Deriving a More Detailed Vegetation Classification from NALCMS for Anchorage, Alaska

Continental-scale medium resolution landcover classifications like the North American Land Change Monitoring System (NALCMS) provide much needed uniform data; however, its categories are too crude to be useful for permafrost or fire modeling in Alaska, which requires the identification of black spruce stands. Field data, on the other hand, is too detailed with often 20 or more unique stand types that are too complex to scale up or model easily.

We used elevation, aspect, slope, and mean growing season temperature to statistically assess the distribution of 30 stand types from an expansive field inventory dataset within the municipal area of Anchorage, Alaska. The results were then used to reclassify NALCMS into more useful categories such as black spruce, white spruce, mixed spruce, deciduous forest/tall shrub, tundra, grassland, bare ground, coastal forest (wetland), urban or developed, and water areas. This improved landcover dataset was then verified with two other field inventories.

Assessing how closely classifications match each other is complicated by the fact that each field inventory, NALCMS, and our classification use different definitions for vegetation categories, but preliminary results indicate accuracies of 60-80% for some landcover categories. After further refinement, the goal is to ultimately develop an approach that allows us to reclassify NALCMS into the desired vegetation categories in several western arctic study areas which will then be used for permafrost modeling and wildfire risk assessment.