comparative urbanism – CityGeographics

World Population Density Map Update with GHSL 2023

May 21, 2023May 22, 2023Posted in City Planning, Mapping, Sustainability, transport, VisualisationTagged comparative urbanism, Density, global cities, Urban Form, Visualisation, web mappingLeave a comment

The European Commission JRC recently released a new 2023 update of the Global Human Settlement Layer (GHSL) data. This update has greatly improved the GHSL data, with a 10 metre scale built-up area dataset of the entire globe which has been used to create a 100 metre scale global population density layer. The level of detail for cities and rural areas is impressive, and it overcomes the limitations of previous releases of the GHSL. I have updated the World Population Density Map website to include this new 2023 data, with both the cartography and statistical analysis now based on the new data.

Improved Level of Detail for Cities and Rural Landscapes
The new GHSL 2023 data has produced a much more detailed 10 metre dataset of built-up area (using recent European Space Agency Sentinel data), and this is the basis for creating the updated population layer. The results are much improved, particularly for complex rural and peri-urban landscapes in the Global South, such as for India shown below. The tens of thousands of small villages are identified and used to more accurately distribute India’s huge population. This is also the case for other key regions such as Sub-Saharan Africa, and China.

The added level of detail also improves the representation of cities, with more accurate density analysis, and improved techniques to differentiate residential from industrial and commercial urban land uses. Previous releases of the GHSL were underestimating urban densities for cities where census data was weaker, but this appears to no longer be the case. The dataset can now be used for more accurate comparisons of population and density for cities across the globe. Example images for Shanghai and New York City are shown below.

Country Density Profiles – the Diversity of Human Settlement
The statistical analysis on the World Population Density Map website has also been updated using the 2023 GHSL data, so you can view the density profiles for all countries around the globe. Some highlights are shown below.

To complement the graph of the population in each density category, this updated version of the World Population Density Map includes Population Weighted Density statistics for each country and city. Population Weighted Density is a measure of the typical density experienced by residents in the country/city, in this case using the 1km² scale GHSL data. The PWD is calculated by weighting each 1km² cell according to the population, summing all the cells for the city/region, and then dividing the sum by the total population of the country/city (i.e. the arithmetic mean). This is a more representative measure than standard population density, which is affected by low density suburban/peri-urban and rural land, even where the population in these areas is relatively low.

China and India have very high density cities, but their large rural populations translate into moderate Population Weighted Density statistics overall. India is 9.9k pp/km² and China is 8.9k pp/km². The table below shows the top 20 countries by Population Weighted Density using the 2020 data-

Rank (by PWD 2020)	Country Name	Population Weighted Density 2020 (pp/km²)
1	Singapore	30.9k
2	Republic of Congo	25.2k
3	Somalia	24.1k
4	Egypt	21.8k
5	Comoros	17.4k
6	Djibouti	17.2k
7	Iran	16.8k
8	Yemen	16.7k
9	Jordan	15.8k
10	North Korea	14.8k
11	Democratic Republic of the Congo	14.2k
12	Bahrain	13.9k
13	Colombia	13.5k
14	Equatorial Guinea	13.5k
15	Turkey	13.5k
16	Morocco	13.4k
17	Bangladesh	13.3k
18	Taiwan	12.9k
19	South Korea	12.7k
20	Western Sahara	12.6k

For comparison, the equivalent Population Weighted Density figure for the UK is 4.1k, France is 3.7k and Germany is considerably lower at 2.7k. The USA is renowned for its low density living and suburban sprawl, and the Population Weighted Density measure for 2020 is 2.2k. This is the lowest figure for any large developed country in the world. Smaller developed countries have similar figures to the USA, including New Zealand, Norway and the Republic of Ireland.

Analysing the World’s Largest City-Regions Using the GHSL
The Built-Up Area and Population layers in the GHSL are used to define a settlement model (GHSL-SMOD) layer, which classifies land into urban and rural typologies. We can use this layer to define the boundaries of city-regions across the globe. This has been done using continuous areas of the highest urban category (urban centres) for the 2020 data. When you hover over cities on the World Population Density website, these city boundaries are highlighted-

This land use based method of defining city-regions produces different estimates of city populations to analyses based on administrative boundaries. The GHSL method generally emphasises large continuous urban regions, such as the megacity region of the ‘Greater Bay Area’ in China shown above, which has formed from the fusion of Guangzhou, Shenzhen, Dongguan and Jiangmen. This is the largest city-region in the world by this measure, with a population of 43.8m in 2020 (rapidly developing from a base of 5.8m in 1980). The top twenty city-regions in the world are shown below-

Rank (by Pop. 2020)	City-Region Name	Population 1980	Population 2000	Population 2020	Pop. Weighted Density 2020 (pp/km²)
1	Guangzhou-Shenzhen-Dongguan-Jiangmen	5.8m	30.9m	43.8m	20k
2	Jakarta	16.1m	26.3m	38.7m	13.4k
3	Tokyo	27m	31.3m	34.1m	10.2k
4	Delhi	8.3m	19.1m	30.3m	29k
5	Shanghai	6.7m	15.1m	27.8m	27.9k
6	Dhaka	6.2m	14.8m	26.8m	47.9k
7	Kolkata	16.3m	22.9m	26.7m	36.4k
8	Manila	11.3m	18.3m	24.8m	27.1k
9	Cairo	9.8m	16.6m	24.5m	44.9k
10	Mumbai	11.3m	18.4m	22.9m	52.4k
11	Seoul	13.3m	19.7m	22.7m	19.8k
12	São Paulo	13.7m	17.4m	19.7m	14.5k
13	Beijing	7.2m	11.6m	19m	20.2k
14	Karachi	5.8m	10.9m	18.7m	48.8k
15	Mexico City	13.8m	17.9m	17.8m	13.2k
16	Bangkok	5.4m	9.2m	17.4m	11.6k
17	Osaka	17.2m	16.8m	15.6m	8.1k
18	Moscow	9.9m	11.9m	14.9m	16.7k
19	Los Angeles	10m	13.1m	14.5m	4.6k
20	Istanbul	6.1m	10.4m	14.3m	25.2k

One of the most impressive aspects of the GHSL is that it is a timeseries dataset going back to 1975. Therefore we can create historical indicators such as the population change data shown in the table above. Many cities have more than doubled, or even tripled in population size since 1980, including Delhi, Shanghai, Dhaka and Karachi. Rates of growth in the USA, Japan and Europe are inevitably much lower, as seen in Tokyo and Los Angeles in the table above. Tokyo is often measured as the world’s largest city (for example in the UN World Urbanization Prospects), though with the GHSL method Tokyo the third largest at 34.1m in 2020. Tokyo is also distinctive in terms of its Population Weighted Density at 10.2k pp/km². While this figure is more than double the density of Los Angeles, Tokyo’s medium density is much lower than cities in China and South Asia. Incredibly, Mumbai’s density figure is five times higher than Tokyo at 52.4k pp/km², and Karachi is not far behind at 48.8k.

