A predicted distribution and biomass model of tamarisk is pictured with previously surveyed polygons. The biomass prediction was created using Landsat 5 TM-derived variables.
Authors: Lane Carter, Jonathan Burnett
Science Advisors/Mentors: Dr. Jeffrey Morisette, Dr. Paul Evangelista, Dr. Tracy Holcombe, Nicholas Young
Abstract: Detection of invasive plant species is vital for management practices and protection of sensitive areas that are affected by their influence on variables such as water tables, native plant communities, soil chemistry, and wildlife habitat. Moderate-resolution remotely sensed data has been utilized in mapping the distribution of various invasive plant species. The aim of this project was to complete this task with Landsat 5 time series analysis for Tamarix spp. (tamarisk) and to move forward by developing a methodology for predicting biomass using the Landsat 5 data and the predicted tamarisk distribution map. Occurrence points for tamarisk were represented by point locations derived from the Colorado Tamarisk Coalition coupled with field data collected by co-authors. Collaborative efforts with local organizations included the U.S. Geological Survey, Colorado Tamarisk Coalition, and the National Institute of Invasive Species Science. We used the acquired Landsat 5 scene to act as the study area boundaries encompassing LaJunta and surrounding areas within southeastern Colorado. We explored various tools and sequence methodologies such as Maximum Entropy (Maxent), boosted regression trees, and a new USGS product, the Software for Automated Habitat Modeling. Tamarisk distribution was successfully predicted using an ensemble method including boosted regression trees, generalized linear model, random forest, and multivariate adaptive regression splines. Biomass was predicted using field collected biomass values coupled with the distribution map and an abundance boosted regression tree model.
Video transcript available here.