Application of Regionalized Variable Theory to Large‐Plot Field Experiments

Abstract

Abstract Spatial soil heterogeneity poses problems on field experiments which use large treatment plots, as in tillage research. Recent studies have established that many soil properties do not vary in a random manner, but often exhibit spatial dependence. Presently available methods to account for spatial effects either do not fully address these problems or they are too complicated to be widely accepted. The purpose of this study was to develop an experimental design which was based on the principles of regionalized variable theory. Using incomplete blocks of size two, the method estimates treatment effects by means of short‐distance comparisons; this prevents inflation of the error term due to spatial soil variability and eliminates biases caused by variable distances between plots. The analysis of variance for the method is simple and well established. A Monte Carlo simulation study using uniformity trial data was performed to compare the proposed method with a randomized complete block design. The study revealed that the use of incomplete blocks reduces the average error mean square by 44%, and the average coefficient of variation by 25% compared to using complete blocks. The proposed method was, on the average, 21% more efficient then a complete block design.