Use of Semivariances for Studies of Landsat TM Image Textural Properties of Loblolly Pine Forests

Jarek Zawadzki1, 4, Chris J. Cieszewski2, Roger C. Lowe3, and Michal Zasada5, 6

Abstract: 

We evaluate the applicability of Landsat TM imagery for analyzing textural information of pine forest images by exploring the spatial correlation between pixels measured by semivariances and crosssemivariances calculated from transects of the Landsat TM images. Then, we explore differences in semivariances associated with images of young, middle- aged, and old, and natural versus planted stands. Finally, we compare semivariances for loblolly pine (Pinus taeda L.) with those of longleaf pine (Pinus palustris Mill.) in Georgia, U.S.A. The results show that, in spite of the low Landsat TM resolution, the semivariances and cross-semivariances may provide useful additional information. Remotely sensed data are inexpensive supplements to ground measurements and are frequently used in forest inventories of large areas due to the cost efficiency and the ability to provide a large amount of information in a short time (Campbell l994, Vogelmann et al. 1998). Most common methods for image classification of remotely sensed images are applied without considering potentially useful spatial information among various pixels. Semivariograms consider the spatial information and have proved useful in analyzing various spatial data (Curran 1988, Woodcock et al. 1988a, 1988b). So far, the semivariograms have been successfully used in forestry applications only with expensive high-resolution data (St.-Onge and Cavayas 1995, Treitz and Howarth 2000). The objective of our study was to evaluate the applicability of the relatively inexpensive, low-resolution Landsat TM7 TM imagery for analyzing the textural information in images of loblolly pine forests (Pinus taeda L.) in Georgia, U.S.A., using geostatistical methods. We analyzed different ages and natural versus planted stands of loblolly pine using semivariograms and cross-semivariograms. To check if semivariograms can discriminate between different species, semivariograms for loblolly pine were compared with those of longleaf pine (Pinus palustris Mill.). We analyzed data from the Thematic Mapper sensor of the Landsat TM7 satellite in combination with ground measurements. We used information from the visible red (RED), the near-infrared (NIR), and the middle-infrared (MIR) bands. The Normalized Difference Vegetation Index (NDVI) as well as the

corrected NDVI (NDVIc) and MIR/RED indices were studied.

Author Keywords:

KeyWords Plus:

Addresses:
1,5Postdoctoral Fellow, 2Assistant Professor, 3GIS Analyst, D.B. Warnell School of Forest Resources, University of Georgia, Athens, GA 30602, USA, biomat@uga.edu.

4Assistant Professor, Environmental Engineering Department, Warsaw University of Technology, Nowowiejska 20, 00–61 Warsaw, Poland, jarek97@yahoo.com.

6Assistant Professor, Department of Forest Productivity, Faculty of Forestry, Warsaw Agricultural University, Rakowiecka 26/30, 02–528 Warsaw, Poland,

zasada@delta.sggw.waw.pl.

Publisher:

IDS Number:

ISSN: