以物種分布模型推估多樣性熱點——評「生物多樣性熱點之推估：以台灣特有鳥種為例」

Using species distribution models to estimate biodiversity hotspots—Comment on 'Estimating Biodiversity Hotspot: A Case Study of Endemic Bird Species in Taiwan'

多樣性分布預測與熱點推估是研擬生物多樣性保育策略的首要工作。呂（2016）提出一套以物種分布模型的分布機率值推估生物多樣性熱點的分析方法，並以臺灣特有鳥種為範例。但是我們發現該方法不論在生態學學理或科學邏輯方面均有嚴重瑕疵，對學術研究或保育政策擬定可能產生嚴重誤導。我們以臺灣繁殖鳥類大調查計畫（BBS Taiwan）的特有鳥種分布資料，重複呂的推估方法。我們也以同樣資料，但以物種的二位元分布模型疊合得到多樣性分布圖，再以之推估熱點。我們以 BBS Taiwan 資料的已知熱點，檢驗兩推估方法得到的熱點之正確性，實證呂方法的謬誤。我們認為呂（2016）的方法不應該再被使用於任何多樣性熱點分析或棲地威脅度評估。 Predicting biodiversity distributions and identifying the hotspots are essential in order to design strategies for biodiversity conservation. Lu (2016) proposed a process to estimate biodiversity hotspots based on the probabilities computed from species distribution models (SDM), using Taiwan's endemic bird species as an example. However, we found that Lu's methods were seriously flawed in ecological concepts and scientific logic, and could be misleading in the formulations of academic research or conservation policies. We used a dataset on endemic bird species from the Taiwan Breeding Bird Survey project to build two biodiversity maps: one based on Lu's process and the other one based on binary (i.e., presence/absence) SDMs, a general method reported in the literature. We tested the predicted hotspots from both maps against a set of known hotspots, and found that Lu's methods severely underperformed compared to the general method. We therefore urge that the methods of Lu (2016) should not to be used anymore in biodiversity hotspot analyses or habitat-risk assessments.

Predicting biodiversity distributions and identifying the hotspots are essential in order to design strategies for biodiversity conservation. Lu (2016) proposed a process to estimate biodiversity hotspots based on the probabilities computed from species distribution models (SDM), using Taiwan's endemic bird species as an example. However, we found that Lu's methods were seriously flawed in ecological concepts and scientific logic, and could be misleading in the formulations of academic research or conservation policies. We used a dataset on endemic bird species from the Taiwan Breeding Bird Survey project to build two biodiversity maps: one based on Lu's process and the other one based on binary (i.e., presence/absence) SDMs, a general method reported in the literature. We tested the predicted hotspots from both maps against a set of known hotspots, and found that Lu's methods severely underperformed compared to the general method. We therefore urge that the methods of Lu (2016) should not to be used anymore in biodiversity hotspot analyses or habitat-risk assessments.