Periodic multiresource surveys and effective management of forest resources would be facilitated by capabilities for accurate detection of disturbances and other changes in land cover. Current techniques are not as effective and efficient as desired. This paper describes a recently developed digital method for change detection using multidate Landsat data. The method employs calculation of spectral change vectors from two different dates, prompting its name -- Change Vector Analysis. The concept and a stratification procedure are described and their features are compared to other approaches. An implementation was tested that utilizes a linear transformation of Landsat data channels and spatial-spectral clustering of multidate data for the definition of spectrally homogeneous stand-like "blobs". Maps of two types of change, harvesting and regrowth, were produced and analyzed for a test site in Northern Idaho.

Date of this Version