| Abstract | This study investigates the proportion of the Mackenzie Delta, Northwest Territories, underlain by near-surface permafrost. The objectives were to assess if the distribution of vegetation can be used to predict permafrost presence and, if so, to determine the proper remote sensing techniques that can be applied to map these vegetation communities with SPOT-5 images. The resulting maps could then be used to estimate the extent of near-surface permafrost. On point bars and on alluvial islands, Horsetail zones in the southern and central delta as well as Willow-horsetail communities throughout the Delta were associated with absence of permafrost in the upper 3 m. Permafrost was present beneath all other vegetation associations as well as at other landforms. NDVI and MSAVI as well as texture information were found to be useful for discriminating between vegetation communities. Three classifiers were tested: Maximum Likelihood (ML), Artificial Neural Networks (ANN), and Linear Spectral Unmixing (LSU). ML achieved the highest overall accuracies of 84.2%, 82.4%, and 83.4% for the southern, central, and northern delta image, respectively. LSU was useful in studying vegetation gradation from one community to another. Zones with presence of near-surface permafrost occupy 91%, 95%, and 96% of the land surface, in the southern, central, and northern delta, respectively. This indicates that the Mackenzie Delta is part of the continuous permafrost zone. The permafrost classification resulting from this research is in agreement with classifications based on climate data but differs with the most recent permafrost map in the Atlas of Canada categorizing the Delta as having discontinuous permafrost. |
