Aspen Cover 2014-2023: This folder contains a map of estimated aspen cover for the period 2014-2023. Of our current products, these maps have the highest out-of-sample predictive accuracy. They were produced using model version 4.6.LS4to9.Ensemble.T02.
Aspen Cover 1984-2023: This folder contains estimates of aspen cover over time, which may be used to assess how aspen cover has changed. Ten- and five-year resolution estimates are found respectively in the subdirectories 10-yr and 5-yr. To minimize sensor-related biases in temporal analyses, we exclusively used data from Landsat satellites 4 to 7. Estimates were produced using model version 4.6.LS4to7.xgb and have a lower out-of-sample predictive accuracy compared to the map produced using model version 4.6.LS4to9.Ensemble.T02. The 5-year resolution maps in this folder have lower out-of-sample predictive accuracy than the 10-year resolution maps. Local-scale estimates of change in cover over time appear to be noisy. Users should exercise caution when interpreting fine-scale estimates of change in cover over time from these models.
To avoid data loss from reprojection, maps are provided in the native USGS Landsat grid projections within our project area, which spans two UTM zones (30-m resolutions, origins at 15, 15). Each map is composed of two GeoTIFF files, with suffixes _z10 and _z11 indicating their UTM zones (Zone 10N and Zone 11N, respectively).
Percent Cover From Above
Percent Cover From Above (PCFA) refers to the proportion of ground area covered by aspen canopies as seen from an aerial perspective. This metric is useful for assessing the extent and density of aspen stands. Understory vegetation cover that is not visible from aerial views due to obstruction by taller trees is not included in PCFA.
Model Name Conventions
Model versions are consistently labeled to reflect the data inputs and algorithms used. For example, 4.6.LS4to9.Ensemble.T02 indicates model version 4.6, Landsat satellites 4 to 9, ensemble modeling approach, and a 2% threshold applied to the xgb component model. See Stewart and Long (2024) for more information.