The Walk Score is an urban metric that gives a comprehensive overview of a city’s walkability and is widely used to assess the accessibility of cities, neighborhoods and specific places based on the essential services availability. Nevertheless, the estimation of the Walk Score metric for a city can be an intense and complex analytical process. In this article, we present the design and development architecture of the Walk Score Toolkit, a Python-based tool for automating the Walk Score estimation. The tool has been tested on over 112 cities to show its effectiveness and computational efficacy, as well as its potential in supporting and analyzing the 15-minute city concept.

Introduction

Walkability in terms of neighborhood services proximity and Walk Score had been assessed in many cities by examining the distribution of the accessibility to essential services (Hall & Ram, 2018). One of the most known applications was conducted in New York City, which was later generalized to around 130 North America cities and it is now available on the website walkscore.com. In this study, the Walk Score is evaluated for each address using all the available route choices to access the proximity services and the score is computed based on the accessibility distances. In this case, the Walk Score is estimated using nine categories, namely groceries, restaurants, shopping, coffee, banks, parks, schools, and books. Some of these categories include multiple amenities, each given different weights according to their importance (Walk Score n.d.).

Similarly, in 2010, a 20-minute neighborhood strategy was implemented in Portland (City of Portland, 2012), considering, in addition to the Walk Score, additional factors related to access and pedestrian conditions. The categories used and additional information in Portland were: grocery stores, commercial stores divided into 2 types, parks, public elementary schools, sidewalks, and frequent service transit stops.

Post-COVID-19 pandemic, the concept of the 15-minute city (Moreno et al., 2021) is becoming popular and getting attention from the international urban planning community. This concept defines a timeframe of 15 minutes of walking or cycling to access the essential services (Abdelfattah et al., 2022). Consequently, part of the research efforts focused on assessing the compatibility of existing cities and urban areas with this concept by evaluating analytically how service accessibility is distributed. This tool aims to join the efforts of using quantitative evidence to support livable cities planning, by recognizing that the overarching concept behind the 15-minute cities is the sustainable and equitable opportunity to access essential facilities.

The process of estimating the Walk Score includes various steps, including i) the definition of the proximity service categories, ii) the collection and mapping of amenities, iii) the walking catchment analysis, iv) the proximity accessibility estimation, and v) the final computation of the Walk Score. Given the need to standardize some of the assumptions required to perform this analysis and the opportunity to streamline and automate an otherwise time-consuming process, the scope of this article is to present the Walk Score Toolkit, a Python tool integrated with QGIS to automatically and quickly assess the Walk Score. The following section discusses the design parameters of the tool.

Next, the Functionalities and Process section describe the detailed structure and role of each of the main functions in the implementation of the tool. Finally, an interactive map showing the results of the Walk Score Toolkit in 112 cities is proposed, to test the scalability and flexibility of the tool in assessing case studies of different size, geographical location and data availability.

Tool Design

Walk Score Toolkit is an end-to-end tool that enables the user to automatically estimate the Walk Score metric over a custom urban region. The design of the tool was meant to be straightforward and efficient for the extensive spatial computation required in the intermediate steps of evaluating the Walk Score. The tool is developed using Python due to its fast and powerful capabilities and ease of use, and for its integration with the QGIS environment. Moreover, many network and geospatial analysis libraries, such as GeoPandas, PySal, OSMnx and networkx available for Python were crucial to develop the Walk Score toolkit.

The design architecture is illustrated in Figure 1 below. Briefly, the tool was built on top of three main functions, addressed in detail in the next section. The design choices for proximity services and the gridding scheme are described in the next paragraphs.

The conceptual framework for designing and implementing the Walk Score Toolkit started with the definition of proximity services; thus, the first step of the project was to identify the essential services and categorize them into eight main categories. Hereby, the following classification is proposed and used in the tool: neighborhood services, healthcare, neighborhood shops, education, sports, cultural entertainment, open leisure and mobility. The corresponding amenities belonging to each of the main categories were defined taking into account the commonly used amenities in literature and practice, as the tool is designed to be suited to different regions with different proximity services and community needs around the world. Below, we are listing the detailed amenities within each of the main categories:

1. Neighborhood services include personal and community services (i.e., post offices, banks, civic and community centers, police services, filling water points, local administration offices, bars and cafes, tobacconists and newsagents);
2. Healthcare: (hospitals, pharmacies, first aid, clinics and local health services);
3. Neighborhood shops: (grocery, retail, bookshops, local markets, restaurants, supermarkets and the local non-food shops);
4. Education: (kindergartens, schools, colleges and universities, professional schools and libraries);
5. Sports: (gym and sports facilities);
6. Cultural entertainment: (cinemas, theaters, art centers and museums);
7. Open leisure: (parks, kid’s playgrounds and gardens);
8. Mobility: (public transportation stops on surfaces, bike parking facilities, bike sharing and car sharing facilities, underground metro stations and railway stations);

In order to fulfill this design goal, OpenStreetMap (OSM) was defined as the primary source of proximity services. We decided to use data from OSM because it is considered the biggest openly licensed collection of geospatial data and the largest Volunteered Geographic Information (VGI) source (Brovelli et al., 2016), moreover OSM constitutes a worldwide available open data source, which is an essential requirement to ensure the scalability of the tool.

