Land‐cover binary change detection methods for use in the moist tropical region of the Amazon: a comparative study

Abstract Many land‐cover change detection techniques have been developed; however, different conclusions about the value or appropriateness of each exist. This difference of opinion is often influenced by the landscape complexity of study areas and data used for analysis. Which method is most suitable for land‐cover change detection in Amazon tropical regions remains unclear. In this paper, 10 binary change detection methods were implemented and compared with respect to their capability to detect land‐cover change and no change conditions in moist tropical regions. They are image differencing (ID), modified image differencing (MID), a combination of image differencing and principal component analysis (IDPCA), principal component differencing (PCD), multitemporal PCA (MPCA), change vector analysis (CVA), vegetation index differencing (VID), image ratioing (IR), modified image ratioing (MIR), and a combination of image ratioing and PCA (IRPCA). Multi‐temporal Thematic Mapper (TM) data were used to conduct land‐cover binary change detection. Research results indicate that MID, PCD and ID using TM band 5 are significantly better than other binary change detection methods and they are recommended specifically for implementation in the Amazon basin. Acknowledgments The authors wish to thank the National Science Foundation (grants 95‐21918 and 99‐06826), the National Aeronautics and Space Administration (grant N005‐334), and Embrapa Satellite Monitoring for their funding support. This project is part of the Large‐Scale Biosphere–Atmosphere Experiment in Amazônia (LBA) program, LC‐09, examining the human and physical dimensions of land‐use and land‐cover change. Notes S_95: significance at the 95% confidence level; S_90: significance at the 90% confidence level; S_85: significance at the 85% confidence level; NS: not significant.