Results and Discussion
Regression outputs
Figure 18 displays the regression models for each of the four species (fireweed, raspberry, aspen, willow) that were applied to the percent cover data from the 1 m2 vegetation quadrats and used to estimate biomass of new growth (kg/ha). Regressions were done separately for treated and untreated cut blocks, but the values used for each model only differ slightly (Table 7, Table 9).
Figure 19 displays the linear regression model used to estimate biomass (kg/ha) for the large willow shrubs measured in the 3.57 m radius plot. Crown volume was used to estimate the biomass of shrubs measured in the 3.57 m radius plot as these willows were too large to use % cover estimates.
1 m2 quadrat biomass regressions using % cover data:
Figure 18. Allometric regression outputs for each species destructively sampled in the 1 m2 biomass quadrats. The regression line for treated blocks is indicated in blue, and the regression line for untreated blocks is indicated in red.
3.57 m radius plot biomass regressions using crown volume (m3/ha) for large willow shrubs:
Figure 19. Linear regression model used to estimate large shrub biomass within the 3.57 m radius plots. The regression used to estimate biomass in treated cut blocks is indicated in blue, and the regression used to estimate biomass in untreated cut blocks is indicated in red.
Regression tables and equation values
1 m2 quadrat plant estimates (ŷ = ln(a) + b*ln(x)):
Tables 7 through 10 display the values of the regression functions used to calculate the estimated biomass for each of the species at the quadrat level. Values b0 and b1 from the allometric regressions (Table 7, Table 9) are displayed in the form of a natural logarithm, so these values were transformed back to the original scale for analysis (Table 8, Table 10). The RMSE values reported here indicate how accurately the model can predict biomass in kg/ha (Table 8, Table 10). Most of the models used for biomass estimation of the 1 m2 quadrats have a good fit, indicated by high r2 values. The models with the lowest average variability between the actual and predicted values (RMSE) are the ones used to predict raspberry biomass in both untreated and treated cut blocks (Table 8, Table 10).
1.78 m radius plot aspen seedling estimates (ŷ = xFW * n):
The mean biomass for aspen seedlings was higher in untreated cut blocks (Table 12). These mean values were used to estimate the total seedling biomass in each of the untreated and treated cut blocks.
3.57 m radius plot aspen sapling estimates (ŷ = xFW * n):
Mean biomass for aspen saplings was higher in untreated cut blocks (Table 12). These mean values were used to estimate the total sapling biomass in untreated and treated blocks.
3.57 m radius plot willow shrub estimates (ŷ = a + bx):
Table 11 displays the values from the linear model used to estimate biomass of large shrubs in the 3.57 m radius plots. These models have low r2 values which could be explained by the high variability in willow biomass across each of the cut blocks. RMSE values for these models also appear to be high when compared to the models used to estimate biomass from % cover. However, these RMSE values are displayed in kg/ha, and for willow shrubs which are all larger than 1.3 m in height (sometimes up to 3 m tall), with large crown volumes, estimates that may occasionally be off by ~100 kg/ha are not too bad.
Table 7. Values from the allometric regression models used to estimate biomass for each of the species in untreated cut blocks. b0 indicates the log transformed "a" value. b1 indicates the log transformed "b" value.
Table 8. Values that were back-transformed into the original units from the untreated quadrat allometric models. RMSE values are displayed in original units (kg/ha).
Table 9. Values from the allometric regression models used to estimate biomass for each of the species in treated cut blocks. b0 indicates the log transformed "a" value. b1 indicates the log transformed "b" value.
Table 10. Values that were back-transformed into the original units from the treated quadrat allometric models. RMSE values are displayed in orginial units (kg/ha).
Table 11. Values from the linear regression model used to estimate biomass of large willow shrubs in the 3.57 m radius plots.
Table 12. Mean fresh weight (g) of aspen saplings and aspen seedlings in each untreated and treated cut blocks. Sapling mean fresh weight was used to estimate biomass of aspen saplings counted within 3.57 m radius plots. Seedlings mean fresh weight was used to estimate biomass of aspen seedlings counted within 1.78 m radius plots.
