| Popis: |
The separability of Ailanthus altissima (tree of heaven) from four native species was investigated using spectral reflectance measurements (350 to 2,500 nm) of leaves collected from 13 May through 24 August 2008. For both the original reflectance and a continuum removed dataset, least angle regression (LARS) and random forest classification were used to identify a single set of optimal wavelengths across all sampled dates, a set of optimal wavelengths for each date, and the dates for which Ailanthus is most separable from other species. Leaf classification accuracy was found to vary with both dates and bands used. Contrary to expectations that early spring would provide the best separability, July and August were also identified as potentially good months for species differentiation. Applying continuum removal generally reduced classification error. Band selection using LARS improved classification accuracy. The optimal spectral bands were selected from across the spectrum, typically including 401 to 431 nm, 1,115 nm, and 1,985 to 1,995 nm. |
