FUENTES DE LOS PROBLEMAS DE INVESTIGACIÓN

4.5.9 Laboratory population oviposition experiment

To determine if there were any oviposition preferences or inhibitions due to possible treatments effects on the intrinsic host quality or desirability of the broccoli plants, an oviposition experiment was conducted using the adult P. xylostella laboratory population. On 9 February 06, three broccoli leaves were cut from each treatment by pressing and gently twisting a petri dish against the underside of a cleaned leaf with a chopping board placed on the opposite side. Each petri dish then contained a whole, topside up, round section of a broccoli leaf. The leaf samples were immediately taken from the field and placed in the centre of the adult moth cage for 24 hours in a randomised arrangement with

three replications (in three trays). A stereo-microscope was then used to count the eggs on each leaf. The procedure was repeated on 14 and 21 February 06.

4.5.10 Semi commercial cover crop experiment 05/06

To determine if the cover crop results from the 04/05 experiment were valid at scales greater than the plot size of 10m2, with the view of commercial implementation, a semi- commercial planting of one hectare of broccoli was established on a farm at Gawler (E 429220, N 5440190) on Tasmania’s northwest coast. This location was 15km west of Forthside Research Station and in a similar environment (climate and soil). The

experimental area was 50m wide and 200m long and divided into four plot pairs, each plot being 25m x 50m. One plot in each pair was randomly designated to have either a cover crop or to be prepared using conventional tillage (that is, bare soil). The rye cover crop was sown on 17 August 05 at a rate of 100kg/ha without fertiliser. The cover crop was sprayed and killed on 15 November 05 using glyphosate (720g ai/ha). Due to time constraints brought on by developmental problems with the one-pass roller/ transplanter, only half the area was planted with broccoli on 5 December, making each plot 12.5m wide and 50m long. Again time constraints, in this instance associated with the management and sampling regime of the Forthside trial, meant that for this experiment 15 randomly selected plants from each plot were sampled once for the presence of P. xylostella eggs and larvae, P. rapae larvae and B. brassicae colonies on 28 December (23 DAT). The trial was terminated on 23 January.

4.5.11 Data analysis 04/05

The P. rapae and P. xylostella larvae and pupae counts from the 04/05 experiment were analysed using a one-way analysis of variance (ANOVA) (Proc GLM, SAS Institute, Cary, NC) for each sampling date. The mean counts from each plot were used as the response variables, while the three replications (blocks) and the four treatments (treatments) were the predictor variables. Treatment means were separated using Fisher’s least significant

difference (LSD) and data were log+1 transformed when necessary to conform to the assumptions of the ANOVA procedure. However, only non-transformed data were reported in the figures and tables.

The B. brassicae colony and parasitism data from the 04/05 experiment were based on the proportion of plants infested. Therefore the data were arcsine square root transformed prior to using the ANOVA procedure. These proportions were used as the response variables, while the predictor variables were also the blocks and treatments.

To determine the effects of different treatments, pairwise contrasts were also planned for all the insect data. These contrasts were performed using the ANOVA model so that the results from the monoculture plots were compared to the results from the strip cropping plots; and the results of the cover crop plots were compared to the bare soil plots.

The pairwise contrasts for 04/05 can be summarised as: 1. Cover crop vs. Bare soil

2. Strip crop vs. Monoculture

An example ANOVA table is presented as an appendix.

4.5.12 Data analysis 05/06

The P. rapae and P. xylostella egg, combined larvae and pupae counts and the P. xylostella vacuum sampling data from the 05/06 experiment were analysed using a one-way ANOVA (Proc GLM, SAS Institute, Cary, NC) for each sampling date. The mean counts from each plot were used as the response variables, while the six columns (block) and the six rows (row) of the Latin square design; and the six treatments (treatment) were the predictor variables.

For the oviposition preference experiment, an ANOVA was also used to analyse the data. The number of eggs oviposited were used as the response variable while each tray, treatment and replication were used as the predictor variables.

For all ANOVA analyses, treatment means were separated using Fisher’s least significant difference (LSD) and data were log+1 transformed when necessary to conform to the assumptions of the ANOVA procedure. However, only non-transformed data were reported in the figures and tables.

To determine the effects of different treatments, pairwise contrasts were planned for the P. rapae egg and larvae data; and for P. xylostella egg, larvae, laboratory population

oviposition preference and vacuum sampling data. These contrasts were performed using the ANOVA model so that the results from the monoculture plots were compared to the results from the strip cropping plots (both rye strips and potato strips); the results from the cover crop plots were compared to the bare soil plots results; and the bare soil monoculture plots results were compare to the two bare soil strip cropping plots (both rye strips and potato strips).

The pairwise contrasts for 05/06 can be summarised as: 1. Cover crop vs. Bare soil

2. Strip crop vs. Monoculture

3. Bare soil strip crops vs. Bare soil monoculture An example ANOVA table is presented as an appendix.

The B. brassicae data from the 05/06 experiment were based on the presence or absence of colonies and parasitised aphids. The use of the presence/absence sampling regime and a low effective sample size (three instances per plot) meant that a logistic regression with a dichotomous response was the appropriate analysis using Proc LOGISTIC in a SAS model (Stokes et al. 2000) in a process summarised by Equation 4.1. The predictor variables were block, row, treatment and sampling date. The odds ratios for each treatment, with respect to the reference level, correspond to the exponential of the logistic regression estimate for that treatment.