Classification of Willow Species Using Large-Scale Aerial Photography

Abstract

The distribution characteristics of willow (genus Salix) can be evaluated with the use of remote sensing, geographic information systems (GIS) technologies, and spatial analysis. This information can be used to better understand willow ecology, such as willow community composition, species relationships, and associations with other landscape attributes (i.e., soils, elevation gradients, water sources, and landscape position). Aerial photographs of a willow-dominated riparian area located in southeastern Oregon were taken in November 1999 at a 1:2 400 scale. Four basic techniques were used to separate willow species from each other in both color and infrared images using ERDAS Imagine® GIS. Willow species included Geyers willow (Salix geyeriana Anderson), Booth willow (Salix boothii Dorn), and Lemmon’s willow (Salix lemmonii Bebb). The techniques used to analyze aerial photographs included image resampling (degrade), image filtering (smoothing), unsupervised classification, and supervised classification. Highest accuracy was obtained using a supervised classification of the color images smoothed with a low-pass convolution filter (84.6% accuracy). Spectral samples were collected using a polygon digitizing method, which had superior results compared to the seed or region growing method. The infrared image was least effective in separating the plants into species classes (58.6% accuracy). This may be due to the lack of the blue band in the infrared image.