Although a regression equation with any utility in predicting suitability for old-growth retentions areas did not materialize, the multiple regression analysis did provide some small clues into a potential relationship between the landscape variables and stand age in this particular study area.

Based on the estimated regression parameters produced in the analysis, a very slight increase in stand age was significantly associated with a southern aspect, lower elevation, shorter distances to large (>500ha) and small lakes (<100ha) and a greater distance to medium lakes (100-500 ha).

These results may be somewhat contrary to what would be expected of a stand age-topography relationship. Due to reduced direct solar radiation, north-facing slopes tend to have lower air and soil temperatures as compared to south facing slopes (Redding et al 2003). It would be expected that these north-facing slopes would thus have less tendency to burn during the fire season and be of an older age-class. The results of this analysis provided evidence to the contrary. In the study area the majority of stands have a very low percent slope (Mean=5.1%). It may be possible that the slope value was not great enough to detect a north-south aspect effect.

The distance to lake features and the apparent relationship to stand age appears somewhat contradictory. Stand age appears to increase slightly in relation to close proximity to large and small lakes while age decreases in close proximity to medium lakes. The dynamics here are somewhat unclear and would require further investigation. It should be noted however that although a relationship between stand age and distance to lakes seems plausible, it is debatable. A recent study in the boreal mixedwood of northern Alberta found no effect of distance to a lake on stand age (Macdonald et al 2004).

Regardless of the significance of these relationships, the estimated regression parameter values are so minute that all five of these variables appear to have very little impact on stand age in the overall regression equation. Of the five significant variables, aspect and elevation have the greatest influence although it is still relatively slight.

The canonical correlation analysis also provided somewhat mixed results. Although two weakly correlated, yet statistically significant, canonical variate pairs were produced, the canonical variates did not represent the late successional stand compositions as desired. The first, and most significant, canonical variate pair provided clues to a relationship between jackpine leading stands with a black spruce component and higher elevations, greater distances to large lakes and closer distances to small lakes and streams. The second canonical variate pair represented white birch leading stands with a trembling aspen component and greater slopes, higher elevations and greater distance to wetlands. This pair also represented stands that were clearly not associated with black spruce.

In many regards these results reflect the typical site characteristics of the focus tree species. For example, a jack pine stand with a black spruce component is often associated with an upland site (higher elevations) and a white birch stand is typically associated with steeper slopes on upland sites. These results do not, however, tell us a great deal about late succesional species associations and our topographic variables. Ideally, a canonical variate pair representing a high jack pine frequency and no black spruce component and a second pair with a high black spruce frequency with a lesser jack pine component would have resulted. An examination into the different topographic characteristics of these species associations may have provided some preliminary clues as to the differences between the late-successional-landscape feature relationship and the early successional-landscape feature relationship.
Overall, the results of this analysis was inconclusive with regards to detecting relationships between old-growth stand characteristics and topographic variables. This could be attributed to a variety factors, first and foremost being rooted in the complicated nature of natural disturbance dynamics. Landscape level age-class distribution is a function of natural disturbance agents including fire, insects, disease and wind. Many factors are known to influence natural disturbance events including local and regional weather patterns, understory and overstory vegetation, topography and soils. There is also a large degree of randomness in natural disturbance patterns due to relatively random events such as lightning strikes. An attempt to estimate the spatial pattern of boreal old-growth based solely on landscape feature is likely an oversimplification. The analysis, as perform in this study, is missing key variables which render the results to be of very little use.