Multifractal analysis of land use pattern in space and time: A case study in the Loess Plateau of China

Abstract

Quantifying patterns of great spatial and temporal complexities are a daunting task for landscape ecologists. Multifractals, with which more detailed information can be obtained from a distribution than mono-fractal does, is especially suited to the description of complex patterns. In our paper, three sites (256 × 256 cells per site, 100 m per cell) from the upper, middle and lower reaches of Yanhe watershed in the Loess Plateau were selected to test the scaling properties of agricultural landscape patterns in 1980, 2000 and 2006 using multifractal Rényi dimensions analysis. We found that: (1) the distributions of cropland, grassland and woodland patterns are multiscaling over nearly four orders of magnitude in scale, whereas water bodies and residential areas are not multifractally distributed, which may result from the resolution of Landsat TM remote sensing images; (2) spatially, the multifractal techniques are capable to uncover subtle differences in land use patterns between different reaches of the watershed that could correspond to their distinct, underlying abiotic and biotic processes; and (3) temporally, the multifractal spectra demonstrate little difference between 1980 and 2000 but a sharp change in 2006, the former depicting the land use patterns in an equilibrium state and the latter the effect of government policies such as large-scale eco-environment conservation programmes starting in the late 20th century. The fact that scaling behaviors exist across a heterogeneous landscape suggests that there may be some simple mechanism that governs agricultural landscape change. Further work on this problem may improve our understanding of the mechanism and enhance our capability to more effectively manage land use.