Robustness of field studies evaluating biodiversity responses to invasive species management in New Zealand

Abstract

Benefits of invasive species management for terrestrial biodiversity are widely expected and promoted in New Zealand. Evidence for this is presented in policy and scientific reviews of the literature, but the robustness and repeatability of the underpinning evidence-base remains poorly understood. We evaluated the design of field-based studies assessing biodiversity responses to invasive species management in 155 peer-reviewed articles published across 46 journals from 2010 - 2019. Each study was assessed against nine principles of experimental design, covering robustness of sampling and avoidance of bias. These principles are important in New Zealand to detect treatment effects from environmental variability driven by underlying gradients such as soil fertility, climate and disturbance. Fifty two percent of studies defined a sampling universe and 68% of studies specified the treatment. Whereas, 54%, 74%, and 50% of studies did not utilise replication, representatively sample the universe, or quantify invasive species, respectively. Ninety five percent of studies quantified biodiversity responses, although a high proportion of these did not representatively sample replicates. Initial conditions and accounting for effects of experimental implementation were not utilised in 57% and 84% of studies respectively. No studies avoided observer/analyst bias using blinding methods, despite this being widely adopted in other fields. Ordinal logistic regression showed these principles varied in how robustly they were applied among categories of biodiversity responses and invasive species. Our findings suggest that greater attention to experimental design principles is desirable: supported by researchers, funding agencies, reviewers, and journal editors. Greater resources is not necessarily a solution to these design issues. Undertaking fewer studies, that are individually more expensive because they better adhere to experimental design principles, is one alternative. Our intent in this article is to improve the robustness of future field studies for at least some principles. Robust designs have enduring value, reduce uncertainty and increase our understanding of when, where and how often the impacts of invasive species on biodiversity are indeed reversible.

The rise and rise of predator control: a panacea, or a distraction from conservation goals?

“…That is, current predation management frequently assumes positive biodiversity outcomes, yet fails to test hypotheses (Betts et al 2021) that would determine the mechanisms behind observed biodiversity responses. Further, experimentally robust assessments of actual outcomes are rare (Allen et al 2023). For example, we found just one published account of long-term monitoring (defined as > 10 years or three or more 1080 cycles) of a rat-vulnerable bird species following repeated aerial 1080 applications (Robertson et al 2019); current landscape-scale projects conducted under the umbrella of PF2050 are yet to publish any quantitative assessments of native biodiversity responses to predator eradication.…”

“…For example, we found just one published account of long-term monitoring (defined as > 10 years or three or more 1080 cycles) of a rat-vulnerable bird species following repeated aerial 1080 applications (Robertson et al 2019); current landscape-scale projects conducted under the umbrella of PF2050 are yet to publish any quantitative assessments of native biodiversity responses to predator eradication. This lack of robust, hypothesis-driven assessment of native species' responses (Betts et al 2021), or adherence to basic principles of good experimental design (Allen et al 2023) makes it difficult to critically test assumptions behind predation-focused management, or to quantify the biodiversity gains it delivers relative to cost.…”

“…However, as these are generally biased towards lowland forests (Innes et al 2019), a different approach would be needed for extensive montane beech forests or dryland ecosystems that currently receive extensive predator control but few if any other interventions. In many mainland locations, hypothesis testing and experimental approaches are needed to assess synergies between predator removal and improvements in habitat quality delivered by concurrent control of ungulates (and a wider suite of predators) (Allen et al 2023).…”

“…Second, experimental manipulations, with supporting ecological models, could be better leveraged to infer impacts of different combinations of pressures on indigenous biodiversity from field observations (Ruscoe et al 2005;Holland et al 2013) and to predict responses to management interventions (Tompkins et al 2013). Large-scale, long-term, and well-funded field experiments, designed to underpin predictive ecological models, are vital components of research needed to support robust biodiversity and ecosystem conservation (Allen et al 2023). Unless such experiments are done and hypotheses tested, there is a significant risk that key drivers of decline may be overlooked or discounted (Caughley & Gunn 1996;Betts et al 2021).…”

“…In general, ecological sciences are strongly single-disciplinefocused in Aotearoa: inter-and trans-disciplinary, crosscutting ecosystem ecology and Mātauranga Māori have been chronically under-funded for decades (Green & Clarkson 2006;PCE 2020;Kukutai et al 2021). A stronger focus on hypothesis testing and good experimental design are also required to understand the mechanisms behind observed ecological responses (Betts et al 2021;Allen et al 2023). As a consequence, the pool of specialist skills needed to advance knowledge on complex ecosystem dynamics, test hypotheses in an experimental framework, measure biodiversity responses to management actions, construct predictive models that can support decision-making, and translate that information into policy is extremely limited.…”