Using multiple natural enemies to manage whiteflies in commercial poinsettia production dataset

Purpose was to determine the effectiveness of using biological control (two natural enemies: a parasitic wasp, Eretmocerus eremicus, and a predatory mite, Amblyseius swirskii) to manage sweetpotato whiteflies (Bemisia tabaci) on poinsettias compared to using regular insecticidal applications in commercial poinsettia production. This trial was conducted at three different commercial greenhouse grower operations in East Texas in 2019, each of which designated one greenhouse as the IPM greenhouse (using biological control) and the other as the conventional insecticide spray program (chemical control). In the biological control houses, natural enemies were released on regular intervals (weekly for E. eremicus and monthly for A. swirskii), whereas growers were able to use any chemical spray programs they were accustomed to in the chemical control house. Fifty randomly selected poinsettias and approximately 50 flagged (and re-visited) poinsettias were inspected weekly (maximum of 20 leaves each) in each greenhouse to quantify the number of whitefly nymphs, pupae, exuviae, and adults, E. eremicus adults and exuviae, and A. swirskii adults. Yellow sticky cards (four per greenhouse) were also established to quantify whiteflies, fungus gnats, or any other problematic pests. A weekly journal was kept to log natural enemy releases and any changes within the greenhouses. In general, the biological control managed greenhouses tended to have higher whitefly densities than their chemically-controlled counterparts; although there was no significant differences in the final density of whiteflies (i.e. last sampling week for biological control vs chemical greenhouses). A similar trend was seen for the proportion of poinsettias infested with whiteflies between the biological control vs chemical greenhouses. High variability in whitefly counts and difference in sampling period length between greenhouses made analysis via repeated measures ANOVA or GLMER unsuitable (i.e., violation of statistical assumptions). To develop a potential binomial sampling plan (i.e., whether proportion of poinsettias infested with whiteflies can predict whitefly density), the relationship between average immature whiteflies observed per plant (log-scale) and proportion of plants infested with whiteflies was modeled. Initial model composed of proportion of plants infested with immature whiteflies as a fixed factor, greenhouse nested within grower, and week number as random factors, and log-transformed mean immature whiteflies per plant as the response variable in a generalized linear mixed model (GLMM) using the lmer (Kuznetsova et al. 2017) function in R Studio (R Studio Team 2015). All random factors were considered non-significant and were removed from the final model. Proportion of poinsettias infested was a strong predictor of both log-transformed average immature whiteflies (p&lt;0.001, adjusted r^2=0.898) and log-transformed maximum number of whiteflies (p&lt;0.001, adjusted r^2=0.777).本试验旨在探究生物防治（biological control）手段——两种天敌：寄生蜂（parasitic wasp）Eretmocerus eremicus与捕食螨（predatory mite）Amblyseius swirskii——用于防治一品红（poinsettias）上烟粉虱（Bemisia tabaci，sweetpotato whiteflies）的效果，并与商业一品红生产中的常规杀虫剂施用方案进行对比。本试验于2019年在德克萨斯州东部的三家商业温室种植基地开展，每家基地均指定一间温室作为有害生物综合治理（IPM）温室（采用生物防治手段），另一间作为采用常规杀虫剂喷施方案的化学防治（chemical control）温室。在生防温室中，天敌按固定周期释放：E. eremicus每周释放一次，A. swirskii每月释放一次；而化学防治温室中，种植者可使用其惯用的任意化学喷施方案。每个温室每周随机选取50株一品红，以及约50株经标记后回访的一品红（每株最多采集20片叶片）进行检视，以量化烟粉虱若虫、蛹、蜕壳（exuviae）及成虫，E. eremicus成虫与蜕壳，以及A. swirskii成虫的数量。每个温室还设置了4张黄色粘虫板，用于量化烟粉虱、蕈蚊（fungus gnats）及其他为害性害虫的种群数量。研究人员按周留存日志，记录天敌释放情况及温室内的各项变动。总体而言，采用生防管理的温室烟粉虱密度往往高于化学防治对照温室；但二者的烟粉虱最终密度（即生防温室与化学防治温室的最后采样周）并无显著差异。生防与化学防治温室中受烟粉虱侵染的一品红比例也呈现相似趋势。由于各温室内烟粉虱计数差异显著，且不同温室的采样周期长度不一，导致采用重复测量方差分析（repeated measures ANOVA）或广义线性混合效应模型（GLMER）开展分析不符合统计假设，因此不适用。为构建可行的二项式抽样方案（即受烟粉虱侵染的一品红比例能否预测烟粉虱密度），本研究对每株植株上观察到的未成熟烟粉虱平均数量（对数尺度）与受侵染植株比例之间的关系进行了建模。初始模型以未成熟烟粉虱侵染植株比例作为固定因子，以种植者嵌套下的温室、周数作为随机因子，以每株未成熟烟粉虱的对数转换后平均数量作为响应变量，使用R Studio（R Studio Team 2015）中的lmer函数（Kuznetsova et al. 2017）构建广义线性混合模型（GLMM）。所有随机因子均被判定为不显著，因此在最终模型中被移除。受侵染一品红比例可显著预测对数转换后的未成熟烟粉虱平均数量（p<0.001，校正后决定系数r²=0.898），以及对数转换后的烟粉虱最大种群数量（p<0.001，校正后决定系数r²=0.777）。