Addison Gensch

Knotweed describes an invasive species group that can be found in its invaded range throughout the United States that is characterized by being difficult to eradicate and, thus, difficult to control. The species in this group are difficult to identify in a field setting due to high variance in morphology. 4 species from this group (Reynoutria sachalinensis, R. japonica, R. compacta, and R. xbohemica) were selected with R. xbohemica being a hybrid of R. japonica and R. sachalinensis and noted as the most invasive amongst the group and the hardest to visually identify. These species were sampled amongst their invasive ranges in Minnesota and northern Wisconsin and scanned using a photospectrometer to collect reflectance data on both the adaxial (sky-facing) and abaxial (ground-facing) sides of their leaves. This data was then analyzed using a partial least squares discriminant analysis in an attempt to create models that can discriminate amongst species. This analysis included models created with full separate adaxial and abaxial data alongside separate adaxial and abaxial data with a limited wavelength to mimic that of a more accessible, cheaper hand-held photospectrometer. Overall, both models created with full wavelength data could identify between R. compacta, R. sachalinensis, and R. xbohemica with moderate accuracy and could identify R. japonica with low-moderate accuracy, with most common misidentification as R. xbohemica. Both models built from limited wavelength data followed similar data trends, however, the adaxial limited wavelength model experienced a notable drop in overall model quality. Findings suggest that models can be created to identify between these species, however there is still some difficulty in identifying between R. xbohemica and R. japonica. Additionally, for models built with a limited wavelength, using abaxial data may improve accuracy.