Two multi-scale contextual approaches for mapping spatial pattern

The majority of landscape pattern studies are based on the patch-mosaic paradigm, in which habitat patches are the basic unit of the analysis. While many patch-based landscape indices successfully relate spatial patterns to ecological processes, it is also desirable to use finer grained analyses for understanding species presence, abundance, and movement patterns across the landscape and to describe spatial context by measuring habitat change across scales. Here, we introduce two multi-scale pixel-based approaches for spatial pattern analysis, which quantify the spatial context of each pixel in the landscape. Both approaches summarize the proportion of habitat at increasing window sizes around each pixel in a scalogram. In the first regression-based approach, a third-order polynomial is fitted to the scalogram of each pixel, and the four polynomial coefficients are used as descriptors of spatial context of each pixel within the landscape mosaic. In the second shape-based approach, the scalogram mean and standard deviation, and the mean slope between forest cover at the smallest window size and each of the larger window sizes are calculated. The values emerging from these two approaches are assigned to each focal pixel and can be used as predictive variables, for example, in species presence and abundance studies. We tested the performance of these approaches on 18 random landscapes and nine actual landscapes with varying forest habitat cover. Results show that both methods were able to differentiate between several spatial contexts. We thus suggest that these approaches could serve as a complement or an alternative to existing methods for landscape pattern analysis and possibly add further insight into pattern-species relations.