The COVID-19 pandemic has had a huge impact on urban mobility, leading to two major questions: (i) how have cities changed during the pandemic? and (ii) which changes will remain as the pandemic subsides? To help answer these questions, this project will build on previous CDRC-funded work to create a spatio-temporal machinelearning model to estimate footfall around Leeds. It will consider the local urban configuration, external factors (such as public holidays and local weather conditions) and, importantly, the impact of various mobility restriction measures. First, the model will be trained using data from the years before the pandemic, and then lockdown restriction conditions like the “rule of six” and closing/opening of non-essential retails will be incorporated. Given the wide spatial distribution in the available footfall sensor data in the case study area (Leeds), the model will be able to estimate the overall change in footfall, as well as the heterogeneous impacts that restrictions will have on different local areas. The model will also allow analysis about city occupation in different scenarios. In summary the project will:
- Develop a suitable open-source footfall machine-learning model
- Calibrate and validate the model using footfall data from the CDRC (SmartStreetSensors) and from Leeds City Council (footfall cameras)
- Develop a dashboard to present maps and related visual outputs to help policy makers to easily explore different scenarios
The project partners, Leeds City Council, have a particular interest in better estimating how footfall in city-centre will vary as the pandemic, and related policies, evolve. Hence a specific case study will be designed around their immediate policy objectives at the start of the project, taking into account current conditions.
Although based on Leeds, the work will be generalisable to other cities that have footfall estimates and, once the model has been trained, could even be applied even where footfall data do not exist. Ultimately we aspire to attract further funding to create a nationwide footfall model, which would represent a great methodological advance as well as a contribution to outcomes like improving public health and creating better living standards. This would also represent an extremely attractive outcome for the CDRC.
This project builds on the prototype machine-learning model developed as part of a previous CDRC-funded internship¹ and will use the data insight gained through a forthcoming project²:
- Predictive data analytics for urban footfall - CDRC-funded internship (2017) attempting to quantify and model ambient populations
- Measuring Ambient Populations during COVID in Leeds City Centre - CDRC-funded internship, starting April 2021
This project also aligns with a number of projects conducted at UoL in collaboration with partners:
- Rapid Assistance in Modelling the Pandemic (RAMP): Urban Analytics. The RAMP:UA project will benefit from an improved understanding about how city centre usage varies under different lockdown scenarios as a means of informing its estimates of population mobility during lockdown. This project therefore offers an opportunity for the CDRC to exploit links with RAMP partners in government and academia.
- Data Assimilation for Agent-Based Models (DUST) requires up-to-date footfall estimates to inform its real-time population mobility models so will benefit from the model produced here.
- Analysing COVID-19 Mobility Responses through Passively Collected App Data - CDRC-funded internship. This LIDA project is analysing smart-phone generated mobility traces from Cuebiq data, and this data can inform this project as well.
- Bringing the Social City to the Smart City (ESRC-Turing) will benefit from a greater understanding about how the dynamics of cities changes under government.
Relevant background publications and research blogs include:
-
N. Malleson et. al. (2017) Predictive data analytics for urban footfall.
-
A. Whipp. et al. (2021) Estimates of the Ambient Population: Assessing the Utility of Conventional and Novel Data Sources. International Journal of Geo-Information 10(3), 131.
-
Ross, S. and Breckenridge, G. (2021) COVID for Christmas? Analysing patterns of Christmas mobility in the UK.
-
Johnson, P., M. A. Andresen, and N. Malleson (2020). Cell Towers and the Ambient Population: A Spatial Analysis of Disaggregated Property Crime’. European Journal on Criminal Policy and Research.
-
Crols, T., and N. Malleson (2019) Quantifying the Ambient Population Using Hourly Population Footfall Data and an Agent-Based Model of Daily Mobility. GeoInformatica.
-
Malleson, N., and Andresen, M.A. (2016) Exploring the impact of ambient population measures on London crime hotspots. Journal of Criminal Justice 46 pp 52-63.
- Indumini Ranatunga
- Nick Malleson
- Vick Houlden
- Patricia Ternes