A new approach for land cover classification and change analysis: Integrating backdating and an object-based method

Abstract

Abstract Accurate information on land use and land cover (LULC) change is crucial for ecosystem monitoring and environmental change studies. Updating/backdating approaches have been increasingly used for LULC classification and change analysis, but mostly based on pixels. Here, we presented a new approach, an object-based backdating approach which integrates the backdating approach with an object-based method, and further compared it with the pixel-based backdating approach. We tested the new approach by using Landsat TM data collected in 2001 and 2009 at the Beijing metropolitan region. We found that: 1) an object-based backdating approach achieved higher accuracy for change detection, LULC classification and change analysis than the pixel-based backdating approach. With the object-based approach, the overall accuracies for the classification and change analysis were 84.33% (versus 69.33% for a pixel-based approach), and 80.00% (versus 70.50% for a pixel-based approach), respectively. 2) The object-based backdating approach greatly increases the efficiency because classification and change analysis are only conducted for locations with changes. The increase in efficiency is particularly important for LULC classification and change analysis conducted at a large area, for example, at the national or global scale.