Great Britain at a glance

Over two years ago, I decided to design a tool that would make it easier to see at a glance how a site or settlement sits within environmental, social and economic geography at the local, subregional, regional and national levels. It wasn’t intended to be the only tool, of course — but the first tool to turn to.

This article shares the final draft of the map that I created (see below), and some detail on the methods that I applied to get there. The most significant areas of discussion relate to Built-up Areas and Travel to Work Regions, which I will probably explore further in a separate article in terms of population density and productivity (and maybe rents versus house prices). I have also indicated throughout the article a few areas that I would like to explore further in future.

Environmental

High ground

Intention and results: I knew that I wanted to show Travel to Work Regions on the map, and that topography has an impact on this, but had not done any research on this before.

In the final draft of the map, the correlation is clearer than I would have expected, so the effort of including this was worthwhile.

Data sources and methodology: I used the ETOPO1 Bedrock GIS data published by the National Oceanic and Atmospheric Administration in the US, which provided a gradient for elevation above sea level. I chose to fade in the data between 275 and 325 m to avoid cluttering the map, focusing on the most significant contiguous areas of high ground.

Key protected areas

Intention and results: Similarly, I knew that I wanted to show population density on the map, and that there is an interrelationship between urban development and environmental and landscape designations — but again, I wanted to avoid cluttering the map with hundreds of smaller non-contiguous areas. I also decided not to display Green Belt designations which sometimes have a stronger social and economic justification than environmental.

Whilst the final draft doesn’t include large areas of high environmental value such as The Wash and along the Humber Estuary, I feel that the National Parks (England Wales) and National Scenic Area (Scotland) designations served the intended purpose.

Data sources and methodology: I reviewed a number of landscape and environmental designations and weighed up the pros and cons of each.

• The Sites of Special Scientific Interest (SSSI) dataset included thousands of smaller and non-contiguous areas, and the extent of designations was limited even in larger areas where development was prohibited.
• There was already a relatively high population density and great extent of urban development in Areas of Outstanding Natural Beauty (AONB) in England and Wales, and I wanted to avoid this overlap on the map.
• For England and Wales, I felt that National Parks designations served the intended purpose most closely, being the least populated areas and protected from development. However, there are only two National Parks in Scotland, and so I reverted to using National Scenic Areas, the broad Scottish Equivalent of AONB’s, which cover a wider area.

The disparity between treatment of England and Wales and Scotland is uncomfortable, and I may produce a version of the map including AONB in future.

Social

Population density and Built-up Areas

Intention and results: As well as showing settlement boundaries, I wanted to be able to distinguish between rural areas where settlements are larger and/or closer together and more remote rural areas.

To aid interpretation:

• West Somerset and New Forest are almost entirely covered by landscape designations and have population densities of around 50 and 250 residents / sq km, respectively
• Chelmsford and York comprise around one-fifth to one-third Built-up Areas and have population densities of around 500 and 750 residents / sq km, respectively
• Basildon and Middlesborough comprise around one-half to three-quarters Built-up Areas and have population densities of around 1,650 to 2,600 residents / sq km, respectively
• Lincoln and Norwich are almost entirely covered by Built-up Areas and have population densities of 2,700 and 3,600 residents / sq km, respectively

The result highlights that the most populous part of Wales is its south coast, the rural nature of Anglia and the South West, as well as the swathe of rural shires from Gloucestershire to Lincolnshire that separate ‘The North’ and ‘The South’.

Data sources and methodology: I suspect that the most detailed and revealing way of achieving the intended purpose would have been to use a fixed geographic grid (e.g. one sq km tiles). However, I opted to use local authority boundaries for simplicity. This will probably also be more useful when drafting reports, as one can easily describe the character of the relevant local authority.

I used the mid-2016 population estimates produced by ONS to calculate a population density per sq km, ranging from 9 residents / sq km in Highland and Na h-Eileanan Siar to 15,524 residents / sq km in Islington, with an average of 279 residents / sq km across Great Britain.

I chose to limit the gradient to four categories, styled to illustrate the interrelationship with landscape designations and Built-up Areas. The thresholds are therefore set such that the top and bottom two categories comprise around two sextiles of local authorities each, and the middle two categories comprise around one sextile each.

The Built-up Area boundaries are created by ONS using an automated approach based on a 50m grid squares. As mentioned before, Lincoln and Norwich are almost entirely covered by Built-up Areas and have population densities of 2,700 and 3,600 residents / sq km, respectively.

Economic

Travel to Work Regions

Intention and results: I wanted the map to reflect the economy of the UK at a regional scale, with the following criteria:

• Derived from commuting patterns, but without requiring programming
• Contiguous areas
• Unconstrained by the old Government Office Regions or national boundaries
• Large enough to have national significance, but not relying on nationally significant cities i.e. may be a cluster of regionally significant cities

The result is 15 regions, each with at least 500,000 workers, quite different to any existing administrative or ceremonial boundaries. As suspected, the boundaries are greatly influenced by high ground such as the Pennines and Brecon Beacons.

Around one-third of the regions (Cambridge-Ipswich-Norwich, Edinburgh-Aberdeen, Plymouth, and Swindon) do not include a nationally significant Built-up Area.

