The 2.0 version of the Fusion Tables challenger CartoDB was released last week. This is great news for all those of you that, for one reason or another, want to break free of Google Maps. CartoDB lets you geocode and plot addresses and other geographical points on Open Street Map tiles, just like Fusion Tables does in Google Maps. With a little help from Quantum GIS you will also be able to make beautiful choropleth maps.
In this tutorial I will walk you through this process. It will hopefully be a good intro to QGIS and geospatial data if you are new to these concepts.
Step 1: Getting the geo data
I’m going to map unemployment data in Swedish muncipalities. So I obviously need a municipality map of Sweden.
Maps come in different formats. For this purpose a basic vector map (in svg format for example) is not robust enough. We need a .shp-, .geojson- or .kml-file. The difference between those formats and svg is that they define data points in geographical coordinates, whereas the svg is defined in pixels. Fortunately the Statistics Bureu of Sweden is generous enough to provide shape-maps of the municipalities.
Now go to QGIS (I think you need to be at version 1.7 to be able to follow this tutorial) and import the downloaded map (Layer >Add vector layer or Ctrl+Shift+V). Sweden_municipality07.shp is the file you want to select. Set the encoding to ISO-8859-1 to get all the Swedish charachters right.
Step 2: Appending data
I got some basic unemployment data from SCB and changed the municipality names to official ids. Here is the whole set.
To join the unemployment data with the map we need to have a column that we can use as a key. A way to patch the two datasets together. In QGIS, right-click the map layer and choose Open Attribute Table to see the embedded metadata.
As you kan see the map contains both names and ids for all the municipalities. We’ll use these ids as our key.
Go back to your Google spreadsheet and save it dataset as a csv-file (File > Download as) and import it in QGIS (again: Layer >Add vector layer or Ctrl+Shift+V).
You will now see a new layer in the left-hand panel.
Confirm that the data looks okay my right-clicking the unemployment layer and selecting attribute table.
To join the two datasets right-click the map layer and choose Properties. Open the tab Joins.
Make sure that you select kommunid and KNKOD (the key columns). To make sure that the merge was successful open the attribute table of the map layer.
Two important things to note here. