A method is presented to determine which of a set of sampling techniques is most nearly optimum for sampling weed abundance. To employ the method, one needs first to develop a set of sampling alternatives. A pilot survey is then conducted to determine time costs and sampling and measurement error variances associated with each technique. We used this method to determine which of three plot sizes (0.6 by 0.6 m, 0.6 by 1.5 m, 1.2 by 2.1 m) and two types of orchard scans is most nearly optimum for estimating weed abundance in apple [Malus domestica (Bork)] orchards. The results indicate that, for weed species that are fairly apparent, orchard scans are optimal when less than 8 h of sampling effort is expended, but when a high degree of precision is required, or when the weed species is more difficult to detect, plot samples may be the better choice. A size of plot ca. 0.6 by 1.5 m size was found to be most nearly optimum.
