Habitat Classification for Ecological Risk Assessment Using Aerial Photography and GIS Data In a Two-Stage Expert System.

There are many challenges in performing a land cover classification that can be used for habitat suitability analysis of multiple species in ecological risk assessments.  The classification scheme must be detailed enough to allow for use to several different habitat requirements, as well as flexible enough to be applicable to different species.  Because habitat suitability often has minimum area requirements, but does not require regular shapes of contiguous cover, high spatial resolution is useful, but imagery of high spatial resolution often lacks the spectral resolution useful for habitat classification.  In this presentation, we will discuss the results of a land cover classification designed to allow for the modeling of habitat suitability for a variety of receptors, such as shrews, robins, and deer.  A hybrid remote sensing / GIS approach was developed that uses both aerial photography and ancillary GIS data.  A two-stage expert system is used to bring together different data sources.  This expert system was designed for and demonstrated on a particular application, but it can be easily adapted to other sources.  The result is a land cover classification with fourteen classes, a minimum mapping unit of 1/20 acre, and distinct polygons that can be as narrow as four feet wide.  Ground reference data from field visits showed an overall accuracy of approximately 80% for qualitatively different classes and 60% for the full classification scheme containing quantitative class divisions, such as those between closed, intermediate, and open canopy cover classes.