Analysis of Active Travel Counts in London

Spatial Analysis
Series of analysis on London’s cycling behaviour based on data from Transport for London.
Published

January 17, 2024

Overview

The first semester at UCL involved quite a bit of data analysis using TfL’s cycling database (quite coincidential). The projects on this page summarises work from my course (and some additional work) involving cycling data analysis.

E-scooter Accident Analysis

I have explored the correlation between the number of accidents and the amount of traffic passing through, showing a weak linear correlation. I have also explored the spatial autocorrelation through local Moran’s I analysis, indicating areas that have a relatively high accident rate compared to the surrounding areas.

Accident density involving e-scooters, aggregated by borough. Central London has more accidents per square km compared to outer London - but given the high density of traffic we cannot immediately conclude these are the unsafe areas.

Local Moran’s I for the frequency of accidents in relation to the traffic density. High value indicates accidents are more frequent than the surrounding areas considering the traffic, and low values indicate fewer accidents.

Some interesting trends were observed, indicating areas that may be potentially safer / more dangerous compared to their surrounding boroughs.

The full report can be found here.

This was done as a final assessment for the Geographic Information Systems and Science module conducted by Dr. Andy MacLachlan. Details of the prompt can be found at the GitHub repository.

Active Travel Count Analysis

We have explored the relationship between the active travel counts and various factors of the built environment.

Active travel count points

We have explored the following characteristics using multiple regression, returning the results shown.

Characteristics explored and their correlation with active travel counts.
Variable Correlation
Zone (Central / Inner / Outer) Negative correlation (more counts in the centre)
Weather (Wet / Dry) No significant correlation
Peak times Positive correlation
Bus lane No significant correlation
Cycle lane Positive correlation
Shared path No significant correlation

One interesting point to note is the immunity of Londoners to bad weather when cycling - needs further investigation to confirm.

This was done as a group project as part of the Quantitative Methods module run by Dr. Huanfa Chen.

Group Members

  • Qiyue Chen
  • Hongkai Ren
  • Yilin Yang

Santander Cycles Usage Data Analysis

The Santander Cycles is the cycle hiring scheme in London, which provides usage data for all journeys taken.

The elevation of Santander Cycles docking stations. The hypothesis was that stations located in higher elevations have more outbound (downhill) flows.

I have explored the relationship of elevation of docking stations and the number of journeys taken, taking away two key findings:

  • When focusing on individual docking stations, the higher the elevation the ratio of departures over arrivals was higher, indicating users taking advantage of downhill journeys.
  • On the other hand, there was no obvious relationship between the height difference from origin to destination when given focus on each journey

Graph showing relationship between departure / arrival ratio and elevation. Weak correlation can be observed.

This seemingly contradictory results provide room for further research, making clear the differences within Santander Cycle usages.

The report can be found at the GitHub repository.

Other Visualisations

A few of my maps from the 30-Day Map Challenge 2023 have been based on cycle usage data.

Map of Santander Cycles Docking Stations in London, colored according to the area that they are claimed to be in.

A map of 10,000 recent trips made by the Santander Cycles, colored by the speed of travel.

The flow of cycles grouped according to the origin and destination zones. A thicker line indicates more flow between these areas, while the circles indicate intra-zone usage.