Chapter 3 Interactions between perennial ryegrass and white clover and their effects on
3.3 Materials and Methods
3.3.8 Data analysis
The total DM yield harvested was adjusted to account for the differences in the post-grazing residual left by the cows among the different cultivars (section 3.4.3 of this Chapter). For this purpose, the difference between the residual post β harvest (1750 kg DM/ha in the first three harvests and 1900 kg DM/ha in the following harvests), and the post-grazing residual from the previous grazing in each subplot, was added or subtracted to the DM harvested, depending if the subplot had been grazed lower or higher than the cutting height. In this way, most of this adjusted DM yield is the actual harvested herbage, but it also includes a small fraction that considers the preference showed by the cows for some of the cultivars, which, if not considered, could bias the results. In the text total DM yield and total adjusted DM yield are used as synonyms.
Chesson-Manly index
The Chesson-Manly (CM) Index (Chesson, 1983; Smit, Tamminga, & Elgersma, 2006; Solomon, Macoon, Lang, Vann, & Ward, 2014) which relates consumption to forage availability as a measure of relative preference, was calculated using the DM estimated with the rising plate meter pre and post- grazing and based on the formula developed by Manly, Miller, and Cook (1972) and Chesson (1983), and used by Smit et al. (2006) and Solomon et al. (2014) (Equation 3).
Equation 3
πΌπ= ln [1 β (consumedi / availablei )] , i = 1,β¦β¦, m
βππ=1ln [1 - (consumedj/ availablej )]
In this formula, consumed i is the amount of consumed herbage of cultivar i and available i is the
available herbage of the same cultivar at the beginning of the grazing event; m is the number of cultivars available to choose (8 in this experiment) and the denominator term is the sum of all numerator terms.
Farming seasons
The first year of the experiment comprised the farming season 2012 β 2013, starting on 1st June 2012
and ending 31st May 2013. The second year comprised the farming season 2013 β 2014, starting on
1st June 2013 and ending 31st May 2014.
Statistical analysis
Analysis of variance was performed on all data using GenStat 17 (VSN International, 2014) with cultivar, nitrogen and clover treatments and their interactions as fixed effects, and block, main plot within block and subplot as random effects. Least significant differences (LSD) at the 5% level were used to declare differences among means. Contrasts among the cultivars were included in the analysis of variance using the COMPARISON function in GenStat with a matrix of 6 contrasts of the
cultivars (Dense versus Open, Prospect AR37 versus Abermagic AR1, Base AR37 versus Bealey NEA2, Mid versus Late, Commando AR37 versus Kamo AR37, One50 AR37 versus Alto AR37); the aim of this analysis was to gain an insight into possible interactions with treatment.
Botanical composition data were analysed before and after angular transformation. Visual assessment of residual plots was conducted; when a transformation was necessary P values and letters (to indicate significant differences) presented in the Tables are from the analysis of transformed data. Percentages and SED from the analysis of untransformed data are included for ease of interpretation.
Tiller and white clover population density data were analysed before and after square root transformation. Visual assessment of residual plots was conducted; when a transformation was necessary P values and letters (to indicate significant differences) presented in the Tables are from the analysis of transformed data. Means and SED from the analysis of untransformed data are included for ease of interpretation.
Regression analyses were conducted between tiller density and seasonal autumn yield within N Γ Clover treatment, for each year using GenStat 17 (VSN International, 2014). Regression analyses were also conducted between seasonal yield for each cultivar in the minus clover treatments and seasonal white clover percentage (and white clover yield) in pastures sown with the same ryegrass cultivar in the plus clover treatments, analysed within N treatment, using GenStat 17 (VSN
International, 2014). The white clover yield was calculated based on the seasonal yield of the mixture and the white clover percentage in the sampling conducted during the same season. Additionally, regression analyses were conducted between tiller density in perennial ryegrass monocultures and white clover percentage (and white clover yield) in mixtures, and between tiller density in mixtures and white clover percentage (and white clover yield) in mixtures, within N treatment, in autumn each year using GenStat 17 (VSN International, 2014). Regression analyses were also conducted between tiller density and white clover growing point density either combined or pooled within treatment, and between white clover growing point density and white clover percentage (and white clover yield) within N treatment, in autumn each year using GenStat 17 (VSN International, 2014).
Moreover, regression analyses were conducted between white clover percentage in each season and seasonal white clover yield within N treatment using GenStat 17 (VSN International, 2014).
Analysis of variance of the adjusted Mean Stage Count (MSC) was conducted before and after square root transformation using GenStat 17 (VSN International, 2014). Visual assessment of residual plots was conducted; untransformed means and SED are presented in the Tables for ease of
interpretation. However, P values and letters are from the analysis of transformed data. The
contribution of reproductive tillers to the total tiller sample expressed as % of the total tiller number and % of the total sample dry weight and their angular transformations were analysed. Visual
assessment of residual plots was conducted; untransformed means and SED are presented in the Tables but P values and letters are from the analysis of transformed data.
Repeated measures analyses were conducted on the adjusted DM yield, white clover percentage, angular transformation of the white clover percentage, tiller density and Chesson-Manly index, using the AREPMEASURES procedure in GenStat 17 (VSN International, 2014). Since there were significant interactions between treatments and season, results of the analysis of variance for individual seasons are presented. Repeated measures analysis was also conducted on the endophyte infection
frequency data using the same procedure, showing no interaction between year and Cultivar. For the post-grazing mass (kg DM/ha), repeated measurements through time were analysed using spline models within the linear mixed model framework as described by Verbyla, Cullis, Kenward, and Welham (1999). Treatment, cultivar, treatment by cultivar interaction, the linear trend of time and the interaction of treatment and cultivar with the linear trend of time were included in the model as fixed effects; block, main plot within block, subplot, linear trend of time within subplot, spline, the interaction of subplot with spline and the interaction of treatment and cultivar with spline were included as random effects. Residual maximum likelihood (REML) in GenStat 16.2 (VSN
International, 2013) was used to fit these models. This method of analysis essentially fits straight lines to the data initially (the linear trends) and estimates the differences in the slopes of these lines for the treatments and cultivars. Curvature in addition to the linear trend is then included in the model (the spline terms) and treatment and cultivar differences in curvature are determined. These are represented by the interactions of the spline term with treatment and cultivar. The fitted curves are hence determined by combining the linear trend and the curvature in addition to this for each treatment cultivar combination.