Quantifying the effects of development projects on biodiversity conservation
Item statusRestricted Access
Embargo end date14/06/2024
There is a global biodiversity crisis where human activities are changing biological communities and driving extinctions. Development activities, such as the construction of infrastructure, conversion of natural habitats to agriculture, and the extraction of natural resources, pose some of the most significant threats to biodiversity. Simultaneously there is a global crisis of human wellbeing, inequality, and inequity. Increased demand for mineral resources, agricultural production, and an infrastructure boom are all predicted in coming decades. Meeting the challenge of addressing human needs while preserving biodiversity requires a better understanding of the interactions between conservation and development. There is a need for research which explores the risk development presents to biodiversity and provides accurate quantification of its impacts. My thesis addresses a number of specific topics within this broad research agenda. The first part of this thesis focuses on the World Bank, a key funder of development in the global south with wide-reaching influence. Chapter 2 examines the environmental safeguard policies of the World Bank, focusing on the Environmental and Social Framework, which replaced their previous safeguard policies in 2018. There are two changes that are particularly relevant for conservation: the wider application of biodiversity offsets, and the use of borrower country frameworks to manage impacts. Due to the substantial flexibility about when offsets or borrowers’ frameworks can be used, and uncertainty around the efficacy of offsets, these changes may amount to a weakening of protections. The project-by-project nature of these mechanisms and the lack of clear criteria may also hinder future efforts to hold the Bank to account. Concerns about these changes were raised by conservation organisations during the consultation process, but the framework's formulation does not fully reflect recommendations made. Although elements of the new policy have the potential to benefit conservation, the flexibility presents a risk to biodiversity. I argue that it is vital for conservation organisations to engage effectively to ensure that any negative impacts which arise do not go unchallenged. Chapter 3 continues the focus on the World Bank and examines the potential risk to biodiversity from development projects funded by the Bank. While the Bank funds a wide range of activities, my analysis focuses on just those which have the greatest potential to cause direct harm to local biodiversity. Using a dataset of World Bank projects funded between 1995-2014, I examine the relationship between potentially harmful project activities and the ranges of globally threatened birds, mammals, and amphibians, as well as Key Biodiversity Areas, protected areas, and biodiversity hotspots. I find that 5 by 5 km cells containing a project activity are more likely than those without to contain a Key Biodiversity Area or biodiversity hotspot and have on average greater richness of globally threatened species. This relationship was statistically significant, even after considering human population and country-level socio-economic effects (except in the case of Key Biodiversity Areas). I also found limited evidence that activities are systematically placed within countries to avoid the ranges of threatened species or Key Biodiversity Areas. By contrast, I found a negative relationship between project activities and protected areas, both globally and within most countries, which may be evidence that potentially harmful activities are placed to avoid protected areas. My findings raise questions about whether the Bank's environmental safeguards have adequately translated into avoidance of highly biodiverse areas. Given the size of the World Bank’s lending portfolio and its role in setting industry best practice, these results are concerning for conservation efforts. In Chapter 4 I shift focus to examine the impact on deforestation of new mining developments in Zambia. Mining is a vital part of the global, and many national, economies. However, mining can have a range of impacts on society, biodiversity, and the local environment. These impacts include the potential to drive extensive land cover change, such as deforestation, with effects reaching far from the mine itself. As much as 49.9 million km2 of land could be considered influenced by mining globally, and the industry is likely an underappreciated contributor to deforestation in mineral-rich countries. Understanding the amount of deforestation associated with mining is important for conservationists, governments, mining companies, and consumers, yet accurate quantification is rare. To address this gap, I applied statistical matching, a quasi-experimental methodology, along with Bayesian hierarchical generalised linear spatiotemporal models to assess the impact of new mining developments on deforestation in Zambia from 2000 to present. Zambia is a globally significant producer of minerals, and mining contributes ~10% of its gross domestic product and ~77% of its exports. Despite extensive deforestation in mining-impacted land, I found no evidence that any of the 22 mines I analysed increased deforestation compared with matched control sites. This is due to control sites also having extensive deforestation, meaning that the counterfactual of no-mining is high rates of forest loss. This result is counter to my predictions and the existing literature both from Zambia and elsewhere. This suggests previous assessments based on correlative methodologies may overestimate the additional deforestation caused by mining. Chapter 5 provides a detailed examination of statistical matching as a methodology for measuring landscape-scale impacts in socio-ecological systems. Using protected area effectiveness studies as a test case, I examine whether commonly used methodologies, determined through a review of the current literature, meet assumptions about unobserved confounders. Using a unique dataset of forest plots from Zambia, I examine whether commonly used methods achieve balance in six confounders, normally unobservable in remote sensing-based analysis of protected area effectiveness. Matching improved balance in unobserved confounders, but was dependent upon subjective choices. Additionally, most datasets did not achieve high levels of balance across all confounders, meaning residual biases often remained. No individual matching covariate or parameter choice improved balance in every instance. These results suggest that, although matching can be a powerful tool, the reduction in unobserved biases is not guaranteed, and careful consideration of its limitations are needed. I argue that to improve rigour in protected area evaluations, it is vital that it becomes standard practice to test for the sensitivity of results to unobserved confounding, the influence of subjective choices in matching algorithms, and the effect of residual spatial autocorrelation. In Chapter 6 I provide a synthesis of the common themes emerging from the thesis. I argue that in order for research into the impacts of development on biodiversity to progress, better data on both biological communities and development activities are needed. These data should include the detailed spatial footprint of projects. I also argue that new approaches which can incorporate quantitative and qualitative research to better capture diffuse as well as direct impacts of development are required to fully understand how development impacts biodiversity. Finally, I argue that future efforts should also focus on understanding the drivers of development to allow for demand-side interventions, in conjunction with site-level management, to better improve biodiversity outcomes while meeting human wellbeing needs.