Skip to body

Center for Spatial Information and Research

Comparative Assessment of Multiplatform and Multistage Remote Sensing for Rural Habitat Assessment Initiatives


The primary goal of this project was to ascertain what the most efficient, effective, and accurate available remote sensing data are for land cover mapping in rural communities. Land cover maps provide baseline data on vegetative cover characteristics and can be used as a digital data input for any type of environmental phenomenon, including wildlife habitat mapping, urban planning, or ecological riparian restoration. A variety of different resolution data are available at a broad scale of costs for users. However, identifying cost versus effectiveness, labor versus accuracy, and processing time versus technical capabilities are still not determined for many common rural planning and management procedures.

The study area that was selected is within part of an ecological restoration project located in Yakima County, Washington north of the community of Wapato along the Yakima River. A hierarchical land cover classification scheme was derived for image processing and field data collection, using 9 primary and 7 secondary classes focusing on vegetative cover and land cover classes based upon the National Wetlands Inventory (NWI). Field data collection was conducted in July 2008 and March 2009 for manual classifications and to obtain training signatures of specific classes for image analysis.

Three multispectral data were used in this analysis and one visible spectrum georectified aerial photo, including a Landsat 5 Thematic Mapper, ASTER, IKONOS-2, and NAIP image. Image preprocessing and analysis took place using ERDAS Imagine and were integrated with ESRI ArcMap. Different image processing methodologies took place to compare methods and results. The different imaging processes included the following three techniques: 1) maximum-likelihood unsupervised classification, 2) a supervised classification using field data for signatures, and 3) data fusion of a Normalized Difference Vegetative Index (NDVI) with band data. Following image processing, error matrices were created to compare and calculate overall accuracy, producer’s accuracy, user’s accuracy, and Kappa coefficients

All information, including the websites where data were acquired free of charge and different methodologies for image processing, are available as a web-based educational module. The module can be accessed at the CWU CSIR website. The module provides an online resource and a visual tutorial for image processing of the above techniques with ERDAS Imagine in addition to a matrix for comparison of what data and resolutions may be the most effective for rural users’ individual needs.