Scaling effects of landscape metrics: a comparison of two methods

Abstract

Scale and scaling are important conceptual and methodological issues in geography and ecology. Grain, one indicator of scale, is widely used to determine the scaling behaviors of landscape spatial patterns represented by landscape metrics. This study compared two strategies, including a high-resolution-data-based resampling (RS) approach and a multisource and multiresolution data (MSMRD) approach, to quantify the grain effects of nine commonly used landscape metrics determined by FRAGSTATS software. Three adjacent, small watersheds in the central Loess Plateau of China were selected as study sites. Using the RS and MSMRD approaches, five and seven landscape metrics, respectively, were detected to have significant and mostly linear grain effects. The MSMRD approach was more effective in detecting grain effects, and landscape areas were found to be important in the grain effects of landscape metrics. Changing grain means changing the spatial analysis or landscape observation scales and, thus, the quality of landscape pattern information acquired, without impacting the ground-surface landscape pattern. In this sense, the analysis of scale effects through the manipulation of grains can be practically implemented in spatial data processing and selection for landscape modeling tasks. This paper also highlights gaps in present scaling research on landscape metrics by their numerical responses to grain sizes, with the conceptual and mathematical formulations largely neglected. The latter, however, is more important in research concerning landscape metrics, such as scaling, correlation, and even the interaction between landscape patterns and ecological processes.