Mapping forest biomass on several pilot regions in Canada with Landsat TM and forest inventory data

摘要

A method has been developed to map forest biomass with Landsat TM and ETM imagery coupled with forest inventory data. The method involves applying an unsupervised classification to a Landsat TM/ETM scene. The unsupervised clusters are labelled according to cover types and forest structure (crown closure and height) using random samples extracted from Inventory datasets as training pixels. Biomass values are assigned to the clusters according to the dominant forest type and structure contained within the clusters. The method was tested on five pilot regions throughout Canada with an objective of evaluating its application in several ecological regions with different species composition and stand structure. This paper concerns (1) the implementation of this method and (2) the comparison of those results between regions. The overall results for classifying forest cover types range from 50 to 60% for the five regions. Correlations between remotely sensed and inventory biomass estimates are variable for different species and pilot regions. However, results show that the method provides a complement to existing inventory-based methods for mapping biomass in managed areas and may constitute an alternative approach for northern areas with weak forest inventory databases.