Thesis Publication:  Photogrammetric Engineering and Remote Sensing. 60(12):1449-1452.

THESIS ABSTRACT

Maintaining and updating road inventory information for GIS applications can be costly in large areas that a managed for timber production. We evaluated the positional accuracy of SPOT (Systeme Probatoire D'Observation de la Terre) HRV 10-metre panchromatic imagery to update road locations in a forested area in northern Idaho. One 7.5 minute quadrangle of SPOT level 1B panchromatic data was precision rectified using a digital elevation model. A 3 by 3 high pass filter was then used to enhance linear features such as roads. The data were then resampled from the original 10-metre pixels to 3.3 metre pixels to increase the precision of on-screen digitizing of roads. The digitized road arcs were compared with reference arcs that had been compiled using an analytical stereoplotter. An 18-metre buffer of allowable error was established around the reference roads. Ninety-six percent of the road arcs digitized from the SPOT imagery were within this error buffer.

The distribution of road arcs outside this error buffer did not differ significantly among slope classes. This may be due to the fact that 1) there is little relief displacement when imagery is acquired from far above the terrain, or 2) the SPOT imagery was rectified using a digital elevation model. We concluded that roads digitized from the displayed SPOT HRV panchromatic image had acceptable positional accuracy for our forest management application. Since one SPOT HRV scene covers approximately 3,600 km2 seamlessly , this method may be more cost-effective than using hundreds of aerial photographs to map the roads from the same area.