HGIS: Evaluation of Techniques for Updating Forestry GIS Databases

Remote Sensing and Image Interpretation

March, 1992

Evalutation of Techniques for
Updating Forestry GIS Databases

Microcomputer GIS and video technology were utilized to update raster and vector forestry databases via Landsat TM images and aerial photography. The study was undertaken in cooperation with the Ontario Centre for Remote Sensing and Spruce Falls Paper and Power Company, Kapuskasing, Ontario.
A subscene of a land cover classification with 100 m pixel resolution was updated using the TM image by a series of processing steps involving frame grabbing, image enhancement, registration and resampling, vector tracing of cutover boundaries with a mouse, and conversion to raster. This methodology (TM Method) was similarly employed to update a 1:20,000 scale vector database. Live video overlay of interpreted 1:20,000 scale aerial photography onto a digital vector file (VCU Method) was also tested to create a cutover map. The control map for evaluation of accuracy of the TM and VCU methods was previously generated by tracing of interpreted boundaries from aerial photography via a Zoom Transfer Scope onto a base map and capturing the vectors via a table digitizer (ZTS Method). Areal accuracy estimates of the TM and VCU vector cutover maps were 91% and 97% respectively. Edge displacement, estimated through commission and omission error polygon shading, was 33 m for the TM Method and 18 m for the VCU Method. Updating small land cover classifications through the TM Method is an economic alternative to digital multispectral classification techniques, but does not achieve the 20 m accuracy required for the GIS Forest Resource Inventory vector database. The accuracy of cutover maps generated through all supplementary methods could be improved by capture of the tertiary road network as control through Global Positioning System (GPS) data.