Relationships betw een tree volume and other factors
3.2.3 Relationships betw een tree volume and other factors
There was a clear relationship between mean plot standing volume and mean plot penetration resistance by several ways of aggregating the penetration profile data. To obtain a single measure of penetration resistance for each plot, we examined the means over a number of depth ranges. The best relationship with plot tree volume was for mean penetration resistance over the depth range of
25 to 100 mm (Figure 23), in other words at rather shallow depth. The straight-line relationship over the range of strength values examined was consistent with the results of many other studies that have shown a decline in root activity and plant vigour between 1.0 to 3.0 MPa. Perhaps the 25 to 100 mm (Figure 23), in other words at rather shallow depth. The straight-line relationship over the range of strength values examined was consistent with the results of many other studies that have shown a decline in root activity and plant vigour between 1.0 to 3.0 MPa. Perhaps the
y = -11.517x + 32.631 5 R 2 = 0.586
Tree volume index (m
Penetration resistance (MPa)
Figure 23. Mean of plot tree volume for the five largest trees versus mean plot penetration resistance over 25-100 mm depth. Snig plots include track area only.
When plot volumes were plotted against plot penetration values, the relationship shown in Figure
23 could not be detected. Although a scatter plot showed a slight suggestion of a relationship, it was not significant (P=0.55 for the five largest trees volume per hectare and P=0.16 for mean of the two largest trees). This further demonstrates the degree of noise to signal in this environment.
Exploratory data analysis of the effects of other variables on tree volume, at both the plot volume and two-largest tree scale, found no significant trends other than for aspect. If any of the measured variables was exerting any influence, it was weak and would have required a much greater number of samples to define any relationship.
Because of the general aspect of the site, plot aspect was mainly distributed between northerly and westerly segments of the compass. Classification and regression tree analysis showed a strong division in the data, with more westerly and southerly aspects (the latter of which was represented by few plots) being associated with lower plot tree volumes. The opposite was the case for northerly and easterly aspects (again with relatively few plots having easterly aspect). Because there is no intuitive basis for a linear relationship between aspect and tree growth, we first re-coded aspect into a categorical variable with two classes – NE and SW. We then examined a two-way analysis of variance with disturbance and aspect being the independent class variables. Although both variables were significant on their own, there was no two-way interaction. Table 10 illustrates the strong influence of aspect on growth and shows that the effect was relatively uniform across most classes (except access tracks and landing snig tracks).
Table 10. Least square mean of five largest trees per plot by disturbance class and aspect class.
Disturbance class
Aspect class
Least square mean plot volume of five largest trees (m 3 ha -2 )
Access road
NE
Access road
SW
Landing snig
NE
Landing snig
SW
7 4.27 Landings NE 5 14.87 Landings SW 7 12.15
Light disturbance
NE
Light disturbance
SW
Major disturbance
NE
Major disturbance
SW
Major snig
NE
Major snig
SW
Minor dist
NE
Minor dist
SW
Minor snig
NE
Minor snig
SW
The lack of a significant aspect by disturbance class interaction suggests that we could have stopped the analysis at this point. However, we re-coded aspect to obtain a variable that might have some linear relationship with tree volume. The aspect values in the northern quadrant (315-45 degrees) were adjusted so that they were gradually rising as they drew closer to 0 degrees (so that 0 became equal to 90). Values within the western quadrant (225-315 degrees) were converted to be gradually decreasing as they drew closer to 270 (270 became equal to 0). Aspects between 45 and 270 were held constant at a value of 45 (there were only around 8 plots within this range). Thus, we created a linear variable for our measured range of aspects that was believed to approximately represent the influence of aspect on tree volume. This may not have been an entirely valid way to treat the data, but the new variable did form a significant linear relationship with plot tree volume (P=0.0003, r-square=0.12). However, it provided no useful improvement to the analysis of variance for disturbance class when used as a covariate. In fact, the only influence of the new aspect variable was to render the entire model non-significant, presumably because it raised the values of the access track and landing snig track classes due to their not sharing the same relationship with aspect as the overall data. This was not entirely surprising given that the aspect by disturbance model was not significant. Nevertheless, we judged that the exercise was of sufficient interest to present here.