The files linked to this reference are the geospatial data created as part of the completion of the baseline vegetation inventory project for Olympic National Park. The vegetation map is a geotiff raster, and at 67MB may be difficult to download. An ArcGIS file geodatabase contains plot data and lookup tables that relate map class units to mapping associations. The geodatabase includes a vegetation Feature dataset with the park boundary and project boundary used in the map.

The map development process was organized around the random forests machine learning algorithm. The modeling used 2,519 plots representing 150 vegetation associations and 50 map classes. Imagery from the National Agriculture Imagery Program and the Sentinel-2 and Landsat 8 satellites, airborne lidar bare earth and canopy height data, elevation data from the U.S. Geological Survey 3D Elevation Program, and climate normals from the PRISM Climate Group were used to develop a variety of predictor metrics. The predictors and the map class calls at each plot were input to a process in which each map class was modeled against every other map class in a factorial random forests scheme. We used the plot-level modeling outcomes and species composition data to adjust the crosswalk between association and map class so that floristic consistency and model accuracy were jointly optimized across all classes. The map was produced by predicting the factorial models and selecting the overall best-performing class at each 3-meter pixel. The final vegetation map, including a buffer surrounding the park, contains 43 natural vegetated classes, seven mostly unvegetated natural classes, and four classes representing burned areas or anthropogenic disturbance.