“Gridify” Hong Kong

View the full-size map HERE

TL;DR

Are there any other ways to describe land utilisation of the city besides merely looking into the ubiquitous land utilisation maps and classification diagrams? I created a layer of hexagonal bins covering Hong Kong, then used it to count different types of land uses. These maps “with a little twist” provide a fresh way to look into land use distribution and presents an uncommon mapping technique of grouping data into regular grids.

Land use maps are, literally, thematic maps that draw the classified land resources/uses of the land. Among the 1100 sq.km. land of Hong Kong, a quarter of the land resources are built-up area, incorporating various land use categories.

Land Utilisation in Hong Kong in 2015
Land Utilisation in Hong Kong in 2015, from http://www.lsgi.polyu.edu.hk/rsl/lsw/vegetation/LULC.html

The cover image (i.e. map on top) is a land use map of Hong Kong with a little twist — the dots shows the percentage of land area of the selected land use categories.

How to read the map?

Before showing any data, explain how your visualization works.

from Alberto Cairo’s blog

The map will look unfamiliar if you haven’t look into some hexagonal grid (hex grid) maps. Maybe we should first talk about how the data are curated and aggregated.

The three steps to create the map

Step I: Tessellation

First thing first, we have a layer of hexagonal mesh, tessellating Hong Kong. Each regular hexagon is 12.5 ha (0.125 sq.km.) in size. In terms of length, the hexagons have a circumradius of 220m (219.35m to be precise). With those trigonometry identities you had learnt (and probably forgot) during high school, you should be able to compute those numbers (The proof is left as an exercise to the reader).

How large is 12.5ha? It’s about 17.5 football pitches, or two-thirds of Victoria Park, or a quarter of HKU campus.

Bins/Grids/Tessellation/You name it

Step II: Computing % of land use in grids

Then, with four major land use categories identified, the percentage land area of each land use category in each grid is calculated.

  • Orange locates the building structure/blocks
  • Purple locates the roads and carriageways
  • Green locates the parks
  • Blue locates the water and other hydrologic features

Let’s show an example using the grid below. Building structures covers 41.0% of the total area. Roads cover 23.3%, parks cover 5.6% and water bodies cover 5.3%.

If you could identify where this grid is located, let me know. I will buy you a drink and wanna have a map chat with you.

Step III: Generating dots with proportional size

After all these tedious data curations, we could visualise the data. In the map, the dots are proportionally sized by the percentage of land area covered. A larger dot means a larger percentage of land belongs to that selected land use category. If the grid does not have any land of the selected type (i.e. 0%), no dots will be drawn. The colour palette above applies.

A neat legend sitting on the bottom-right of the maps

Finally, overlaying the layers and adjusting the opacity of the dots would make the map shown in the cover photo.

Binning into hex grid — why?

But you may want to take a step backward and ask:

Why hex grid is used? Why you choose not to use district boundaries to summarise the data?

This is a common question rose when people looking into hex-grid maps. Here’s the short answer:

The aggregation of incident point data to regularly shaped grids is used for many reasons such as normalizing geography for mapping or to mitigate the issues of using irregularly shaped polygons created arbitrarily (such as county boundaries or block groups that have been created from a political process).

from https://pro.arcgis.com/en/pro-app/tool-reference/spatial-statistics/h-whyhexagons.htm

To be specific, this is the discussion of the “unit of analysis” problem in the spatial analysis field.

Notes on choosing the spatial unit of analysis

NOTE: You could probably skip this long section and jump directly to the results (i.e. the final map).

The spatial unit of analysis refers to the spatial boundaries used to aggregate data from scattered single data points to “merged” results. Undoubtedly, the most common spatial analysis unit of Hong Kong is the District Council boundary.

