An England Walkability Index at LSOA level

Introduction

The primary need to create an England Walkability Index (WI) is to support the subsequent analysis which is related to the mismatch between the walkability and heat exposure in Birmingham. Walkability measures how friendly or supportive an area is for walking and is commonly influenced by the access to transport, land-use mix and the street-network connectivity. The construction of WI provides insights to local authorities, planners, public health teams and communities to understand what makes their community walkable or not.

The availability of a dedicated England-wide WI is severely limited. The Access to Healthy Assets & Hazards (AHAH) is an England-wide, publicly available data set maintained by the Geographic Data Service for Great Britain [1]. It measures the accessibility of health-promoting and health-damaging features of the built environment. Their index combines retail environment, health services, physical environment and air quality. However, AHAH is not designed specifically to measure walkability and does not directly capture some of the core features of a walk-supportive environment as mentioned earlier.

The London walkability index (Whitehall II) measures street connectivity, land-use mix, dwelling density and destination, which is the closest to what is needed here but it is unusable here because it covers only London and the data are outdated [2]. Alternatively, Space Syntax has developed the WI database that covers every building in Great Britain [3]. However, no publicly downloadable and reproducible national data set was identified from the available documentation and access seems to be provided for commercial purposes.

Given these data availability limitations, an England WI was developed for this project using publicly accessible data sets with a reproducible methodology. The England WI adopts the method used in the US Environmental Protection Agency (EPA) National Walkability Index, which provides an established and detailed methodology to follow [4].

Method

The US EPA National Walkability Index

The US EPA National Walkability Index (US-NWI) is built from four variables. The EPA index is calculated for US census block groups from the Smart Location Database. The four variables are intersection density (D3b), proximity to transit (D4a), employment mix (D2b_E8MixA) and employment/household mix (D2a_EpHHm). Each variable is converted into 20 quantiles, where each rank holds 5% of block groups. Ranks from 1 to 20 are assigned where 20 denotes the most walk-supportive and 1 is the least walk supportive. The final US-NWI is a weighted function of the four variables' rank scores:

and the final NWI is classified into four categories regarding to the level of walk supportiveness:

Data and components

Although there are no exact English equivalents to the original US datasets, the same underlying building blocks of walkability can be constructed using data available for England. All four components are calculated for England at lower super output area level (LSOA) using 2021 boundaries. The openly accessible sources are listed as follows:

D3b — street network intersection density

The measure was constructed from the 2026 England OSM road network. Firstly, obvious pedestrian ways (footway, pedestrian, steps) or multimodal highway types (residential, living street, tertiary, secondary, primary, living_street) were initially retained. Ambiguous OSM types (path, cycleway, bridleway, track, service) were included with the inspections of the additional tags such as the foot, access and service so that private and unsuitable driveways and parking aisles were excluded. Retained network segments were classified as po and mm, implying pedestrian-oriented and multimodal roads respectively. This allowed different types of intersections to contribute differently to the D3b score.

To prevent creating false intersections never existed due to different vertical levels of road networks, OSM tags like bridge, tunnel and layer were used to assign each network segment to a vertical level. This ensured a road crossing over a bridge was not, in plan view, incorrectly identified as an intersection.

Since running England road network as a whole would be computationally intensive. Instead, England was divided into 20 km processing tiles, with a 500m buffer added around each tile to avoid losing intersections close to tile boundaries. The network subdivision was performed using the sfnetworks R package and nodes connected to at least three or more network edges were retained as intersections. Intersections were given additional weights according to the presence of one-way and two-way road segments as follows:

Each intersection was then spatially joined to the England LSOA 2021 boundaries. Within each LSOA, the D3b score was calculated as a weighted sum, according to their classification (mm or po) and number of intersections. The formula is:

where mm3 and po3 refer to three-link intersections, while mm4 and po4 refer to intersections with four or more connected links. The final D3b score for each LSOA was then divided by its area in square kilometres to derive the final D3b per km².

D4a — public transport accessibility

D4a measures how easily residents within their LSOA can reach a transit stop on foot. It is calculated as the walking distance to the nearest transit stop from the population-weighted centroid of an area. Public transport stop data were taken from NaPTAN and filtered to stop types that were relevant:

The stop coordinates were then converted to an sf object and transformed to British National Grid so that distance could be measured in metres. Instead of identifying all transit stops within three-quarter mile straight-line radius (approximately a 15-minute walk), the five nearest transit stops to each LSOA population weighted centroid were identified using k-nearest neighbour search on Euclidean distance. This methodological departure was motivated by computational cost of the EPA search method. Once the five candidates had been identified, the walking-network distance was calculated with OSRM. A locally hosted OSRM engine with a pedestrian routing profile within Docker app was used for faster calculation. Routing this many origin-destination routing with public API services would have been considerably slower and subject to user limits. Finally, the candidate with the shortest walking distance for each LSOA was selected.

D2b_E8MixA — employment mix

BRES 2021 employment counts are mapped from 2-digit SIC divisions to eight EPA-style groups as closely as possible: retail, office, industrial, services, entertainment, education, health and public administration. The mix is the normalised Shannon entropy of the employment shares:

where p_k is the share of the LSOA's employment in group k. The full broken down formula can be found in the technical documentation and user guide of Smart Location Database [5].

D2a_EpHHm — employment and household mix

The same BRES data are mapped to five broader employment groups (retail, office, industrial, services, entertainment) and households from Census 2021 are added as a sixth category. Entropy is calculated the same way, normalised by the log of the number of categories actually present in the LSOA:

The full broken down formula can be found in the technical documentation and user guide of Smart Location Database [5].

Final index calculation

As mentioned earlier in this method section, each component is converted to a 20-quantile rank across all English LSOAs and the index is calculated as a weighted sum of each component. A higher score indicates a more walk-supportive environment, vice versa. For easier interpretation, the index was then grouped into four categories representing different levels of walk-supportiveness.

The code to derive each component and the final index can be accessed on the author's GitHub. For further details, please read the EPA NWI methodology and user guide and the technical documentation and user guide of Smart Location Database [4], [5].

Results

The result of the England WI is presented in the following map:

Conclusion

This project developed a reproducible England Walkability index with publicly available data at LSOA level by adapting the rigorous and transparent EPA US-NWI method. Although it closely follows the conceptual structure of the EPA method, the English adaptation involves departures in some calculations and data substitutions. The results have not been peer-reviewed or independently validated and should be interpreted with caution and should not be treated as an officially validated measure.

References

1. Geographic Data Service. Access to Healthy Assets and Hazards (AHAH), version 5.1. 2026.
2. Stockton JC, Duke-Williams O, Stamatakis E, Mindell JS, Brunner EJ, Shelton NJ. Development of a novel walkability index for London, United Kingdom. BMC Public Health. 2016;16:416.
3. Space Syntax. Walkability Index.
4. Environmental Protection Agency. National Walkability Index Methodology and User Guide. 2021.
5. Chapman J, Fox EH, Bachman W, Frank LD, Thomas J, Reyes AR. Smart Location Database: Technical Documentation and User Guide. U.S. Environmental Protection Agency. 3rd ed. 2021.