Multi-Scale Effect on Landscape Pattern Analysis Using Satellite Data with a Range of Spatial Resolutions

In recent years, identifying the relationship between pattern and scale has emerged as a central issue in ecology and geography. Scale has been defined by grain or resolution but bias in results will occur if the scale is wrongly selected relevant to the landscape evaluation. In this research, satellite data of varying resolution, QuickBird (2.5m), ALOS/AVNIR-2 (10m), Terra/ASTER (15m) and Landsat/ETM+ (30m), were employed to analyze the scale effects of grain size. The research was implemented at Azeta, a typical rural landscape located in Sakura City, central Japan. Land-cover classifications were first implemented using the Maximum Likelihood Method on satellite data of varying resolution. Based on the results of these classifications, a number of landscape metrics imbedded in the FRAGSTATS were extracted for landscape pattern analysis. The results indicate that most landscape patterns show some degree of consistency and scaling relations such as power-law among the various satellite resolutions. The applicability of these various satellite data resolutions for landscape analysis in the target area was also evaluated.