Developing a Remote Sensing-based Mapping Method for Swidden Land Use Detection:

Case Studies in Two Karen Villages of the Bago Mountains, Myanmar

抄録

The dynamic and complex nature of swidden agriculture makes it difficult to detect the area from satellite images and Landsat-related methodologies in many studies. Thus, mapping methods for current agricultural land use and swidden land use detection methods from freely available time series satellite images were developed, integrated with detailed information of ground truths and interview surveys in order to contribute to sustainable land use management for legally unrecognized swiddens in Myanmar. Satellite images were classified by using the NDVI break value method and maximum likelihood method, which were then specified with the multiple criteria of ground truths such as the elevation and size of swidden fields. The accuracy of these methods was examined with actual ground truth data collected in 2016. As the maximum likelihood method showed better accuracy to detect the swidden fields, it was, therefore, selected. The accuracy test revealed that over 70 % of the actual swidden plots were detected and the failures of such detection were mainly caused by incomplete burned fields and small-sized swidden plots as well as the topography. However, the developed method is applicable for swidden land use detection at the plot level with preferable accuracy in undulating hilly regions.