全球变化知识资源中心

MaxEnt, a commonly used approach of species distribution modelling, is widely used to predict plant invasion at the large spatial scale based on occurrence records and environmental variables. However, the number of occurrence records, number of environmental variables, and spatial scales have a large potential to affect the ability of MaxEnt to predict invasive plant distributions. In this study, we used the area under the curve (AUC) of the receiver operator characteristics as an indicator of MaxEnt performance, and evaluated the effects of the number of occurrence records, number of environmental variables, and spatial scales on MaxEnt distribution modelling of invasive plants based on 1015 cases of invasive plants. Next, we suggested improvements for model performance. We found significant relationships between the AUC and the above-mentioned modelling parameters. Furthermore, we determined the relevant threshold values for the available MaxEnt models (i.e. AUC >0.7). We suggested using an appropriate number of occurrence records and environmental variables (e.g. >5) and covered cell sizes of 5.0 arc-min to model the distributions of invasive plants on the global scale. Our study provides practical references using MaxEnt to prevent and control plant invasion under global changes and contributes to the exploration of species distribution modelling mechanisms.