Abstract: Objective To build a temperature difference-based CO₂ mosquito lamp sampling model, and to provide evidence for CO₂ mosquito lamp monitoring frequency. Methods A total of 229 CO₂mosquito lamps were set to monitor the density of Culex pipiens pallens every ten days (defined as one period) from April to November in 15 districts of Shanghai, China. The monitoring data of 2019 were used as the training set, and those of 2020 were used as the test set. A function of mean and standard deviation was established based on Taylor’s power law, and the sample size formula for mean comparison between two samples was substituted into the function to establish a sampling model based on density difference. A linear regression model was used to establish a regression equation of the density difference between two adjacent monitoring activities and the temperature difference (during 10 days before monitoring) between two monitoring activities. The density difference-based sampling model was substituted to construct a temperature difference-based sampling model. The Wilcoxon rank-sum test was used for mean comparison between two samples. The model was validated using accuracy, recall, and F-measure. Results In the density difference-based sampling model, for both 2019 and 2020, the difference in mean mosquito density was significant between two adjacent periods(all P<0.05), between two periods with an interval of one period(all P<0.05), and between two periods with an interval of two periods(all P<0.05), accounting for 34.78%, 59.09%, and 76.19% in 2019, respectively, and 21.74%,59.09%, and 66.67% in 2020, respectively. In the validation of the temperature difference-based sampling model with the data of 2020, the accuracy was 0.563, the recall was 0.720, and the F-measure was 0.632. Conclusion The temperature difference-based sampling model is practical, which can estimate the optimal frequency of CO₂ mosquito lamp monitoring based on temperature difference. CO₂ mosquito lamp monitoring frequency in Shanghai from April to November can be adjusted from once every 10 days at present to once every 20 days, and can be increased based on temperature changes.

Key words: CO₂mosquito lamp, Time-frequency, Culex pipiens pallens

摘要： 目的 建立基于温度差的CO₂诱蚊灯抽样模型，为CO₂诱蚊灯监测频次的确定提供科学依据。方法 在上海市15个区，于2019和2020年的4－11月每旬设置229个CO₂诱蚊灯监测1次淡色库蚊密度。以2019年的监测数据为训练集，2020年的监测数据为测试集。通过泰勒幂法则建立均数与标准差间的函数，代入两样本均数比较的样本量公式，建立基于密度差的抽样模型。使用线性回归模型建立临近2次CO₂诱蚊灯监测结果密度差和该2次监测前1旬的温度差的回归方程，代入基于密度差的抽样模型，建立基于温度差的抽样模型。两样本均数比较使用Wilcoxon秩和检验，模型验证使用准确率、召回率和调和平均值（F-measure）进行评价。结果 2019年相邻2旬、间隔1旬、间隔2旬淡色库蚊监测密度比较，均值差异均有统计学意义（Wilcoxon秩和检验，均P<0.05），占比分别为34.78%、59.09%和76.19%；2020年相邻2旬、间隔1旬和间隔2旬密度比较，均值的差异有统计学意义（均P<0.05），占比分别为21.74%、59.09%和66.67%。以2020年数据验证基于温度差的抽样模型，准确率为0.563，召回率为0.720，F-measure为0.632。结论 基于温度差的抽样模型具备实用意义，可以根据温度差估算CO₂诱蚊灯监测的最佳监测频次。目前上海市的CO₂诱蚊灯监测每年4－11月每旬监测1次，建议间隔1旬开展监测，并根据温度变化适当增加频次。

关键词: CO₂诱蚊灯, 时间频次, 淡色库蚊