Statistical Methods to Address Spatial Variation in Pasture Evaluation Trials

Abstract

The primary breeding objective for white clover improvement in easternAustralia is more reliable persistence under summer moisture-stress while maintainingherbage production (especially winter growth). To identify elite germplasm, evaluationof candidate lines is done in situ, under field conditions for which the prospectivecultivar is targeted. These conditions (e.g. moisture stress, nutrient infertility, grasscompetition) may impose spatial influences within the field site that as artefacts mask oralter the uniform expression of response to environment.A significant component of artefact variation which affects interpretation of thedata is the spatial variation due to plot position. Failure to correct for this may giveunreliable comparisons as some lines inadvertently (through randomisation) areallocated to favourable or unfavourable plots.Our methodology was developed at two sites (Glen Innes , Armidale in northernNew South Wales, Australia) in an experiment which sought to identify superior linesamong 20 candidate entries. Cultivars of reputed stability and spanning the expectedperformance range, are replicated at regular intervals through the field site. Thedifference amongst replicates within cultivar (that is a reduced G X E) is used toestimate spatial trends. Elite lines are identified from temporal profiles of plantperformance related to the breeding objectives.We describe (i) design and analysis which allows estimation of the spatialeffects, and (ii) the temporal profiling of plant performance. These statistical proceduresare seminal in identifying superior breeding lines that subsequently proceed to cultivardevelopment.