Methods for efficient estimation of rust incidence in coffee plantations

Two systematic patterns were evaluated for efficiency in estimating coffee rust incidence in lots of Caturra and Bourbon. True disease incidence was d etermined in the middle third of each tree in the selected lot. The differe nt sampling patterns were compared on the basis of the bias and standard er ror of the estimated incidences obtained with computer simulations of diffe rent combinations of sample sizes, average incidence and patterns of spatia l dispersion for the disease. The parallel row pattern was at least as effi cient as the traditionally recommended W pattern. In order to attain better estimates of coffee rust incidence, it seems more important to sample more leaves per tree than more trees per field. The recommended protocol for es timating rust incidence in coffee plantations is described.