Submitted to: Ecology and Management of Rangelands
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 25, 2010
Publication Date: September 1, 2010
Citation: Hunt, E.R., Gillham, J.H., Daughtry, C.S. 2010. Improving potential geographic distribution models for invasive plants by remote sensing. Ecology and Management of Rangelands. 63(5):505-513. Interpretive Summary: It is generally more cost effective to deal with the initial stages of plant invasion with a few plants, compared to when the invasive plant cover is large enough to be detected by remote sensing. Leafy spurge (Euphorbia esula L.) is a noxious invasive weed that degrades large areas of rangeland. The flower-bracts of leafy spurge are distinctively yellow green, which can be remotely sensed either with high spatial resolution sensors or hyperspectral sensors. Potential distribution models, such as the Weed Invasion Susceptibility Prediction (WISP) model, use geographic information to predict the favorable locations of invasive plants. However, it is difficult to adequately test these models with ground data. We compared the areas predicted to have leafy spurge by the WISP model with a classified image of flowering leafy spurge obtained from NASA’s Airborne Visible Infrared Imaging Spectrometer (AVIRIS), and the WISP model predicted were not significantly different from chance. By examining each variable with the classified AVIRIS image, the predictive ability of the WISP model was highly significant, primarily by reducing the potential area where it is predicted to grow. This reduces the areas that need to be monitored on the ground, so invasive plants could be more efficiently controlled at the initial stages of invasion.
Technical Abstract: Remote sensing is used map the actual distribution of some invasive plants such as leafy spurge (Euphorbia esula L.), whereas geospatial modeling can indicate the potential distribution over a landscape. Geographic data layers were acquired for Crook County, Wyoming, and the potential distribution of leafy spurge presence or absence were predicted using the Weed Invasion Susceptibility Prediction (WISP) model. Hyperspectral imagery and field data were acquired in 1999 over parts of the study area. Leafy spurge presence or absence was classified using the Spectral Angle Mapper with a 74% overall accuracy. However, the user accuracy was 93% showing that where leafy spurge was indicated in the image, leafy spurge was found at that location. Using kappa analysis, there was no agreement between WISP model predictions and either the field data or the classified image. Kappa analysis was then used to compare predictions based on single geographic data layers, to increase the power to detect subtle relationships between independent variables and leafy-spurge distribution. Many of the categorical parameters used in the WISP model were negatively associated with leafy spurge distribution and other categorical parameters not in the model were positively associated with leafy-spurge. A second model for leafy spurge was generated from the kappa analyses, and only a few variables contributed to predicting the distribution of leaf spurge. With both the field data and classified image, the second model had significantly increased accuracy, primarily by reducing the areas predicted to have potential for invasion. It is generally more cost effective to deal with the initial stages of plant invasion with a few plants, compared to when the invasive plant cover is large enough to be detected by remote sensing (including imaging-spectrometer data). By reducing the potential area that needs to be monitored, management of invasive plants could be more efficiently performed by field crews.