Semiwariancje tekstury obrazów satelitarnych Landsat TM obszarów leśnych

The paper presents selected applications of the geostatistical methods for remote sensing studies of forest ecosystems. The texture of the remote sensing images of the forests can be highly complex, but at the same time it is very useful for interpretation of these images. We explored differences on semivariances calculated from transects of the Landsat TM images associated with images of young, middle−aged, old, and natural versus planted stands, as well as compared semivariances for loblolly pine (Pinus taeda L.) and longleaf pine (Pinus palustris Mill.) in Georgia, USA. We analyzed the imagery data in conjunction with ground measurements. The results showed that in spite of the relatively low Landsat TM resolution (30m), the semivariances might give useful additional information about the above−mentioned classes. The semivariances for loblolly pine depend on the age as well as on the origin of the forest stand (natural vs. planted). On the other hand, because of the low−resolution, the classical parameters of a range of semi−variances are rather difficult for classification. However, we found important differences for sill values as well as for semivariance values for the distances of a few lags. These parameters seem to be useful for forest stand discrimination. Significant differences were also found between semivariances calculated from loblolly pine and longleaf pine stands.