Identifying robust policies against deforestation
Agriculture is one of the main drivers of deforestation, which contributes to carbon emissions and biodiversity loss. Finding ways to increase the productivity of agriculture on the same amount of land while protecting areas of existing forest can bring economic benefits and reduce deforestation. My work focuses on understanding agricultural production systems, particularly for high-value cash crops, and encoding these systems in computational simulations. It is possible to use these simulations to test the effect of different policies such as increasing fertilizer use and mechanization affect the profitability of agriculture and the pattern of deforestation. The simulations can also test the influence of uncertainties such as commodity prices on policy outcomes. As a result, decision makers can identify optimal policies which are robust to uncertainties. The advantage to a simulation is that it lets decision makers test these policies in advance before making financial commitments, and avoid potentially costly mistakes. I am also exploring how new data from satellites, drones, and sensors, as well as the application of machine learning algorithms to these data can inform these simulations. See the videos below for an example simulation and more about the methods.
This work is funded by the Green Templeton College DPhil Scholarship.
Group members involved:
Adam Formica, Richard Bailey, Richard Grenyer
Optimization of anti-poaching measures
Poaching of African elephants Loxodonta africana is one of today’s most
pressing and widely-publicized conservation issues. The international
demand for ivory – a substance so valuable it is sometimes referred to as
‘white gold’ – has led to the rapid decline of many elephant populations.
There are numerous controversies and uncertainties surrounding elephant
poaching and the ivory trade, and debates on how to mitigate the poaching
crisis continue to the present day.
Agent-based models are promising tools to both inform decision-making
and improve our understanding of elephant poaching, because they acc-
ount for the complex and dynamic interplays between elephants, poachers,
and law enforcement. This project explores how various policy and manage-
ment interventions affect levels of poaching. With this model, we hope to
structure the debate around this complex and controversial issue, and help
guide future research or policymaking. This model could also be adapted
in the future to fit the specific conditions of a particular park or country, thus
providing greater insights into the potential outcomes of different
Group members involved: Emily Neil, Richard Bailey, Ernesto Carrella, Jens Koed Madsen
Stability of mutualistic ecological systems
Conserving ecosystem function and associated services requires deep understanding of the underlying basis of system stability. While the study of ecological dynamics is a mature and diverse field, the lack of a general model that predicts a broad range of theoretical and empirical observations has allowed unresolved contradictions to persist. In this work we provide a general model of mutualistic ecological interactions between two groups and show for the first time how the conditions for bi-stability, the nature of critical transitions, and identifiable leading indicators in time-series can be derived from the basic parameters describing the underlying ecological interactions. Strong mutualism and nonlinearity in handling-time are found to be necessary conditions for the occurrence of critical transitions. We used the model to resolve open questions concerning the effects of heterogeneity in inter-species interactions on both resilience and abundance, and discuss these in terms of potential trade-offs in real systems. This framework provides a basis for rich investigations of ecological system dynamics, and may be generalisable across many ecological contexts.
For further details see Feng & Bailey (2018).
This project was funded by the University of Oxford John Fell Fund.
Project now complete - Group members involved
Wenfeng Feng, Richard Bailey, Kirsty McGregor
Belief persistence in social networks
Social networks include structures such as social media and Internet forums. While they are a tremendous boon for sharing information and generating social contacts, they are also subject to the spread and maintenance of misinformation, manipulation, and alternative facts. This type of information represents a serious problem, as these beliefs can spread, solidify and persist in social networks.
The project explores how these beliefs can emerge, spread and be maintained through social network interactions. We approach this through rational processes (such as Bayesian belief revision) and Agent-Based Models. Given the process approach, we can test possible interventions to influence beliefs and behaviour in social networks. The project targets belief diffusion in diverse areas such as political campaigns, the climate change debate, and conspiratorial thinking.
Group members involved:
Jens Madsen, Richard Bailey, Toby Pilditch
Agent-based models are notoriously difficult to estimate because they contain many parameters, model many competing mechanisms and output a cornucopia of summary statistics which are difficult to weight. While many algorithms work for this or that estimation, no estimation algorithm seems to be inherently better than the others.
This project tries to collect a set of common "reference-table" estimation algorithms that can be quickly deployed to estimate agent-based models. While these may not always be the "best" algorithms they produce useful confidence intervals and are easily testable which is important to notice under-identification and explore which parameter may need more sensitivity analysis.
Group members involved:
Ernesto Carrella, Richard Bailey, Nicolas Payette