Sensitivity of landscape pattern indices to input data characteristics on real landscapes: implications for their use in natural disturbance emulation

Abstract

Resource management strategies have begun to adopt natural landscape disturbance emulation as a means of minimizing risk to ecosystem integrity. Detailed understanding of the disturbance regime and the associated spatial landscape patterns are required to provide a “natural” baseline for comparison with the results of emulation strategies. Landscape pattern indices provide a useful tool to quantify spatial pattern for developing these strategies and evaluating their success. Despite an abundance of indices and tools to calculate these, practical knowledge of interpretation is rare. Quantifying changes in landscape pattern indices and the meaning of these changes is confounded by index sensitivity to input data characteristics such as spatial extent, spatial resolution, and thematic resolution. Sensitivity has been examined for simulated landscapes but rarely using real data for large areas as real landscapes are more difficult to manipulate systematically than simulated data. While simulated data offer a control, they do not provide an accurate portrayal of reality for practical applications. Our goal was to test the sensitivity of a suite of landscape pattern indices useful for disturbance emulation strategy development and evaluation to spatial extent, spatial resolution, and thematic resolution using current land cover data for a case study of the managed forest of Ontario, Canada. We also examined how sensitivity varies spatially across the study area. We used Landsat TM-based land cover data (> 45.5 million ha), controlling spatial extent (2,500 to 2,560,000 ha), spatial resolution (1 to 16 ha), and thematic resolution (2 to 26 classes). For each index we tested a hypothesis of insensitivity to changes in each input data characteristic using a combination of ANOVA and regression and compared our results with previous studies. Of the 18 indices studied, significant (p< 0.01) effects were found for 17 indices with changes in spatial extent, 13 indices with changes in spatial resolution and 18 indices with changes in thematic resolution. A significant (p < 0.01) linear trend accounted for the majority of the variance for all of the significant relationships identified. Most of the mean index responses were consistent with those interpreted from previous studies of simulated and real landscapes; however, sensitivity varied greatly among indices and over space. We suggest that variation in sensitivity to input data characteristics among indices and over space must be explicitly incorporated in the design of future natural disturbance emulation efforts.