Natural landscapes are composed of a variety of biotic and abiotic elements that interact in multiple ways and across scales, resulting in dynamic systems transitioning through complex states. The state of a landscape at an instance in time can be represented on a regular grid, where the value at each cell represents a mutually exclusive land cover class. When the parameters that define the representation are fixed then it is possible to describe the resultant pattern in terms of its composition and configuration.

Features of these landscapes are often subject to fragmentation, the division of a class into smaller and more isolated patches by any number of natural or anthropogenic processes. This process is of concern because landscape fragmentation is often linked to habitat loss, increased isolation, and loss of connectivity; all of which can lead to a decrease in biodiversity or decrease landscape functionality. The quantification and monitoring of landscape fragmentation is challenging, often context specific, but central to landscape ecology.

Numerous metrics to quantify landscape fragmentation have been developed. These metrics are based on various measures related to spatial patterns of landscape features: their land cover composition and configuration. Although they can be effective for quantifying fragmentation, I have been examining the possible improvement offered by a process-based method for quantifying landscape fragmentation. The goal of my study was to compare metrics relying on a process constrained to a given landscape with static metrics based on composition and configuration alone. Two-thousand binary landscapes were simulated using a stochastic landscape simulator within the R environment. The landscapes randomly varied by relative class composition (low impedance and high impedance classes) and degree of spatial autocorrelation and then each landscape was summarized by 20 conventional pattern-based metrics. Least cost path analyses were performed 50 times on each landscape using randomized start and end points on the west and east margins respectively. The resulting 5 summaries provided process-based metrics for these same landscapes, together with their variability, for comparison.

In my proposed presentation I will present the results of my study. Both the pattern-based metrics and process-based metrics were less sensitive to land cover configuration (spatial autocorrelation) than land cover composition (proportion of impedance in traversing a landscape). All pair-wise comparisons were tested, and the most effective fragmentation indices were ranked.