Connecting the dots: Assessing landscape connectivity algorithms for biodiversity conservation

Abstract

Biodiversity is currently facing threats from both climate change and habitat fragmentation. With climate zones (conditions) shifting pole-wards, the ability of populations to track these changes in fragmented landscapes are crucial for biodiversity conservation, while the ability of scientists to assess these dynamics is crucial for conservation planning. To simulate the movement of organisms in a fragmented landscape, scientists often use connectivity matrices - tables that contain the probabilities of successful dispersal between different pairs of habitat patches.

While mechanistic, individual-based correlated random walk (CRW) models are commonly used to estimate these probabilities, simpler alternatives exist based on interpatch distance and patch size which are easier to develop, use and understand. Depending on their scientific credibility, these simpler alternatives could be more practical to use in decision making processes.

In this context, our objective is to validate simpler models as viable alternatives for complex models and to understand their limitations. Specifically, how good are simple algorithms in mimicking the CRW model? To address this question, we compared the connectivity matrices of ten simple algorithms to those of a CRW model across 36 contrasting landscape-disperser combinations. We used the coefficient of determination (R²) and Akaike information criterion (AIC) to rank the algorithms.

Our results show that the frequently used exponential algorithm (EXP), in which the connectivity decays exponentially with the interpatch distance, did not perform well, with a mean R² of 0.745 and a minimum R² of 0.185 between the connectivities of the EXP model and the CRW model. On the other hand, the CRD-lim model - which uses a constant•radius/distance relation within a maximum interpatch dispersal distance (d_max) - performed best, with a mean R² of 0.918 and a minimum R² of 0.809.

Furthermore, the CRD-lim algorithm emerged as a robust alternative to random walk or the exponential decay models when evaluating connectivity matrices for specific landscapes and species. Notably, this approach spans a wide range of spatial scales, offering a combination of simplicity and power that makes it suitable for conservation planning in the face of climate change and habitat fragmentation across a range of species.