Vegetation change detection with a Laplacian filter using Landsat data: a case study in a boreal forest-tundra transition area

A change detection method combining linear-weighting average-smoothing filtering for detecting vegetation changes was tested in a boreal forest-tundra transition area in northern Finland. Zero position of the second derivative of the normalized difference vegetation index (NDVI) was defined as a boundary of vegetation and compared using two Landsat MSS images taken in 1972 and 1973. Boundaries in both images agreed well in treeless tundra. While numerous unmatched boundaries, which were defined as boundary changes, appeared in mountain birch forests because of the difference of small peaks in the NDVI images. However, uncertain changes could be removed by smoothing the derivative images and by referring the local amplitude of the smoothed NDVI images. Most of detected boundary change areas showed reasonable pattern changes of the NDVI, though mainly phenological vegetation changes seemed to appear