應用最大熵物種分布模式與衛星影像預測嘉義地區沙氏變色蜥之分布

沙氏變色蜥(Anolis sagrei)列名在已入侵外來種處理分級名單中的優先管理防制物種，指出沙氏變色蜥可能會是未來快速擴張的入侵物種。物種分布模式(Species distribution models, SDMs)對於探索未知的生物分布上，展現強大的空間預測能力；而遙測影像偵測空間異質性，在管理生物資源上有重要價值。本研究預期繪出嘉義地區海拔500公尺以下沙氏變色蜥的潛在棲息地，並從選擇的變數中解釋沙氏變色蜥偏好的棲地特性。 2018年7月在嘉義縣水上鄉三界埔的樣區內，收集了20個沙氏變色蜥出現點位作為訓練樣本，並用全球生物多樣性資訊機構下載的出現記錄作為樣區的測試用資料。環境變數採用ASTER數值高程模型、WorldClim氣候資料、Landsat 8衛星影像所產生的28個空間圖層，作為MaxEnt最大熵物種分布模式的參數。最後選擇高程值、亮度溫度、土地覆蓋分類、NDVI、坡度作為模式的預測變數。訓練樣本模式AUC值0.989，顯示模式具有非常好的判別力。本研究從DEM的結果，推論沙氏變色蜥偏好高溫而較乾燥的環境；大氣層頂亮度溫度顯示沙氏變色蜥偏好高溫環境；土地覆蓋的結果顯示沙氏變色蜥在所有類型的棲地都可能成攻入侵；NDVI發現植被覆蓋不良與茂盛，都會使沙氏變色蜥的出現機率下降。本研究結果提供沙氏變色蜥的空間分布圖，理解沙氏變色蜥所偏好的棲地特性，可以做為後續入侵生物經營空間管理所需的基礎。另一方面，遙測影像對於在物種分布模式上有改進效果，可以作為未來高解析度的物種分布研究的參考。

關鍵字

最大熵物種分布模式；衛星影像；入侵種；沙氏變色蜥

並列摘要

The brown anole (Anolis sagrei) has been rapidly expanding in Taiwan in the last two decades. However, its potential distribution is still not clear. Most of the lizard distribution studies use interpolated climatic data which present average states in spatial and temporal scale. We use remote sesning data to keep spatial heterogeneity that can’t be shown in interpolated climatic data. The first aim of this study is to determine the potential habitats of the brown anole less than 500 meters above sea level in Chiayi area. Second, we are wondering the properties of the habitat which affect the probability of occurrence. We collected 20 the brown anole occurrences with stratified random sampling based on land cover classification in Santzepu, Chiayi County in July, 2018. 29 environmental coverages are assembled from ASTER GDEM, WorldClim climatic variables, Landsat 8 image. Predictions are generated in MaxEnt and tested by GBIF occurrences t. The final model used elevation, thermal infrared, land cover, NDVI and slope. The AUC value is 0.989 which means the predictions is robust. The result shows that the brown anole is limited less than 100 meters above sea level, which means it prefer higher temperature and less rainfall. The thermal band also shows the brown anole prefer higher temperature. Land cover and NDVI performs that the brown anole is a generalist, but prefer open vegetation sites. In conclusion, the result can be the spatial basis of the invasive species management. We also prove the remote sensing data can provide the instant land surface states such as the temperature and the vegetation cover that can refine the predictions of species distribution modeling in lizard species.