Modelling and Predicting Footfall in Natural Spaces Using Statistical Models
Year: 2026
Type: conference-paper
Abstract: Understanding visitor footfall to green spaces, such as nature reserves is important for planning, infrastructure management, the design of sustainable transport systems and measuring and evaluating the impact of interventions. Traditional surveys are often expensive, provide limited temporal coverage, and lack detailed information at the individual monitoring site level. In contrast, automated people counters provide continuous digital footfall data, creating new opportunities to analyse and predict pedestrian movement on paths, but such counters are expensive and limited in number. This study aims to model and predict digital footfall on paths in natural spaces using two statistical approaches: a linear model and Bayesian Additive Regression Trees (BART). We used automated people counter data to train the model and validate the predictive performance of the models and quantify the associated uncertainty in the predictions. While both models were able to reproduce the temporal dynamics of footfall, the BART model generally achieved higher predictive accuracy than the linear model across most monitoring sites. While these results are encouraging for many locations, we also identify areas where model performance could be improved. These insights help inform the selection of suitable locations for future monitoring sites and support planning decisions regarding infrastructure provision and resource allocation.
Institutions: University of Exeter, Natural England
Language: English
Topic: Urban Green Space and Health
Subfield: Urban Studies
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7204794288