Tracking Phragmites australis expansion in the Bear River Migratory Bird Refuge using AggieAirTM aircraft data

Document Type


Journal/Book Title/Conference

Remote Sensing and Hydrology 2010 Symposium, Jackson Hole, Wyoming, USA

Publication Date



This research examines the use of unmanned air vehicles (UAV), a cutting edge technology developed at the Utah Water research lab for acquiring airborne imagery using drones for the assessment of abundance of an invasive species Phragmites australis in a wetland vegetation setup. These UAV’s acquire multispectral data in the visible and near-infrared bands with a spatial resolution of 0.5 meters. The study area is the Bear River Migratory Bird Refuge (MBR) which lies in northern Utah, where the Bear River flows into the northeast arm of the Great Salt Lake. The Refuge protects the marshes found at the mouth of the Bear River; these marshes are the largest freshwater component of the Great Salt Lake ecosystem. A common reed, Phragmites australis, is a tall (1.5-4.0 m) coarse perennial grass found primarily in brackish and freshwater wetlands, growing at or above mean high water. The methodology is to build Bayesian statistical supervised classification model using relevance vector machine (RVM) employing the inexpensive and readily available UAV data. The UAV images of the bird refuge are processed to obtain calibrated reflectance imagery. Thereafter, the isodata clustering algorithm is applied to classify the multispectral imagery into different classes. Using ground sampling of the species, pixels containing the Phragmites australis are deduced. The training set for the supervised RVM classification model is prepared using the deduced pixel values. A separate set of ground sampling points containing the Phragmites australis are kept aside for validation. The distribution of Phragmites australis in the study area as obtained from RVM classification model is compared to the validation set. The RVM model results for tracking of Phragmites are encouraging and the new technique has promising real-time implementation for similar applications.

This document is currently not available here.