Imperial scientists solve conservation problems with new mathematical techniques

Three multidisciplinary Imperial College London research teams have secured over £50,000 in funding from the Turner Kirk Trust (TKT) Sprint Challenge.

The TKT Sprint Challenge is a research initiative that joins researchers from the Centre for Environmental Policy (CEP), Department of Life Sciences (DoLS) and Department of Mathematics (DoM) to solve environmental challenges with new mathematical approaches.

This year’s funded projects include:

Understanding collective animal behaviour for conservation

One of the winning teams is focused on predicting collective animal behaviour, using modern scientific computation methods. The project aims to analyse how groups of animals, such as sheep and birds, exhibit self-organisation and patterns like herding and flocking under the presence of UAVs and climatic events.

Dr Dante Kalise from the DoM and Professor Vincent Savolainen from the DoLS will collaborate to develop mathematical models that quantify the interactions among individual animals within these groups.

Collective animal behaviour may be characterised by the need to minimise energy expenditure or avoid predators, from which predictable collective behaviour arises.

"In applied mathematics, we have advanced computational methods to quantify collective animal behaviour but in order to develop meaningful models, we need to collaborate with life scientists and behavioural ecologists, like Vincent, to guide us on the relevant interactions and features to be analysed,” said Dr Kalise.

These models could be transformative for conservation efforts. By predicting how animal populations will respond to environmental disruptions or human interventions, such as the presence of drones for monitoring, conservationists may design better field experiments.

Climate variability impacts on deforestation in Nepal

Dr Matthew Clark from the CEP, along with Dr Adam Sykulski from the DoM, will use advanced remote sensing techniques and statistical models to analyse deforestation patterns in Nepal.

As climate change drives more extreme weather events, communities may be forced intensify deforestation to meet their daily needs.

"We aim to understand how changes in rainfall and temperature affect deforestation, particularly in community-managed forests," said Dr Clark.

Our goal is to move from reactive to proactive conservation, using these forecasts to implement strategies... Dr Matthew Clark Centre for Environmental Policy

The team will apply remote sensing and spatiotemporal statistical methods to satellite imagery of Nepal, which they will use to analyse and forecast changes in forest cover at the resolution of individual trees.

By collaborating with local conservation groups and integrating socio-economic data, the team hopes to create actionable insights for both policymakers and conservationists to better protect these vital ecosystems.

"Our goal is to move from reactive to proactive conservation, using these forecasts to implement strategies like farm insurance to mitigate economic pressures driving deforestation," Dr Clark said.

Using AI and big data to predict community conservation efforts

Predicting local community engagement in conservation is critical to achieving global commitments to expand area-based conservation to 30% of the world’s surface by 2030.

Dr Thomas Pienkowski from the CEP and Dr Andreas Joergensen from the DoM aim to address this challenge by applying AI and big spatial data to forecast patterns of community conservation engagement.

“Meeting these global goals will require substantial scaling up of community-conservation measures,” said Dr Pienkowski, “However, it’s often hard for conservation agencies to know which communities are most likely to engage with these measures and, thus, where they should focus their efforts.”

The project will use graph neural networks to predict community engagement. These can analyse different characteristics of communities to classify and predict engagement.

Our approach here is quite novel compared to other studies trying to understand scaling in conservation... Dr Tim Pienkowski Centre for Environmental Policy

The dataset includes information on 2.5 million settlements across Uganda, Malawi, Kenya, Tanzania and Zambia, integrating data on existing conservation initiatives and factors like wealth levels and proximity to tourism hotspots.

“Our approach here is quite novel compared to other studies trying to understand scaling in conservation,” said Dr Pienkowski, “We’re using cutting edge AI techniques to see if we can make accurate predictions on who might engage.”

By enhancing predictive capabilities, the project may help agencies strategically allocate resources to improve the scaling of conservation efforts.