Once the main categories were defined, it was needed to identify the correct space discretization to implement. In this case, it was deemed coherent to adhere to another industry standard, which is the H3 hexagon grid; a hierarchical geospatial index based on hexagons developed by Uber. The selection of this gridding schema was driven by two main reasons: firstly, the worldwide consistency and availability, and secondly, the hierarchical structure, which allows the users to select the appropriate detail level according to their analysis needs. Moreover, the H3 hexagon scheme enables the integration of results with different spatial data. The design of the tool, however, allows the user to pick the zoom level of Walk Score estimation by choosing the H3 hexagon level (see Figure 2).

Functionalities and Process

As mentioned earlier, the Walk Score toolkit has three main functions, in this section the specifics of each of the three functions is presented.

1. Isochrone generation

As the Walk Score is based on the accessibility to the essential services through the pedestrian network, an extraction is performed from OpenStreetMap Highway data, which includes exclusively segments suitable for walking. Once the walkable network is extracted, the H3 hexagons’ centroids are assigned to the nearest nodes in the network; this step is aimed at linking each centroid of the cells to the network, enabling the calculation of the walking accessibility within the network. Furthermore, accessibility to services is calculated in terms of walking minutes to reach the service, starting from the centroid of each hexagon. In order to correctly estimate the walking time, 15 isochrones, one for each minute interval, are generated over the network. This step is critical as it needs to be consistent among networks of different size, granularity and structure, moreover it can become resource intensive in case of large analysis extents. This process was optimized through OSMnx (Boeing, 2017) and networkx Python libraries and the generation of convex-hull vector-based polygons.

2. Services data extraction

After the travel time analysis, the tool filters and downloads the amenities data for each of the predefined categories within the study area, by querying the OpenStreetMap database. In addition to this, in this version of the tool, an alternative function is implemented, enabling the user to either add custom POI data to the original OSM data, or to input a complete set of custom POIs (see Figure 2).

3. Walk Score computing

This function performs an intersection between the proximity services data with the generated isochrones. In this tool, each amenity is weighted based on its proximity to the cell centroid (i.e., the needed time to reach the amenity from the starting point). This is done by implementing an exponential distance decay function due to its ability to reflect actual pedestrian walking behavior. This approach is different from the traditional Walk Score (Walk Score, n.d.), which by design assigns different weights for different categories. For the scope of this tool, the Walk Score from each grid cell considers the nearest amenity from each category without implementing additional weighting schemes keeping a consistent importance among all the amenity types.

The process makes the weighting values of each category range from 0 to 1 based on its closeness to the centroid and consequently, the Walk Score is initially calculated as a sum of the weighted categories. This step keeps the initial values of the Walk score in a range between 0 (reflects no essential service within 15-minutes) and 8 (fully served zone). An additional process is performed to scale the values of the Walk score to be between 0 and 100, which is the preferred visualization scale.

Along with the Walk Score layer (see Figure 3), additional diagnostic features were implemented in the tool to be able to dissect the results and identify specific service deficiencies. The diagnostic report (i.e., a layer that provides information about the density of each of the category services throughout the city) enables a quick overview of the density of the mapped services within the city.

Conclusions and future works

In this article, it is presented the Walk Score Toolkit, a Python tool integrated with QGIS to assess the urban metric of Walk Score. Moreover, the focus is given to the design choices, the technologies and the libraries used, and the design architecture to develop the tool, as well as a detailed description of the main functionalities of the tool. As an evaluation of the tool, the Walk Score of 112 cities selected to join the EU Mission: 100 Climate-neutral and smart cities by 2030 around the world, were estimated. The results can be accessed through this interactive map, and show the advantages of the tool in the optimization of proximity services analysis for walkability planning and in support of the 15-minute city concept.

The future versions of the tool are intended to extend the functionalities with the idea of building a more scalable and versatile toolkit. In particular, the development will focus on the utilization of custom pedestrian networks as alternative data input and add functionalities to highlight zones covered by the different proximity services in the model of the 15-minute city.

---

Acknowledgments

The analyzed data were treated according to the GDPR-General Data Protection Regulation (EU, 2016/679). This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

References

Abdelfattah L., Deponte D., & Fossa G. (2022). The 15-minute city as a hybrid model for Milan. TeMA – Journal of Land Use, Mobility and Environment, 71-86. https://doi.org/10.6093/1970-9870/8653

Boeing, G. (2017). OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks. Computers, Environment and Urban Systems, Volume 65, 2017, Pages 126-139, ISSN 0198-9715. https://doi.org/10.1016/j.compenvurbsys.2017.05.004 

Brovelli, M. A., Minghini, M., Molinari, M. E., & Zamboni, G. (2016). Positional accuracy assessment of the OpenStreetMap buildings layer through automatic homologous pairs detection: the method and a case study. In: The International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Prague, Czech Republic, Vol. XLI-B2, pp. 615-620. https://doi.org/10.5194/isprs-archives-XLI-B2-615-2016

City of Portland (2012). The Portland Plan. Retrieved February 26, 2024, from https://www.portland.gov/bps/planning/documents/portland-plan/download

Moreno, C., Allam, Z., Chabaud D., Gall C., & Pratlong, F. (2021). Introducing the “15-Minute City”: Sustainability, Resilience and Place Identity in Future Post-Pandemic Cities. Smart Cities. 2021; 4(1):93-111. https://doi.org/10.3390/smartcities4010006

Hall, C. M., & Ram, Y. (2018). Walk score® and its potential contribution to the study of active transport and walkability: A critical and systematic review. Transportation Research Part D: Transport and Environment, 61, 310-324. https://doi.org/10.1016/j.trd.2017.12.018

Walk Score (n.d.). Walk Score Methodology. Retrieved February 16, 2024, from https://www.walkscore.com/methodology.shtml