5th Symposium on Fire and Forest Meteorology and the 2nd International Wildland Fire Ecology and Fire Management Congress

Monday, 17 November 2003
Factors affecting vegetation cover mapping for Landfire
James Vogelmann, SAIC and USGS, Sioux Falls, SD; and C. Huang, B. Tolk, and Z. Zhu
Poster PDF (1.4 MB)
The LANDFIRE project is a joint effort between USDA Forest Service and Department of the Interior agencies to provide the spatial data and predictive models required for characterizing fuel conditions and fire regimes and for helping to evaluate fire hazard status. A significant component of the research involves development of a detailed land cover classification data layer that can be used in conjunction with other spatial data layers for input to various fire fuels and fire characterization models. In the current investigation we are conducting a pilot project in central Utah to develop and fine-tune land cover generation methodology to meet LANDFIRE requirements. Three dates of Landsat 7 Enhanced Thematic Mapper Plus (ETM+) satellite data, digital elevation model data and biophysical settings information, are the primary spatial data sets being used to generate the vegetation map information. Extensive field information (6172 plot locations) has been made available to this project from the US Forest Service Forest Inventory and Assessment (FIA) and Utah State University. Digital information from each data layer is extracted for each plot location, and decision tree analysis is being used to generate the land cover classifications. The relative importance of the various data layers for vegetation mapping and determination of the number of field plots necessary for optimal mapping are under investigation. First-order accuracy values are being determined using cross-validation approaches. Research thus far indicates that some vegetation communities, such as pinion pine-juniper, are very distinct and are relatively straightforward to map. However, others, such as Douglas-fir and white fir, are spectrally very similar and represent potential classification challenges. We are working to optimize the use of the available field information with the spatial data in order to generate the best large region land cover products possible in an operational framework.

Supplementary URL: