Investigating habitat use and preferences of a threatened species can be challenging, especially if wild populations have decreased to such low numbers that they occupy only fractions of their former natural range. Hence, assessing habitat suitability of a potential release site for a threatened species before a reintroduction attempt can be difficult because frequently no comparable baseline data are available. In these instances, post-release monitoring data can inform about habitat use and preferences of a reintroduced species. Here, we use monitoring data of an endangered endemic island bird, the Chatham Island black robin Petroica traversi, to investigate habitat preferences and the temporal change in distribution patterns across 26 years following a reintroduction. We show that densities and distribution of black robin pairs at the reintroduction site have changed significantly over the years. Spatial distribution of pairs is clustered, and this clustering has intensified as the population increased. We used the maximum entropy method MaxEnt to model habitat suitability on the island, showing that black robins clearly prefer forested areas inland that are within 70 m to the forest edge at lower elevations (<40 m a.s.l.) and on slopes that have a N-NE aspect. The model also identified one area on the island that comprises suitable habitat, but is currently uninhabited. Applying maximum observed densities to the available area of each suitability class, carrying capacity is estimated as 170 nesting pairs across the island, highlighting the need to find further appropriate habitat urgently. Overall, these topographical and habitat preferences considerably restrict this species' potential distribution, a constraint that has serious conservation implications for future population growth of current populations and (re)introductions to new locations. This study demonstrates that post-release data can reveal relevant limitations to habitat use of highly threatened species.

Abstract For years researchers have recognized the need to consider environmental and contextual variables in the social and behavioral sciences. Multilevel models have grown in popularity in large part because they provide a means to explicitly model the influence of context on many individual level processes. However, in applications of these and other statistical models that incorporate context into the analysis, rarely is physical location or distance between entities considered. In this paper we discuss a variety of spatial analysis techniques and their applications in educational and psychological research. We provide examples with the SAS software package and other more specialized spatial analysis software.