- 學位論文
以人工智慧模式評估在氣候變遷影響下對台灣區域河川流量之衝擊
Use Artificial intelligence model to assess the Taiwan region’s river flows impact under the influence of climate change
摘要
氣候變遷為目前世界各國所重視的議題，根據IPCC第四次之評估報告，氣候變遷對氣候及降雨型態帶來變化，相對的水資源也會受到衝擊影響。而台灣之水資源大都來自於河川，因此以河川流量做預估對象，藉此評估對未來水資源之衝擊。 研究中利用台灣北、中、南、東四區中央氣象局雨量及溫度資料和水利署之流量資料，以輻狀基底類神經網路(Radial Basis Function Neural Network，簡稱RBFNN)為核心結合基因演算法建立區域流量評估模式(Regional Flow Impact Model，簡稱RFIM)，再將降尺度後之氣候資料輸入RFIM模式，模擬分析未來氣候變遷對區域流量之衝擊，並利用拔靴法(Bootstrap Method)重複建立模型來評估RFIM模式之不確定性，提供給決策者做為規劃未來水資源應用之參考。 模擬分析結果顯示未來各區流量上下界差異不會因情境上之不同而影響極大，約為正負2億立方公尺之間。北區上下界流量差異平均13億立方公尺，中區平均21億立方公尺，南區平均17億立方公尺，東區平均10億立方公尺。在任一情境下北區豐枯水期流量差異約為每年40億立方公尺，中區豐枯水期流量差異約為每年88億立方公尺，南區豐枯水期流量差異約為每年93億立方公尺，其東區受情境影響較大，在A1B情境下豐枯水量差異約為每年30億立方公尺，在B1及A2情境下約為每年67億立方公尺。在流量趨勢方面，在A2情境短期、A1B情境短期、A1B情境長期、B1情境中期，四區皆有流量增加之趨勢。
並列摘要
The climate changes are the most currently important issue around the world. According to the forth assessment report from IPCC, the climate changes will result in the climate and rainfall variations; it will also impact on the water resources. The water resources almost come from the river flow in Taiwan. For this reason, the study estimates the impact of river flow in the future. In this research, the related rainfall and temperature data from the Central Weather Bureau and river flow data from the Water Resources Agency, including the northern, the middle, the southern and the eastern region of Taiwan, were collected first. Then, the Regional Flow Impact Model (RFIM) is constructed by combining the Radial Basis Function Network (RBFNN) and the Genetic Algorithm (GA). The downscaling climate data are inputted to the calibrated RFIM to investigate the impact on regional river flow which offers the references for decision- makers as planning the applications of future water resources. Meanwhile, the Bootstrap sampling method is used estimate the uncertainty of RFIM. Simulation results show that there exist slight differences in river flow under different scenarios with upper and lower bounds about 0.2 billion cubic meters between positive and negative. The average difference between the upper and lower flows is 1.3 billion cubic meters in the northern region, 2.1 billion cubic meters in the middle region, 1.7 billion cubic meters in the southern region, and 1.0 billion cubic meters in the eastern region. For any scenarios, the river flows between the wet season and dry season are about 4.1 billion cubic meters in the northern region per year, 8.8 billion cubic meters in the middle region per year, 9.3 billion cubic meters in the southern region per year. However, the river flow in the eastern region is highly influenced by the scenarios. In A1B scenarios, the flow between the wet and dry season are about 3.1 billion cubic meters in the eastern region per year. In B1 and A2 scenarios, the difference in flow is about 6.7 billion cubic meters. Increase in river flow of the four regions under the scenarios of A2 short-term, A1B short-term, A1B long-term and B1 mid-term can be observed.