Data from: No silver bullets in correlative ecological niche modeling: insights from testing among many potential algorithms for niche estimation

The field of ecological niche modeling or species distribution modeling
has seen enormous activity and attention in recent years, in light of
exciting biological inferences that can be drawn from correlational models
of species’ environmental requirements (i.e., ecological niches) and
inferences of potential geographic distributions. Among the many methods
used in the field, one or two are in practice assumed to be ‘best’ and are
used commonly, often without explicit testing. We explore herein
implications of the “No Free Lunch” theorem, which suggests that no single
optimization approach will prove to be best under all circumstances: we
developed diverse virtual species with known niche and dispersal
properties to test a suite of niche modeling algorithms designed to
estimate potential areas of distribution. The result was that (1) indeed,
no single ‘best’ algorithm was found, and (2) different algorithms perform
in very different manners depending on the particularities of the virtual
species. The conclusion is that niche or distribution modeling studies
should begin by testing a suite of algorithms for predictive ability under
the particular circumstances of the study, and choose an algorithm for a
particular challenge based on the results of those tests. Studies that do
not take this step may use algorithms that are not optimal for that
particular challenge.

生态位模型（ecological niche modeling）与物种分布模型（species distribution modeling）领域近年来受到广泛关注与大量研究投入，这是因为通过物种环境需求（即生态位）的相关模型，可获取极具价值的生物学推论，并据此推断物种潜在的地理分布范围。该领域现有诸多建模方法，但实际应用中常默认一至两种方法为“最优”，且往往未经过明确验证便被频繁使用。
本文探讨了“没有免费的午餐”（No Free Lunch）定理的相关启示，该定理指出不存在适用于所有场景的通用最优优化方法：我们构建了一系列生态位与扩散特性已知的虚拟物种（virtual species），以此测试一组旨在估算物种潜在分布范围的生态位建模算法。
研究结果显示：其一，确实不存在通用的“最优”算法；其二，不同算法的表现因虚拟物种的具体特性差异显著。
据此得出结论：生态位或物种分布建模研究应首先针对研究的具体场景，测试一组算法的预测性能，并基于测试结果为特定研究任务选择合适的算法；未遵循该流程的研究，可能会选用不适用于该特定任务的算法。