This study seeks to answer the research question: "How did movement patterns in London change during the COVID-19 pandemic, and were certain categories of locations affected differently?" By analyzing human mobility trends across different Polygons of Interest (POIs), such as major transport hubs, parks, and event venues, and time intervals, this project aims to uncover spatial and temporal variations in movement behaviors and assess the broader implications for urban planning and crisis management.
The analysis delves into the impact of lockdown phases, the relaxation of restrictions, and the subsequent recovery of movement patterns. By applying spatial and temporal clustering techniques, such as ST-DBSCAN and Moran’s I, the project provides insights into shifts in urban activity, the emergence of localised hotspots, and the varying effects across different location types. This project not only offers a comprehensive understanding of movement trends during the pandemic but also serves as a foundation for future studies in urban resilience and mobility planning.
The foundation of this project is the 2020 movement data of London, which provides daily aggregated metrics of anonymized human activity. The data is provided by mapbox and can be retreived here.
To enhance spatial analysis, the dataset is complemented with additional geospatial information, including:
Preprocessing involved extensive data cleaning, mapping, and aggregation, with a particular emphasis on temporal trends. This approach allowed for an in-depth examination of mobility changes before, during, and after key COVID-19-related events, such as lockdown announcements and subsequent easing measures.
By integrating spatial clustering techniques and time-segmented analyses, this project aims to provide a clear understanding of how mobility patterns evolved throughout the pandemic.
The movement dataset used in this study provides a numerical representation of human mobility across London during 2020. The data is normalised against the UK national average, with a value of 1.0 indicating mobility equal to the national mean. Values above 1.0 signify higher activity, while values below 1.0 indicate reduced movement levels.
The dataset enables detailed analysis of fluctuations in urban activity, particularly in response to key pandemic events such as lockdowns and the easing of restrictions. Visual representation of the movement data further assists in identifying areas of high and low mobility throughout different time periods.
For this project, six categories of POIs were chosen to assess movement variations across different urban locations:
These POIs were selected based on their significance in urban mobility and their varying degrees of impact during the pandemic. Their inclusion enables an in-depth comparison of mobility changes across different location types.
The Temporal Data section highlights how mobility patterns in London evolved throughout the different phases of the COVID-19 pandemic in 2020. By dividing the year into distinct intervals aligned with policy changes and restrictions, the short- and long-term effects of lockdowns, re-openings, and other interventions on human activity are analyzed.
To better understand mobility changes, the year 2020 has been divided into seven key intervals, each corresponding to COVID-19 measures or milestones. The time intervals are based on governmental documents that can be found here.
The Events Dataset focuses on specific dates where notable events and policy decisions occurred. These events include football matches, concerts, and public gatherings, particularly in POIs such as football stadiums and large event venues. By examining mobility patterns during these events, a deeper understanding of the short-term fluctuations in movement across London can be retreived.
The table below summarizes these events, their corresponding dates, and their places:
Before diving deeper into the spatial (temporal) analyses, the following figures provide an initial overview of how human mobility evolved in London during 2020. They show the overall trends of the movement data in relation to the COVID-19 cases (see Fig.1), as well as a regional classification of London, to get a better overview about the average mobility trends in different regions. Therefore, London's boroughs were divided into different regions, which are also shown in the map (see Fig.2) below.
Figure 1: Comparison of movement data with COVID-19 cases across different lockdown phases.
This visualization highlights the correlation between movement trends and COVID-19 cases in London. The sharp decline in mobility during the lockdown phases is clearly visible, followed by gradual recovery as restrictions eased. One notable sight is the small drop (red points) in the summer months. Surprisingly the COVID-19 (blue line) cases per day were at an all-time low at that time. This could be explained by the fact that there were summer holidays and people could travel in UK.
Around the second lockdown there is a major rise of COVID-19 cases and in the second lockdown the COVID-19 cases drop fast. In the Tier 4 System towards the end of the year the COVID-19 cases rise rapidly again. This also correlates with the movement data, which decreases with the beginning of the Tier 4 System.
