Assessing bird habitat occupancy from gradient-based landscape metrics and principal polar spectral indices in the colombian andean region

Abstract

ContextThe patch-mosaic model (PMM) is the most common way to describe the landscape in ecological research. Despite this, the gradient model (GM) was proposed as a more accurate representation of the heterogeneity of landscapes; however, little has been explored on the behavior and performance of continuous variables and surface-based metrics from GM under different analytical scenarios.Objectives We address the question: which landscape metrics, patch-based or surface-based, best explain habitat occupancy patterns of six bird species with different ecological preferences?MethodsWe generated detection histories for six bird species in a fragmented Andean landscape from Colombia. We obtain patch-based metrics from a land cover classification and surface-based metrics from the principal polar spectral indices (PPSi) to describe the landscape. Finally, we fitted dynamic occupancy models using variables derived from landscape models and compared their performance using quasi-AIC for each species.Results We obtained 909 detections for the six selected bird species. We found that PPSi and surface-based metrics were more informative when assessing occupancy patterns for five of the six species studied. In addition, surface-based metrics allowed to detect interspecific differences between species beyond an affinity for a particular cover type.Conclusions Surface-based metrics can be an alternative for assessing species response to landscape heterogeneity, particularly those that may be more sensitive to fine-scale changes in vegetation cover. However, there is no single “best” model to describe the landscape for all cases. PPSi can be very useful for land cover analysis in landscape ecology studies as an alternative to more popular vegetation indices.