優勢度於樣點資料對物種分布模型之影響

The Effects of Dominance in Sampling Data on Species Distribution Modelling

林政道(Cheng-Tao Lin)

指導教授 ： 邱祈榮

國立臺灣大學/生物資源暨農學院/森林環境暨資源學研究所/碩士(2009年)

https://doi.org/10.6342/NTU.2009.02590

摘要

近年來許多研究報告指出物種分布模型會受到資料以及物種特性的影響。而一個物種的優勢度往往表示這個物種的族群能夠成功建立更新的指標，此外低優勢度的棲地也許對於分布模型來說具有雜訊，並影響物種分布模型的表現。因此在本篇研究中，我們將探討以下兩個問題：將較低優勢度資料移除是不是會增加物種分布模型的表現？在取樣資料中的優勢度對於物種分布模型來說是不是一個具有影響力的因素？

臺灣鐵杉在臺灣主要是分布在海拔兩千至三千公尺廣泛分布的物種，而本研究中將臺灣鐵杉選定為目標物種。在本篇研究中，總共有兩個測試模式，第一個測試模式使用重要值指數(IVI)來選擇出現點位，並根據優勢度將樣點切分為遞增及遞減兩個累積相對優勢度資料集。第二個測試模式則是根據對數胸高斷面積來選擇出現點位，並根據優勢度將樣點切分為遞增及遞減兩個累積相對優勢度資料集。切分完測試模式後，使用 GAM 和 MAXENT 來進行物種分布模式的測試。

在第一個測試模式中，在遞減累積相對優勢度資料集內若逐漸去除較高的優勢度資料，兩個物種分布模式的 AUC 值會逐漸下降。相對於遞減累積相對優勢度資料集，遞增累積相對優勢度資料集則無此趨勢。在第二個測試模式中，不管是遞增或遞減的資料集則無明顯的差異。儘管不同的優勢度之下，整體來說 MAXENT 的表現比 GAM 還要來的好一些。我們的研究結果並指出在樣點資料的優勢度會對物種分布模型的表現造成影響。

關鍵字

優勢度 ； 臺灣鐵杉 ； GAM ； MAXENT ； 物種分布模型(SDM)

並列摘要

It has been reported that the performance of species distribution models are related with properties of data and species traits. The dominance of a species in a habitat represents the successfulness of regeneration of a population there and thereby may be associated with the probability of species occurrence. Habitats with low dominance of a species may be a noise for modelling, which might reduce the accuracy of SDMs. Here we would like propose two questions: Does removal of low dominance data increase the accuracy of SDMs? Is species dominance an influential factor for SDMs?

Tsuga chinensis var. formosensis, a native conifer species which is widely distributed in habitats ranging from 2000 m to 3100 m above sea level in Taiwan, was selected for modelling. Two scenarios were evaluated for testing the dominance effects in sampling data. The first scenario used IVI to select presence data according to the dominance and the sampling plots were divided into ascendant and descendant accumulative datasets. The second scenario used logarithm basal area to select presence data and the sampling plots were also divided into ascendant and descendant accumulative datasets. GAM and MAXENT were both used for building the models.

In the first scenario, AUC values of the two models decrease while gradually removing higher dominance datasets in the descendant accumulative datasets. In contrary, removal of low dominance data in ascendant accumulative datasets does not increase the accuracy of the two models. Similarity, in the second scenario, there are no significant differences amongst ascendant and descendant datasets of the two models. Regardless of various dominance levels of data, the accuracy of prediction of MAXENT is slightly higher than that of GAMs. Our result shows dominance in sampling data would affect the performance of species distribution modelling.

並列關鍵字

dominance ； Tsuga chinensis var. formosensis ； GAM ； MAXENT ； species distribution modelling(SDM)

參考文獻

Araujo, M. and Williams, P. (2000). Selecting areas for species persistence using occurrence data. Biological Conservation, 96(3):331–345.

Araujo, M. B. and Guisan, A. (2006). Five (or so) challenges for species distribution modelling. Journal of Biogeography, 33(10):1677–1688.

Austin, M. (2007). Species distribution models and ecological theory: A critical assessment and some possible new approaches. Ecological Modelling, 200(1-2):1–19.

Austin, M. P., Belbin, L., Meyers, J. A., Doherty, M. D., and Luoto, M. (2006). Evaluation of statistical models used for predicting plant species distributions: Role of artificial data and theory. Ecological Modelling, 199(2):197–216.

Congalton, R. G. and Green, K. (1999). Assessing the accuracy of remotely sensed data: Principles and practices. page 137. Lewis Publishers, Boca, Raton.