基于最大熵模型预测青海省莱姆病的地理盼布

Prediction of potential geographic distribution of Lyme disease in Qinghai province with Maximum Entropy model

目的应用最大熵模型（MaxEnt）预测青海省莱姆病的分布。方法查阅1990年以来青海省人群莱姆病的血清检测数据，共收集到6个县（互助、泽库、同德、大通、祁连、循化）的血清学结果，将互助、泽库、同德县的血清学检测结果以及青海省环境和人为活动数据[包括海拔、人口足迹、归一化植被指数（NDVI）、温度等]导入MaxEnt软件，分析环境以及人为条件适宜的莱姆病潜在地理分布，然后以大通县、祁连县、循化县血清学数据作为验证数据，与模型预测结果进行比较。结果MaxEnt预测结果显示，青海省存在3个莱姆病的热点区域，主要分布在青海省东部林区。在相关影响因子当中，NDVI对于模型的贡献最大；其次是人口足迹。用于验证模型的大通县、祁连县和循化县均分布于青海省东部。其中，循化县位于热点区域Ⅱ中，而大通县紧邻热点区域Ⅲ，位于热点区域Ⅲ的北部地区，祁连县不在预测的热点区域中，且莱姆病血清阳性率在3个调查县中为最低。模型运行良好[曲线下面积（AUC）=0．980]。结论实际的人群莱姆病血清学数据与模型预测结果基本吻合。MaxEnt模型可用于莱姆病风险分布的预测。植被和人为活动可能与莱姆病传播有关。

Objective To predict the potential geographic distribution of Lyme disease in Qinghai by using Maximum Entropy model （MaxEnt）. Methods The sero-diagnosis data of Lyme disease in 6 counties （Huzhu, Zeku, Tongde, Datong, Qilian and Xunhua） and the environmental and anthropogenic data including altitude, human footprint, normalized difference vegetation index （NDVI） and temperature in Qinghai province since 1990 were collected. By using the data of Huzhu Zeku and Tongde, the prediction of potential distribution of Lyme disease in Qinghai was conducted with MaxEnt. The prediction results were compared with the human sero-prevalence of Lyme disease in Datong, Qilian and Xunhua counties in Qinghai. Results Three hot spots of Lyme disease were predicted in Qinghai, which were all in the east forest areas. Furthermore, the NDVI showed the most important role in the model prediction, followed by human footprint. Datong, Qilian and Xtmhua counties were all in eastern Qinghai. Xunhua was in hot spot area II, Datong was close to the north of hot spot area III, while Qilian with lowest sero-prevalence of Lyme disease was not in the hot spot areas. The data were well modeled in MaxEnt （Area Under Curve：0.980）. Conclusions The actual distribution of Lyme disease in Qinghai was in consistent with the results of the model prediction. MaxEnt could be used in predicting the potential distribution patterns of Lyme disease. The distribution of vegetation and the range and intensity of human activity might be related with Lyme disease distribution.