Playing God with guppies – informing tough conservation decisions using a model experimental system

With human-induced extinctions and population declines accelerating worldwide (Dirzo et al., 2014), it is more important than ever to foster dialogue among researchers, conservationists, and the public at large about the costs and benefits of potential management actions. We are therefore grateful for the insightful commentaries provided by Mills (2017), Grueber (2017) and Tallmon (2017) on our article in this issue of Animal Conservation (Kronenberger et al., 2017), in which we test the demographic effects of divergent immigrants on small laboratory populations of Trinidadian guppies Poecilia reticulata. Using replicated mesocosm populations, we found that divergent immigrants had a positive effect on population fitness compared to no immigrants at all. This positive effect was larger when immigrants were adaptively similar but genetically divergent (as opposed to adaptively divergent but genetically similar). As reflected by the commentaries, our study adds to the mounting evidence that demographic and genetic rescue are plausible management strategies in certain cases and provides some empirical insight into the outcome of non-ideal translocation scenarios. However, much work is still needed to accurately predict the outcome of translocations, especially those between adaptively and/or genetically divergent populations. When the individuals introduced into a vulnerable population are divergent, there is a greater likelihood of negative outcomes such as genetic swamping with maladaptive alleles and outbreeding depression (Frankham et al., 2011). It is therefore generally agreed that, if translocations are the only option for re-establishing gene flow (Mills, 2017), donor and recipient populations should be adaptively and genetically similar. Indeed, when divergent immigrants are avoided, translocations designed to induce genetic rescue are almost always successful (Frankham, 2015; Whiteley et al., 2015). But as we proceed further into the Earth’s sixth mass extinction (Dirzo et al., 2014), the reality is that often the only potential donor populations that remain are either adaptively divergent, genetically divergent, or both (e.g. Funk et al., 2016). If translocation is to be an option in these cases, we will need more precise information about when divergent immigrants will and will not have the desired positive effect on population fitness. This was the primary motivation for our study. The primary merit of our study, we believe, is that it was experimental. Tallmon (2017) reminds us that a long-standing issue in conservation biology is ‘a lack of replication, randomization, and experimental controls needed to ascribe the effects of applied treatments to observed outcomes’. This issue arises out of the nature of the field itself; because our goal is conservation, we are almost exclusively dealing with threatened species that cannot be ethically or practically manipulated, or large, charismatic species that are not amenable to study across multiple generations (Tallmon, 2017). Utilizing case studies to generate recommendations for when and how to carry out translocations (e.g. Frankham et al., 2011) is a valuable approach, but it should be complemented with insights gleaned from other theoretical and empirical research, including manipulative experiments. Case studies minimize risk as they strive to provide populations with the best possible chance of rescue. Manipulative experiments, in contrast, are able to embrace risk and test the boundaries of theory by simulating more extreme translocation scenarios, which offers a powerful means of testing and refining previously established guidelines. Given our experimental design, we were unfortunately unable to tease apart the relative impacts of demographic and genetic rescue and, because population monitoring lasted <2 generations, we could not be certain whether the rescue effect would have persisted in later generations. Future experiments should aim to rectify the shortcomings of our study by, for example, including a treatment with immigrants from the same population to control for strictly demographic effects, extending population monitoring for multiple