SIMPLE CONNECTIVITY MEASURES IN SPATIAL ECOLOGY

TLDR

Connectivity is a core concept in spatial ecology, yet most recent measures focus only on the nearest neighbor or a limited neighborhood of the focal patch. The study compares simple connectivity metrics for predicting colonization events in two large, high‑quality empirical datasets. The analysis shows that nearest‑neighbor distance is an inferior predictor, buffer measures perform better but depend on radius, and the most robust metric incorporates focal patch size and all source populations, while nearest‑neighbor methods fail to detect significant effects across many dataset sizes, demonstrating that simplicity does not compensate for poor performance.

Abstract

Connectivity is a fundamental concept that is widely utilized in spatial ecology. The majority of connectivity measures used in the recent ecological literature only consider the nearest neighbor patch/population, or patches within a limited neighborhood of the focal patch (a buffer). Meta-analysis suggests that studies using nearest neighbor connectivity measures are much less likely to find statistically significant effects of connectivity than studies that use more complex measures. Here we compare simple connectivity measures in their ability to predict colonization events in two large and good-quality empirical data sets. The nearest neighbor distance to an occupied patch is found to be an inferior measure. Buffer measures do much better, but their performance is found to be sensitive to the estimate of the buffer radius. For highly fragmented habitats, the best and most consistent performance is found for a measure that takes into account the size of the focal patch and the sizes of and distances to all potential source populations. When experimenting with reduced data sets, it was discovered that nearest neighbor measures fail to find a statistically significant effect of connectivity for a large range of data set sizes for which the more complex measures still detect a highly significant effect. We conclude that the simplicity of a nearest neighbor measure is not an adequate compensation for poor performance.