Credits and Links
The Global Human Settlement Layer is published by the European Commission Joint Research Centre. All the GHSL layers are available as open data and can be downloaded on their website. The GHSL population data uses as an input the Gridded Population of the World data by CIESEN at Columbia University.

The World Population Density Map website has been created by Duncan A. Smith at CASA UCL. The mapping data is hosted on a tileserver at CASA UCL kindly set up by Steven Gray. The interactive mapping layers are hosted on Carto.

Pandemic Geographies and Challenges with the 2021 England & Wales Census Results

August 16, 2022December 15, 2022Posted in City Planning, Housing, London, MappingTagged census, Census 2021, comparative urbanism, demographics, Great Britain, Housing, LondonLeave a comment

The Census is the most comprehensive demographic survey in the UK, providing detailed data for government and researchers in many fields, from health and education, to planning and transport. The 2021 Census has a unique context, as the 2021 census day (21st March 2021) occurred when the UK was still in the 3rd national lockdown which began on the 6th of January 2021. The lockdown will likely have various impacts on the census results, particularly on groups who may have changed their residence during lockdown, such as students (many of whom were studying remotely) and employees in the hardest hit sectors, such as retail, arts and hospitality.

The issue is not that the census will be inaccurate per se (indeed the Census has a very thorough survey methodology) but rather that the period in time captured of March 2021 will have aspects unique to the pandemic. These aspects are likely to be temporary as society returns to something more like normality in 2022 and beyond. While Scotland chose to delay its 2021 census for a year (which may prove to be a sensible decision), researchers in England and Wales will need to be make the most of the 2021 results and be made aware of any unusual aspects.

At present only the early population results have been released for the 2021 Census, so more detailed breakdowns of population groups will have to wait for further releases later this year. The following analysis compares the Census 2021 local authority totals to the ONS mid-year population estimates for 2020 to check how the census population results compare to the next most recent population estimate.

The differences between the 2020 data and 2021 Census are likely to reflect several factors-

The higher accuracy of the census methodology. The ONS mid-year estimates can have some errors due to limited data on some groups, such as international migrants, which are better represented in the census. Potentially Brexit could have increased the degree of error in the mid-year estimates, given changes in international migration.
Temporary pandemic changes to places of residence. These could include for example students working remotely from home during term time (including international students not coming to the UK), younger populations returning to live with parents as jobs furloughed/ended/changed to remote working, and wealthier residents choosing to live in second homes.
Longer term pandemic changes to residential preferences. This could reflect changing residential preferences towards larger houses with more space/gardens following a dramatic rise in remote working during the pandemic.

Right now the extent of these different factors is not known, and it is very difficult to separate them without more analysis and data. So the following discussion is speculative in nature.

Comparing the Census 2021 Populations to the 2020 ONS Mid-Year Estimates
The map below shows the percentage differences between the 2020 mid-year population estimates, and the 2021 Census. Blue areas show where the census 2021 population is lower than the 2020 estimates, and red areas where the census 2021 population is higher than the 2020 estimates. The differences are substantial. In South East England there is a strong geographical pattern with Inner London populations down dramatically (Camden and Westminster both have 24% lower populations in the census results). London as a whole has a population of 8.8 million in the 2021 census, which is 200k lower than the estimated 2020 total. In contrast, commuter towns and the home counties surrounding London have distinctly higher populations of around 5-10%. This pattern very much looks like a pandemic geography of Inner London residents leaving during the lockdown. Analysis by the GLA using PAYE income data confirms this general conclusion, and also points to this population drop being concentrated in young adults (note also the GLA analysis shows this population largely returning to Inner London by 2022). It is possible however that other factors such as post-Brexit emigration and very high rents are also reducing Inner London populations, and could have produced errors in the 2020 mid-year estimate data.

It is not just in the South East where there are differences between the 2020 and 2021 data. The South West and the Midlands are also areas where generally 2021 Census populations are higher than the 2020 data. The higher populations are mainly in more rural authorities, as well as some urban areas including Leicester, Lincoln, Derby, Worcester and Swindon, while Coventry and Nottingham have lower 2021 populations (university related?). There is no simple pattern here, and there are likely some 2020 mid-year population errors here in addition to any pandemic related changes. There also appear to be higher populations in areas within 1-2 hour journey times to London, possibly linked to changing residential preferences following the rise in flexible working.

North-West England has a mixed pattern with higher census populations in Cheshire to the South and in Burnley, but not in central Manchester or Liverpool. In Yorkshire and Humberside, Leeds and Hull have higher populations in the 2021 Census, while Sheffield is lower. The North East and Wales generally have a much closer alignment between the 2020 and 2021 data. The higher than expected populations in many rural and smaller town authorities fits with pandemic related patterns, but the mixed picture for many cities implies that the situation is complex, and may include both pandemic changes and errors in the 2020 data.

London and the South East
As mentioned above, the 2021 Census data for London and the South East does look to have been significantly influenced by the pandemic, with much lower than expected populations in Inner London, and higher populations in towns surrounding Greater London, and those with longer distance rail connections, such as Peterborough, Milton Keynes and Reading. We can look in more detail at some of these patterns.

As well as higher populations in commuter towns surrounding London, there are also higher population results recorded in the Outer West London boroughs of Ealing and Hounslow. This is quite an outlier compared to the rest of Greater London, and it is not clear why pandemic or mid-year population error factors would affect these boroughs in particular. In relation to the student population argument, it is interesting that both Oxford and Cambridge have higher than expected 2021 Census populations, likely because Oxbridge colleges insisted on students being on campus in 2021, and likely because the 2020 data has underpredicted wider population increases.

