Outcomes of the Imperial-TKT Sprint Challenge round two projects

Three multidisciplinary Imperial College London research teams have reported outcomes as part of the Turner Kirk Trust (TKT) Sprint Challenge.

The Sprint Challenge was launched in 2023 in partnership with Imperial. It aims to address unsolved global conservation challenges by funding experimental, multidisciplinary research projects that may struggle to attract conventional funding.

Following the success of the first round, which provided a total of £50,000 in funding to three research teams, the second round launched in 2024. This round is now complete, with three new projects tackling critical conservation problems through novel mathematical approaches. The success of these projects is due to the generous technical and financial support received from TKT.

You can read about the outcome of each project below.

Understanding collective animal behaviour for conservation

Hurricanes, wildfires, and other extreme weather events driven by climate change threaten animal populations, yet the mechanisms by which collective behaviour influences survival under such events remain poorly understood.

The team - Dr Dante Kalise (Co-PI, Mathematics), Professor Vincent Savolainen (Co-PI, Life Sciences), and Sara Bicego (former PhD student in Mathematics) - selected the macaque population on Cayo Santiago - where 98% survived Hurricane Maria in 2017 - as a case study, with comparative fieldwork at Longleat Safari Park. They developed a two-population agent-based model distinguishing leaders (who could anticipate storm trajectory) and followers, incorporating endogenous social forces (attraction, repulsion, alignment) and exogenous environmental forces (wind influence, repulsion from storm center).

Computational simulations tested scenarios with and without leader anticipation, and drone footage collected in December 2024 provided further empirical data.

The results showed that leadership and anticipation are critical to animal survival.

The team produced a validated modelling framework and simulator, and generated empirical datasets (drone footage) for further analysis. It also catalysed interdisciplinary connections between mathematics, conservation science, and AI, laying groundwork for journal publications, future field studies, and generalisation of the methodology to other species and climatic threats.

Climate variability impacts on deforestation in Nepal

While community-based conservation is proposed to enhance social and environmental resilience, its effectiveness under climate-induced shocks is not well quantified, particularly regarding small-scale resource harvesting patterns.

Focusing on Nepal’s long-standing community forest management program, the team of researchers (Dr Matthew Clark from the Centre of Environmental Policy and Dr Adam Sykulski from the Department of Mathematics), produced high-resolution country-scale tree-cover change data (3 m resolution, Jan 2018 - Dec 2023). They collated daily rainfall data for 3,983 municipalities and devised a novel 'rainfall irregularity' metric via functional data analysis to detect climate shocks. Community forest registrations were matched to municipalities, and additional covariates (economic, demographic, topographic, road density) were assembled. A Bayesian hierarchical regression assessed how harvesting patterns responded to precipitation anomalies in community- versus non-community-managed forests.

The analysis revealed that, under baseline conditions, forest loss rates were similar across management types, but during climate shocks, municipalities with community forestry maintained stable harvesting rates while those without experienced up to 4x more tree loss.

This suggests that community forests actively buffer against climate shocks, demonstrating that devolving management to communities can build environmental resilience

Using AI and big data to predict community conservation efforts

Conservation planning often reacts to past patterns; there is a need for predictive approaches to anticipate where and when community-based initiatives are likely to emerge, especially given complex socio-ecological drivers.

The study, led by Dr Andreas Joergensen (Mathematics) and Dr Thomas Pienkowski (Centre for Environmental Policy), targeted Community Forestry engagement in Zambia by assembling data for 539,000 rural settlements, covering factors including forest cover, food insecurity, road access, and literacy.

This proof-of-concept demonstrates that combining AI methods with 'big' spatial data can reliably forecast community engagement patterns, surprising given real-world complexities.

The findings lay a foundation for scenario-based simulations to predict future emergence of initiatives and guide targeted facilitation, with dissemination planned via NGO networks and further funding exploration for AI applications in conservation science.