Abstract: Objective To explore the technique of maximum entropy model for extracting Oncomelania hupensis snail habitats in Poyang Lake zone. Methods The information of snail habitats and related environment factors collected in Poyang Lake zone were integrated to set up the maximum entropy based species model and generate snail habitats distribution map. Two Landsat 7 ETM+ remote sensing images of both wet and drought seasons in Poyang Lake zone were obtained， where the two indices of modified normalized difference water index （MNDWI） and normalized difference vegetation index （NDVI） were applied to extract snail habitats. The ROC curve， sensitivities and specificities were applied to assess their results. Furthermore， the importance of the variables for snail habitats was analyzed by using Jackknife approach. Results The evaluation results showed that the area under receiver operating characteristic curve （AUC） of testing data by the remote sensing?based method was only 0.56， and the sensitivity and specificity were 0.23 and 0.89 respectively. Nevertheless， those indices above?mentioned of maximum entropy model were 0.876， 0.89 and 0.74 respectively. The main concentration of snail habitats in Poyang Lake zone covered the northeast part of Yongxiu County， northwest of Yugan County， southwest of Poyang County and middle of Xinjian County， and the elevation was the most important environment variable affecting the distribution of snails， and the next was land surface temperature （LST）. Conclusions The maximum entropy model is more reliable and accurate than the remote sensing?based method for the sake of extracting snail habitats， which has certain guiding significance for the relevant departments to carry out measures to prevent and control high?risk snail habitats.

Key words: Oncomelania hupensis； Poyang Lake； Maximum entropy model； Remote sensing

摘要： 目的 探索最大熵模型在提取钉螺孳生地方面的应用。 方法 根据鄱阳湖地区采集的钉螺孳生地数据和相关环境因素构建最大熵生态位模型并生成鄱阳湖地区钉螺孳生地分布图；利用鄱阳湖地区丰水期和枯水期的Landsat7 ETM+遥感图像，根据改进的归一化水体指数及归一化植被指数的相关阈值提取钉螺孳生地；通过受试者工作特征曲线、灵敏度和特异度指标对两种方法的提取结果进行评价和比较，并应用刀切法对影响钉螺孳生因素的重要性进行分析。 结果 遥感方法的受试者工作特征曲线下面积值仅为0.56，灵敏度为0.23，特异度为0.89，而最大信息熵生态位模型的受试者工作特征曲线下面积值为0.88，所提取钉螺孳生地分布图的灵敏度为0.89，特异度为0.74。鄱阳湖地区钉螺孳生地主要集中分布在永修县东北部、余干县西北部，鄱阳县西南部以及新建县中部，而影响钉螺孳生的最主要环境因子是高程，其次是陆地表面温度。 结论 基于最大熵模型的钉螺孳生地提取方法较基于遥感技术的提取方法更为准确可靠，对于相关部门有针对性地选择钉螺孳生重点区域开展防治措施有一定指导意义。

关键词: 湖北钉螺；鄱阳湖；最大熵模型；遥感