How niche mismatches impair our ability to predict potential invasions@en

Accurately anticipating potential invasions is a crucial step to deal with invasive species. The most commonly used method to identify suitable areas for invasion is to build Ecological Niche Models (ENMs); nevertheless, the accuracy of these models might be highly overestimated. We explored how biogeographical, biological and methodological factors might affect our ability to identify suitable areas for invasions. We created virtual species to control and explore disagreements between the fundamental and the available niche in the native region. First, we explored how differences in species characteristics (climatic tolerance and dispersal capacity) might hinder our ability to predict invasions with ENMs. We also evaluated how different algorithms behave under those differences. Furthermore, we measured how prediction accuracy varies throughout the world, by evaluating the degree of niche mismatch in each zoogeographic region. In general, ENMs predictions were more accurate for species with broad tolerance, while species with narrow tolerance had high levels of overprediction. We were able to distinguish algorithms by how much they under or overestimate species fundamental niche. Some zoogeographic regions are indeed more error-prone, as levels of climatic incompleteness and fundamental niche representation within the distribution vary throughout the world. We demonstrate that predicting potential invasions with ENMs might incur in wrongful niche estimation, therefore, missing potential locations for invasion. Biological factors such as species tolerance and dispersal capacity must be considered when discussing model uncertainty. Besides, algorithms had different predictions, with high levels of under or over prediction. Therefore, when it comes to invasions predictions, we must be meticulous, considering that the fundamental niche might not be adequately expressed within the native region.精准预判潜在入侵事件，是应对外来入侵物种的关键环节。当前用于识别物种潜在入侵适宜分布区的主流方法，是构建生态位模型（Ecological Niche Models, ENMs）；然而此类模型的预测精度往往存在被严重高估的问题。本研究探讨了生物地理、生物学以及方法学等多类因素，对我们利用ENMs识别入侵适宜分布区能力的影响。我们通过构建虚拟物种，实现对原生分布区内基础生态位与有效生态位之间偏差的控制与探究。首先，我们分析了物种特征（气候耐受范围与扩散能力）的差异，如何阻碍ENMs对入侵事件的预测精度。同时，我们还评估了不同建模算法在上述物种特征差异下的表现情况。此外，我们通过评估各动物地理区内的生态位错配程度，量化了全球范围内ENMs预测精度的空间异质性。总体而言，气候耐受范围较广的物种，其ENMs预测精度更高；而耐受范围较窄的物种则容易出现严重的过度预测现象。我们可根据不同算法对物种基础生态位的高估或低估程度，对其进行性能区分。部分动物地理区确实更容易出现预测误差，这是因为全球范围内各区域的气候数据完备度以及分布区内基础生态位的表征程度存在显著差异。本研究证实，利用ENMs预测潜在入侵事件时，可能会出现生态位估计偏差的问题，进而遗漏真正的潜在入侵分布区。在探讨模型预测不确定性时，必须纳入物种耐受范围、扩散能力等生物学因素。此外，不同建模算法的预测结果存在显著差异，部分算法会出现严重的预测不足或过度预测问题。因此，在开展入侵物种预测研究时，我们必须保持严谨，需考虑到原生分布区内的基础生态位可能无法得到充分表征这一关键前提。