Mapping successional boreal forests in interior central Alaska

Abstract

It is necessary to develop new satellite methods to monitor boreal forests responses to climate warming. Remotely sensed maps derived from hyperspectral Advanced Visible/Infrared Imaging Spectrometer (AVIRIS) data were developed and compared for the four conifer forest ecosystems at the Bonanza Creek Experimental Forest, a long-term ecological research site on the Tanana River flood plain near Fairbanks, Alaska. The site was first stratified into montane, lowland alluvial plain, and flood plain zones based on topography. A classification of six forest and three non-forest types was created from AVIRIS images and compared on a pixel-by-pixel basis to a published vegetation map, a classified SPOT (Satellite Pour l'Observation de la Terre) image, and a hybrid SPOT image and digital elevation model classification. A comparison of AVIRIS with SPOT results showed that the AVIRIS classification was consistently more accurate (74, 43, and 43% overall accuracy, respectively). Hyperspectral classification methods have promise for mapping forest ecosystems in other boreal regions when little or no ground data are available for validation. The time difference between the creation of these maps show that substantial ecosystem changes have occurred over the past 15 years, demonstrating the need for developing a capability to obtain cost-effective landscape characterization.