The big question on the geography of the South East is to what extent these pandemic related changes are a temporary lockdown phenomena, or may relate to longer term trends in residential preferences. The analysis by the GLA using PAYE data pointed to the population decreases in Inner London being a short term trend for younger adults, which in turn could have pushed up populations in the wider South East in 2021. However, an argument can also be made that some of the patterns observed fit trends of households looking for more spacious residences, and adapting to flexible working patterns that do not require daily attendance at the office. Areas beyond 1 hour travel to London with more affordable housing become much more attractive in this context (and have seen big house price increases). The map above shows a ring of local authorities surrounding London with higher than expected populations in 2021 that stretches beyond the South East into the Midlands and South West. We will need to wait for more data to see whether this is a trend beyond the immediate residential changes during the pandemic.

Age Profile Comparison, 2011 and 2021
In the comparison above, it is very difficult to separate out errors in the mid-year estimates from genuine population changes. Another approach is to look at the age profiles in 2011 and 2021 for those areas with significant population differences in the 2021 census. Firstly for Inner London boroughs with lower than expected populations, you can see very clearly in the charts for Westminster and Camden that the lower populations are focussed on younger adults, 20-40. This fits with the temporary pandemic residential changes argument. There are however other factors aside from the pandemic, such as increased rents and post-Brexit visa issues, that could also lower the population of younger adults.

For London as a whole, there is a modest drop in the population in their 20s, and increase in nearly all other age groups, with the average age increasing overall. The comparison between the 2020 population data with the 2021 census above did not pick up unexpectedly lower populations in other large English cities apart from London. Looking at other English cities in terms of age profiles, generally there does not appear to be this fall in the proportion of younger adults. Leeds is a fairly typical example shown below. Manchester on the other hand has a pattern a bit more like London, and perhaps this signals more pandemic related changes here, or maybe more similarity to London in terms of international migration.

Turning to those cities with higher than expected population increases in the 2021 census, we can also look at their age profiles. The examples below of Peterborough and Milton Keynes show really big increases in populations in their 30s and to a lesser extent 40s. Many of these households will have kids, and so there are similar jumps in the population of young children (though this does not appear in the age 0-4 group). This pattern looks very much like these towns are attracting families looking for more affordable housing, and the 2020 data has underestimated this trend. It is possible the pandemic has further encouraged this, but it looks overall like a longer term trend. Note that other towns growing rapidly in the South East such as Reading and Bedford has similar age profile charts (as does Ealing in London). The big outlier is Cambridge, where the population increase is geared more towards adults in their 20s.

Summary
This analysis has found some significant differences between the new 2021 census data, and mid-year population estimate data from 2020. It is very difficult to know whether this is due to errors in the 2020 data, or alternatively pandemic factors affecting the population in March 2021. Some of the biggest differences are in London, and it does appear that London experienced a drop in the younger adult population during the pandemic, particularly in Inner London. Manchester also has signs of a similar trend. GLA analysis indicates this drop was temporary in London, though there are longer term factors such as high rents which could also be playing a role.

Another big difference between the 2020 and 2021 data is much faster growth in many towns and small cities in the South East. Places like Milton Keynes and Bedford have growth of around 17% between 2011 and 2021. The age profile data shows this is driven mainly by adults in their 30s and 40s, often with children. The population differences look more like errors in the 2020 data here, though it is possible that the pandemic has accelerated families moving to more affordable towns to purchase larger housing.

Overall is not straightforward to separate out errors in the ONS mid-year estimates from pandemic changes, or to separate temporary pandemic changes from any longer term trends that are emerging. When the full data is released it will likely be possible to filter out certain demographics (e.g. students, younger populations) more affected by the pandemic. But it does look like the census 2021 data is going to be less certain than usual, particularly for London, and maybe for other large cities. Given that the census is traditionally used as a basis for investment in public services, more caution will be needed when using the 2021 census results (indeed London Councils have already responded that the 2021 census is underpredicting London’s population).

Table of Local Authorities with Greatest Increases Between 2020 ONS and 2021 Census Populations

Name	2011 Census Population	2020 Mid-year Population	2021 Census Population	Population Change 2011-2020	Population Change 2011-2021	Difference Between 2020 ONS and 2021 Census
Cambridge	123867	125063	145700	1.9	17.6	16.5
Reading	155698	160337	174200	3.2	11.9	8.6
Ealing	338449	340341	367100	0.3	8.5	7.9
Oxford	151906	151584	162100	0.9	6.7	6.9
Harlow	81944	87280	93300	6.2	13.9	6.9
Peterborough	183631	202626	215700	9.8	17.5	6.5
Milton Keynes	248821	270203	287000	8.1	15.3	6.2
Bedford	157479	174687	185300	10.7	17.7	6.1
Hounslow	253957	271767	288200	6.6	13.5	6.0
Cherwell	141868	151846	161000	6.7	13.5	6.0
Burnley	87059	89344	94700	2.7	8.8	6.0
Slough	140205	149577	158500	6.3	13.0	6.0
Watford	90301	96623	102300	6.6	13.3	5.9
Rushmoor	93807	94387	99800	0.0	6.4	5.7
Luton	203201	213528	225300	4.9	10.9	5.5
Crawley	106597	112474	118500	5.1	11.2	5.4
Swindon	209156	222881	233400	6.3	11.6	4.7
West Northampton.	375101	406733	425700	8.2	13.5	4.7
Merton	199693	206453	215200	2.9	7.8	4.2
Basingstoke and Deane	167799	177760	185200	5.5	10.4	4.2
Leicester	329839	354036	368600	7.4	11.8	4.1
Pendle	89452	92145	95800	2.9	7.1	4.0

Table of Local Authorities with Greatest Decreases Between 2020 ONS and 2021 Census Populations

