基于组合模型和GIS的白纹伊蚊广东省适生区预测

Predicting the potential distribution of Aedes albopictus in Guangdong Province based on the combined models and GIS

【目的】白纹伊蚊入侵力极强,也是广东省登革热传播的主要媒介,预测其在广东省的适生区可为制定防疫策略提供科学依据。【方法】针对传统方法没有考虑环境因子权重的问题,本研究构建了4种适生区预测组合模型。首先采用4种因子权重模型对相关性分析筛选出的环境因子进行权重划分,然后分别与相似离度值公式结合,最后基于GIS技术对白纹伊蚊在广东省的适生区进行预测。【结果】精度验证和适生区预测分布表明,地理探测器与相似离度模型组合的模型预测精度最高,AUC平均值为0.944,标准差为0.008,预测的白纹伊蚊入侵低风险地区处于广东省北部,占广东省总面积的4.05%,绝大部分地区处于中风险地区和中高风险地区,占广东省总面积的85%以上,而广东省中部的广州、佛山和东莞等地处于高风险地区,占广东省总面积的8.77%。【结论】与不考虑因子权重的相似离度模型相比,考虑因子权重的组合模型能有效提高适生区预测精度,其中地理探测器模型通过探究空间异质性划分因子权重,比传统统计学模型效果好,其组合模型适生区预测精度最高。

【Aim】Aedes albopictus is a highly invasive mosquito and the main medium of dengue contagion in Guangdong Province,South China;its potential distribution in this region can provide scientific evidence for the establishment of epidemic prevention strategies.【Method】Traditional methods do not take into account the relative importance of environmental factors.To address this problem,four combined models for predicting potential distribution were constructed:combinations of the four factor⁃weighting mod⁃els(geographic detector,multivariable linear regression,principal component analysis,and factor analysis)and the analogy devia⁃tion model.The weights of environmental factors screened by correlation analysis were first divided by the four factor⁃weighting mod⁃els and then combined with the analogy deviation value formula.Finally,the potential distribution areas of A.albopictus in Guang⁃dong Province in southern China were predicted using GIS technology.【Result】The accuracy verification and prediction of the four combined models showed that the accuracy of the combination of the geographic detector model and the analogy deviation model(GDM⁃ADM)was the highest,with a mean AUC of 0.944 and a standard deviation of 0.008.Meanwhile,the GDM⁃ADM model prediction showed that the areas with a lower risk of invasion of A.albopictus were in the northern part of Guangdong Province,ac⁃counting for 4.05%of the total area,and most areas were under medium or medium⁃high risk,accounting for more than 85%of the total area,while Guangzhou,Foshan,and Dongguan in central Guangdong Province were at high risk,accounting for 8.77%of the total area.【Conclusion】Compared with the analogy deviation model,which does not consider the weights of the factors,the com⁃bined models that consider factor weights can effectively improve the accuracy of the prediction of potentially suitable areas for mos⁃quitoes.The geographic detector model divides the weights of the factors by exploring spatial heterogeneity,which is more effective than traditional statistical models,and its combined model has the highest accuracy for the prediction of the potential distribution ar⁃eas of A.albopictus.