排序(ordination)是探討生物群聚與環境之間關係常用的多變數統計方法。本研究以模擬數據及鳥類群聚資料,探討主成分分析(principal components analysis, PCA)、對應分析(correspondence analysis, CA)、以及降趨對應分析(detrended correspondence analysis, DCA)三種常用的排序法在不同群聚變異梯度長度下的排序表現。結果發現三種排序法適用的群聚梯度長度不同。DCA運處結果的軸長可作爲群聚梯度長度判斷的參考;梯度軸長在2SD以上,且CA因爲物種沿環境梯度的非線性分布而産生明顯拱形效應(arch effect)時,以DCA的結果較佳;但若樣點在CA的空間未有明顯拱形排列時,則以使用CA較適當。車長在2SD以下時,以使用CA及PCA較佳,DCA則可能因爲降趨(detrending)及重新刻劃(rescaling)的程序,而嚴重扭曲數據原本蘊含的生態意義。排序方法的選擇應考慮生物群聚之梯度長度,以獲得合理的結果。
Ordination is the collective term for multivariable techniques that elucidate the variation in communities and detect relationships between community gradients and environmental factors. We used artificial and avian communities to investigate the performance of 3 ordination techniques of principal components analysis (PCA), correspondence analysis (CA), and detrended correspondence analysis (DCA) for different gradient lengths of communities. Results suggest that applicability of ordination methods varies according to gradient length. If the lengths of the ordination axes are greater than about 2 standard deviations (SD), the results of the CA may display the arch effect, and one should consider using the DCA. However, if the configurations of site points of the CA do not display this effect, one should use the CA. If the ordination lengths are less than about 2SD, the DCA may destroy the ecologically meaningful in formation by detrending and rescaling, and therefore one should consider using the PCA of CA. The length of the gradient strongly influences the correctness of the ordinations; investigators should carefully choose an appropriate method according to gradient lengths.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。