Unravelling the effects of individual heterogeneity on connectivity estimates: Insights from Spatial Capture-Recapture modelling
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Spatial Capture-Recapture (SCR) models, parametrized with least cost path distance, provide a unifying framework for explicitly estimating landscape connectivity and population size from individual detection data. However, we frequently encounter individuals with larger apparent space use than the rest of the population. To avoid biased population size estimates, it is common practice to remove these outliers from the analysis. Yet such individuals are likely to be very important when the aim is to estimate connectivity. We therefore investigated how unmodelled individual heterogeneity in space use affects population size and connectivity estimates from SCR models.
We first conducted a simulation study. We varied the proportion of the population whose space use is most constrained by landscape structure. We fit four SCR models, increasing the degree of modelled individual heterogeneity to assess its effect on parameter estimates. Secondly, we applied these models to the Pyrenean brown bear population, examining the effects of removing outliers or modelling explicitly heterogeneity on parameter estimates.
Our simulation study showed that population size was underestimated when no individual heterogeneity was modelled and increased with increasing the level of individual heterogeneity modelled. Moreover, unmodeled individual heterogeneity led to biased connectivity estimates towards the group, which is the most detected. In our case study, modelling individual heterogeneity was key to unravel the patterns in population space use. Female brown bear movements were restricted by road density (estimated resistance: α 2 = 0.27 [0.01, 0.43]), whereas a small group of males were not affected by road density ( α 2 = - 0.16 [-0.64, 0.24]).
Our study provides valuable insights for optimizing the application of SCR models. When the primary objective is to accurately estimate population size, we recommend removing outlier individuals with larger home range sizes to avoid the risk of underestimating population size. Conversely, when the aim is to assess landscape population-level connectivity, modelling individual heterogeneity is essential. Our findings underscore the importance of tailoring the SCR model to the specific research goal, ensuring more precise and meaningful ecological inferences.
