Conference on Applied Statistics in Agriculture

Using a Generalized Linear Mixed Model Framework to Account for Spatial Variability in a Comparison of Orchard Sprayer Efficacy

Uniform application of pesticides in vineyard and orchard systems can be difficult to achieve due to variability in the density and structure of the crop canopy. Depending on the equipment used and environmental conditions, applications can result in poor spray coverage, spray drift, and wasted spray which, in turn, are manifested as a combination of poor pesticide efficacy, economic losses and potential environmental problems for the grower. A study was therefore designed and carried out to test new sprayer equipment aimed at addressing these issues. Statistically, the study presented a unique replicated three dimensional spatial design which captured response variability (coverage) both within and across trees in an orchard setting. Application of a generalized linear mixed model framework allowed comparison of sprayer designs in terms of their application efficiencies while accounting for the intra- and inter-tree correlation of the coverage response. Examples demonstrating various models and their associated correlation structures are given and the resulting interpretations discussed.