The methodology used to produce the national forest type database combined ground-truth (from FIA plot data) with multi-date imagery and variety of other spatially continuous geospatial data. The predictor data themes include,
- Elevation, slope, and aspect
- Unified Ecoregions
- MODIS Vegetation Indices such as EVI, NDVI.
- - MODIS Vegetation Continuous Fields
- - Percent tree cover, percent herbaceous cover, and percent bare ground
- MODIS fire points for developed from the MODIS Active Fire Maps
- - MODIS 8-day composites
- - Multiple dates from the spring, summer, and fall
- All MODIS 32-day composite images that are cloud-free between the years 2001-2003
- Alaska Vegetation Classification
Statistical models developed in Rulequest's See5 data mining software link the FIA plot variables with the imagery and geospatial data. See5 creates classification trees, which have the advantage of not assuming parametric properties within the predictor data and are thus are more appropriate for the multi-scale, multi-source data, which are being used.