Regression Tree Analysis
What Went in?
In the initial analysis, these variables were included in the regression tree:
rpart(SC_HT~LWIDTH+SOLAR+AGE+ADJ_SP1+SOIL_NUT+SOIL_MOIS+SC_UTIL+ADJ_TYPE+ADJ_LC,method="anova",data=Rdat)
rpart(SC_HT~LWIDTH+SOLAR+AGE+ADJ_SP1+SOIL_NUT+SOIL_MOIS+SC_UTIL+ADJ_TYPE+ADJ_LC,method="anova",data=Rdat)
Age matters.
The initial regression tree analysis shows that the year since disturbance accounts for the most variation in canopy height online compared to other variables. So much in fact, that it determines every branch of the tree. At the first node, time since disturbance accounts for 23.6% of the variation. If a seismic line was cleared less than 18.5 years prior then average canopy height on the line is 0.900 m. If the line was cleared for more than 18.5 years, mean canopy height is 2.822 m. The greatest heights were achieved (mean=4.07 m, Fig. 1) if the line was cleared over 26 years ago. The lowest mean height was 0.54 m and this occurred if the line had been cleared for less than 8.5 years.
It is logical that there would be a strong relationship between recovery and time (Fig. 2). It is interesting that it takes at least 18 years to reach a mean height over 2 m naturally on the seismic line after clearing.
It is logical that there would be a strong relationship between recovery and time (Fig. 2). It is interesting that it takes at least 18 years to reach a mean height over 2 m naturally on the seismic line after clearing.
While it is helpful to have an understanding of the length of time natural recovery occurs, planners interested in restoration and silviculture treatments to aid recovery need to know more. I ran the regression tree analysis a second time and removed "time since disturbance" (AGE) from the predictor variables.
THis needs further investigation!
The modified regression tree results in the adjacent stand composition accounting for the most variation in canopy height online (4.52%, Fig. 3). If the adjacent stand to the seismic line is coniferous (instead of mixedwood or deciduous), the mean canopy height is 1.01 m. This indicates that there is limited recovery in coniferous forests (Fig.4). There is also an impressive 3260 samples that are coniferous of the total 4995. The majority of seismic lines in the region occur in coniferous forests. This presents a daunting outlook on the situation for recovery of seismic lines.
There were "runners up" for important factors for the first split. The next factors were leading conifer (3.3%), the dominant adjacent species (3.1%) and soil nutrient (2.8%) and moisture (2.0%) availability. Clearly, this shows that ecosite and adjacent stand have a role in recovery, albeit a small one. Surrogate splits in the tree include line width and line utilization. The regression tree analysis supports the relationship observed in the data exploration for line width. If the width is less than 5.5 m there is less recovery than between 5.5 and 7.5 m. Mean canopy height doubles if there is no presence of vehicle tracks on the line (continued disturbance). |
Back to the hypothesis.I hypothesized that increasing availability of moisture and nutrients, dominant adjacent species and low disturbance frequency would be most important in determining recovery rates.
The study found that contrary to the hypothesis time since disturbance was by far the most important determinant of recovery. In concordance with the hypothesis, low disturbance frequency (ie. no ATV or truck usage) and the species present in the adjacent stand did have a role in recovery when time since disturbance wasn't considered. |
Conclusion: where to restore?
Seismic lines over 18 years old, in deciduous or mixedwood forests (especially white spruce, jack pine, aspen, balsam fir, or poplar dominated), have no evidence of frequent ATV or vehicle usage, with a width between 5.5-7.5m are likely to regenerate naturally without restoration.
Seismic lines that are coniferous, less than 18 years old, with poor nutrient availability, high soil moisture content, width greater than 7.5 m or less than 5.5 m, and have frequent disturbance from ATVs and other vehicles are unlikely to regenerate even with expensive restoration treatments.
Restoration efforts should be focused on those seismic lines that lie in between these two site conditions. Appropriate silviculture treatments could be tested on these sites first. Unfortunately, trees in the northern boreal forest grow slowly and any restoration will inevitably take many years.
Seismic lines that are coniferous, less than 18 years old, with poor nutrient availability, high soil moisture content, width greater than 7.5 m or less than 5.5 m, and have frequent disturbance from ATVs and other vehicles are unlikely to regenerate even with expensive restoration treatments.
Restoration efforts should be focused on those seismic lines that lie in between these two site conditions. Appropriate silviculture treatments could be tested on these sites first. Unfortunately, trees in the northern boreal forest grow slowly and any restoration will inevitably take many years.
Implications for conservation.
This study shows that natural recovery of seismic lines is limited. Assisted recovery through silviculture treatments may also have limited success because generally, besides time, none of the factors that could be used to develop restoration methods appear to account more than 5% of the variation in seismic line regrowth. Realistically, the amount of time it takes to restore these seismic lines does not meet the rapid increase in industrial development in northern Alberta. Estimates predict that threatened woodland caribou will be extirpated from Alberta within the next century (Schneider et al. 2010). Since time is essential, for both line recovery and boreal-specialists, perhaps the most important thing we can do is conserve the intact old-growth boreal forests that are still remaining. In terms of restoration, planners may have to do their best to aid recovery and wait 20 years for results.