A model for predicting boreal vegetation dynamics and management requirements on electric transmission right-of-ways, Interlake region, Manitoba

Thumbnail Image
Walker, David John
Journal Title
Journal ISSN
Volume Title
The objective of this study was to develop a predictive model of vegetation dynamics and management requirements on a HVDC electrical transmission corridor through the boreal forest of the Interlake region, Manitoba. The model was based on correlations between forest vegetation (obtained through ground truthing) and LANDSAT TM image analysis. Thirty sites along the transmission right-of-way (ROW) were selected for ground truthing. At each site, transects were established in the forest and adjacent ROW, and vegetation (species percent cover in quadrats) and edaphic information collected. The sites were classified using cluster analysis, based on the forest vegetation data. Three recently burned sites were obvious outliers, and were treated separately in subsequent analyses. Three forest vegetation groups were recognised: dry coniferous, wet coniferous, and mixed forest. Ordination methods were used to reduce the dimensionality of the data so as to simplify subsequent analyses. Discriminant analysis of the forest and ROW vegetation indicated statistically significant discrimination of the three vegetation groups in two-dimensional ordination space. Correspondence between the forest and adjacent ROW vegetation was tested using canonical correlation analysis. The results showed a statistically significant correlation (R2 = 0.77, p < 0.01) between vegetation of the forest and that of the adjacent ROW. The strength of this relationship suggested that a model to predict ROW vegetation based on forest vegetation could be developed. Tree recruitment on the ROW was summarized for each vegetation group. Wet coniferous sites had the highest tree density, and dry coniferous sites the lowest. Vegetative propagation (suckering or layering) was the predominant recruitment method in the wet coniferous and mixed forest sites. The results also indicated that black spruce was not affected by current management techniques designed to reduce ROW tree density. At the recently burned ROW sites, post-fire recruitment of jack pine was high. LANDAT TM spectral reflectances (bands 3, 4 and 5) for 25 of the 30 forest sites were analyzed to determine their correlation with the forest vegetation. Multiple discriminant analysis indicated statistically significant discrimination of the vegetation groups based on the three LANDSAT bands. Based on these results, a predictive discriminant classification model based on spectral reflectances in bands 3, 4 and 5 was developed. Using a Mahalanobis distance criterion, the model correctly classified 2l of the 25 forest sites. A test of model robustness was undertaken by classifying an independent random sample of 40 sites. All but two of these 40 sites were successfully classified at the a = 0.05 probability level. The predictive model is based on observed high correlations between forest and ROW vegetation, and between forest vegetation and LANDSAT TM spectral reflectances (bands 3, 4, 5). Using the model, sites along the HVDC right-of-way in the Interlake region can be classified into one of three vegetation groups. Once classified, ROW vegetation and tree recruitment at the site is predictable, which in turn suggests specific management requirements. The model could also be used to predict specific management requirements for a newly developed ROWs in the region.