Data
Data table processing
The data collected in the field was entered into a combined database with separate tabs for each of the nested plots. The data recorded in each of these separate tables was recorded in a different format, so some processing needed to be done to merge the tables into one suitable for analysis. R Statistical Software (v4.3.1; R Core Team, 2021) was used to reformat and analyze the data. An example of the database raw data can be seen in Table 1 and Table 2. Raw data from the field was split into "vegetation" sample plots (plots that were re-measured over multiple years) and "biomass" sample plots (plots that were destructively sampled).
The quadrat data contained multiple columns to list the plant species identified in each of the 1 m2 quadrats throughout the cut block (Table 1). This had to be rearranged into individual rows and then differentiated between north and south by adding an additional column containing N or S. Then, everything except for fireweed (C. angustifolium), raspberry (R. idaeus), willow (Salix spp.), and aspen (P. tremuloides) had to be filtered out in order to create a data table only containing the four species we are estimating biomass for (Table 3). Seedling data contained all conifer and deciduous species identified in the 1.78 m radius plot. All species were filtered out except for P. tremuloides. Sapling data contained a tally of each species identified in the 3.57 m radius plot. All species were filtered out except for P. tremuloides. Shrub data contained measurements of the crown width and height of each shrub within the 3.57 m radius plot (Table 2). A new table containing only Salix spp. was created (Table 4).
Table 1. Sample of raw vegetation data recorded from the north and south quadrats at each plot in a cut block.
Table 2. Sample of raw shrub data recorded from the 3.57 m radius plot at each plot in a cut block.
Examples of the output tables can be seen below:
Table 3. Sample of the output table for quadrat plants used for further analysis.
Table 4. Sample of the output table of willow shrubs.
Biomass data processing and exploration
The data collected from destructive sampling of each of the four species (C. angustifolium, R. idaeus, Salix spp., P. tremuloides) was used to create regression equations in order to estimate total biomass in kg/ha for the entire cut block for each of the 38 sites. An example of the raw biomass data can be seen in Table 5.
Through the exploration of the biomass data, it was determined that simple linear regression would not be suitable for the biomass estimates of the quadrat percent cover data (Figure 1). The linear function created using simple linear regression does not fit the data well, as the y-intercept passes through a negative value. This resulted in negative values being returned when the function was applied to the vegetation percent cover data for biomass estimates. Therefore, an allometric regression was used (log-log scale) and values were later converted from natural logarithms back into plain numbers we could use for further analysis.
For the seedling and sapling data, only a count of the number of seedlings was recorded in the vegetation plots, therefore only the means of the seedling and sapling fresh weights from the biomass data were used to estimate total aspen seedling and total aspen sapling biomass across each of the openings.
The shrub data contained measurements of crown width and plant height. Radius was then calculated from averaging the two crown widths, and crown volume was calculated assuming a cylinder. Simple linear regression of the shrub biomass data was then used to estimate total biomass for each of the shrubs recorded in the vegetation plot data.
Table 5. Sample of raw data form a destructively sampled biomass quadrat.
Figure 5. Example of a simple linear regression model created during the data exploration phase for aspen within the 1 m2 quadrats. The y-intercept can be seen passing through a negative value.
After biomass estimates were calculated for each of the nested plots, data was combined into one table that could be used for further analysis (Table 6).
Table 6. Output of combined datasets. Columns "D","E","F", and "G" contain the estimated browse biomass for each of the nested vegetation plots (quadrat, seedling, sapling, and shrub). The combined biomass of the nested plots in kg/ha was calculated and recorded in column "L". Columns "M" and "O" contain the weights of the vegetation samples which were sent to the University of Guelph for glyphosate analysis. Column "N" contains the glyphosate concentrations in ppm received from the University of Guelph.
Exploratory graphics
The residual plots below were created to evaluate the accuracy of the regression models used to estimate biomass. Most of the models displayed a good fit with the exception of the 1 m2 quadrat treated willow model, and the untreated willow shrub model.
During the first year after herbicide application all four species of interest displayed reduced biomass production and a recovery of biomass can be seen in year 2 (Figure 6). Interestingly, the median biomass of willow in year 2 of recovery appears to have exceeded the initial willow biomass of the untreated blocks.
Residual glyphosate concentration also decreased across all species over the two-year period (Figure 7). Raspberry plants had the highest initial glyphosate concentration, and displayed the lowest biomass recovery over the two-year period.
Figure 6. Box plot of the changes in biomass over a two-year period. "0" indicates blocks that were untreated at the time of measurement. "1" is one year after herbicide application, "2" is two years after herbicide application.
Figure 7. Box plot of the glyphosate trend from year 1 to year 2 in each of the species.
Figure 8. Plot of the residuals from the regression that was used to calculate fireweed biomass across 1 m2 quadrats within treated cut blocks.
Figure 9. Plot of the residuals from the regression that was used to calculate fireweed biomass across 1 m2 quadrats within untreated cut blocks.
Figure 10. Plot of the residuals from the regression that was used to calculate raspberry biomass across 1 m2 quadrats within treated cut blocks.
Figure 11. Plot of the residuals from the regression that was used to calculate raspberry biomass across 1 m2 quadrats within untreated cut blocks.
Figure 12. Plot of the residuals from the regression that was used to calculate aspen biomass across 1 m2 quadrats within treated cut blocks.
Figure 13. Plot of the residuals from the regression that was used to calculate aspen biomass across 1m2 quadrats within untreated cut blocks.
Figure 14. Plot of the residuals from the regression that was used to calculate willow biomass across 1 m2 quadrats within treated cut blocks.
Figure 15. Plot of the residuals from the regression that was used to calculate willow biomass across 1 m2 quadrats within untreated cut blocks.
Figure 16. Plot of the residuals generated from the regression that was used to calculate willow biomass across 3.57 m radius shrub plots within treated cut blocks.
Figure 17. Plot of the residuals generated from the regression that was used to calculate willow biomass across 3.57 m radius shrub plots within untreated cut blocks.