Detecting an Invasive Shrub in Deciduous Forest Understories Using Remote Sensing

Abstract

Remote sensing has been used to directly detect and map invasive plants, but has not been used for forest understory invaders because they are obscured by a canopy. However, if the invasive species has a leaf phenology distinct from native forest species, then temporal opportunities exist to detect the invasive. Amur honeysuckle, an Asian shrub that invades North American forests, expands leaves earlier and retains leaves later than native woody species. This research project explored whether Landsat 5 TM and Landsat 7 ETM+ imagery could predict Amur honeysuckle cover in woodlots across Darke and Preble Counties in southwestern Ohio and Wayne County in adjacent eastern Indiana. The predictive abilities of six spectral vegetation indices and six reflectance bands were evaluated to determine the best predictor or predictors of Amur honeysuckle cover. The use of image differencing in which a January 2001 image was subtracted from a November 2005 image provided better prediction of Amur honeysuckle cover than the use of the single November 2005 image. The Normalized Difference Vegetation Index (NDVI) was the best-performing predictor variable, compared to other spectral indices, with a quadratic function providing a better fit (R2 = 0.75) than a linear function (R2 = 0.65). This predictive model was verified with 15 other woodlots (R2 = 0.77). With refinement, this approach could map current and past understory invasion by Amur honeysuckle.