Kepler.gl is an open source mapping tool that claims to work for large scale datasets.
It has been developed by Uber, where they have developed an in-house solution based on open source components which they use to analyse their data. Luckily for us, they decided to make their solution open source and available to us.
Kepler.gl works within your browser, which is a nice feature as it means you retain control of your data, which could be important if you wanted to map data which could contain sensitive data.
To try the system out I downloaded our 2011 Census Headcounts, in particular the file called UK postcode data and supporting metadata for 2011 frozen postcodes, which is a zip file.
I unzipped this, ready for me to load into Kepler.gl. I chose this dataset as I know it contains latitude and longitude information, as well as population and deprivation data.
Uploading data was pretty straight forward. There’s an option to browse for your data file or drag and drop the file into the browser.
A slightly annoying bit for me was that map opens focused on San Francisco, when I know the data I added was for the UK. But it was easy to refocus the map on the UK using the standard grab-and-pull functionality.
To map the data, I needed to add a layer and choose the type of data.
For this data I knew it was point data. I also entered a name of the layer. I called it 2011 Census Postcodes. It’s possible with Kepler.gl to add more than one layer so giving your new layer an meaning full name is useful.
It next asked for the fields that contain the Lat(latitude) and Lng(longitude).
In our data I discovered that we mislabelled them, so the field names were the opposite of what they should be (I’ll get this corrected).
You’ll notice that there is the option to add a field to represent the Altitude. For this initial visualisation, I left that blank.
This now created a map showing UK postcodes, but (to be honest) it was a bit boring.
Kepler.gl has the option to colour the postcode points based on the value of a field.
In this data, were the UK Townsend Deprivation scores as quintiles calculated at the output area level, so I used this field to colour-code the points. I also sized the points based on the number of people living in that postcode.
The finished map of the UK shows a very mixed view, but if you zoom into a town and city you can then see the differences between postcodes.
For example, here’s a map of Belfast showing differences in deprivation between postcodes. Dark red is less deprived and yellow is most deprived.
Overall I found this web app easy to use, but it may give some issues for people unfamiliar with mapping.
However, as a free tool to map data without sending it back to a server it presents an option to map more personal data without the worry of having this data hosted some where you don’t know.
Rob Dymond-Green is a Senior Technical Co-ordinator for the UK Data Service, working with aggregate census and international data.