Forecasting the effects of climate change on species and populations is a fundamental goal of conservation biology, especially for montane endemics which seemingly are under the greatest threat of extinction given their association with cool, high elevation habitats. Species distribution models (also known as niche models) predict where on the landscape there is suitable habitat for a species of interest. Correlative niche modeling, the most commonly employed approach to predict species' distributions, relies on correlations between species' localities and current environmental data. This type of model could spuriously forecast less future suitable habitat because species' current distributions may not adequately represent their thermal tolerance, and future climate conditions may not be analogous to current conditions. We compared the predicted distributions for three montane species of Plethodon salamanders in the southern Appalachian Mountains of North America using a correlative modeling approach and a mechanistic model. The mechanistic model incorporates species‐specific physiology, morphology and behavior to predict an annual energy budget on the landscape. Both modeling approaches performed well at predicting the species' current distributions and predicted that all species could persist in habitats at higher elevation through 2085. The mechanistic model predicted more future suitable habitat than the correlative model. We attribute these differences to the mechanistic approach being able to model shifts in key range‐limiting biological processes (changes in surface activity time and energy costs) that the correlative approach cannot. Choice of global circulation model (GCM) contributed significantly to distribution predictions, with a tenfold difference in future suitability based on GCM, indicating that GCM variability should be either directly included in models of species distributions or, indirectly, through the use of multi‐model ensemble averages. Our results indicate that correlative models are over‐predicting habitat loss for montane species, suggesting a critical need to incorporate mechanisms into forecasts of species' range dynamics.

中文翻译：

---

天空中消失的岛屿？气候变化下山地蝾螈范围动态的相关性和机制预测比较

预测气候变化对物种和种群的影响是保护生物学的一个基本目标，特别是对于山地地方性动物而言，鉴于它们与凉爽、高海拔的栖息地有关，它们似乎面临着最大的灭绝威胁。物种分布模型（也称为生态位模型）可预测景观上的何处适合感兴趣的物种栖息。相关生态位建模是预测物种分布的最常用方法，它依赖于物种位置与当前环境数据之间的相关性。这种类型的模型可能会虚假地预测未来不太适合的栖息地，因为物种当前的分布可能无法充分代表它们的耐热性，而且未来的气候条件可能与当前条件不相似。我们使用相关建模方法和机械模型比较了北美阿巴拉契亚山脉南部三种山地蝾螈物种的预测分布。机械模型结合了物种特定的生理学、形态学和行为来预测景观的年度能量收支。这两种建模方法在预测物种当前的分布方面表现良好，并预测所有物种都可以在更高海拔的栖息地中持续到 2085 年。机械模型比相关模型预测了更多的未来合适的栖息地。我们将这些差异归因于机械方法能够模拟关键范围限制生物过程的变化（表面活性时间和能量成本的变化），而相关方法则不能。全球环流模型 (GCM) 的选择对分布预测做出了显着贡献，基于 GCM 的未来适宜性差异为 10 倍，表明 GCM 变异性应直接包含在物种分布模型中，或者通过使用多模型间接包含整体平均。我们的结果表明，相关模型过度预测了山地物种的栖息地丧失，这表明迫切需要将机制纳入物种范围动态的预测。