Dr. Bruce A. Roundy
Dr. Steven L. Petersen, Brigham Young University, [email protected]
Ryan Jensen, Brigham Young University, [email protected]
Dr. Steve Bunting
Dr. Christopher Neale, Utah State University, [email protected]
Dr. Richard Miller
Dr. Robin Tausch
Dr. Jeanne Chambers
SageSTEP Study Plots:
All SageSTEP woodland sites except for Five Creeks. Click here to see site locations.
Study Design and Objectives:
Research is needed to quantify the relationship between vegetation data acquired from remote sensing imagery (RS) and geographic information systems (GIS) with ground-truth data. Utilizing RS and GIS technologies to characterize the spatio-temporal patterns of pinyon-juniper woodlands, we can define cost-effective indicators and develop efficient methods for rangeland assessment and monitoring such as tree density and tree canopy cover. These data can provide critical information that supports management and conservation of the sagebrush biome.
- Evaluate the accuracy of data obtained from RS imagery and GIS for characterizing pinyon and juniper tree cover and density data across multiple locations and phases using 1-m resolution, RGB NAIP imagery and ground-truth field data.
- Develop and refine methods of RS and GIS estimation of tree cover and density to allow rangeland managers to use NAIP imagery to rapidly assess woodland ecological conditions to better support management decisions.
RS and GIS data used in the study include National Agriculture Imagery Program (NAIP) images (2006-2010) of several of the SageSTEP woodland study sites. Additionally, images were collected in June 2009 using fixed-wing manned aircrafts. Images were collected at multiple scales using a variety of sensors and wavelength bands (color, color-infrared, hyperspectral, and synthetic aperture radar). Digital elevation data will also be included to investigate multiple-topographic attributes contributing to fuel loads and vegetation differences.
Image processing and statistical analysis will be conducted at the Brigham Young University Geospatial Habitat Analysis Lab. Programs used to analyze data will include Feature Analyst (ERDAS 9.3), Feature Extraction (ENVI 4.2), Definiens Developer (eCognition), and Global Mapper. By applying a variety of image analysis methods, researchers will build multiband images to describe surface fuels and distinct classes and patterns of vegetation structures.
Feature Analyst will be used to classify NAIP images for establishing the relationship between features sensed remotely and ground truth data. Feature Analyst uses image characteristics, such as color, size, shape, texture, pattern, shadow, and spatial association to identify and delineate unique feature attributes. Feature Analyst uses an automated feature extraction procedure by incorporating software agent technology to identify features like hydrology, vegetation and other land cover features based on user-specified examples (shapefiles).
An accuracy assessment will be conducted for each site to determine statistical reliability of classified data using ERDAS Imagine 8.4 software. Sample size will be based on the binomial probability theory due to binary information (pinyon/juniper vegetation versus everything else). Using binomial distribution with an expected accuracy of 95% at an allowable error of 5%, the number of sample points necessary for reliable results is 76 fixed area plots (SageSTEP subplots). These plots will be randomly sampled across the network of woodland sites. Fifty random points within each of the subplots will be user classified and analyzed producing overall accuracy and a Kappa statistic. The Kappa statistic is a measurement of agreement or accuracy between the remote sensing derived classification map and ground-truth data. By using Kappa statistics, we can then better determine errors of omission and commission that will enhance the credibility of using remote sensing derived land cover information.
Results from this study will contribute to the development of methods for applying RS/GIS data to rapidly and accurately assess fuel loads that will help managers evaluate landscapes for fire risk, plan site treatments that will help avoid catastrophic wildfire events, and improve vegetation management on broad spatial scales. Rapid assessment methods will allow managers to strategically select stands for fuel reduction treatments and predict effects of treatments on fuel loads and vegetation response following management treatments.
Hulet, A., B. Roundy, S. Petersen and S. Bunting. 2010. Applying Remote Sensing and Definiens eCognition Developer 8 to Estimate Tree, Shrub, and Intercanopy Begetation Cover of Pinyon-Juniper Woodlands in the Great Basin. 63rd SRM and 50th WSSA Annual Meeting: Working Landscapes, Denver, CO, February 7-11, 2010. Poster available here.