MaxEnt模型参数设置对其所模拟物种地理分布和生态位的影响——以茶翅蝽为例

【目的】生态位模型被广泛应用于入侵生物学和保护生物学研究,现有建模工具中,MaxEnt是最流行和运用最广泛的生态位模型。然而最近研究表明,基于MaxEnt模型的默认参数构建模型时,模型倾向于过度拟合,并非一定为最佳模型,尤其是在处理一些分布点较少的物种。【方法】以茶翅蝽为例,通过设置不同的特征参数、调控倍频以及背景拟不存在点数分别构建茶翅蝽的本土模型,然后将其转入入侵地来验证和比较模型,通过检测模型预测的物种对环境因子的响应曲线、潜在分布在生态空间中的生态位映射以及潜在分布的空间差异性,探讨3种参数设置对MaxEnt模型模拟物种分布和生态位的影响。【结果】在茶翅蝽的案例分析中,特征参数的设置对MaxEnt模型所模拟的潜在分布和生态位的影响最大,调控倍频的影响次之,背景拟不存在点数的影响最小。与其他特征相比,基于特征H和T的模型其响应曲线较为曲折;随着调控倍频的增加,响应曲线变得圆滑。【结论】在构建MaxEnt模型时,需要从生态空间中考虑物种的生态需求,分析模型参数对预测物种分布和生态位可能造成的影响。

Effects of model parameters in MaxEnt modeling of ecological niche and geographic distribution: case study of the brown marmorated stink bug, Halyomorpha haly

[Aim] Ecological niche modeling is widely used in studies of biological invasion and conservation. Among available models, MaxEnt is the most popular and widely used model. Recent studies, nonetheless, showed that MaxEnt models tend to overfitting, and it may not be the ideal model when using default setting, especially for species with small sampling size.[Method] We investigated the influence of three parameters (i.e. feature combination, regularization multiplier, and maximum number of background points) on MaxEnt modeling of ecological niche and geographic distributions using the marmorated stink bug (Halyomorpha halys, Heteroptera) as a case study. Native niche models were calibrated for H. halys in eastern Asia and transferred worldwide for comparison and evaluation. Response curves of the species to environmental variables, projections of potential distribution in ecological spaces, and the variation of these potential distributions were used to investigate the effects.[Result] Feature combinations had significant effects on MaxEnt modeling outcomes of ecological niches and geographic distributions, followed by the regularization multiplier; the maximum number of background points had the least effect. We found that response curves based on features of hinge and threshold were more complicated than that based on the others; the response curves became more smooth as the regularization multipliers increased.[Conclusion] Ecological requirement of the species in its ecological space should be considered when fitting a MaxEnt model, the influence of the Maxent parameters on modeling ecological niches and geographical distributions should also be considered, rather than simply based on default settings.