Thursday, 28 May 2020

Communicating big numbers

This is a very short post, inspired by the recent 'An Incalculable Loss' piece in the New York Times that printed 1,000 names of people who succumbed to Coronavirus. I found this very powerful, and it served as a great reminder of the human toll of what's happening right now. Sometimes it's hard to see the humanity in a chart or map, as useful and as powerful as they can be.

This is a cropped version of the image below

There have been some excellent written pieces and visualisations over the past few months, and these have helped the public understand better what is happening but at the same time I think it is easy to lose perspective with something so big. In the United Kingdom, it was recently reported that the total figure for 'excess deaths' passed 60,000. Some pointed out that this is close to one in a thousand. To me, that doesn't really resonate because it's the absolute value here that really hits home. Rates are very useful for comparisons, but sometimes the raw numbers are what we really want to get a sense of.

Therefore, it makes more sense to me, in terms of understanding the scale of the pandemic in the UK, if I think of it as the entire population of a place like Scarborough, or Corby, or Livingston, or Barry, or Macclesfield. All substantial towns of around 60,000 residents. 

So, to try to get my head round the scale of what's happening I plotted 60,000 points on a page, then visualised them using human figures. That is what you see below. 

Open in new tab/window to see full size

The largest version of this image can be found here. Once it loads in your browser you will be able to zoom and scroll but the idea is that either zoomed or unzoomed this helps shed light on how big the figure is.

Wednesday, 29 April 2020

Population density in Europe

Population density is a subject I've been writing about for a while, so I decided to create a few more renders of European population density using the EU's GHS_POP data, which is freely available. The maps below use 1km x 1km data and the height of the bars represents the number of people living in any one square. The big squares are 50km x 50km (about 30 miles) and are there to provide sense of scale. The highest 1km densities are found in Spain and France, and Madrid, Barcelona and Paris in particular where you get values of more than 50,000. You can read more on that in a previous piece I did on the topic. Anyway, enough words for now - see below for the six different renders I created; I've tried to create a few interesting perspectives here. Scroll below the images for a bit more technical information about the process and the data.








Why do this?
I did these because I find it a useful way of understanding wider patterns and the bird's-eye view gives a nice sense of perspective over a large area. But of course with 3D mapping it's always a bit of a balancing act because turn the map one way and you inevitably obscure something or somewhere of interest to people. Although, in this case, it is more experimental and aesthetic than analytical, which I think is okay sometimes. 

The data for my area look wrong!
This may be the case because after all the quality of any kind of visualisation like this depends upon the data inputs. Anyway, the data come from the EU's Global Human Settlement project and are available for different years. I'm using the 2015 data here, and you can read more about it and get the data yourself at their excellent new download pages - and do read the documentation, which is also great.

Data processing and mapping
I downloaded the entire global 1km dataset and then in QGIS I clipped out an area focusing on Europe and then extracted it as a rendered tif. If that means nothing to you, don't worry! It should be helpful to anyone who wants to replicate this. I then imported the data into Aerialod and spent some time tweaking lots of different settings and I wanted to give Europe a nice hopeful glow effect. 

Should I give away my trade secrets on what settings and colours I used here? Of course I should - see below.




This is a difficult time right now, and I'm always looking for alternatives to doomscrolling, so I hope you find these interesting to look at - I found them interesting to make!

Monday, 30 March 2020

Making 3D landscape and city models with Aerialod

This is my second post on Aerialod - the interactive path tracing renderer for height maps, by @ephtracy. It's available in 32 and 64-bit versions on Windows and it's super-lightweight but you can do some really amazing things with it. My previous post covers more of the basics, including controls and interface settings, so here I'll share some more information on how to tweak the settings (of which there are many) to create nice looking vistas. But as a quick reminder, hold down your right mouse button to rotate the map and use the space bar plus the left mouse button to move it around. Scroll wheel to zoom in and out. Keyboard shortcuts are here.

First I'll show you some new visuals I created in Aerialod, then I'll show you some of the settings I used to create them - in screenshots and in a small slide set. All the images I've posted below are in a shared folder, as well as a sample Lidar dataset for Cheddar Gorge in England that you can just drop in to Aerialod and then tweak the settings as you wish. You can of course create these kinds of images in other software (notably Blender) but Aerialod is much easier to use, though can be confusing at first.

So, let's begin with the highest peaks in Scotland, England and Wales, plus one other view that I like.

I did this using a 5m DTM (not open data)

See the curved horizon? - that's the SG lens setting

A very nice looking mountain

I've added Glen Etive because it's so lovely

But of course we don't always have to map things like mountains. If we have good quality Lidar data, as we do in much of the UK, we can create quite interesting cityscapes, as you can see below.


