Abstract
Spatial heterogeneity, in the wide sense of pattern complexity and variability, appears to be one of the dominant influences of ecological functions and processes. The most widely used quantification of heterogeneity in environmental studies is based on derived Shannon information theory indices. Yet, such heterogeneity index is not able to quantify the surface patterns of a landscape with a fully explicit scheme at each place and scale of the landscape. The objective of this work is to propose a methodology to bypass this disadvantage and to capture the local and scaling variations of the landscape. The method developed here combines a multiscale heterogeneity map (MHM) with a heterogeneity profile (HP) averaging at each place (respectively, at each scale) the spatial information contained in the landscape. Confidence levels have been estimated on the basis of a simple neutral landscape model to detect significant non-random heterogeneity fluctuations. This methodology is illustrated on both categorical (patchy) and continuous (pixelised) real landscapes: a heterogeneous agricultural landscape in Brittany and a spring phytoplankton bloom in north Atlantic. Heterogeneity analyses of both image types can simultaneously reveal subtle quantitative and qualitative features (gradients, boundaries, random or patchy patterns at successive scales) linked to ecological processes. (c) 2007 Elsevier B.V. All rights reserved.