透過您的圖書館登入
IP:18.118.254.94
  • 學位論文

以多維度經驗模態分解應用於台灣各城市溫度上升以及極端事件調查之研究

Spatiotemporal Trend and Variability of Warming and Extreme Weather Events in Taiwan Based on Multi-dimensional Complementary Ensemble Empirical Mode Decomposition (MCEEMD)

指導教授 : 蔡宛珊

摘要


全球的氣候正在面臨嚴重的暖化現象,然而台灣在過去的100年間暖化的速度則高於全球增溫的平均值,可見暖化對台灣氣候而言一直都非常的顯著,如何因應對台灣氣溫不斷上升的趨勢是對災害防範非常重要的一個課題。然而傳統計算趨勢的統計方法,例如: 直線趨勢法或Mann–Kendall趨勢檢測法,無法有效地顯示出不同空間尺度上的暖化隨著時間的變化。因此在本研究中,根據時頻分析希伯特黃轉換中的經驗模態分解(Empirical Mode Decomposition),引入了多維度經驗模態分解法Multi-dimensional Complementary Ensemble Empirical Mode Decomposition (MCEEMD)應用於台灣長時間的網格化觀測資料來檢測全台灣於各個城市過去58年(1960~2017)暖化的趨勢。本研究發現自2000年開始大部分的區域除了東部地區皆有變暖的現象,然而台北,台中以及高雄市等較高度發展的都市其暖化的速度則又更加快速。故根據以上區域差異的特性,調查了溫度上升與影響都市熱島效應的參數進行相關性分析。 上述各城市暖化現象以及熱島效應伴隨極端天氣事件頻繁發生的衝擊越來越明顯。本研究係以不同極端氣候的門檻值來分析台灣於過去58年熱浪、極端低溫與極端降雨事件發生的趨勢,發現熱浪發生的頻率在暖化較嚴重的地區皆有明顯的上升,然而在大部分地區之極端低溫發生的事件數則皆隨著時間下降。 在極端降雨方面,根據會產生不同程度災害警戒的門檻值來評估大雨(日降雨>80mm)以及豪雨事件(日降雨> 200mm)的發生在暖化情況下的趨勢。除此之外,本研究也根據CMIP5模式的資料調查了暖化最為嚴重的台北地區進行未來熱浪事件在夏天(7~9月)的推估。在RCP4.5暖化情境下,可看出熱浪到了21世紀末頻率以及延時在全台皆有明顯的增強。然而在暖化程度最強的情境RCP8.5下,全台到了21世紀末熱浪發生的頻率下降,反之延時則增長到20天以上,可看出在夏季一次熱浪事件可能會持續一個月以上。 然而熱浪會引發許多熱相關的傷害以及疾病,在暖化以及都市熱島效應下,使得高度發展以及人口集中的城市為高度風險區。為了預防這類相關的災害使當局能夠做出相對應政策,預測未來極端高溫的準確性是非常重要的。故本研究最後也引入了MCEEMD-RBFNN預測模型來對台北市未來7天的日高溫進行預測。預測的結果於相關係數(R) 、均方根誤差(RMSE)以及平均絕對百分比誤差(MAPE)表現上皆顯示此模型在預測未來7天的日高溫誤差非常的低,並在前三天完美地捕捉了極端的日高溫值。

並列摘要


The climate of Taiwan has experienced a temperature increase of 1-1.4 °C in the past century, which is higher than the global average. Such a warming trend corresponding to climate change increases the probability of extreme weather events. However, the spatial and temporal characteristics of such warming are difficult to evaluate using traditional trend detection techniques. Therefore, a method of Multi-dimensional Complementary Ensemble Empirical Mode Decomposition (MCEEMD) is introduced in this study to deal with the gridded climate data. The spatial and temporal evolution of temperature from year 1960 to 2017 is revealed. Noticeable warming in recent decades occurred in highly developed cities such as Taipei and Kaohsiung city. Based on the results, the correlation between the accumulated warming and the essential urban indicators are quantified at urban and rural stations in five representative cities. On the other hand, the characteristic timescale of daily maximum/minimum and precipitation data is identified. The spatial and temporal trend of extreme weather events, including heat wave, extreme cold events and extreme precipitation, are also examined from 1960 to 2017. Meanwhile, the heat wave trend in the future in Taiwan is estimated by applying the CMIP5 model data. Under the RCP4.5 warming scenario, the variation of frequency and duration of heat wave events is demonstrated in the mid and end of the 21st century. Under the RCP8.5 warming scenario, all regions of Taiwan may experience a month of heat wave at the end of the 21st century. Moreover, in order to prevent heat-related hazards, the Radial Basis Function Neural Network (RBFNN) model coupled with the MCEEMD algorithm, is proposed for the next 7 days' daily maximum temperature forecasting in Taipei. Considering two cases of 7 and 14-days input values, the number of neural nodes in the hidden layer is determined. The forecasting results by the MCEEMD-RBFNN model perform much better in capturing the extreme value of daily maximum temperature than the predictions by RBFNN model.

參考文獻


Broecker, W. S. (1975). Climatic change: are we on the brink of a pronounced global warming? Science, 189(4201), 460-463.
Broomhead, D. S., Lowe, D. (1988). Radial basis functions, multi-variable functional interpolation and adaptive networks. Complex Systems, 2, 327-355.
Chen, S., Grant, P., Cowan, C. (1992). Orthogonal least-squares algorithm for training multioutput radial basis function networks. Paper presented at the IEE Proceedings F (Radar and Signal Processing).
Chen, V. C., Qian, S. (1996). Time-frequency transform vs. Fourier transform for radar imaging. Paper presented at the Proceedings of Third International Symposium on Time-Frequency and Time-Scale Analysis (TFTS-96).
Chen, C. M., Wang, F. R. (2000). 台灣地區長期暖化現象與太平洋海溫變化之關係. 大氣科學, 28(3), 221-241.

延伸閱讀