| Summary: | Natural resource management in northern boreal forests requires tree species identification for improved decision making. Satellite remote sensing provides a more cost-effective and time-efficient way to obtain this information in these large, remote, inaccessible areas. However, satellite signals are highly mixed due to increased tree shadows and visible understory vegetation in higher latitude, lower density open forests. Thus, methods used in southern forests are largely unsuitable. Therefore, spectral mixture analysis (SMA) was tested as it separates these signal components (trees, understory, shadow) at sub-pixel scales, allowing improved forest information. In this study, SMA was used to identify the dominant species near Fort Providence NWT using Landsat-5 Thematic Mapper imagery. An accuracy of 79 % was achieved for four species validated against 48 ground plots using multiple-date imagery acquired at different stages of the growing season. These positive results indicate SMA’s capability to retrieve species information of highly mixed open stands.
|
|---|