As part of the Tombolo project, we’re helping cities to make better, information-driven decisions. To ensure that we’re helping them to make the decisions that they care about, we’re actively consulting and collaborating with Greenwich, Leeds & Milton Keynes, and have been for the past 18 months.
In consultation with our partner cities it became clear that understanding the access that residents have to services is of real interest. In discussions with the Royal Borough of Greenwich in particular, we found that understanding access to services such as town centres, GP’s, hospitals and education centres was a priority. This prioritising of town centres and healthcare services also happens to align closely with the goals of the NHS executive who have expressed the wish that in the UK ‘patients will get fair access to consistently high quality, prompt and accessible services right across the country’.
Knowing that cities and the NHS want to better understand the levels of access that UK residents have to services, we decided to investigate access a little further.
On first inspection, the notion of access to a service might sound obvious, but when trying to constrain a definition it becomes clear that it can be hard to be specific. If someone asked you to define access what would you say?
As it turns out, we’re not the first people to consider what access is and we can look to academia for some suggestions. Perhaps unsurprisingly, access to town centres and particularly health services has been a priority in academic research and the Tombolo team have considered this to guide our thoughts around access.
It became clear in the 1980’s that not having a definition for describing access makes it hard to measure and to use as a metric in decision-making (Penchansky and Thomas 1981). As a result, Penchansky and Thomas decided to define access as the ‘degree of fit between the individual user and the service’. More specifically, they divided the idea of access into what I like to call the Five A’s:
|Dimension of Access||What is it?|
|Availability||Does the service exist and does it have the capacity for the number of people wanting to use it?|
|Accessibility||Can people get to the service given their transport resources or the time and cost of transport?|
|Accommodation||Can people use the services even if they can get to them / do they understand how to?|
|Affordability||Can people afford to use the service when they get to it?|
|Acceptability||Will people want to use the service given the characteristics of the service provider?|
Even with these divisions, it is clear that there are many features of access and that access can be perceived from a range of perspectives. As a result, the nature of accessibility is still being examined over 20 years later (for an example see Guilford et al., (2001)).
To constrain our research and make it as useful to the cities as possible, we focused on the accessibility feature of access. Accessibility is perhaps the one of the ‘Five A’s’ that is most synonymous with access and it is an area where we have the skills and open data to provide useful insights as part of the Tombolo project.
However, even the dimension of accessibility has multiple components depending on what perspective you take. For example, physically getting to a service could include walking, cycling, any type of public or private transport or even any number of combinations of these. Incorporating accessibility options for mobility related impairments further increases the dimensions that must be considered. As a result, there are a huge number of travel times and cost permutations available to citizens. The process of experimentation that we discuss below and the permutations that we investigate are a consequence of our curiosities and are primarily driven by the data that we have available.
General Connectivity in Greenwich – our first step
The Tombolo team started investigating accessibility within Greenwich on a broader scale (ignoring services) using what is known as the Space Syntax method. The Space Syntax method examines the properties of a street network to model the patterns of movement potential in a region (such as the borough of Greenwich (Figure 1)). For example, some streets in the borough may get more use than others because they are better connected and provide more direct routes to other areas.
When investigating patterns of movement potential in Greenwich we found streets with varying levels of connectivity. Though connectivity is not strictly accessibility, using the space syntax method enables us to identify which streets in Greenwich are most spatially accessible when getting from A to B and this serves as an interesting starting point for investigating accessibility further.
Figure 1: Movement potential in Greenwich. Dark blue streets have high levels of movement potential on both regional and local scales. Pale blue streets have high levels of movement potential on regional scales only and pink streets have high levels of movement potential on local scales only. Other large-scale streets outside the borough of Greenwich used in the calculation are also shown.
Accessibility to town centres at different granularity – our second step.
Once we’d understood a little about the connectivity of Greenwich, our next step was to look at potential travel times to town centres. To keep things simple, we decided to investigate the estimated driving time from every building in Greenwich to the nearest town centre (Figure 2). The notion of a ‘town centre’ is open to debate so the Tombolo team used town centre boundaries provided in documentation from the borough of Greenwich (Royal Borough of Greenwich, 2014).
To estimate the driving time from each building we:
- Identified a building
- Identified the nearest street to that building
- Assigned driving speeds to each street in the network (GOV.UK speed Limits, 2016)
- Minor roads = 429.1 meters/min
- A-roads & B-roads = 643.7 meters/min
- Motorways = 1287.4 meters/min
- Calculated the quickest route (time) from that building’s point on its street to the nearest town centre boundary.
In this analysis we did not consider disabled access or account for things like getting into the car or finding parking.