Name	2011 Census Population	2020 Mid-year Population	2021 Census Population	Population Change 2011-2020	Population Change 2011-2021	Difference Between 2020 ONS and 2021 Census
Camden	220338	279516	210100	27.0	-4.6	-24.8
Westminster	219396	269848	204300	22.9	-6.9	-24.3
City of London	7375	10938	8600	47.6	16.6	-21.4
Islington	206125	248115	216600	20.3	5.1	-12.7
Coventry	316960	379387	345300	19.7	8.9	-9.0
Kensington and Chelsea	158649	156864	143400	-0.9	-9.6	-8.6
Hackney	246270	280941	259200	13.7	5.3	-7.7
Richmondshire	51965	53732	49700	0.8	-4.4	-7.5
Tower Hamlets	254096	331969	310300	29.7	22.1	-6.5
Kingston upon Thames	160060	179142	168000	11.7	5.0	-6.2
Gwynedd	121874	125171	117400	3.0	-3.7	-6.2
Isles of Scilly	2203	2226	2100	0.1	-4.7	-5.7
Canterbury	151145	166762	157400	10.7	4.1	-5.6
Sheffield	552698	589214	556500	6.8	0.7	-5.6
Brighton and Hove	273369	291738	277200	6.9	1.4	-5.0
Newcastle-under-Lyme	123871	129610	123300	4.6	-0.5	-4.9
Guildford	137183	150352	143600	9.3	4.7	-4.5
Blaenau Gwent	69814	70020	66900	0.3	-4.2	-4.5
Nottingham	305680	337098	323700	10.9	5.9	-4.0

Global Prospects for a Post-Car World

September 26, 2017September 26, 2017Posted in UncategorizedTagged comparative urbanism, Publications, Sustainability, transportLeave a comment

Earlier this year I worked on some charts and maps for a Greenpeace report authored by sustainable transport academic Robin Hickman, exploring the impacts of automobile dependence and the prospects for a post-car world. The report is online here.

The much debated phenomenon of ‘peak-car’ can be observed in many countries in the global north, in terms of a levelling off of private car use and increases in public transport, as shown in the graphs below. Many theories have been put forward to explain this trend, from the growth and densification of cities, to economic crises, fuel tax changes, to declining car use by younger demographics, and behavioural changes related to the internet. Given the many negative impacts of automobile dependence, from substantial GHG emissions, to air pollution, millions of road deaths annually, and contributions to the global obesity epidemic, clearly this behavioural trend is a great opportunity to develop more sustainable city forms much more widely across the globe.

This process is also observable at the level of individual cities, using data compiled by Newman and Kenworthy. Unfortunately this data is only available up to the year 2000.

The picture is of course different for newly industrialised countries, many of which are experiencing substantial growth in car ownership. I mapped data on vehicle sales, highlighting rapid growth in China and in Asia more generally, compared to static and declining markets in Europe and North America-

Thus the global sustainability challenge is to accelerate peak-car trends in the global north, and to try to curb the mistakes of automobile dependence being repeated in the developing world. Note that while the chart above suggests that the global south is increasingly responsible for GHG emissions, the picture from per-capita emissions is quite different as shown below, with the highest per-capita emissions in North America, the Middle East and Australia-

Global Urban Constellations

December 26, 2016January 3, 2020Posted in UncategorizedTagged comparative urbanism, Density, global cities, interactive mapping, Urban Form, Visualisation7 Comments

Geographers have long grappled with the complex and ever changing configurations of global urbanism. Many terms have been coined to describe new 20th and 21st century urban forms: conurbations (Geddes, 1915), multi-nuclei cities (Harris & Ultman, 1945), megalopolis (Gottman, 1961), world cities (Hall, 1966), desakota (McGee, 1991), fractal cities (Batty & Longley, 1994), network cities (Batten, 1995), postmetropolis (Soja, 2000), splintering urbanism (Graham & Marvin, 2001), polycentric mega-city regions (Hall & Pain, 2006)…

These concepts are diverse, coming from different perspectives with different methods and archetypal case studies. But there are shared themes: a focus on more diffuse and polycentric urban forms; recognition of city connections across multiple scales; and the rise of ever larger urban regions embedded in thicker global networks.

Representing and exploring the diversity of contemporary global urban forms is a challenge for cartographers. We often focus on mapping the amazing richness and diversity of dominant global cities like London and New York. Yet this is clearly a very biased lens from which to frame the vast majority of the globe, as researchers have noted. Postcolonial critiques like Robinson’s ordinary cities (2006) argue for a much more representative and cosmopolitan comparative urbanism. From a different angle, provocative research like Brenner’s (2014) ‘planetary urbanism‘ has critiqued the contentions of a universal urban age, arguing that urban/rural distinctions are no longer meaningful where capitalist networks reach to every corner of the globe.

I recently released an interactive map of the new Global Human Settlement Layer (GHSL) produced by the European Commission JRC and CIESIN Columbia University. This dataset makes several advances towards an improved cartography of the diversity of global urbanism. Firstly it is truly global, representing all the world’s landmass and settlements at a higher level of detail, down to 250m. Secondly the population density and built-up layers are continuous: there are no inherent city boundaries or urban/rural definitions (the GHSL includes an additional layer with urban centres defined, but the user can ignore these and create their own boundaries from the underlying layers). Thirdly the dataset is a time-series, including 1975, 1990, 2000 and 2015. Finally the data layers and the methods used to create them are fully open.

Diversity and Structure of Global Urban Constellations
The complexity and scale of the GHSL data is both beautiful and beguiling. In China and India there are continuous landscapes of connected urban settlements with hundreds of millions of people, scattered across many thousands of square kilometres. The cartographic appearance of these regions is like constellations of stars coalescing in vast nebulae of diffuse population. Densities of South and South-East Asian towns and small settlements in semi-rural regions exceed many major cities in Europe and North America. These are complex evolving landscapes at a scale and extent unprecedented in the history of urbanism.

Similarly there are unique trends in other major regions of urbanisation such as Latin America. Here major centres are very high density, but the extent of diffuse rural populations is far less prevalent. As a result countries like Colombia and Brazil have some of the highest urban population densities in the world.

The recognition of this global diversity does not mean abandoning global theories of urbanism. Even amongst such complexity and diversity, we can still observe shared spatial patterns and connections. Clearly we are observing landscapes heavily influenced by our current era of intense globalisation, as well as retaining inherited patterns from previous eras. Spatial logics of globalisation are apparent across the globe, though differentiated between regions, economies and societies.

