic approach. Geomorphology, 350, 106861.

Motta, M., de Castro Neto, M., & Sarmento, P. (2021). A mixed approach for urban flood prediction using Machine Learning and GIS. International Journal of Disaster Risk Reduction, 56, 102154.

Mourato, S., Fernandez, P., Marques, F., Rocha, A., & Pereira, L. (2021). An interactive Web-GIS fluvial flood forecast and alert system in operation in Portugal. International Journal of Disaster Risk Reduction, 58, 102201.

Nam, D. H., Mai, D. T., Udo, K., & Mano, A. (2014). Short‐term flood inundation prediction using hydrologic‐hydraulic models forced with downscaled rainfall from global NWP. Hydrological Processes, 28(24), 5844-5859.

Nazir, H. M., Hussain, I., Faisal, M., Elashkar, E. E., & Shoukry, A. M. (2019). Improving the prediction accuracy of river inflow using two data pre-processing techniques coupled with data-driven model. PeerJ, 7, e8043.

Niu, Z., Zhong, G., & Yu, H. (2021). A review on the attention mechanism of deep learning. Neurocomputing, 452, 48-62.

Nyrup, R., & Robinson, D. (2022). Explanatory pragmatism: a context-sensitive framework for explainable medical AI. Ethics and information technology, 24(1), 13.

Park, S., & Yang, J.-S. (2022). Interpretable deep learning LSTM model for intelligent economic decision-making. Knowledge-Based Systems, 248, 108907.

Pouyanfar, S., Sadiq, S., Yan, Y., Tian, H., Tao, Y., Reyes, M. P., Shyu, M.-L., Chen, S.-C., & Iyengar, S. S. (2018). A survey on deep learning: Algorithms, techniques, and applications. ACM Computing Surveys (CSUR), 51(5), 1-36.

Rözer, V., Peche, A., Berkhahn, S., Feng, Y., Fuchs, L., Graf, T., Haberlandt, U., Kreibich, H., Sämann, R., & Sester, M. (2021). Impact‐based forecasting for pluvial floods. Earth's Future, 9(2), 2020EF001851.

Teunissen, P. (2007). Least-squares prediction in linear models with integer unknowns. Journal of Geodesy, 81(9), 565-579.

Tian, X., Schleiss, M., Bouwens, C., & van de Giesen, N. (2019). Critical rainfall thresholds for urban pluvial flooding inferred from citizen observations. Science of The Total Environment, 689, 258-268.

UNDRR and WMO. (2022). Global status of multi-hazard early warning systems: Target G. United Nations Office for Disaster Risk Reduction: Geneva, Switzerland.

Wang, H., Zhao, Y., Zhou, Y., & Wang, H. (2021). Prediction of urban water accumulation points and water accumulation process based on machine learning. Earth Science Informatics, 14, 2317-2328.

Wang, Y., Li, C., Liu, M., Cui, Q., Wang, H., Jianshu, L., Li, B., Xiong, Z., & Hu, Y. (2022). Spatial characteristics and driving factors of urban flooding in Chinese megacities. Journal of Hydrology, 613, 128464.

Wei, J., Hang, R., & Luo, J.-J. (2022). Prediction of Pan-Arctic Sea Ice Using Attention-Based LSTM Neural Networks. Frontiers in Marine Science, 918.

Wu, Z., Zhou, Y., Wang, H., & Jiang, Z. (2020). Depth prediction of urban flood under different rainfall return periods based on deep learning and data warehouse. Science of The Total Environment, 716, 137077.

Xu, K., Ba, J., Kiros, R., Cho, K., Courville, A., Salakhudinov, R., Zemel, R., & Bengio, Y. (2015). Show, attend and tell: Neural image caption generation with visual attention. International conference on machine learning, 2048-2057. PMLR.

Xu, Y., Hu, C., Wu, Q., Jian, S., Li, Z., Chen, Y., Zhang, G., Zhang, Z., & Wang, S. (2022). Research on particle swarm optimization in LSTM neural networks for rainfall-runoff simulation. Journal of Hydrology, 608, 127553.

Yan, C., Tu, Y., Wang, X., Zhang, Y., Hao, X., Zhang, Y., & Dai, Q. (2019). STAT: Spatial-temporal attention mechanism for video captioning. IEEE transactions on multimedia, 22(1), 229-241.

Yang, Z.-b., Zhang, J.-p., Zhao, Z.-b., Zhai, Z., & Chen, X.-f. (2020). Interpreting network knowledge with attention mechanism for bearing fault diagnosis. Applied Soft Computing, 97, 106829.

Yaseen, Z. M., Kisi, O., & Demir, V. (2016). Enhancing long-term streamflow forecasting and predicting using periodicity data component: application of artificial intelligence. Water resources management, 30(12), 4125-4151.

Yu, W., Nakakita, E., Kim, S., & Yamaguchi, K. (2015). Improvement of rainfall and flood forecasts by blending ensemble NWP rainfall with radar prediction considering orographic rainfall. Journal of Hydrology, 531, 494-507.

Zhai, M., Xiang, X., Zhang, R., Lv, N., & El Saddik, A. (2019). Optical flow estimation using channel attention mechanism and dilated convolutional neural networks. Neurocomputing, 368, 124-132.

Zhang, S., & Pan, B. (2014). An urban storm-inundation simulation method based on GIS. Journal of Hydrology, 517, 260-268.

Zhou, Q., Leng, G., Su, J., & Ren, Y. (2019). Comparison of urbanization and climate change impacts on urban flood volumes: Importance of urban planning and drainage adaptation. Science of The Total Environment, 658, 24-33.