Factors affecting success of conservation translocations of terrestrial vertebrates: A global systematic review

Abstract

Translocation—moving individuals for release in different locations—is among the most important conservation interventions for increasing or re-establishing populations of threatened species. However, translocations often fail. To improve their effectiveness, we need to understand the features that distinguish successful from failed translocations. We assembled and analysed a global database of translocations of terrestrial vertebrates (n = 514) to assess the effects of various design features and extrinsic factors on success. We analysed outcomes using standardised metrics: a categorical success/failure classification; and population growth rate. Probability of categorical success and population growth rate increased with the total number of individuals released but with diminishing returns above about 20–50 individuals. Positive outcomes—categorical success and high population growth—were less likely for translocations in Oceania, possibly because invasive species are a major threat in this region and are difficult to control at translocation sites. Rates of categorical success and population growth were higher in Europe and North America than elsewhere, suggesting the key role of context in positive translocation outcomes. Categorical success has increased throughout the 20th century, but that increase may have plateaued at about 75% since about 1990. Our results suggest there is potential for further increase in the success of conservation translocations. This could be best achieved by greater investment in individual projects, as indicated by total number of animals released, which has not increased over time.

Methods

This data was compiled from the literature and a questionnaire. How this data was collected and processed is detailed in the paper. The attached code will run all analyses detailed in the paper.

Usage Notes

To replicate the results in their entirety you will likely need a supercomputer. A representative version of the results can be created by simply reducing the number of times the dataset it split (say from 1000 to 10).