Abstract: Objective　To examine the impact of climate, land use and their interaction effect on the population density of Oncomelania hupensis. Methods　O. hupensis snail surveillance data in Yunnan Province from 2013 to 2022 were obtained from the National Information System for Parasitic Diseases Prevention and Control of Chinese Information System for Disease Control and Prevention, and the density of living snails was calculated in each county⁃level administrative district. Climate data were obtained from the China Meteorological Forcing Dataset version 2.0, and land use data were captured from the China Land Cover Dataset. The raster climate data and land use data were clipped to the administrative border of Yunnan Province in the R package version 4.5, and the climate data were resampled to a spatial resolution of 30 m × 30 m consistent with the land use data using bilinear interpolation. Annual meteorological indicators were extracted from each county⁃level schistosomiasis surveillance administrative area in Yunnan Province from 2013 to 2022, including average temperature, maximum temperature, minimum temperature, temperature difference, average precipitation rate, maximum precipitation rate, minimum precipitation rate, downward shortwave radiation flux, downward longwave radiation flux, air pressure, wind speed, specific humidity and relative humidity. Annual land use structure and landscape pattern index in each county⁃level schistosomiasis surveillance administrative area in Yunnan Province from 2013 to 2022 were extracted and calculated, including the proportion of areas occupied by each type of land use, patch density, largest patch index, landscape separation index, landscape fragmentation index, landscape segmentation index, contagion index, patch cohesion index, edge density, and shape index. Land use entropy index was introduced to represent the degree of habitat fragmentation and ecological niche breadth. The density of living snails and environmental factors were subjected to rank correlation analysis, and variables with no statistically significant correlation (P > 0.1) were excluded. Then, a correlation coefficient matrix was constructed, and ∣rs∣ of 0.85 and greater was defined as the threshold of high collinearity. For highly correlated variables in each group, indicators with a more biological explanatory power or a better statistical efficiency were retained. A generalized additive model (GAM) was constructed with O. hupensis snail density as a response variable and screened environmental variables as independent variables, and the model performance was evaluated using the variance explained. Results　The area occupied by O. hupensis snail habitats was 1 056 to 1 680 hm2 in Yunnan Province from 2013 to 2022, appearing an overall tendency towards a decline; however, the density of living snails appeared a tendency towards a slight rise. Based on the correlation coefficient matrix, 14 core explanatory variables were retained, including farmland, water bodies, bare land, impervious surface, landscape separation index, landscape fragmentation index, edge density, shape index, land use entropy index, downward longwave radiation flux, downward shortwave radiation flux, maximum temperature, minimum precipitation rate and specific humidity. Two types of GAM were constructed, including the main⁃effects model without interaction terms and the interaction model with interaction terms. The corrected Akaike information criterion (AICc) of the optimal main⁃ effects model was -285.223, and the core variables included bare land, impervious surface, landscape separation index, edge density, shape index, land use entropy index and downward shortwave radiation flux. Following introduction of second⁃order interaction terms, the AICc value of the optimal interaction model was -345.526, and the significant interaction combinations included: farmland × impervious surface, farmland × landscape separation index, farmland × shape index, farmland × land use entropy index, impervious surface × landscape separation index, impervious surface × edge density, and shape index × downward shortwave radiation flux. The proportion of the main⁃effects model explaining the variation of O. hupensis snail density was 39.8%, and the explanatory power of GAM increased to 79.0% following introduction of interaction terms. Conclusion　Landscape separation index, edge density, shape index and other landscape pattern indicators pose significant impacts on the intensity and direction of changes in O. hupensis snail density in Yunnan Province.

Key words: Oncomelania hupensis, Density, Climate, Land use, Landscape pattern, Interaction effect, Generalized additive model, Yunnan Province

摘要： 目的　分析气候、土地利用及其交互效应对云南省钉螺种群密度的影响。方法　于国家寄生虫病防治信息管理系统获取2013—2022年云南省钉螺监测数据，计算各县级行政区活螺密度。自中国区域地面气象要素驱动数据集（China Meteorological Forcing Data）v2.0获取云南省气候数据，自中国陆地覆盖数据集（China Land Cover Dataset）获取该省土地利用类型数据。在R 4.5软件中将栅格格式气候与土地利用数据裁剪至云南省行政边界范围，随后采用双线性插值法将气候数据重采样至与土地利用数据一致的30 m × 30 m空间分辨率。提取2013—2022年云南省县级血吸虫病监测行政区逐年关键气象指标，包括平均温度、最高温度、最低温度、温差、平均降水率、最大降水率、最小降水率、向下短波辐射通量、向下长波辐射通量、气压、风速、比湿和相对湿度。提取并计算2013—2022年云南省各县级血吸虫病监测行政区年度土地利用结构及其景观格局指数，包括各土地利用类型面积比例、斑块密度、最大斑块指数、景观分离指数、景观碎片化指数、景观分割指数、蔓延度指数、斑块凝聚性指数、边缘密度、形状指数。引入土地利用熵指数以表征生境破碎化程度与生态位宽度。对活螺密度与环境因子进行秩相关分析，剔除相关性无统计学意义（P > 0.1）的变量，随后构建相关系数（rs）矩阵，设定∣rs∣ ≥ 0.85为高度共线性阈值。在每组高度相关的变量中，保留生物学解释力更强或统计效能更佳的指标。以钉螺密度为响应变量，筛选后的环境变量为自变量，构建广义可加模型（generalized additive model，GAM），采用方差解释度对模型进行评价。结果　2013—2022年，云南省钉螺面积为1 056 ~ 1 680 hm2，整体呈下降趋势，但活螺密度略有回升。基于rs矩阵，保留14个核心解释变量，包括农田、水域、裸地、不透水面、景观分离指数、景观碎片化指数、边缘密度、形状指数、土地利用熵指数及向下长波辐射通量、向下短波辐射通量、最高温、最小降水率和比湿。本研究构建了未加入交互项的主效应模型和加入交互项的交互模型2种GAM。最佳主效应模型修正赤池信息准则（corrected Akaike information criterion，AICc）值为−285.223，核心变量为裸地、不透水面、景观分离指数、边缘密度、形状指数、土地利用熵指数及向下短波辐射通量。加入二阶交互项后，最佳交互模型AICc值降至−345.526，显著交互组合包括农田 × 不透水面、农田 × 景观分离指数、农田 × 形状指数、农田 × 土地利用熵指数、不透水面 × 景观分离指数、不透水面 × 边缘密度、形状指数 × 向下短波辐射通量。主效应模型解释钉螺密度变异的比例为39.8%，引入交互项后GAM解释力升至79.0%。结论　景观分离指数、边缘密度、形状指数等景观格局指标对钉螺密度变化强度和方向发挥重要影响。

关键词: 钉螺, 密度, 气候, 土地利用, 景观格局, 交互效应, 广义可加模型, 云南省