Since this boundary is too well-known, it has become more or less the de-facto unit of spatial analysis. You could find this unit of analysis in those ubiquitous reports written by NGOs or academia when they try to visualise the data with maps. The common headline for these reports would be “Yuen Long district has the highest percentage of juvenile crime rates among 18 Districts”, “Kwun Tong District has the highest average building age” or “All the new Dengue fever cases this week appear in Wong Tai Sin District”.

However, choosing the spatial unit incorrectly could greatly affect the analysis result. When you use different spatial units for analysis, your results would greatly differ, and could also significantly affects the interpretation of findings. I have already written some related topics in my previous blog. And the jargon describing this phenomenon is called Modifiable areal unit problem (MAUP).

Choosing a suitable unit greatly depends on the nature of the project. Then, how should one choose a suitable spatial unit?

Administrative Boundaries — why?

Using District Council Boundary allows readers to relate the demographics quickly. If you are making some policy-related maps, using administrative boundaries like District Council boundary would be preferred as it is easier to policymakers to relate the average socio-economic status of each census unit. And it is easier to take actions as the working unit are also divided by the administrative boundaries.

Urban planners in Hong Kong must be used to think boundaries in terms of tertiary planning unit (TPU) and street block (SB). They are drawn for master-planning matters. They are some neat unit of analysis related to urban planning matters as the census also aggregates data in street block level. This is exactly the case in my travel pattern mapping project. I used street blocks as the unit of analysis since 1. census data are in street block level and 2. it is easier for readers to understand the socio-economic background of each “block” and thus take related actions.

Administrative Boundaries — why not?

Yet, Be aware that the administrative census tracts are artificially made by the authorities. They are not always the best bin to aggregate things. You know the boundaries of the 18 districts well, or even the detailed boundary of each election area. But have you aware of the evil gerrymander?

Hex bins — why?

In this project, census data and demographics are not the main dishes. I am just focusing on the overall morphology of the city.

With equal-area hex bins, I could compare the grids on the same basis as they mostly have the same landmass. With points, you could also easily use the size of the point as an additional variable for visualising data. This helps to make the overall presentation neat and easy to read.

Tessellated bins also allow direct comparison of the topic across cities. Keep in mind that the size of administrative boundaries across cities varies greatly! What if I want to apply the same token to visualise the difference between Hong Kong, Tokyo and Singapore? If the maps are aimed to be simple and neat, hexagonal grids should be the silver bullet. The result maps are clean and neat for you.

Administrative boundaries are also not suitable for some data visualisation techniques (e.g. scale the size according to the quantity of the data). What happens if you enlarge or shrink irregular administrative boundaries to present the results like the dots? Cartograms could be hard to read when you just pump up/scale down those strange, gerrymandered boundaries. In my project on double ageing, I merely colour the street blocks to visualise the data, yet not adjusting the size of the street block boundaries¹.

Binning into grids is of course not a perfect solution — you still have to face the problem that every grid has a different “nature”. Some of them have a part of them falling on the sea; Some of them are located on the steep slopes that are mainly uninhabitable.

Yet, I have to say again: The choice of unit of analysis greatly depends on the project’s nature. You are choosing the best one for your project, not the perfect one. I would like to make a map unified and “visually pleasing” to read. So? Hexagonal grids!

Land Use, “gridified” and counted

Back to the map again. We could start investigating the patterns of the map after the not-so-short discussion above. When it comes to uncommon mapping methods, time spent on explaining the maps may take much more time.

Here I will discuss the selected land use categories one by one.

Building Structure

Architects are somehow obsessed with figure ground drawings, which is just a fancy term to describe a map showing building footprints in the cities. Figure grounds are commonly used for baseline studies to examine the urban morphology and the context of travel/walking network.

A figure ground, from https://www.citymetric.com/horizons/figure-ground-diagrams-tell-stories-about-cities-2359

So, let’s count each grid by the percentage of land covered by buildings. Here the term “building” refers to all types of buildings block and structures.