In addition to the caveats discussed above, we realised that the granularity of our results could dramatically impact their interpretation. For example, when making decisions relating to city or borough resources, information is often provided on a Lower Super Output Area (LSOA) level where a single LSOA contains around 1500 persons. When we compare information at LSOA level (by averaging the travel times for all the buildings with a centroid in a given LSOA) to the initial analysis at building level, we see how small changes in how the data are represented can make large differences to the potential interpretations. This raises interesting questions around the value of data at different scales and brings into question the actionable value of different types of analyses.
Figure 2: Travel time by car to the nearest town centre. Pink lines demonstrate the town centre boundaries used in the method described above. We show two layers, travel time at a building scale and travel time at an LSOA scale. Click on the different travel times to see how long it takes to get to the town centres from different locations.
Accessibility to town centres looking at different methods – our third step.
Once we’d aggregated travel times to LSOA level we were also able to compare the initial accessibility methodology to methodologies used elsewhere. For example, the Department of Transport (DfT) also provides estimates of accessibility to services which includes travel time by car (Department for Transport, 2015). Understanding how differences in methodology can lead to small differences in inference can also provide useful context when making data driven decisions (Figure 3). When we compared our investigations with those from DfT we found some interesting differences in estimated driving time from an LSOA to the closest town centre:
To generate their travel time estimates the DfT:
- Identified the population weighted centroid of each output area (OA) (an output area is a region containing ~300 people).
- Estimated the walking time from each centroid to the nearest road
- Set a minimum driving time of 5 minutes (to account for getting in / out of a vehicle)
- Assigned driving speeds to the network:
- Motorway = 1325 metres/min
- Urban Motorway = 1325 metres/min
- A road 711.67 metres/min
- B road = 693.33 metres/min
- Minor road = 613.33 metres/min
- Local Street = 320 metres/min
- Private road – restricted access = 283.33 metres/min
- Private road – public access = 246.67 metres/min
- Estimated the driving time from that point on the road to the population weighted centroid of the nearest town centre.
- Calculated the population weighted average of the travel time for all Output Areas to generate data at an LSOA level.
Figure 3: Comparing our methodology (left) with that of the Department for Transport (right). The methodology on the left used the town centre boundaries (black lines) and on the right, the town centres (black points). Missing LSOA’s on the right are due to the differences in LSOAs available from the DfT analysis. This is intentionally highlighted to demonstrate how small changes in open geospatial data can influence analyses. We note how the five-minute minimum journey time set by DfT dominates the results.
As we can see, the general idea between the two methodologies is the same but small changes in assumptions lead to notable changes in the results. Such small changes include:
- The selection of what constitutes a town centre. For example, we used the town centre boundaries supplied by a Greenwich planning document, but DfT used population weighted centroids of the town centres. There were also differences in what we considered as town centres (see table A1).
- The assumptions surrounding vehicle speeds on each road.
- The use of OA centroids as an origin instead of the buildings in the LSOA.
- Differences in the transport network.
- Differences in the methodology relating to minimum journey times and walking to and or from a vehicle.
- Changes in the spatial layout of LSOAs.
In our opinion, the assumption of a minimum journey time of 5 minutes is probably the most notable difference that changes the inferences that can be made using the results. Both approaches are valid and each uses reasonable assumptions. Which data would you use?
Future analyses and lessons learned
This process has made it clear that asking a reasonably well constrained question such as “how long will it take Suzy to get to her nearest town centre” is not as easy as it may sound. We are fortunate in this situation that the geospatial and infrastructure data exist for us to address this question numerically, however, even with reasonable assumptions, two similar methodologies yielded very different insights.
We have found this process valuable and when seeing our results, Greenwich expressed interest in understanding access to other services which will likely be one of our next steps in development. As part of Tombolo project, we are developing software to make these types of analyses easier to complete. The goal being that cities and boroughs can ask more questions and make better information driven decisions to make their spaces better places to live. You’re already on the Tombolo website so to learn more about the Tombolo project and the other ways we’ve been integrating with cities take a look around!
|Town Centre||Tombolo – town centre boundaries (paths)||Dft – population weighted centroids|
Table A1: Town centres used in the different methodologies
Gulliford, M., Figueroa-Munoz, J., Morgan, M., Hughes, D., Gibson, B., Beech, R., & Hudson, M. (2002). What does’ access to health care’ mean?.Journal of health services research & policy, 7(3), 186-188.
Penchansky, R., & Thomas, J. W. (1981). The concept of access: definition and relationship to consumer satisfaction. Medical care, 19(2), 127-140.
GOV.UK speed limits. (2016). Retrieved September 31, 2016, from https://www.gov.uk/speed-limits
Department for Transport (2015). Journey time statistics guidance. Retrieved October 31, 2016, from https://www.gov.uk/government/publications/journey-time-statistics-guidance
Royal Borough of Greenwich. (2014). Royal Greenwich local plan: Core strategy with detailed policies. Retrieved September 28, 2016, from http://www.royalgreenwich.gov.uk/info/1004/planning_policy/869/local_development_framework/2