德國的榛子與帚石楠在暖化下的始花日期趨勢

近期以來，暖化使部分溫帶地區的開花物候有提早出現的趨勢；然而，不同物種對暖化的反應並不相同，本研究以詳盡地分析方法驗證這件事。本研究分析德國的榛子(Corylus avellana L.)與帚石楠(Calluna vulgaris L.)在不同測站的始花日期(First flowering dates，FFD)與溫度之距平序列(Anomaly series)，分析時間皆為1951至2015年(基底時間為1961至1990年)。本研究同時分析兩物種測站平均的FFD和該區域的溫度距平序列。此外，本研究檢驗榛子的FFD與其對應溫度距平序列兩者與北大西洋震盪(North Atlantic Oscillation)的相關性。考量物候資料常為非線性(Nonlinear)與非平穩(Nonstationary)，本研究使用總體經驗模態分解法(Ensemble empirical mode decomposition)萃取FFD與溫度距平序列的趨勢。為進一步瞭解趨勢的呼應情形和未來可能的演變，本研究估計所萃取趨勢的速度和加速度。最後，本研究使用最大熵值重複置還取樣法(Maximum entropy bootstrap)建立趨勢與這些估算的信賴區間。結果顯示，榛子的FFD距平序列大多具提前的趨勢，其對應溫度距平序列具上升的趨勢，兩相呼應。不僅如此，所估計的榛子FFD與溫度趨勢之速度與加速度也呼應。相對的，帚石楠的FFD距平序列的趨勢並不呼應溫度距平序列的趨勢。然而，榛子的FFD距平序列的趨勢顯著提前的時間早於與其對應溫度距平序列的趨勢顯著上升的時間，這可能是因為榛子的FFD對溫度變化較敏感。最後，榛子的FFD與其對應溫度距平序列分別和平均一至三月的北大西洋震盪指數呈負與正相關。暖化若持續，榛子的FFD距平序列的趨勢在大部分測站會持續提前。然而，位於最高海拔的測站，其暖化與FFD距平序列提前的趨勢可能趨緩。

關鍵字

暖化；始花日期；總體經驗模態分解法；最大熵值重複置還取樣法

並列摘要

Over the past several decades, flowering phenology of many species in some temperate regions has advanced in response to warming. However, species response to warming differently. This thesis intends to analyze the complex relationships between warming and plant flowering. This study analyzed the first flowering dates (FFD) of hazel (Corylus avellana L.) and heather (Calluna vulgaris L.) and the corresponding temperature anomalies at different phenological observation stations in Germany. The observation period was from 1951 to 2015 (base period 1961- 1990). This study also analyzed the averaged across stations FFD and the corresponding regional temperature anomalies of both species. In addition, this study examined the correlations between the North Atlantic Oscillation and the FFD of hazel and their corresponding temperature anomaly series. Because phenological data are usually nonlinear and nonstationary, this study used ensemble empirical mode decomposition to extract temperature and FFD anomaly trends. In order to thoroughly understand the correspondence between the two anomaly trends and their possible future directions, this study estimated the velocities and accelerations of the extracted trends. Finally, this study used maximum entropy bootstrap to establish confidence intervals of the trends and their estimated velocities and accelerations. The results showed that the advancing trends of FFD anomaly series of hazel corresponded well to the rising trends of the temperature anomaly series. Moreover, the estimated velocities and accelerations of the extracted trends of FFD and temperature anomaly series also corresponded well. In contrast, FFD anomaly trends of heather did not correspond to the temperature anomaly trends. This study also found that time lags existed between the extracted trends of the temperature and FFD anomaly series in hazel with the FFD trends leading. The reason maybe that hazel’s FFD is more sensitive to temperature variations. Lastly, the FFD anomalies of hazel and their corresponding temperature anomalies were negatively and positively correlated with the average January to March North Atlantic Oscillation, respectively. Based on the estimated velocities and accelerations, FFD anomaly trends of hazel will keep advancing in most of the stations if warming continues. As for the highest station, both the warming and the advancing FFD anomaly trends of hazel are likely to slow down in the future.