Discrimination of Leafy Spurge (Euphorbia esula L) and Purple Loosestrife (Lythrum salicaria L) Based on Field Spectral Data

Abstract

Leafy spurge (Euphorbia esula L.) and purple loosestrife (Lythrum salicaria L.) are invasive weeds that displace native vegetation. Herbicides are often applied to these weeds during flowering, so it would be ideal to identify them early in the season, possibly by the leaves. This paper evaluates the spectral separability of the inflorescences and leaves of these plants from surrounding vegetation. Leafy spurge, purple loosestrife, and surrounding vegetation were collected from sites in southeastern North Dakota and subjected to spectral analysis. Partial least-squares discriminant analysis (PLS-DA) was used to separate the spectral signatures of these weeds in the visible and near-infrared wavelengths. Using PLS-DA, the weeds were discriminated from their surroundings with R2 values of 0.86 to 0.92. Analysis of the data indicated that the bands contributing the most to each model were in the red and red-edge spectral regions. Identifying these weeds by the leaves allows them to be mapped earlier in the season, providing more time for herbicide application planning. The spectral signatures identified in this proof of concept study are the first step before using ultra–high resolution aerial imagery to classify and identify leafy spurge and purple loosestrife.