Abstract
In recent years, identifying the relationship between spatial pattern and scale has emerged as a central issue in ecology and geography. Scale has been defined by grain, or resolution here. Bias in the results will occur if the scale is wrongly selected for landscape evaluation. In particular, with remote sensing becoming widespread, monitoring and detection on landscape dynamics from local to global scale has become available, selecting a satellite with appropriate spatial resolution for research objective scale is thus becoming essential. Moreover, it is necessary to understand scaling among different satellite data, which is important for improving the efficiency of remote sensing. In this research, satellite data of various 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 Yatu landscape located in Sakura City, Chiba prefecture. Land cover classifications were first implemented using the Maximum Likelihood Method on satellite data of various resolution. Based on the results, classification maps from each satellite image were systematically resized from their original pixel size and a series of coarser resolution maps were created through the majority rule. Finally, nine of landscape metrics imbedded in the FRAGSTATS were derived from these aggregated categorical data for landscape pattern analysis. The results indicated that most landscape metrics obviously increased/decreased in scaling relations such as power-law and logarithmic among the satellites having various resolutions.
