Semivariogram analysis of Landsat 5 TM textural data for loblolly pine forests

Abstract

The objective of this study was to evaluate the applicability of Landsat 5 TM images for analysing the textural information on pine forest stands in western Georgia, United States. Analysing spatial correlations between pixels measured by semivariances and cross-semivariances (cross-correlation between two radiometric bands) calculated from transects of Landsat TM images, we explored differences between semivariances associated with images of stands of various ages, origins (natural vs. planted) and species (loblolly pine - Pinus taeda L. - versus longleaf pine - Pinus palustris Mill.). We analysed both ground measurements and the satellite images using the visible, the near infrared, and the middle-infrared bands. We also analysed semivariances and cross-semivariances calculated from the Normalized Difference Vegetation Index and the Ratio Vegetation Index. The results showed that in spite of the relatively low Landsat TM spatial resolution (30m) the semivariograms and cross-semivariograms provided potentially useful information about the above-mentioned classes. The semivariances and cross-semivariances calculated from Landsat TM images of loblolly pine stands depend both on the age and the stand origin. In particular, large differences exist in semivariance and cross-semivariance sills. Significant differences also exist between semivariances calculated from stands of loblolly and longleaf pine.

Keywords

• semivariance
• textural classification
• remote sensing
• loblolly pine
• landsat tm