Not a surprising result that the largest dots are located along the two sides of the Victoria Harbour. It could also be observed that the grids in new towns are generally smaller. The difference of the development history and planning of the new towns are possibly the causes. The permitted building density (in terms of site coverage) in the core urban area is usually higher than that of new towns.

Road

Road accounts for 46 sq.km. land area (4.1%) in Hong Kong, which is more about the same as the amount of land occupied by building blocks and structures! Here is the gridded road map.

The patterns here are quite obvious, as you could observe those large circles forming the highway routes. Route 4 sits on the north-tip of Hong Kong Island, Route 5 wraps the southern Kowloon Pensisula, Route 9 encircles the New Territories and Route 8 cuts through north-tip of Lantau Island. You could see those super-wide carriageways with 4–10 lanes takes the majority of the land area in the grids they pass through.

For reference, the width Island Eastern Corridor could reach 50m, while Kwun Tong By-pass flyover is in general 25m wide. In some mega-scale interchange like Kwai Ching Interchange, the total width of all highways and flyovers (Tsing Kwai Highway and Kwai Ching Road) could add up to a width of about 150m. This is what made those mega-size dots next to Kwai Tsing Container Terminals.

Parks

How about urban green spaces? Parks are the lungs of the cities. Parks and other recreation facilities in total take up 28 sq.km. of land. Yet, here I am interested in those passive open spaces like parks and sitting-out area only.

Keep in mind that this map only accounts for urban parks. The countryside and country parks (which are the real lungs of the city) are excluded here.

Remember I said a hex grid roughly equals to 2/3 of Victoria Park? I believe you could easily spot that large dot representing Victoria Park and its proximity. Same as other territorial open spaces.

Water Bodies

Finally, all the hydros. Seas here are excluded in the dataset — I mean, the hex grids are tessellated within the enclosed land body.

The reservoirs stand all the way out. I will not waste time to pinpoint the reservoirs one by one. You could also observe the river flowing along Tai Po and Shatin New Town. The tiny dots are the stream courses, with natural water falling from the top of mountains to the gather grounds on the periphery of the countryside.

Hiatus

These series of maps are not mainly for inferencing new insights, nor interpreting new results for planning. Those basic spatial layout of the land use and percentage of land for each type of land cover are readily available on the Planning Department’s website.

Instead, these maps are casting the light on 1. providing a fresh way of looking into land uses and 2. discussing some less-common cartographic map methods. There are more than one hundred reasons to make a map. Sometimes you want to tell some new insights with your spatial data. Sometimes, you are just trying to play with some not-so-common visualisation methods.

There are many other ways to present spatial data. Hope this method should give you a more fascinating view of the ubiquitous land utilisation map. Or, at least I hope it will be less boring.

References

More on choosing spatial units for analysis:

Land use data:

  • Land use open data of Hong Kong provided by the Planning Department (in raster format)
  • 1:50 000 Digital Topographic Map provided by the Planning Department (in raster format)

[1] No, you do not want to see the map with those street blocks scaled strangely.

--

--

--

Urban Data Science Enthusiast | Urban Planning | GIS | Maps | Data Visualisation | kennethwong12.netlify.app

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Do We Even Need Experts Anymore?

Introduction to the EBSCO platform using CINAHL

GA COVID-19 Report December 13, 2020

Sales Forecasting for Rossmann Drug Stores

Parties and Sex Crimes in NYC: What’s the correlation?

From Excel To Databases with Python

Keep Your Credibility Intact, Don’t Cite Bad Data

GA COVID-19 Report November 1, 2021

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Kenneth Wong

Kenneth Wong

Urban Data Science Enthusiast | Urban Planning | GIS | Maps | Data Visualisation | kennethwong12.netlify.app

Meet the Women Mapping the World: Part 2

Why Microsoft Excel needs to allow table-based referencing in custom conditional formatting rules.

Sample: conditional formatting using custom rules in Microsoft Excel

How Do I Add OpenStreetMap to WordPress?

Mobile GIS