Ozdemir I., Mert A., SENTURK O.

JOURNAL OF ENVIRONMENTAL ENGINEERING AND LANDSCAPE MANAGEMENT, vol.20, no.2, pp.168-176, 2012 (SCI-Expanded) identifier identifier


The aim of this study was to predict landscape structural metrics using the features extracted from the ASTER multispectral satellite imagery with 15 m spatial resolution. The landscape structural metrics were calculated on the basis of forest map polygons generated from 1:15000 scaled aerial photos by photo-interpretation technique. The landscape metrics and corresponding image features that are texture parameters and segmentation polygons were determined for four different landscape extents. A stepwise multiple linear regression analysis was carried out to identify the most significant image-derived predictors of landscape metrics for each extent. The regression models established for the landscape metrics including the Number of Patches (NUMP), Edge Density (ED), Shannon's Diversity Index (SDI) and Patch Richness (PR) performed moderately with adjusted R-2 values of 0.50 and 0.53 (P < 0.01), indicating that 50-53% of the total variation in these landscape metrics can be explained by image-derived features. By contrast, the regression analyses showed that there were weak relationships between the image features and Interspersion Juxtaposition Index (IJI), and Shannon's Evenness Index (SEI) (adj. R-2 is varied from 0.12 to 0.30, P < 0.01). According to the results of model evaluation, the Entropy measures based on Grey Level Co-occurrence Matrix (GLCM) calculated from the infrared and red bands of ASTER were found as the most correlated parameters with the landscape metrics. Besides, the window size (extent) of 81-144 ha (between 900x900 and 1200x1200 m) might be recommended in estimating the landscape metrics using the ASTER or similar satellite imagery. It can be concluded that the 15 m resolution satellite data used for estimating landscape spatial structure cannot replace aerial photos or very high resolution satellite imagery for local-level inventories. However, it might have potential for predicting landscape heterogeneity for large-scale inventories.