images here show georegistration supplied by vendor. road overlay is first indication
that it is not accurate, and tree overlay shows the consequences of this.
A two-pronged approach was taken to remedy the situation. first, an additional
 gcp based registration was done on each scene.  In an urban setting, this was easy
to do because of distinct features (road intersections).  More difficult in rural areas (Lake
Anne and Menominee). With distinct canopies, we were able to then fine-tune tree locations by
looking at plot center and individual trees to see if we could make adjustments to get tree
canopy outlines aligned with imagery.  Where canopies were not distinctly identifiable, we
had to rely on doing our best with the gcp registration, and we also coarsened the resolution and limited the
dataset to larger dominant trees to insure that we had tree/imagery coincidence.