The pull of coastal areas for global trade is an obvious spatial pattern. The importance of port cities is also applicable to historic periods of ancient civilisations, and indeed to globalisation in the 18th and 19th centuries. But the difference in the 20th and 21st centuries appears to be the more intensive links between major ports and global megaregions of production and manufacturing. We can observe this in the huge megaregions of China: the Pearl River Delta and Yangtze Delta (both with around 50m population depending on where the boundary is drawn), which are China’s leading manufacturing centres.

It also applies to Europe, with the higher density spine of the ‘blue banana’ linking low country ports to manufacturing centres in western Germany and north-eastern France, and more loosely to south-east England and northern Italy. As well as the manufacturing roles, it is clear that most major global financial centres are closely linked to megaregions, either at their core (e.g. Shanghai, New York, Tokyo) or within a couple of hours travel (e.g. Hong Kong, London, Paris). These centres provide the capital and business services that embed megaregions in global networks.

The importance of ports is also evident in South Asia. Port cities in South Asia are amongst the fastest growing in the world, such as Dhaka, Mumbai, Karachi, Kolkata and Chennai. But megaregions here appear as yet to be less extensive and well connected. Latin American cities are even more spatially separated and precisely defined in density terms, though there are signs of increasing connections between for example the two great Brazilian metropolises, Sao Paulo and Rio de Janeiro, and in the north between Venezuelan and Colombian port cities.

Another fascinating pattern relates to large previously rural areas of population in developing countries that are urbanising in more diffuse and bottom-up patterns. McGee used the term desakota (village-city) to describe patterns of disperse rural development in Java Indonesia. There appear to be similar patterns emerging across regions of China and India, including many areas of the vast Ganges plain, and along the great rivers of China. One of most striking features in China is the concentration of semi-rural and urban populations radiating south-west from Beijing towards Shijiazhuang and then south towards Zhengzhou (this follows one of China’s oldest rail routes, built 1903 and is nearly 600km long).

There are several areas of sub-Saharan Africa where desakota-like patterns seem to be apparent. The west coast around Nigeria and Ghana is one such area. Another is the many developments around Lake Victoria in Uganda, Kenya and Tanzania. Clearly the cultural and geographical diversity is very high in these regions, and my own knowledge of these countries is very limited. But the similar density patterns is still of interest.

Population and Density Statistics
The World Population Density map includes density statistics at national and city scales, with population totals classified into density groups (turn on the Interactive Statistics button at the top left). These help to identify differences in patterns of settlement, and how city densities relate to national distributions.

If we view the world’s highest density cities, we can see the clear links to the above discussion of urbanisation in South Asia and East Asia, and major global port cities. Note however there are many issues with defining and measuring density, which need to be borne in mind when interpreting such statistics. These are measures of residential density, and results will likely be affected by the scale and accuracy of the underlying census data. It would also be better statistically to measure peaks as the 95th or 99th percentile to prevent a single square km cell skewing the results, as there are some outliers in the results.

Highest peak density cities GHSL 2015 1km scale-

City Name	Country	Peak Density (000s pp/km2)	Mean Density (000s pp/km2)	Population (millions)
Xiamen-Longhai	China	330.5	6.3	4.75
Peshawar	Pakistan	228.9	3.3	7.54
Dhaka	Bangladesh	197.8	9.1	24.83
Daegu	South Korea	189.4	8.5	2.58
Maunath Bhanjan	India	177	38.4	0.77
Cairo	Egypt	175.5	5.1	37.84
Kolkata	India	173.5	5.8	26.87
Baharampur	India	166.1	38	1.25
Bahawalpur	Pakistan	136.9	29.6	1.06
Xi’an	China	135.4	7.1	6.04
Kabul	Afghanistan	132.7	18	4.36
Nanjing	China	130.1	6.7	6.6
Guangzhou-Shenzhen	China	128.3	5.6	46.04
Hangzhou-Shaoxing	China	127.6	4.4	7.81
Manila	Philippines	127	9.9	22.45

We can also consider the highest population city-regions based on the GHSL urban centre boundaries. These are defined as continuous built-up areas, with polycentric regions linked into single cities. This leads to quite different results for world’s largest cities, with the Pearl River Delta measured as the world’s biggest urban agglomeration at 46 million (and that’s not including Hong Kong or Macao). It is interesting to compare this to results from the UN World Urbanisation Prospects data, which keeps these regions as separate cities and identifies Tokyo as the world’s largest city-region.

Highest population urban centres GHSL 2015 1km scale-

City Name	Country	Peak Density (000s pp/km2)	Mean Density (000s pp/km2)	Population (millions)
Guangzhou-Shenzhen	China	128.3	5.6	46.04
Cairo	Egypt	175.5	5.1	37.84
Jakarta	Indonesia	20.4	6.1	36.4
Tokyo	Japan	23	6.2	33.74
Delhi	India	68	11.1	27.63
Kolkata	India	173.5	5.8	26.87
Dhaka	Bangladesh	197.8	9.1	24.83
Shanghai	China	104.4	7.5	24.67
Mumbai	India	49.5	13.9	23.41
Manila	Philippines	127	9.9	22.45
Seoul	South Korea	103.1	8.8	22.13
Mexico City	Mexico	42	8.2	20.09
SÃ£o Paulo	Brazil	38.7	8.9	20.02
Beijing	China	84.8	6.6	19.9
Osaka	Japan	13.4	5	16.53

Future Cartography of Global Urbanism
Population density is clearly a very useful base from which to understand urbanisation and patterns of settlement. But we can also see its limitations too in the World Density Map if urbanisation is viewed only in terms of density. Many US city-regions are very low density, much lower than semi-rural parts of Asia and Africa, but these US cities are amongst the most affluent and highly urbanised areas of the globe.

Clearly a more comprehensive cartography of global urbanism would combine population density with measurements of development and economic activity, and the flows of people, goods, energy and information that describe the dynamics of how cities and networks function. The development of open global datasets like the GHSL will greatly help in these endeavours.

Another important issue is improving the sophistication of spatial statistics to include multiple urban boundaries and limit Modifiable Areal Unit effects. This would be possible with the GHSL dataset, and I have tried including national and city statistics, but clearly MAUP effects remain when using fixed city boundaries. Something along the lines of my colleagues’ research testing statistics for multiple boundaries simultaneously and showing their influence would be a good avenue to explore.

