Grazing management studies within the Temperate Pasture Sustainability Key Program: experimental design and statistical analysis

Abstract

Long-term agricultural experiments such as the Temperate Pastures Sustainability Key Program (TPSKP) present significant challenges in the areas of planning and design, conduct, analysis and reporting. This paper concentrates on 2 aspects, namely, the experimental design and the statistical analysis.

For long-term agricultural experiments which examine the effects of management strategies over time, an enumeration of the initial biodiversity is essential and permits the allocation of treatments to plots in such a way that potential bias in the estimation of treatment effects due to lack of uniformity in experimental units (plots) is reduced in the covariate analysis. Spatial replication is considered essential and the design should include at least 2 starting dates for management strategies so that the possible interaction between the year of start and the management strategy can be described.

The data resulting from repeated measurement of herbage mass of major individual species or species groups represent a longitudinal data set with complexity due to the staggered commencement of treatments and also in part due to the nature of some of the strategies (closure and cuts). The analysis presented is the cubic smoothing spline approach of Verbyla et al. (1999) which integrates cubic splines, random coefficients, covariance modelling and estimation of systematic deviation. This approach, based on linear mixed models and using residual maximum likelihood (REML) has the flexibility to cope with the staggered imposition of management strategies and permits the partitioning of trends into smooth and non-smooth components, thereby quantifying species persistence and seasonal influence under each management strategy.