A Novel Approach to Mapping Vegetation Cover in Alaska Using the Multi-angular Imaging SpectroRadiometer (MISR)

Over the last half-century, shrub density has been increasing in Arctic Alaska, and forest cover across Interior Alaska has been significantly altered by wildfires, disease, and succession. As the climate continues to warm across the Arctic, there is reason to expect continued changes in forest and shrub cover. Comprehensive, quantitative inventories of shrub density (i.e., cover and height) across Arctic Alaska and forest density throughout Interior Alaska do not currently exist but are necessary to provide a baseline for land use- land cover change studies. Most existing large-area land cover classifications in the region have been derived from Landsat imagery and suffer from two major limitations stemming from the relatively infrequent and inconsistent acquisition of Landsat scenes over Alaska: (1) the inability to characterize the entire region of interest for one single year or growing season, and (2) inconsistent classification of similar vegetation types located in different component scenes within the scene mosaic due to inter-scene differences in vegetation phenology. In order to overcome these limitations and produce consistent forest and shrub density maps that characterize conditions for a single year, it is necessary to utilize sensors that have collected imagery more regularly and more consistently across Alaska. The Multi-angular Imaging SpectroRadiometer (MISR) and the Moderate-resolution Imaging Spectroradiometer (MODIS), both on NASA’s Terra spacecraft, have been consistently imaging Alaska since early 2000. The objective of this study was to assess the capabilities of these two sensors for mapping shrub and forest cover in Arctic and interior Alaska. Preliminary mapping efforts using regression trees indicate that, while both MISR and MODIS can detect shrub and forest cover, MISR-derived cover maps are consistently more accurate than MODIS-derived maps, and rival the accuracy of Landsat-derived maps. The multi-angular data available from the red spectral band from MISR, as well as the higher spatial resolution of the green and blue spectral bands (275 m for MISR versus 500 m for MODIS) appear to give MISR an edge for forest and shrub cover mapping applications in Interior and Arctic Alaska. Preliminary results also indicate that forest cover maps derived from late winter MISR scenes with snow cover are more accurate than maps derived from MISR scenes acquired at any point during the summer. This is encouraging, as late winter is one of the most consistently cloud-free periods in Interior Alaska and the production of cloud-free late winter mosaics should be feasible for most years, potentially allowing for the development of annual forest cover and height products. Future research will explore the potential for using MISR to map shrub and forest height (in addition to density) in Arctic and Interior Alaska.