World Population Density Interactive Map

December 14, 2016January 3, 2020Posted in UncategorizedTagged comparative urbanism, Density, global, global cities, interactive mapping, Sustainability, Urban Form, Visualisation3 Comments

A brilliant new dataset produced by the European Commission JRC and CIESIN Columbia University was recently released- the Global Human Settlement Layer (GHSL). This is the first time that detailed and comprehensive population density and built-up area for the world has been available as open data. As usual, my first thought was to make an interactive map, now online at- https://luminocity3d.org/WorldPopDen/

The World Population Density map is exploratory, as the dataset is very rich and new, and I am also testing out new methods for navigating statistics at both national and city scales on this site. There are clearly many applications of this data in understanding urban geographies at different scales, urban development, sustainability and change over time. A few highlights are included here and I will post in more detail later when I have explored the dataset more fully.

The GHSL is great for exploring megaregions. Above is the northeastern seaboard of the USA, with urban settlements stretching from Washington to Boston, famously discussed by Gottman in the 1960s as a meglopolis.

Europe’s version of a megaregion is looser, but you can clearly see the corridor of higher population density stretching through the industrial heartland of the low countries and Rhine-Ruhr towards Switzerland and northern Italy, sometimes called the ‘blue banana’.

The megaregions of China are spectacularly highlighted, above the Pearl River Delta including Guangzhou, Shenzhen and Hong Kong amongst many other large cities, giving a total population of around 50 million.

The Yangtze Delta is also home to another gigantic polycentric megaregion, with Shanghai as the focus. Population estimates range from 50-70 million depending on where you draw the boundary.

The form of Beijing’s wider region is quite different, with a huge lower density corridor to the South West of mixed industry and agriculture which looks like the Chinese version of desakota (“village-city”) forms. This emerging megaregion, including Tianjin, is sometimes termed Jingjinji.

The term desakota was originally coined by McGee in relation to Java in Indonesia, which has an incredible density of settlement as shown above. There are around 147 million people living on Java.

The intense settlement of Cairo and the Nile Delta is in complete contrast to the arid and empty Sahara.

Huge rural populations surround the delta lands of West Bengal and Bangladesh, focused around the megacities of Kolkata and Dhaka.

There is a massive concentration of population along the coast in South India. This reflects rich agriculture and prospering cities, but like many urban regions is vulnerable to sea level changes.

The comprehensive nature of the GHSL data means it can be analysed and applied in many ways, including as a time series as data is available for 1975, 1990, 2000 and 2015. So far I have only visualised 2015, but have calculated statistics for all the years (turn the interactive statistics on at the top left of the website- I’ll post more about these statistics later). Change over time animations would definitely be an interesting approach to explore in the future. Also see some nice work by Alasdair Rae who has produced some excellent 3D visualisations using GHSL.

New Paper- Online Interactive Mapping: Applications and Techniques for Socio-Economic Research

February 23, 2016March 11, 2016Posted in MappingTagged CASA, comparative urbanism, interactive mapping, open data, Visualisation, web mappingLeave a comment

I have a new paper published in Computers Environment and Urban Systems- Online interactive thematic mapping: applications and techniques for socio-economic research. The paper reviews workflows for creating online thematic maps, and describes how several leading interactive mapping sites were created. The paper is open access so you can download the pdf for free.

Figure_04 — Global Metro Monitor by Brookings- http://www.brookings.edu/research/reports2/2015/01/22-global-metro-monitor

The paper features web mapping sites by Oliver O’Brien (http://www.datashine.org.uk), Kiln (http://www.carbonmap.org) and Alec Friedoff at Brookings (http://www.brookings.edu/research/reports2/2015/01/22-global-metro-monitor). Many thanks to these cartographers for agreeing for their work to be included in the paper, particularly Ollie O’Brien who also kindly provided comments on the paper draft. Also many thanks to Steven Gray at CASA who set up the hosting for the LuminoCity3D site.

Here’s the paper abstract-

Recent advances in public sector open data and online mapping software are opening up new possibilities for interactive mapping in research applications. Increasingly there are opportunities to develop advanced interactive platforms with exploratory and analytical functionality. This paper reviews tools and workflows for the production of online research mapping platforms, alongside a classification of the interactive functionality that can be achieved. A series of mapping case studies from government, academia and research institutes are reviewed.

The conclusions are that online cartography’s technical hurdles are falling due to open data releases, open source software and cloud services innovations. The data exploration functionality of these new tools is powerful and complements the emerging fields of big data and open GIS. International data perspectives are also increasingly feasible. Analytical functionality for web mapping is currently less developed, but promising examples can be seen in areas such as urban analytics. For more presentational research communication applications, there has been progress in story-driven mapping drawing on data journalism approaches that are capable of connecting with very large audiences.

And here are some example images from the mapping sites reviewed in the paper-

Datashine by Oliver O’Brien and James Cheshire- http://www.datashine.org.uk

Luminocity — LuminoCity3D by Duncan Smith- http://luminocity3d.org

Figure_06 — The Carbon Map by Kiln- http://www.carbonmap.org

An Urban Renaissance Achieved? Mapping a Decade of Densification in UK Cities

September 9, 2013August 24, 2015Posted in UncategorizedTagged comparative urbanism, Density, Great Britain, LuminoCity, Planning, Population Change, UK Census 2011, Urban Form, VisualisationLeave a comment

It’s been 14 years since the landmark Urban Task Force report, which set the agenda for inner-city densification and brownfield regeneration in the UK. Furthermore we’ve seen significant economic and demographic change in the last decade that’s greatly impacted urban areas. We can now use the 2011 census data, mapped here on the LuminoCity GB site, to investigate how these policies and socio-economic trends have transformed British cities in terms of population density change.

The stand-out result is that there’s a striking similarity across a wide range of cities, with overall growth achieved through high levels of inner-city densification (shown in lighter blue to cyan colours) in combination with a mix of slowly growing and moderately declining suburbs (dark purple to magenta colours).

We can see this pattern in the growing urban regions of Manchester, Birmingham, Leeds and Sheffield above. Manchester has the fastest population growth after London, with 8.1% growth in the city-region, and a massive 28% growth in the core local authority. Average densities in Manchester have gone up by 28% (+35 residents per hectare), but it’s not a uniform growth. There are new development sites at a very high 300 or 400 residents per hectare, contrasting with low density surrounds and the extensive remaining brownfield sites. There is a patchy nature to the current urban fabric of Manchester, indicating that much further development could still take place.