Here's a little Salford Sunset to get things rolling

A Newcastle-Gateshead vista, along the Tyne

The example data I've put in the shared folder is of Cheddar Gorge in the south west of England, and it looks like this once you fiddle with a few settings.

Sun low in the sky, dusky effect

I've tweaked some of the settings here to make it glow

Different angle, light filtering through the gorge

A foggier, early morning effect

So how do you do all this?
If you've not used Aerialod before then you'll really need to read my first blog post on it in order to get to grips with the controls, etc. Once you've done this, look closely at the screenshots below as the settings in them show you how I achieved the effects in the four images above. Study them closely and then see further below for a short slide set with more annotation on the settings options in Aerialod.

Take some time to Google some of the different terms and they'll begin to make a lot more sense - e.g. the Rayleigh setting refers to Rayleigh scattering, which relates to the blue colour we see in the sky. So, using the default Rayleigh setting in Aerialod you'll see a blue sky but if you reduce the number to, say 10, it will become more blue and if you put it up to 90, for example, it will look a lot less blue and instead more like a lovely glowing orange/yellow sunset. 

This relates to the first Cheddar Gorge image

This also relates to the first image

Notice the glowing light at the corners here

This is the fourth Cheddar Gorge image above

In addition to reading this, it's a good idea to check out the @ephtracy Twitter account for other tips, plus the #aerialod hashtag on Twitter. There are also now a few good video tutorials online, including this one by Steven Scott.

Here is the small set of slides, with annotated screenshots, that I made in order to help you get to grips with the settings a little better - direct link here.




All this, including the data, can be found in the Aerialod tutorial folder I made for this short blog post. Just drag and drop the .asc file I provided into Aerialod and start playing around.

I've put everything in here

The information above should be all you need to create more realistic, impressive 3D landscape or city models. It all depends of course upon being able to get good quality data at a high resolution - I've provided an example in the folder, but you can get a lot more on the Defra Lidar download page for England. You can use the OpenDem searcher to find suitable data for other parts of the world.

To end, I'm just going to post a few more images that I haven't shared anywhere else before, before a few final comments.


This is done by keeping the 'Map' option on in Grid settings

In the example above, I have all the options on in the Grid settings (Map, Ground, Vertical) and then I have used a value in the 'Step' settings on the right to match the resolution of the underlying data (in this case 5 metres) so that I get a blocky kind of effect - see below.


This works if you want to Minecraft your map

The Western Highlands of Scotland

In the image above, I just used a 1:50,000 Land-Form dataset shared by Carol Blackwood to render the Western Highlands of Scotland - I added the labels using GIMP. You can get the individual tiles as open data from Ordnance Survey as well (OS Terrain 50).


This city is home to two football teams, as you can see here

In the example above, I've just taken some Lidar open data of Liverpool and created a view looking down on Anfield and Goodison - home to Liverpool and Everton respectively. They are very close together and the densely packed terraced housing nearby also makes for an interesting example use case with this kind of data.

And, last but not least, St Kilda - a very wild and remote archipelago off the Western Isles of Scotland that is now a World Heritage Site.


In this example, I've added some fog

Notes: you can add a single file to Aerialod (it can handle PNG, JPG, TIF, IMG and ASC formats) or you can add multiple files at once using the little button on the top right that looks like a folder. That is covered in my previous blog post. If you try to load humongousbytes of data then it may crash. Sometimes it won't crash but just won't load. Generally I find it just works and the Scotland 1:50,000 terrain model above is over 600MB and worked fine for me. Just remember that when you hover over any of the tools in Aerialod you will get information on what it is and what it does at the bottom of the window. 

Citation: Blackwood, Carol. (2017). Scotland Land-Form PANORAMA® DTM, [Dataset]. EDINA. https://doi.org/10.7488/ds/1929.

Wednesday, 11 March 2020

45 Minute Cities

Have you ever wondered what a city would look like if it only included areas within 45 minutes of the centre? Of course you have, otherwise you wouldn't be here. Well, even if you haven't, please read on and be sure to check the nerd notes at the bottom of the page. The idea here was to produce a set of maps, based on a central point in 26 different British cities, and a 45 minute time cut off. More specifically, from everywhere shown on the maps below you should be able to reach the centre (in this case I chose a main railway station) within 45 minutes of leaving home (note, not from when you get on a train or in your car). I included travel by public transport, driving, or driving plus train in my model and an arrival time of 08:45 on a Monday morning. I then calculated the population within each area and the maps you see below are the result. These '45 minute cities' are of course not actual cities from an administrative point of view, although they are a kind of functional entity that makes more sense from an economic point of view. But really this is just a first-stage experiment and I thought the results were interesting so I'm sharing them here.





























