Accuracy assessment of vegetation community maps generated by aerial photography interpretation: perspective from the tropical savanna, Australia

摘要

Aerial photography interpretation is the most common mapping technique in the world. However, unlike an algorithm-based classification of satellite imagery, accuracy of aerial photography interpretation generated maps is rarely assessed. Vegetation communities covering an area of 530 km~(2) on Bullo River Station, Northern Territory, Australia, were mapped using an interpretation of 1:50,000 color aerial photography. Manual stereoscopic line-work was delineated at 1:10,000 and thematic maps generated at 1:25,000 and 1:100,000. Multivariate and intuitive analysis techniques were employed to identify 22 vegetation communities within the study area. The accuracy assessment was based on 50percent of a field dataset collected over a 4 year period (2006 to 2009) and the remaining 50percent of sites were used for map attribution. The overall accuracy and Kappa coefficient for both thematic maps was 66.67percent and 0.63, respectively, calculated from standard error matrices. Our findings highlight the need for appropriate scales of mapping and accuracy assessment of aerial photography interpretation generated vegetation community maps.