The West Midlands Conurbation is the third fastest growing city-region at 7.3%, with a higher 10% growth in the core city authority Birmingham. Density increases are more modest here (+13 residents per hectare) but the same general pattern remains. Similar patterns of high density inner-city growth are also clear in Leeds (5% growth) and Sheffield (8% growth).

The trend applies to medium size cities also. Those cities with the highest growth rates like Leicester (+18%), Nottingham (+14%), Cardiff (+13%) and Bristol (+12.5%) show fewer signs of suburban depopulation-

Scottish cities have a stronger tradition of high density inner-city living. With compact cores already in place, Edinburgh (+6.5%) and Aberdeen (+5%) have been expanding the inner city into Leith and Old Aberdeen-

Meanwhile the UK’s former industrial powerhouses of Glasgow, Liverpool and Newcastle display a more problematic variation on this pattern. City centre intensification is still much in evidence, with core city authority populations growing at 8% in Newcastle, 6% in Liverpool and 4% in Glasgow. But this growth is in combination with outright decline in some surrounding towns and suburban areas, particularly around Glasgow. These patterns are linked to major programmes to overhaul poor inner-city housing stock, but are also inevitably linked to weaker economic growth in Glasgow and Liverpool. The picture is better in Tyne & Wear, where there are more positive employment signs (8% growth in workforce jobs 2001-2011).

What is driving this urban dynamic?

In addition to planning policy shifts, a series of economic and demographic changes are contributing to the pattern of central growth and struggling suburbs, as commentators have variously been observing in the UK and US (e.g. gentrification researchers, Erenhalt, Kochan). Demographic aspects include more students, immigrants, singles and childless couples. Economic aspects include city-centre friendly service and knowledge economy jobs, as well as increased costs of petrol. For these trends to occur over a wide range of demographically and economically diverse cities in the UK and beyond, clearly there are multiple factors pulling urban populations and growth in similar directions.

London Extremes

We’ve avoided the gigantic outlier of London so far. It’s a city apart in many ways- much larger (8.1 million in the GLA area) and faster growing (+14% 2001-2011). It’s also massively higher density, with average residents per hectare 50% higher (nearly 200 residents per hectare) than the next most dense city-region in GB. The biggest changes have been in Inner East London. Tower Hamlets (where Canary Wharf has boomed) is 1st on every indicator- highest population change (+28.8%), highest employment change (+50%!!), highest population density (324 residents / hectare). The pressures for growth in London are so high that there is little surburban decline in population terms (although employment has been declining significantly in Outer London).

Yet the high rate of densification in London has come nowhere near meeting housing demand. London is the midst of a massive housing shortage and crisis, with some of the world’s highest property prices. The debate is currently raging about what needs to be done to accelerate construction, with advocates of transforming more land to community ownership (e.g. Planners Network UK), relaxing planning regulations such as the green belt (e.g. LSE SERC), and implementing an array of measures simultaneously (e.g. Shelter Report). We can see London’s challenges in the maps, such as the failure thus far of the flagship housing expansion programme, the Thames Gateway, to deliver. Some high profile development sites like Stratford and Kings Cross have only recently opened for residents and so do not show in the 2011 data.

London2 — The Thames Gateway- aside from Woolwich, little housing has been delivered.

Another more surprising result is the fall in the population of Inner West London, particularly Kensington and Chelsea. While this finding does need some context- K&C is still the forth most densely populated local authority in the country- it’s still an amazing trend given the extreme population pressures in London. It is in line with arguments that the most expensive properties in London have become investments for international capital rather than homes for living. Such trends push prices up, cut supply and bring questionable benefits to the city. Addressing this issue would require tax changes, and macro economic factors like the value of the pound and yields on alternative investments are also clearly influential.

London3 — Inner London- expansion in the East and decline in Kensington & Chelsea

Summary- an Ongoing Renaissance and Suburban Challenges

Well to state the obvious GB cities are, with only a few exceptions, growing significantly. That’s not to be sniffed at given the history of widespread urban decline throughout the second half of the 20th century. And secondly the pattern of growth in density terms is clear- densifying inner cities, and fairly static or declining suburbs. The scale of London and the severe housing crisis has it’s own unique dynamics, while Glasgow and Liverpool are still dealing with significant population loss in many areas of the city region. But on the whole, the pattern is surprisingly consistent across cities in Great Britain.

Clearly this review prompts a series of further questions analysing the economic, demographic, gentrification, deprivation and property market processes inherent in this urban change, and what future city centres and suburbs will be like. Hopefully this mapping exercise should is a useful context for the ongoing research.

Launching LuminoCity GB: Urban Form and Dynamics Explorer

September 4, 2013March 11, 2016Posted in UncategorizedTagged comparative urbanism, Density, Great Britain, LuminoCity, Urban Form, VisualisationLeave a comment

Our cities have been changing dramatically in recent years, with the intensification of urban centres, redevelopment of old industrial spaces, new demographic trends, and the pressures of a volatile global economy. The aim of the LuminoCity website, which launches in beta today, is to visualise urban form and dynamics to better understand how these trends are transforming cities in Great Britain. Explore the site for yourself here- luminocitymap.org.

LuminoCity_PopDensity03 — London Population Density by Built-up Area 2011

LC_GlasgowJobDen01 — Glasgow Jobs Density by Built-up Area 2010

LC_ManchesterPopChan02 — Manchester Population Density Change 2001-2011

The visual style developed for LuminoCity combines urban activity data with built-form. Density values are calculated by dividing fine-scale (LSOA) employment and population data by built-up area, and then mapping the results to the same building footprint data (Ordnance Survey VectorMap). The result is a novel city perspective on common demographic indicators like population and employment density, with links between density and the texture of the built-environment clearly highlighted. So for example in the London map above, we can see the patchwork pattern of recent high density developments in Docklands (along the river to the east), and high density clustering around major rail stations like Paddington.

