應用BIOMOD2模擬臺灣水青岡之分布

Using BIOMOD2 to model the species distribution of Fagus hayatae

物種分布模擬（Species Distribution Modeling, SDM）是利用物種分布點資料與環境預測圖層去量化物種－－環境關係，其中，模擬物種分布之方法在近年來被大量地發展使用。本研究以臺灣水青岡為材料，應用7項環境預測變數，在BIOMOD2 （BIOdiversity MODelling 2）平臺上同時完成10種模擬方法之分析，TSS、ROC、KAPPA指標評估結果顯示以GAM、RF、MAXENT方法模擬效果較佳，而FDA、SRE模擬效果較差，但實繪10種方法之預測機率圖，則發現不同方法所得之大致輪廓是相似的且符合現今臺灣水青岡之分布，同時為降低單一SDM方法的誤差，本文對預測較佳之模型進行整體模擬，藉由比較臺灣水青岡現今分布與潛在適合生育地之差異，有助於未來探討其與常綠闊葉樹之競爭、氣候變遷影響等問題。 Species distribution modeling (SDM) is a numerical tool that combines species occurrence data with environmental predictor layers to explore speces-environment relationship. The SDM methods had greatly advanced in recent years. In this study, we use 10 SDM methods on BIOMOD2 platform to model the distribution of Fagus hayatae with 7 environmental variables. The results reveal that GAM, RF, and MAXENT gain the best performance, whereas FDA and SRE have the worst performance based on the TSS, ROC, and KAPPA evaluation. The general patterns of 10 SDM predictive maps are same and coincident with the current distribution of F. hayatae. For reducing the uncertainty of individual model, we perform the ensemble modeling by BIOMOD2. To compare the current and predicted distribution of F. hayatae could facilitate the future works such as species competition and climate change effects.