Methodology
Study area
The data collection was conducted in the BERA study area in the upper Athabasca oil sands region in the summer 2022 (form June to August). I sampled 30 sites in 15 cut blocks distributed in five sample areas from Fort McMurray in the North and Lac la Biche in the South (figure 5a, 5b, 5c and 5d).
Figure 5a: Sampling plots in the Fort Mackay area.
Figure 5b: Sampling plots in the Grayling Creek and Anzac areas.
Figure 5c: Sampling plots in the Lac la Biche area.
Figure 5e: Sampling plots in the Conklin area.
Study design
Table 2: List of species sampled with the regeneration survey.
Site selection was conducted using the Geographic Information System QGIS 3.22.10 to isolate the intersections between Conventional seismic lines (more than 3m width; ABMI 2018), Mesic Upland Ecosites and cut-blocks. Out of 30 sites, 15 represented the sampling where the clearcutting occurred (“experimental site”) and 15 the control units in the uncut forest (“control sites”), (figure 6). For each experimental site, I set up a 20 m2 plot inside the seismic line (“on-footprint plot”) and one on the adjacent forest (“off-footprint plot”), (figure 7). For control sites, I was interested only in "on-footprint" plots. Only cut blocks that have been cut after the seismic line, at least 300 m distant to each other and with the same maximum age (10 years) have been retained. At the end of this process, I obtained 15 replicates for each of the following combination of disturbances: "line" (on-footprint plots in control sites), "cut-block" (off-footprint plots in experimental sites) and "lines in cut-block" (on-footprint plots in experimental sites).
To estimate the regeneration rate for the different disturbance scenarios, I counted all non-germinant and alive tree and shrub stems by species and height class inside each plot. The species sampled withthis regeneration survey (table 2) were selected according to BERA recommendations, including species that can slow animal movements, by increasing horizontal and vertical structure of the line. I present the stem density as both the number of stems counted inside the 20 m2 and the number of stems per hectare (multiplying the abundance from the plot by 500). The number of stems per hectare are used to compare my results with the highest density class in Alberta's reforestation standards (5,000 stems/ha; Alberta Agriculture and Forestry 2018). Height categories have been selected according to BERA recommendations as: (1) < 120 cm for deciduous species and < 80 cm for coniferous species and (2) ≥ 120 cm for deciduous species and ≥ 80 cm for coniferous species. The assumption at the base of this choice is that only stands in height category 2 are capable of efficiently interrupt the connectivity of the line.
The hypothesis formulated a priori were tested investigating two main aspect of seismic lines regeneration:
(i) How forest harvesting affects the total trees regeneration inside the line?
This analysis was focused on investigating whether forest harvesting promotes the ability of seismic lines to reach an appropriate stocking level. This aspect was investigated analyzing the total trees density (i.e., no shrubs) from each disturbance scenario with the two height categories considered separately.
(ii) How forest harvesting affects the species composition inside the seismic line?
This analysis was focused on investigating how stems regeneration changes qualitatively in species composition based on the presence of forest harvesting. This aspect was addressed analyzing separately each species group and single species and checking for differences in stem density between the disturbance scenarios. I included both trees and shrubs species and I considered only stem densities in height category 2.
Figure 6: Schematic representation of two sampling sites.
Figure 7: Schematic representation of two sampling plots.
Data analysis
To visualize and estimate differences in stem density between different disturbance scenarios, I plotted the means of stem densities against the presence/absence of cut blocks and seismic lines. This initial analysis was performed for both question (i) and (ii), and it was focused on estimating the effect size of the differences between disturbance scenarios on raw data. Secondly, to test the significance of the effect of forest harvesting and seismic lines on stems density, and to test the two hypotheses formulated a priori (table 1), I fitted generalized linear mixed models. I chose these models to account for the spatially grouped nature of data. Sample areas and site IDs are included as nested random effects to account for the possible spatial autocorrelation. The effect of the two disturbances is addressed creating binary dummy variables and representing the presence (1) and absence (0) of seismic lines and clearcutting. The response variable was represented by the stem density per 20 m2 (total or for each species/species group depending on the question) for both question (i) and (ii). Models are performed using the function glmmTMB() from the “glmmTMB” package. All analyses were performed using R, version 4.2.0 (R Core Team 2022).