When dividing London into different regions (see Fig.2), distinct mobility patterns across the city can be observed. In the city center, where the movement is much higher than in the outskirts (see Fig.2), the impact of lockdowns has the biggest impact. But the city center is always higher than the outskirts and also the overall mean (see Fig.1).
Figure 2: Regional classification of London alongside average mobility trends in different regions.
By employing advanced spatial and temporal analytical methods, this chapter explores the deeper effects of the COVID-19 pandemic on movement patterns across the city. From broad temporal trends to localized clustering phenomena, the findings offer insights into how Covid 19 transformed human activity. Specific analyses, including Moran's I and ST-DBSCAN, highlight variations across different periods and POIs, unveiling trends and anomalies. This chapter combines both visual and quantitative approaches to deepen the understanding of mobility behavior during the pandemic year.
The "Interval analysis" page allows users to interactively analyze human activity patterns in London during the different COVID-19 intervals. By selecting and comparing maps from different phases, such as Pre-Corona, First Lockdown, and Four-Tier Restriction Phase, users can visually assess the impact of lockdowns and reopening policies on urban mobility.
This ST-DBSCAN was applied to the Greater London Area, specifically Westminster, Greenwich, and Regent's Park in London. It examines how mobility changed before and during the COVID-19 pandemic to identify hotspots. The study also discusses methodological limitations and offers recommendations for future research to enhance clustering accuracy and expand the scope of analysis.
The Moran's I analysis reveals spatial patterns in human activity during COVID-19 intervals. Global Moran's I highlights the clustering of activity, with stronger patterns during the Pre-Corona and Relaxation phases, while lockdowns showed decreased clustering. Local Moran's I provides a view, uncovering contrasts near roads and POIs.
The comparison across POI categories reveals consistent declines in activity across most locations during COVID-19. Bus stops, subway stations, football stadiums, sightseeing spots, Royal Parks, and large event venues all saw reduced visitor numbers compared to pre-pandemic levels. These decreases reflect the widespread impact of restrictions on commuting, tourism, and large gatherings, altering urban mobility patterns.
This discussion evaluates the key findings of our study, highlighting the implications of mobility shifts in London during the COVID-19 pandemic. We reflect on spatial and temporal patterns, assess the impact on various POIs, and address methodological limitations. Additionally, we explore possible extensions that could enhance future research on urban mobility in crisis scenarios.
The analysis confirmed that mobility patterns in London underwent significant changes during the COVID-19 pandemic. The first lockdown saw the most drastic decline in movement, particularly in central London and at major POIs such as subway stations and event venues. While some redistribution of movement occurred towards residential areas and parks, the overall decline was substantial.
The Moran’s I analysis revealed a clear shift in spatial clustering. Pre-pandemic movement was highly concentrated in central districts, whereas lockdown periods led to a more dispersed movement pattern. The reduction in movement at transport hubs and commercial areas coincided with increased activity in residential zones and parks, reflecting behavioral adaptations to restrictions.
During relaxation phases, the movement gradually returned to pre-pandemic clustering patterns, but full recovery was not achieved by the end of 2020. The persistence of lower activity levels in business districts and tourist locations suggests long-term behavioral changes, potentially influenced by shifts in remote work policies and continued travel restrictions.
The analysis also highlighted variations between different boroughs. While central London saw the sharpest declines and slowest recovery, outer boroughs exhibited more stable mobility patterns. This difference underscores the varying levels of dependence on commuting and tourism across different parts of the city.
Different POI categories experienced varying levels of impact. Public transport hubs, such as subway and bus stations, saw sharp declines in movement, with subway stations experiencing the most significant drops due to remote work policies and public transport avoidance. While bus stops also saw reductions, they recovered more quickly, likely due to their role in essential daily commuting.
Football stadiums and event venues were among the hardest hit, suffering from prolonged inactivity due to restrictions on large gatherings. Movement data shows near-total inactivity during strict lockdowns, with only minor recovery towards the year’s end when limited-capacity events were permitted. The absence of major sporting and entertainment events significantly altered mobility patterns in their surrounding areas.
Sightseeing locations, including museums, landmarks, and tourist hotspots, followed a similar trend. Visitor numbers plummeted during lockdowns, and while some recovery was observed during relaxation phases, overall foot traffic remained below pre-pandemic levels. The gradual return of tourism-dependent movement was hindered by continued travel restrictions and reduced international visitors.
