In the paper, we described the methodology enabling automatic land cover pattern change analysis, using satellite data.
We relied on the post-classification comparison technique with the classification process based on a supervised
approach, joining image segmentation, knowledge-based rules to extract a training set, and the maximum likelihood
decision rule on two hierarchical levels. The Polish Carpathians has been studied at a medium scale level, over a period
of 19 years (1987-2006). The information about land cover was obtained from Landsat TM and ETM+ images.
The overall accuracy received in this approach amounted 88%.
This study focuses on forest monitoring at landscape level on the basis of a methodology combining satellite data mapping and image morphological processing. It aims to contribute to reporting on trends of forest fragmentation and connectivity, by using forest spatial pattern metrics. The Carpathians were selected as a study area.
For five case study areas single-date forest - non-forest maps derived from Landsat images acquired in the 1980s and 2000s were an input for the post-classification change detection. Morphological image processing was applied then to map forest spatial pattern into six classes (core, patches, edges, perforation, connectors and branches). Further, connectivity and fragmentation processes were assessed on the basis of the proportion of forest pattern classes.
We found a general trend of forest increase over the last decade. An increase of forest fragmentation and connectivity was noticed for four case study areas and a decrease for one case study area. The increase of forest cover may indicate the decline of importance of mountain agriculture, while changes of forest fragmentation and connectivity are probably related to the transformation of forest management practices in the 1990s in the region. We conclude, that the proposed methodology allows assessing trends in forest fragmentation and connectivity at approximately 1 ha minimum mapping unit.