Incorporating sampling precision into an action threshold for monitoring ant (Hymenoptera: Formicidae) population levels in vineyards

This study presents an approach that can be used to improve sampling precision for determining when ants should be controlled before they promote mealybug infestations to injurious levels in vineyards. Data on ant infestation were collected in a two-season study (2005/2006 and 2006/2007) from two main vine-growing areas in the Western Cape Province using a presence-absence sampling system. On each of three farms, seven blocks, each consisting of 20 plots of 5 vines each, were inspected biweekly. A regression analysis of observed versus binomial variance using dummy variables to test differences due to season and location was performed. None of the differential coefficients in the full model were statistically significant for stems, shoots and leaves. Therefore, the linear regression of the observed versus binomial variance was not affected and a reduced model was used. For bunches, significant differences were found and a full model was applied. The models were used to estimate the sampling error as well as the probability of correctly deciding to take chemical control action against ants on each of the vine sections. A sampling error of 22% was obtained which remained constant even when more than 20 plots were sampled. Taking or initiating control action against ants when 20% of the vines were infested would not result in under reacting in 95% of the cases.