Measuring and predicting the impact of private protected areas

Biodiversity is in decline globally, and protected areas (PAs) are a cornerstone of the policy response for reducing this decline, with calls to protect 30% of the world’s ecosystems by 2030. While most PAs are public, such as national parks, private protected areas (PPAs) are an important contributor to biodiversity conservation, as many threatened species and habitats exist on privately owned land. The number and size of PPAs have been increasing globally and they will likely play a significant role in meeting the 2030 targets. However, the efficacy of PPAs in reducing biodiversity loss is not well understood. This can be measured by estimating the 'impact' of a PPA: the observed outcome under protection relative to the counterfactual without the protection. Strategically locating PPAs in areas where they will have a greater impact will result in greater reductions in biodiversity loss — however, it is first necessary to robustly estimate the impacts of existing PPAs. This thesis aims to advance understanding of PPA impacts in reducing biodiversity loss, drawing on a diverse range of literature ranging from conservation science through to machine learning, statistics, and impact evaluation. Specifically, I develop a novel framework that incorporates individual and program-level PPA impacts, as well as estimating both past and future impacts. I analyse the extent to which biases in the locations of PPAs, relative to threatening processes, differ from PAs, and demonstrate an approach for integrating impact into conservation planning. All research was conducted in Australia, which is an ideal testbed for studying PPAs, since PPAs are scattered throughout the territory, and it is home to some of the oldest PPAs in the world. In addition, Australia has high-quality spatial data available to support the research objectives of this thesis. First, I analyse how the location of PPAs nationally are associated with covariates that represent biodiversity and the likelihood of areas being cleared for agriculture and urban development, and examine how that compares to public PAs. Previous work has shown that public PAs tend to occur in biassed locations (e.g., low land capability and remote areas). It is not clear the extent to which PPAs are similarly biassed, and I examine this question at the level of the Australian continent. I find that the PPAs tend to target areas of high threatened species richness; however, they are, on average, placed in areas that have lower risk of being cleared compared to if they were placed randomly on private land. I find that this bias towards unproductive land is greater in PPAs compared to public PAs. Next, I demonstrate a transparent and robust workflow using synthetic control design, statistical matching, and time-series data of woody vegetation cover to estimate impact for individual PPAs in the state of Victoria, Australia. Typically, conservation programs are assessed by estimating average impact, but this masks variations among individual PPAs. By combining individual impacts, I also estimate the program-level impact using a meta-analytic approach. Using this framework, I demonstrate that, while the covenants in the Goldfields region of Victoria have, on average, led to improved woody vegetation cover (0.3-0.8% per year), there is significant variation among the individual covenants (-4 to +7% change per year). This variation could be explained by ecological factors, and differing management requirements and landholder behaviours. Understanding both the past and likely future impacts is crucial for a comprehensive assessment of the conservation value provided by PPAs. While estimating past impact helps us to estimate the efficacy of existing programs, future impact enables us to anticipate the potential future benefits; combining the two provides a longer baseline over which to measure impact. However, there is limited research that looks at past and future impact in an integrated framework. I develop a method combining predictive modelling, scenario analysis, and statistical matching to assess past and future impacts of PPAs in the North Coast Bioregion of New South Wales (NSW), Australia. I evaluate the past (2010-2020) and future (2020-2030) impacts of 119 PPAs in preventing woody vegetation loss. I find a positive but marginal effect of PPAs in avoiding past and future woody vegetation loss, with an average annual avoidance rate of 0.10% in the past and 0.12% in the future. Systematic conservation planning is a method for spatially locating PAs to identify cost-effective areas to protect biodiversity. Here, I compare four strategies for prioritising PPAs and evaluate their performance based on return on investment (ROI) of prioritised areas, calculated as the ratio of impact and cost. Using data on habitat suitability of 306 threatened species, a spatial layer of predicted risk of woody vegetation loss, and land acquisition costs in the North Coast Bioregion of NSW, Australia, I undertake prioritisations based on: (i) representation of threatened species; (ii) representation and cost; (iii) impact of avoiding habitat loss; and (iv) impact and cost. I calculate the landscape-level ROI and find, as expected, the strategy combining impact and cost outperforms other strategies — but, surprisingly, only by a factor of 1.3 compared to representation with cost. Impact and cost improves on impact only by a factor of 3.6 and representation only by a factor of five. Examining species-level ROI shows that using impact and cost also provides greater conservation benefits for rare species compared to other strategies. This case study shows how incorporating impact and cost may have significant benefits to prioritisations, but shows that adding cost alone has a relatively greater benefit than moving to impact alone. PPAs contribute significantly to the global conservation effort, but there is limited information on the extent to which PPAs have achieved conservation impact. The research in this thesis has contributed to the advancement of PPA impact evaluation. However, the approaches developed for measuring impact are also applicable to public PAs and other effective area-based conservation measures. This thesis helps us to understand the impact that PPAs have achieved to date, and supports the planning and implementation of more impactful PPAs, thus supporting the 2030 framework for expanding the global PA network.