Rapid global change has increased interest in developing ways to identify suitable refugia for species of conservation concern. Correlative and mechanistic species distribution models (SDMs) represent two approaches to generate spatially-explicit estimates of climate vulnerability. Correlative SDMs generate distributions using statistical associations between environmental variables and species presence data. In contrast, mechanistic SDMs use physiological traits and tolerances to identify areas that meet the conditions required for growth, survival and reproduction. Correlative approaches assume modeled environmental variables influence species distributions directly or indirectly; however, the mechanisms underlying these associations are rarely verified empirically. We compared habitat suitability predictions between a correlative-only SDM, a mechanistic SDM and a correlative framework that incorporated mechanistic layers (‘hybrid models'). Our comparison focused on green salamanders Aneides aeneus, a priority amphibian threatened by climate change throughout their disjunct range. We developed mechanistic SDMs using experiments to measure the thermal sensitivity of resistance to water loss (r_i) and metabolism. Under current climate conditions, correlative-only, hybrid and mechanistic SDMs predicted similar overlap in habitat suitability; however, mechanistic SDMs predicted habitat suitability to extend into regions without green salamanders but known to harbor many lungless salamanders. Under future warming scenarios, habitat suitability depended on climate scenario and SDM type. Correlative and hybrid models predicted a 42% reduction or 260% increase in area considered to be suitable depending on the climate scenario. In mechanistic SDMs, energetically suitable habitat declined with both climate scenarios and was driven by the thermal sensitivity of r_i. Our study indicates that correlative-only and hybrid approaches produce similar predictions of habitat suitability; however, discrepancies can arise for species that do not occupy their entire fundamental niche, which may hold consequences of conservation planning of threatened species.

中文翻译：

---

将生理学整合到相关模型中可以改变对受威胁两栖动物在气候变化下栖息地适宜性的预测

快速的全球变化增加了人们对开发方法为受保护物种确定合适避难所的兴趣。相关和机械物种分布模型 (SDM) 代表了两种生成气候脆弱性空间明确估计的方法。相关 SDM 使用环境变量和物种存在数据之间的统计关联生成分布。相比之下，机械 SDM 使用生理特征和耐受性来识别满足生长、生存和繁殖所需条件的区域。相关方法假设模拟的环境变量直接或间接影响物种分布；然而，这些关联背后的机制很少得到经验验证。我们比较了仅相关 SDM 之间的栖息地适宜性预测，一个机械 SDM 和一个包含机械层（“混合模型”）的相关框架。我们的比较集中在绿色蝾螈上Aneides aeneus，一种在其分离范围内受到气候变化威胁的优先两栖动物。我们通过实验开发了机械 SDM，以测量抗失水性的​​热敏感性 ( r _i) 和新陈代谢。在当前的气候条件下，仅相关的、混合的和机械的 SDM 预测栖息地适宜性的类似重叠；然而，机械 SDM 预测栖息地的适宜性将扩展到没有绿蝾螈但已知有许多无肺蝾螈的地区。在未来变暖情景下，栖息地适宜性取决于气候情景和 SDM 类型。相关模型和混合模型预测，根据气候情景，被认为合适的面积将减少 42% 或增加 260%。在机械 SDM 中，能量适宜的栖息地随两种气候情景而下降，并且是由r _{i的热敏感性驱动的}. 我们的研究表明，仅相关和混合方法对栖息地适宜性产生了相似的预测；然而，对于不占据其整个基本生态位的物种，可能会出现差异，这可能会对受威胁物种的保护计划产生影响。