Biomass (kg/ha) over a two-year period
Over the two years after operational herbicide application, all four of the indicator plant species displayed a decline in mean biomass in year 1, and a recovery of biomass in year 2, with the exception of aspen (P. tremuloides) which did not appear to recover at all (Figure 20, Table 13). The reduction in aspen biomass during year 2 is most likely due to the mortality of saplings following herbicide application. Fireweed (C. angustifolium) and willow (Salix spp.) displayed much greater recovery than raspberry (R. idaeus), with willow having the highest biomass recovery (Table 13).
As seen in the exploratory graphics, glyphosate persisting within plant shoot tissues declined over the two years following herbicide application (Figure 7). Raspberry plants displayed the highest median concentrations of glyphosate in year 1 (Figure 7) and had low recovery of biomass over the two-year period (Table 13).
Figure 20. Dot plot containing means and 95% confidence intervals for each of the four indicator species. Means were calculated using the emmeans() package in R Statistical Software (v4.3.1; R Core Team, 2021).
Table 13. Emmeans() values as displayed on Figure 20.
Discussion and Take-home messages
Table 14. Paired t-test results indicating biomass recovery from year one to year two after herbicide application. Species with significant recovery are highlighted in blue.
Although willow (Salix spp.) biomass declined during the first year after herbicide application, there was a significant recovery of biomass in year 2 (p-value=0.011) (Table 14). The higher availability of willow browse biomass in these cut blocks after the second year presumably makes these blocks preferential for foraging moose and increases the potential exposure to glyphosate. In the context of this study, a hypothetical model of moose browsing patterns can be created to evaluate potential exposure (Table 16). Moose that choose to browse in these operationally treated cut blocks have the potential to consume up to 0.1942 mg of glyphosate per ha, if the consumption rate of willow is 20%.
Table 15. One-sided t-test results. Probabilities of new shoots containing <0.1, <0.08, and <0.05 ppm (mg/kg) of glyphosate one year after herbicide application and two years after herbicide application.
There are few studies on the effects of acute glyphosate exposure to ungulates, however, the maximum residue limit allowed for human food consumption in Canada is currently 0.1 mg/kg (Kolakowski et al, 2020). Although this cannot be directly compared to moose exposure potential, the residue limit of 0.1 mg/kg can serve as a general reference for the data of this study. Raspberry had the highest concentrations of glyphosate recorded in year 1 (Figure 7), and there is a 29% chance that any individual raspberry plant will contain more than 0.1 mg/kg of glyphosate after the first year (Table 15). This is much higher than the other three indicator species (fireweed, aspen, willow), which have less than a 1% chance of containing 0.1 mg/kg of glyphosate after the first year.
Of the four species, aspen has the lowest chance overall of containing high amounts of glyphosate (Table 15). Presumably, this is due to the majority of aspen dying after the initial application of herbicide, rendering it impossible for glyphosate to move into new tissues. However, for all species collected in this study, the probability of glyphosate remaining higher than 0.1 mg/kg after the second year of recovery is less than 1% (Table 15).
Table 16. Amount of browse biomass consumed at 5%, 10%, and 20% consumption rates, and the corresponding consumption of glyphosate.
The average glyphosate residue in willow shrubs across the treated cut blocks was 0.0216 mg/kg (Table 16). However, greater consumption rates risk a higher potential for exposure for moose (and other wildlife) foraging in operationally treated cut blocks.
Take-home messages:
Recovery of biomass in fireweed, raspberry, aspen, and willow plants varied across cut blocks. Overall, fireweed and willow showed the most resilience, regaining a large portion of their biomass, indicating a higher availability of browse for moose or other ungulates. This creates the potential for higher exposure to glyphosate as these cut blocks are preferential to foraging moose. With a very low density, or no large trees, and higher availability of browse, wildlife is more likely to select these harvested blocks. However, the concentration of glyphosate remaining within new shoots two years after application may be inconsequential to browsing animals. Although the long-term effects of exposure to glyphosate are not well documented, we can be >99% certain that levels of glyphosate within these four studied species will not exceed 0.1 mg/kg after two years of recovery within these cut blocks, greatly reducing the exposure rate to ungulates or other wildlife foraging here. These results can be used by forest managers and stakeholders to determine the best course of action when making decisions about how to responsibly regenerate our harvested cut blocks in Alberta.