| Summary: | To assess relationships within the biogeoclimatic ecosystem classification(BEC) system, 102 sample plots were established in immature western hemlock [Tsuga heterophylla (Raf.) Sarg.] stands distributed within the sub montane very wet maritime variant of the Coastal Western Hemlock Zone in coastal British Columbia. Using the methods described in the BEC system, plant associations, and field derived soil nutrient regimes (SNR) and soil moisture regimes (SMR) were identified. Direct measures of SNRs, i.e., soil chemical measures of the forest floor and mineral soil expressed on a concentration basis, and site index (height at a reference age) were determined for each stand.
Despite a lack of understory species, both plant associations and 6diagnostic species were linked to an underlying nutrient gradient. For the former, the use of canonical discriminant analysis on the principal components analysis (PCA) scores of the soil chemical measures showed a definite but overlapping trend. This trend was correlated most positively with nitrogen, and negatively with the C:N ratio and potassium. For the latter, canonical correlation analysis of 6 diagnostic species with 4 forest floor chemical measures resulted in the 4 chemical canonical variates explaining 37% of the variance in the species domain. Generally, oxylophytic species varied negatively with total and mineralizable nitrogen, and positively with available potassium and magnesium. The relationships with nitrophytic species were reversed.
A PCA ordination showed the soil nitrogen measures to be the most important factors in accounting for the variation in the soil chemical data. Discriminant analysis, used to see how well the nitrogen properties could distinguish SNRs, correctly classified 91% of the plots into their source group. However, this high success rate was not repeatable; on a validation set, the discriminant function correctly classified only 54% of the plots.
All regression models reported, relating site index (m/50 yr) to indirect and direct variables, showed significant (p<0.05) results, and had adjusted R2 values ranging from 0.35 to 0.81. The standard errors of estimate (SEE) were relatively high, ranging from 4.5 m to 8.8 m. The best fit regression equation, having the highest adjusted R2 value and lowest SEE, was the categorical model which related site index as a function of SMRs and SNRs. The best fit analytical model related site index as a function of nitrogen positively and potassium negatively (adjusted R2 = 0.67,SEE = 5.61m). However, this model failed to adequately predict, within 3meters, the site index of the data set from which it was derived. Applied to a test data set, the prediction results improved but the equation tended to underestimate site index. Multi collinearities among the soil nutrient properties were noted and therefore, PCA regression used to supplement the interpretation. Examination of the loadings of the axes of the final regression equation (adjusted R2 = 0.63, SEE = 5.89) indicated that the strongest relationships with site index were positively related with sulphur and logarithmic transformations of nitrogen, and negatively with the C:N ratio and potassium. Supplementary relationships were also noted.
It was concluded that there exists a relationship between nitrogen measures and field derived SNRs. Further, the SNRs can be combined with SMRs to predict western hemlock site index. As well, soil chemical measures can also predict site index. However, there was a relatively large variation associated with predictions made using the regression equations.
|
|---|