What's this all about, then?
Partly it's about me using a new tool more regularly - TravelTime platform's QGIS plugin - and partly it's about some other work I've been doing over the past few years relating to functional urban areas, polycentricity, functional regionalisation, and whatnot. You can read a bit more about one of the off-shoots of this work here, which is all about a little graphic I tweeted in February 2020. It also relates to some of my previous research, including this work on transport-related barriers to employment for the Joseph Rowntree Foundation.

But for now, in this case, it's mostly about curiosity and me wondering 'how big are these cities if we see them as 45 minute commuter-sheds based on arriving at some central point with enough time to get to work for 9am on a Monday morning?'.


What about Marchetti's constant?
The idea that on average we travel 30 minutes each way to work each day is of course well known, but I thought a 45 minute time frame, from leaving the house to arriving at the workplace (and including interchanges, walking, and the actual realities of the daily commute) would be more interesting and also realistic. My door-to-desk journey these days takes about 20 to 25 minutes but in the past it used to take anything from 50 minutes to an hour and a half when I commuted between Liverpool and Manchester each day.


How does this compare to travel-to-work-areas?
I'm glad you asked, because I had the same question early this year so I added up the 2018 population data for most TTWAs and the results are shown below, as well as on this Twitter thread. This is particularly interesting when we look at, say, Leeds and compare it to the Leeds map above. Or even the Warrington one above. But of course TTWAs are based on a commuting self-containment threshold and my maps here are just experimental outputs created as part of a little curiosity project (for now at least).

I only had English and Welsh data for this

The numbers for my city look wrong
This may be the case, but it may be a bit more complicated than that. The specific point of arrival I have chosen (the central railway station in each city - or three in London) and the specific time I chose can have an impact on the results. However, I've been through the maps and I think they look right, or at least plausible for sure. In some cases it may very well be possible to get from a location off the map to the central point I chose, but the idea here is it needs to be reliable and without rushing for connections and allowing enough time to get to and from the mode of transport at either end. It's a whole journey commute time rather than a single point-to-point, best case scenario trip. So, you may be able to do better yourself but this is based on a daily journey that most people can realistically make within 45 minutes.


'Why is Warrington so big?', and other FAQs
The population of Warrington isn't 4 million, so let's all calm down. But maybe it is and nobody had noticed. Okay, maybe not. Yet at the same time it is highly accessible, with the M6 and M62 plus the Liverpool-Manchester railway line passing through as well as the West Coast mainline. I'd say Warrington has the Connectivity Double Whammy (CDW) nailed down pretty well. I used to travel through Warrington each day on my day to work and I definitely had the 'if I lived here I'd be home by now' thought more than once. In this sense, then, Warrington may be the Ultimate 45 Minute City.

You might also ask why Leeds has such a high population in this little 45 minute city experiment, when we compare it to the TTWA population above. I suspect it's due to a mix of possibly unique not-very-good internal connectivity and other things Tom Forth knows a lot more about. Also probably has a bit to do with overlapping job markets nearby - e.g. Manchester, Sheffield and the like.

What's the deal with Glasgow vs Edinburgh? Well, if you take the combined populations in the 45 minute city for both of them, you get a good chunk of Scotland - more than half. But why is Glasgow so much bigger than Edinburgh? Again, a mix of things but Glasgow has excellent suburban rail, a subway, motorways and so on, whereas Edinburgh has epic congestion and a good few buses and not too much other stuff, despite the new-ish tram system. Plus the wider Glasgow area just has more people in it.

Why did you choose the central railway station? Just because it is generally very central and close to the jobs. I could have put the arrival point somewhere else in the city centre without it making too much of a difference but I wanted to be consistent and there is no agreed 'this is definitely the city centre' point in each city anyway.

I've spotted something wrong, who do I tell? Feel free to get in touch. You should probably treat this in the spirit it is intended - i.e. an experimental take on functional urbanism - but if you do see something that looks egregiously wrong please let me know. I can also feed back to the people who make the TravelTime platform tool as they're always trying to improve their models and are very keen on feedback.


Nerd notes
I mapped this in QGIS 3.4 using the TravelTime platform plugin. I created a little model (see below) to automate the process than ran a batch process on it for the 26 cities I chose. The 26 include a few that may seem like odd choices but I did try to include places far and wide. For the population counts, this is based on the most recent LSOA (England and Wales) or data zone (Scotland) population estimates from mid-2018. The boundaries here do not align perfectly with the 45 minute isochrone so I suggest you consider the specific numbers in each case as being 'roughly' or 'about' rather than exact. I did use the population-weighted centroids though so having said that I expect the population figures are close to the real numbers. The TravelTime platform api used real-world transit information and if you want more details on it just check out their website.

This is the QGIS model I made