优于单一最大熵生态位模型的混合烟粉虱生境风险评估模型

预测入侵生物的潜在地理分布、快速评估其高脆弱性区域是实现入侵生物前瞻性风险预警的重要手段。MaxEnt生态位模型是目前应用最广泛的生境风险评估方法，操作简单，预测精度较高，但模型对数据的质量和数量过分依赖。以烟粉虱Bemisia tabaci为对象，引入地理探测器显式描述评价因子的空间关联规律和贡献度，结合MaxEnt生态位模型，提出一种混合生境风险评估模型(Geo-MaxEnt),并与单一MaxEnt生态位模型进行对比。结果表明：(1)地理探测器模型显示，海拔(0.56)、土地利用(0.43)、最热月最高温度(0.36)和年平均温度(0.30),对烟粉虱的空间分布具有显著影响，各因子对烟粉虱生境的影响存在显著的差异。海拔和土地利用PD值最高，是影响烟粉虱生境的主要驱动因子。环境因子的交互作用强化了各个因子的影响力。(2)单一MaxEnt生态位和所构建的模型总体精度分别是94.86%(AUC 0.91)和98.13%(AUC 0.94),相较之下，所构建的模型精度略高，表明所构建的模型是合理的，具有高度的可靠性；(3)对于高风险区，混合模型优于MaxEnt模型，但两者在空间分布方面高度...

A habitat risk assessment model for mixed Bemisia tabaci that outperforms a single MaxEnt

Predicting the potentially geographical distribution of invasive species and rapidly assessing their high vulnerability areas are important means to realize the prospective risk warning of invasive species. At present, the MaxEnt niche model is the most widely used habitat risk assessment method with simple operation and high prediction accuracy, but it is too dependent on the quality and quantity of data. This study took Bemisia tabaci as the object, introduced geographic detectors to explicitly describe the spatial correlation and contribution of evaluation factors, and combined with the MaxEnt niche model to propose a mixed habitat risk assessment model (Geo-Maxent) and compare with the single MaxEnt niche model. The results showed that:(1) Geo-detector model showed that altitude (0.56), land use (0.43), the max temperature in the warmest month (0.36) and annual average temperature (0.30) had significant effects on the spatial distribution of Bemisia tabaci, and there were significant differences in the effects of various factors on the habitat of Bemisia tabaci. In addition, the PD(Factor detection) values of altitude and land use were high, indicating that altitude and land use are the main driving factors affecting the habitat of Bemisia tabaci. Besides, the interaction of environmental factors strengthened the influence of each single factor. (2) The overall accuracy of the single MaxEnt niche and the constructed model was 94.86% (AUC is 0.91) and 98.13% (AUC is 0.94), respectively, indicating that the accuracy of the constructed model was higher. (3) As for the high risk regions, the mixed model is better than the MaxEnt model, but the spatial distribution is highly consistent, mainly in the eastern region. As for the unsuitable regions, the results of the two models are consistent, and the unsuitable regions of the MaxEnt model are slightly larger than the mixed model. Geographic detectors can explain the interaction and synergistic effects of invasion driving factors, and can better express the ecological significance of habitat suitability and candidate factors of invasive insects, which is more effective in evaluating the habitat risk of invasive insects.