Aspen
Seismic line density appears to have only a small effect on bird community composition in aspen stands. If composition was strongly influenced by seismic line density, stations belonging to the same density category would be close together in the NMDS ordination but instead there is no discernible pattern (Figure 18). Figure 18 shows the three-dimensional solution of the NMDS (stress = 0.22), as the stress level was high when only two dimensions were used (stress = 0.32). When stress levels are above 0.3, presentation of the data may be poor and an increase in the number of dimensions should be considered (Zuur et al. 2007). Vectors representing four bird species (mourning warbler, Connecticut warbler, red-breasted nuthatch, and black-throated green warbler) are also shown in the plots. To more closely examine trends in abundance of individual species in relation to treatment type, I created graphs in which points were sized by the abundance of the species such as the mourning warbler (Figure 19). These graphs confirmed that there was little effect of treatment type on abundance for most species. Despite the lack of a pattern in the NMDS, the MRPP indicated that there were significant differences between the four density categories (P < 0.0001). Pairwise comparisons suggested that this was largely due to a difference between the "no lines" category and the other three categories. Therefore, seismic lines seemed to have a small effect on bird community composition in aspen stands, but did not result in large shifts in species composition.
Because seismic line creation in northern Alberta is expected to continue to accelerate in the future, further research on songbird community composition at higher line densities than those included in this study (~>8km/km2) could be informative. There may be a threshold at which a larger number of species respond strongly to seismic line density. For example, ovenbirds in aspen stands show no response to seismic line densities less than 8.5km/km2, but decline in abundance at a rate of 19% per 1km/km2 increase in line density above this threshold (Bayne et al. 2005). In Alberta, there is currently no regulation of the cumulative density of lines in an area, so extremely high densities may well become a reality.
Figure 18. NMDS plots of a) dimension 2 versus dimension 1 and b) dimension 3 versus dimension 1 for aspen abundance data. Points are colour-coded by treatment (green = 0 km/km2, blue = 0-3 km/km2, black = 3-6 km/km2, red = >6 km/km2). Orange vectors represent trends in four species (MOWA = mourning warbler, CONW = Connecticut warbler, RBNU = red-breasted nuthatch, BTNW = black-throated green warbler).
Figure 19. NMDS plots of a) dimension 2 versus dimension 1 and b) dimension 3 versus dimension 1 for the aspen dataset. Points are sized by abundance of mourning warblers and colour-coded by treatment (green = 0 km/km2, blue = 0-3 km/km2, black = 3-6 km/km2, red = >6 km/km2).
Black spruce
Seismic line density did not appear to affect songbird community composition in black spruce stands. Sampling units belonging to the different treatments did not cluster together in the NMDS plot, as would be expected if treatment influenced bird community composition (Figure 20). Stress was again high when only two dimensions were used (stress = 0.32), and consequently the three-dimensional solution is shown in Figure 19 (stress = 0.23). Vectors of trends in abundance of four bird species (chipping sparrow, dark-eyed junco, Tennessee warbler, and Swainson's thrush) are also shown in the plots. Plots of points sized by abundance were again created to better visualize the data but are not shown, because no clear patterns were evident and they were therefore similar to Figure 19. The MRPP confirmed that there were no significant differences between the three seismic line treatments (P = 0.18). In studies such as this one where no significant effects are found, it is possible that the sample size used was too small to allow detection of a difference between treatments. However, a simple power analysis in PopTools (Hood 2006) indicated that there was sufficient power to detect a difference in abundance of at least 20% given the sample size and amount of variability in the data. This suggests that the sample size in this study was large enough to detect a difference between treatments.
This lack of a response to seismic line density does not mean that line creation can continue unabated without impacting bird community composition because, as discussed above, threshold effects may occur. In addition to examining areas with higher densities, research on the effects of 2-3 m wide low-impact seismic lines is also necessary. New technology has recently allowed these lines to become a viable alternative to the conventional 8-10 m lines, and they are likely to be more common in the future. Although they are narrower, the low-impact lines are spaced more closely together (~200 m) than conventional lines and therefore their effects on songbird communities may differ.
Figure 20. NMDS plots of a) dimension 2 versus dimension 1 and b) dimension 3 versus dimension 1 for songbird abundance in black spruce stands. Points are colour-coded by treatment (green = no seismic lines, blue = one seismic line, black = two seismic lines). Brown vectors represent trends in four species (CHSP = chipping sparrow, DEJU = dark-eyed junco, SWTH = Swainson's thrush, TEWA = Tennessee warbler).
Site-level results |
Figure 21. NMDS plot of mean abundance per treatment per site in the black spruce stands. Points are colour-coded by treatment (green = no seismic lines, blue = single line, black = two lines). Brown vectors represent four species (CHSP = chipping sparrow, DEJU = dark-eyed junco, SWTH = Swainson's thrush, TEWA = Tennessee warbler). |
Detectability
NMDS plots of mean density and mean simulated density per treatment per site both showed no relationship between treatment type and bird community composition in the black spruce data (Figure 22). As described previously, calculation of densities allowed me to correct for differences in detectability between species. In the simulated density dataset, I examined the effects of adjusting for detectability differences between treatment types as well, by increasing the effective detection radius for each species by 10% in the single line treatment and a further 10% in the double line treatment. In other words, it was assumed that birds would be easier to detect in habitats that were more open (i.e. with more seismic lines). In both NMDS ordinations, the two-dimensional solutions had acceptable levels of stress (stress = 0.28). There did not appear to be any substantive differences between these plots and the NMDS plot of the abundance data (Figure 21). The MRPPs were also not significant for either the density (P = 0.345) or simulated density data (P = 0.0718). These results suggest that adjusting for differences in detectability between species or treatments by calculating densities may have little effect on the results in these types of multivariate analyses. However, given the significant effects that accounting for detectability can have in univariate analyses (Thompson 2002), further investigation of the sensitivity of multivariate results to detectability is needed before any general conclusions can be made.
Figure 22. NMDS plots of a) mean density and b) mean simulated density per treatment per site in black spruce stands. Points are colour-coded by treatment (green = none, blue = single, black = double seismic). Brown vectors represent four species (CHSP = chipping sparrow, DEJU = dark-eyed junco, SWTH = Swainson's thrush, TEWA = Tennessee warbler).