There are three layers included in the beta version of LuminoCity-

Each layer provides a complementary angle on urban form, with Employment Density showing business agglomeration patterns, and Population Density Change highlighting where intensification is occurring and where population losses are found. Examples of these three layers for major cities are shown above. The Population Density Change is particularly interesting in light of clear patterns of city centre growth and static or declining suburbs in many British cities, such as Manchester above. There is also in London a distinct pattern of population loss in the western inner-city, likely due to international capital speculation leaving under-occupied housing (see image below). These trends will be discussed in a further post later this week.

London Population Density Change 2001-2011

Multi-Scale Interactive Statistics

As well as browsing the map you can also click on particular locations to get a set of core statistics and rankings of that area for the current map layer. The statistics are at three spatial levels- City Region, Local Authority and LSOA. This feature shows how typical a particular area is compared to the wider city-region and the country as a whole. It also helps to communicate the variation in density measurements according to scale.

Location Statistics for Manchester, one of Britain's fastest growing cities — Location Statistics for Manchester, one of Britain’s fastest growing cities

Site Credits

The data used for the LuminoCity site is Crown Copyright Office for National Statistics, National Records of Scotland and Ordnance Survey. Cartography and site design by Duncan A Smith. The map layers were produced using the excellent TileMill software by MapBox.

The site concept was partly inspired by Ollie O’Brien’s ‘New Booth’ Map of Deprivation for Great Britain.

Datasets Used

The population data comes from the UK 2001 and 2011 Census, published by Office for National Statistics and National Records of Scotland. The employment data is derived from the Business Register and Employment Survey 2009-2011, also published by Office for National Statistics. The building footprint and urban area data is from the Ordnance Survey Vector District and Meridian products. These datasets have been published by the OS as Open Data, which is a fantastic recent development enabling sites like this to happen.

Spatial Analysis Method Details and Errors

All socio-economic mapping contains a degree of error, and the building footprint density approach used here introduces some issues. The Lower Super Output Area zone geography at which the population and employment data is published is fine scale but is not at the individual block level. Each LSOA zone represents groups of adjacent city blocks. The density results are therefore an average of the adjacent blocks in each zone. The results are affected by a particular version of the Modifiable Areal Unit Problem, and represent the density of fine-scale city neighbourhoods rather than of particular buildings. You can view the specific geography of the LSOA zones by turning on the ‘Admin Boundary’ layer on the LuminoCity site to see how blocks are aggregated.

Additionally the analysis does not consider building use (there are several technical and copyright challenges with this) and so population and employment density measures include all buildings rather than distinguishing residential and commercial property densities.

Finally, the ONS has not yet published census 2001 and 2011 population counts at the same LSOA geography, and a proportional spatial join method by building area was used to convert the 2001 LSOA census data to 2011 LSOAs for the Population Density Change layer.

Feedback and Comments

If you like the site or have any feedback or comments then you can tweet me @citygeographics, or email duncan2001@gmail.com. The site is in beta at the moment, and I plan to add more layers and interactivity in future releases. I’ll be blogging here in more detail about what the visualisations reveal about the changing geography of British cities over the coming weeks.

ESRI Urban Observatory- the right model for city crowdsourcing?

July 23, 2013August 26, 2013Posted in UncategorizedTagged comparative urbanism, ESRI, global cities, open data, Smart CitiesLeave a comment

This month ESRI made an interesting move into the field of global city data with the launch of Urban Observatory (TM). The site has some great interactive visualisation ideas with simultaneous mapping of three interchangeable cities, linked navigation and indicator selection. It provides an intuitive interface to explore the diverse forms of world cities-

Furthermore this ambitious project is intended to be an extensible platform. Jack Dangermond (billionaire founder of ESRI) and Richard Saul Wurman (founder of TED with a long-standing interest in city cartography) discuss in the introductory video how they want many more cities to join in, to crowdsource city data from around the world, using the ArcGIS online platform.

So is this project going to be the answer for all our global urban and smart city data needs? Well I think despite the great interface, as a city crowdsourcing model ESRI’s urban observatory is not going to work. But it’s interesting to explore why, particularly in relation to the bigger questions of whether the open city data revolution is going to be truly global and inspire a new era of urban analysis and comparative urban research.

ESRI’s site states that “information about urbanization does not exist in comparative form”. In reality comparative urban analysis is a growing trend across many sectors, from international organisations like OECD, EU and UN (including the original UN Habitat Urban Observatory); to environmental organisations like ICLEI and C40; to economically focussed organisations like the World Bank and Brookings; to global remote sensing providers like the USGS; to major commercial data producers in transport and telecoms; to the many urban academic research centres around the world (including the two London based centres I’ve worked for, CASA and LSE Cities).

Global cities data example- GaWC Network at Loughborough

Brookings MetroMonitor- comparison of US cities' economic performance — Brookings MetroMonitor- comparison of US cities’ economic performance

LSE Cities- over a decade exploring comparative urbanism

There’s a rich and growing field of data providers and analysis techniques to draw on for comparative urban analysis. Indeed the ability to gather and analyse urban data is absolutely central to the whole Smart City agenda. But there are clearly many challenges. What do cities gain by opening up their data? Who then owns the data and controls how it is presented? Who selects what data is included and excluded?

I believe the natural platform for civic data (and subsequently for the international comparison of urban data) will be an open platform with wiki features to encourage civic engagement. This provides the answers to the above questions- citizens gain from better access to data and institutional transparency; citizens own the data and have a say in what is included and how it’s presented. This is the model for current successful open data sites like the London Datastore, where anyone can access the data, and Londoners can request new datasets (backed by freedom of information legislation). Unfortunately the governance situation is of course much more complicated for the international comparison of cities, and this has limited progress.

As the world’s leading provider of GIS software, ESRI are in a strong position to integrate global datasets, and have clear commercial interests in amassing urban data for their clients. But it’s much harder to answer questions about who owns and controls data in their urban observatory project. Arguably this will limit the number of cities volunteering to take part, and limits the project’s ability to respond to the diverse demands of global cities and their citizens.

A further huge challenge in comparative urbanism is in developing the right analytical techniques and indicators to answer key urban questions. This will inevitably require more sophisticated analysis tools than a set of thematic maps, and needs to draw on the many research strands developing the most relevant analytical tools.

Overall there will be some exciting competition in the coming months and years in the expanding market of international urban data integration and visualisation, with different models from commercial, government and academic contexts. ESRI’s urban observatory is an innovative project, and should stimulate further advances.