| 摘要: | 氣候異常改變在上個世紀還是一個尚待確認的議題，但近二、三十年來，自然環境不斷透露出相關訊息後，至今各領域之學者專家無一不承認，大量的溫室氣體排放，直接或間接造成了全球氣候的變化。因此1992年在巴西里約日內盧所簽訂的「氣候變化綱要公約」，以及1997年在日本京都所簽訂的「京都議定書」，都是國際間為了削減大氣中的溫室氣體所做的一些協定。但是根據IPCC（2001）的報告指出，即使最樂觀的策略也很難在本世紀末將二氧化碳的濃度穩定在二倍工業革命之前的水準（280ppm），因此氣候變遷在本世紀已是一不可避免的態勢。 台灣地區的降水量帶給全區的天賦水量約為九百餘億立方公尺，但由於人口密度高，地形上的限制以及地理與季節分布的不平均，每個國人可以利用的水資源量其實很有限。近來由於產業結構的變動，連帶牽動標的用水量的供需失調問題，加以本世紀確已存在的天候的異常狀況，使水資源不敷使用的窘況益發突顯，從而引發了本研究的出發動機。一般在時間序列的處理分析上，大多採用傅利葉分析，但在氣候變遷的影響下，傅利葉分析並不能有效地提供所分析資料的訊息，為此本研究嘗試使用小波轉換來確立資料的資訊。 小波轉換的觀點是由法國學者Morlet首先提出，其除了可分析平穩特性的資料外，還可以對於非平穩與非週期性的資料進行分析，並且有能力表現出局部的分析，如資料趨勢、異常的不連續點…等。本研究首先以小波轉換針對中央氣象局在台灣所設的21個測站進行資料分析，時間從1971至2003年共33 年的資料，分析的氣象參數有平均溫度、降水量、相對濕度、與降水天數，比較各區域之氣象趨勢與地域之差異性，再從北部區域集中討論石門集水區的水庫操作策略，嘗試找出在氣候變遷下，水庫操作的實際可行的因應策略，以能落實水資源的永續發展。 本研究獲致的主要成果，有： 1. 本研究針對全台之水文氣象資料作一長時間分析，確立各區域在不同氣象參數之趨勢，從而提出警告與建議；並且在小波理論的應用上，可以看出隱藏於資料背後的資訊，尤其是未來週期的變化，對於水資源調配與政策的擬定有一定的助益。 2. 利用差值分析比較，容易發現各區域的平均表現，若採用傳統數學迴歸來作為線性預測，迴歸之相關性都表現的不好，能獲得準確的值效果有限，如由差值的灰生成趨勢可瞭解整體變化趨勢，配合簡單的灰預測，更可約略預測未來可能的走向，在應用層面上可以落實並執行。 3. 本研究以小波理論中的小波轉換與多尺度分析來進行資料處理，獲致良好的結果，在一大堆資料中所隱藏的訊息，都隨著分析圖形而顯露無遺，從結果判斷可提供訂定政策者真實且確切的資訊，另對於長期趨勢及週期表現更可抓住其脈動，窺其變化。 4. 在本研究中，對於石門水庫的規線的功用，進行個別探討與分析，從不同的調整中，瞭解規線變動對後續操作所帶來的影響，在這樣的過程中，可得到調整規線的經驗，可供給其他水庫在模擬操作，或調整規線時的一個參考，而且不同時期（入流豐富、入流枯旱）的需求也不儘相同，如此的規線調整，可真實符合需求。 5. 從本研究的討論中，提出「彈性規線」的作法是確實可行，尤其在氣候變遷的世代，特殊的操作環境是需要被考慮的，如何能將降水的資訊納入規線內，應是可以著力一個方向，針對降水預測的改善，對於規線操作是絕對的幫助的，如能有更精準的預測模式，規線之操作結果將更趨完善。 6. 對於在台灣採年計法方式的操作，應隨著氣候變遷的影響，有所調整，本研究提出以“季”為操作期距，是一拋磚引玉的作法，因案例所在之集水區，降水情況恰巧以3~4個月為一期距，因此在石門水庫的操作期距，依本研究的結論更改為以“季”為操作間距；另外有關防洪操作，由本研究的經驗中，採反饋式類神經網路頗具成效，對於其他水庫也可嘗試利用此方法，並配合適當的演算法，也應可有不錯的結果。 Climate abnormality still is an issue under defining until twentieth centenary. But natural environment continued to reveal some information about it recently. Experts in various fields began to admit the phenomenon due to heavy greenhouse gases emissions. So the Framework Convention on Climate Change (FCCC) signed in Rio, Brazil in 1992 and “Kyoto Protocol” signed in Kyoto, Japan in 1997, both are the agreements for controlling in principle the CO2. Then, the report of Intergovernmental Panel on Climate Change (IPCC, 2001) told that it is impossible to reduce CO2 emissions to double (=280ppm) of quantity before Industrial Revolution in twenty-first centenary. The effect of climate change will keep on going in this centenary. Yearly precipitations can produce over 90 billions cubic meter of water in Taiwan. Because of high density of population, limits of the geography, different locality and season, everyone in Taiwan can use water within limits. However, the changes of industrial types make the unbalance for supply and demand of water resources. In addition to all of the above, the allocation of water resources becomes extreme important under climate change. The motivation of this study is developed from the above reasons. Generally analyzing time series data, Fourier analysis would be applied. Moreover, the consequences from climate change block to utilizing Fourier analysis. This study tried to apply wavelet theorem for making sure the information of data. The mechanics of wavelet transfer was developed by Morlet in France. It not only can investigate the stable data, but also treat unstable and non-periodic data. From series, it can find the trend, localities, abnormal discontinuities…etc. First, this study applied wavelet theorem to analyze the data of the CWB 21 stations from 1971 to 2003. The climate parameters to examine are temperature, precipitation, relative humidity, and the numbers of rainy days. From results, we can get differences and trends of the data. Second, we made the comparisons between the northern of Taiwan and Shin-Men basin from analyzing data. We tried to look for the best rule curves under climate change, and confirmed that rule curve could operate in reality. The main fruits form this study include that: 1. This study made a long term analysis on climate data to establish the trends of series. From wavelet applications, we got some hidden information of the time series, and the change of the climate period in the future. All of the information is able to aid policy-making and the allocation of water resources. 2. With the comparisons of the differences, it is easy to be showed local results in various regions. The linear regression is not appropriate for analyzed data, because the R-square is bad. Hence, with grey generating, we can see the trends of the changes, and predict the next value in the series into real operations. 3. We received better result to observe the climate data by means of wavelet transfer and multi-resolution analysis. There is much useful information displayed in all of random and confused dada after data processed. Those outputs can be referred to the decision-makers for observe long term trends and period changes. 4. In this study, we discussed the functions of the rule curves individually. By different trials, we can understand the influences of the modified rule curves. The conclusions may be supplied to others to simulating reservoir operation or adjusting local rule curves. 5. The concept of “flexible rule curve” can be worked in real situations form conclusions. In times of climate change, particular operations must be needed. To combining precipitation and rule curves is an interesting topic that is worth of researching. 6. The period of a reservoir operation should change from “year” to “season” from our conclusions in Shin-Men reservoir. Other reservoirs can change theirs depending on real precipitations. Then, flood operation in reservoir can add recurrent neutral network into consideration and better ends will be expected with this framework. |