Parks demonstrated the highest resilience among POI categories. While movement slightly decreased during the initial lockdown due to stay-at-home orders, it recovered swiftly and even increased in some cases as residents sought outdoor recreational spaces. This trend highlights the importance of accessible green spaces in maintaining public activity levels during crisis periods.
While this study provides valuable insights into mobility trends during the COVID-19 pandemic, several limitations should be acknowledged. One major limitation is the temporal scope of the dataset, which covers only the year 2020. This restricts the ability to analyze long-term behavioral changes and the full extent of post-pandemic recovery. Future studies incorporating data from subsequent years could provide a more comprehensive understanding of how mobility patterns evolved beyond the initial crisis.
Another limitation is the spatial resolution of the movement dataset, which is aggregated into 100-meter grid cells. While this level of granularity is useful for general trends, it does not capture finer-scale variations in movement, particularly within densely populated areas or specific points of interest. A dataset with higher spatial resolution could offer more precise insights into localized mobility shifts.
Additionally, the study does not differentiate between different user groups, such as commuters, tourists, and residents. Since movement patterns vary significantly between these groups, distinguishing them could enhance the interpretation of findings. Incorporating demographic or socioeconomic data, such as income levels or employment types, could further refine the analysis and provide deeper insights into movement motivations.
The ST-DBSCAN clustering method used in this study also has inherent limitations. The choice of parameters, such as the search radius and minimum cluster size, significantly influences the results. Due to computational constraints, extensive parameter tuning was not feasible. Future studies could employ automated optimization techniques to refine clustering accuracy and better identify mobility patterns.
Lastly, the analysis does not account for external factors such as weather conditions, policy changes, or economic fluctuations, all of which can influence mobility. Integrating additional contextual data sources could enhance the robustness of future research by controlling for these variables and providing a more comprehensive picture of mobility dynamics.
The findings of this study highlight the significant impact of COVID-19 on urban mobility patterns in London. Especially the first lockdown led to drastic reductions in movement, particularly in commercial and transport hubs, while parks and residential areas exhibited more resilience. The recovery of movement was uneven across different POI categories, with public transport hubs and event venues experiencing prolonged declines, whereas parks quickly rebounded.
The analysis also revealed that while movement trends partially recovered during relaxation phases, they did not return to pre-pandemic levels by the end of 2020. This suggests long-term behavioral shifts, potentially driven by remote work policies, changes in public transport usage, and evolving urban dynamics.
While this study provides a fitting analysis for the research question, there are several areas for future research remain. Expanding the dataset to include post-2020 mobility data would allow for an analysis of long-term recovery patterns and permanent behavioral changes resulting from the pandemic.
Another key extension would be the integration of additional data sources, such as economic indicators, weather patterns, and demographic data, to provide a more comprehensive understanding of movement drivers. Machine learning techniques could also be applied to enhance clustering and predictive analysis.
While the study successfully analyzed mobility changes in London during the COVID-19 pandemic, several extensions could improve the depth and explanation of the findings. A key improvement would be the refinement of POI categorization. Instead of broad categories such as sightseeing locations or transport hubs, a more granular classification could differentiate between commuter-heavy subway stations and tourist-oriented stations, museums and historical landmarks, as well as shopping centers and business districts. This would allow for a more precise understanding of how different types of locations were affected.
Another important extension involves integrating additional external data sources to better explain mobility shifts. For example, incorporating weather data could reveal how conditions such as rain, temperature, and seasonal changes influenced outdoor activity. Socioeconomic indicators like income levels, employment rates, and work-from-home adoption could provide deeper insights into movement variations across different demographic groups. Additionally, real-time transport network data from Transport for London (TfL) would allow for a distinction between pedestrian movement and public transit usage, enhancing the interpretation of mobility dynamics. Overall, these extensions would improve the study’s accuracy and usability.
Finally, comparative studies across multiple cities could offer valuable insights into how different urban environments responded to the pandemic. Such an approach could help urban planners develop more resilient mobility strategies for future crises.
