Data
Data Table
First, the species cover data was checked and cleaned by removing columns with all zeroes and standardizing species identifications across the three years of data collection (Table 1).
Individual species cover was also grouped into plant functional groups and diversity metrics (Table 2).
Table 1. Abbreviated cover table. Columns A-C are metadata. The experimental units were 3 x 7 m plots (PLOT) of grassland plant species to which each a unique treatment was applied. Columns D-G are predictor variables. There were three manipulated predictor variables, all categorical: Time of treatment [spring (S) or fall (F)], Rate of herbicide treatment [none (0x), half the recommended rate (0.5x), the recommended rate (1x) and two times the recommended rate (2x)], and Year [2019, 2020, or 2021]. Columns H-R show 11 of 37 plant species response variables (% cover of different plant species; species names are abbreviated into the first four letters of the genus and three letters of specific epithet except for "Bromus" which was a combination of Bromus squarrosus and Bromus tectorum).
Table 2. Response variables also included groupings of plant functional groups and diversity metrics.
Exploratory Graphics
Cover Boxplots
​Boxplots were used to get an idea of the distribution of data for specific species and plant functional groups of interest. The boxplots also helped to identify outliers and errors. Outliers were checked. The boxplots for brome cover are shown. The same was done for perennial grass and forb cover.
Site x Year
I compared bromegrass cover at both the Pinhorn and Ross sites in each year of data collection (Figure 1). In the pretreatment year, the two sites had similar brome composition and there was a low spread of data. In the growing season immediately after treatment, there was more spread but the median brome cover was similar. However by 2 years after treatment brome cover had large spread at both sites and was positively skewed. This indicates that 2 years after treatment there was likely more treatment effect, since brome cover would be decreasing in the treated plots and there was a greater frequency of low values and a lower frequency of higher values.
Figure 1. Brome cover by site for each of the years prior to or following indaziflam treatment. YAT = year after treatment.
Rate x Year and Time x Year
Since site seemed to have little effect on brome cover but year did, I combined the sites and looked at how rate and time of indaziflam application affected the distribution of brome cover in the post treatment years.
Figure 2. Boxplots of brome cover by rate of indaziflam (0x, 0.5x, 1x, and 2x the recommended rate) at 1 and 2 years after treatment.
Figure 3. Boxplots of brome cover by time of application of indaziflam at 1 and 2 years after treatment.
1 year after treatment, the boxplots had similar median and spread for all rates and times (Figures 2 and 3). At 2 years after treatment, despite a decent amount of variance and skew, the 1x and 2x rates had a lower median than the control (0x), and fall application time was lower than the control.