The map below shows a high degree of self-containment within the regions; at least 80% of workers are also resident within their region (excluding workers resident outside Great Britain) in all but 2.6% of local authorities. The 10 exceptions are:

• Hambleton sits within the Tyneside region, although 19% of its workers are resident in the Leeds-Bradford region.
• Warrington sits within the Liverpool region, although 19% of its workers are resident in the Manchester region.
• Stafford and East Staffordshire have complex commuting patterns spanning the Birmingham, Manchester and Sheffield-Nottingham-Leicester regions.
• Daventry, Harborough, Rugby and Rutland have complex commuting patterns spanning the London, Birmingham and Leeds-Bradford regions.
• Crawley and Waverley have a high proportion of workers resident in both the London and South Hampshire-Brighton and Hove regions.

Data sources and methodology: I was originally inspired by ONS’s analysis of Travel to Work Areas by rail, managing to identify around 14 contiguous regions. However, having reflected upon how few people travel to work by rail, and how little commuting happens across regional boundaries, I decided to look at all methods of travel together.

I set up a model in Microsoft Excel using 2011 Census data that allowed me to initially assign each local authority area to a region based on the Travel to Work by rail areas, and then manually reallocate authorities to increase the levels of self-containment, then checking that they were still contiguous. The way that I set up the model happened to exclude workers who were resident outside Great Britain from the analysis, but this was not intentional and does make the methodology harder to explain.

As the process was so manual, I am sure that there will be alternative combinations that achieve acceptable – or even higher – levels of self-containment. I have an idea how I might optimise this in future with a few lines of code using random numbers at a local authority level. However, I suspect that this method would be too inefficient to look at lower geographies or find the definitive optimal solution and so require much more programming.

Whilst there were some self-contained economic areas around Carlisle, Bournemouth/Poole and Middlesborough (and probably others), these had fewer than 500,000 workers. Imposing this minimum size threshold results in 15 regions ranging from 521,000 to 7.10m workers (Swindon and London regions, respectively).

The regions are named according to their most significant Built-up Area(s). Where a continuous urban area was more commonly known as separate settlements, Built-up Area Sub Division name(s) were used instead (e.g. Birmingham rather than West Midlands and Leeds-Bradford rather than West Yorkshire). Tyneside and South Hampshire were used for brevity.

Key Built-up Areas

Intention and results: It was pretty obvious that size would be the key determinant for any classification of Built-up Areas, but I decided to look at workplace population and Gross Value Added alongside residential population factors to highlight BUA’s like Cambridge, Ipswich, Northampton, Oxford, Reading, and Swindon that ‘punch above their weight’ in terms of resident-worker ratio and productivity per resident.

I then added pass/fail location criteria to reduce the number of BUA’s that featured on the map. These criteria were arrived at after looking at the initial results —essentially, making subjective judgement calls about economic significance or how cluttered the map should be. I may do some class analysis using the Jenks method in future.

By way of comparison, whilst I was working on this article, a Parliament Houses of Commons Library Briefing Paper was published classifying: 12 “Core cities”; 24 settlements with resident population over 175,000 as “Other cities”; 119 with resident population 60,000–174,999 as “Large towns”, and; 270 with resident population 25,000–59,999 as “Medium towns”.

The 18 service centres that do feature on the map are the obvious ones, with Carlisle, Lincoln and York amongst the large, and; King’s Lynn, Perth and Truro amongst the small. I’d like to do a deep dive on these settlements at some point in the future, to understand what factors have constrained their growth and assess the strength of the economic case for infrastructure investment or other economic stimulus there.

Data sources and methodology: The methodology evolved over many months, and was severely limited by the lack of GVA data at lower than NUTS3 region scale. I’d like to explore methods for approximating these by BUA in future (probably taking the NUTS3 region average and then weighting based on household income and sectoral mix at lower geographies).

Birkenhead, Newport and Southend-on-Sea are the three BUA’s that are regionally significant in size, but do not have the minimum resident-worker ratio of 40% to be eligible for inclusion on the map. Despite this, their characteristics don’t feel significant enough to create a ‘Subregionally significant’ category.

The map below shows the location of the 183 service centres that are ineligible for inclusion on the final map:

• 42 fall below the resident-worker threshold (seven large and 35 small). Many of these are located in Kent and Essex — probably reflecting high levels of out-commuting to London. Remote BUA’s such as Barrow-in-Furness, Kendal and Workington are also in this category — and I’ve yet to determine why.
• 31 fall below the distance threshold of 40km from an upper tier BUA (17 large and 14 small). The four largest of these (Crawley, Doncaster, Slough and Warrington) have resident populations over 150,000 and workplace populations over 75,000 despite being located within 40km of an upper tier settlement. These don’t feel like characteristics of service centres and it feels odd that they do not feature on the map — but equally they don’t feel significant enough to create a ‘Subregionally significant’ category.
• 110 fall below both the resident-worker and distance thresholds (22 large and 88 small).

And so, the final map shows Great Britain at a glance…

As always, the views expressed above are my personal ones, not those of any organisation that I am associated with. Comments and queries are welcome, particularly around the three areas that I’d like to explore further in future articles: factors affecting the growth of large service centres into regionally significant settlements; optimising self-containment in Travel to Work regions; and; approximating GVA at smaller geographies.