A Statistical Evaluation of LANDSAT TM and MSS Data for Mapping Forest Cutovers

Abstract

The cutover area and boundary placement mapping accuracies from currently used air photo techniques and LANDSAT TM and MSS imagery were determined for Alberta forestry conditions. Overall area accuracies for the air photo techniques. LANDSAT TM, and MSS images were 86.7%, 89.5%, and 86.9%, respectively. The influence of increasing accuracy with increasing cutover size, however, must be kept in perspective. Although interpreted cutover areas were generally smaller compared with their actual areas, the deviation was systematic and could therefore be corrected by regression techniques. Overall boundary placement errors for air photo techniques. LANDSAT TM, and MSS images were 30.1, 24.9, and 38.3 m, respectively. Most of the LANDSAT TM errors were in the 20-m range and were within 5 m of the target 15-20 m desired by the Alberta Forest Service. Improvements to this may be achieved by using larger scale satellite images. The base map scale of 1:20 000 was a limiting factor to this study, however, as it was too small with both plotting errors and pen line widths contributing to existing map errors. LANDSAT TM images offered a 12:1 cost savings in data acquisition and are more efficient in comparison with conventional air photos. Use of either medium scale air photos or LANDSAT TM images can be recommended depending on the level of timber commitment accuracy requirements and is presented in the paper as an operational cutover update mapping strategy. Keywords: Forest inventory, update, LANDSAT, change detection, mapping, accuracy analysis, remote sensing.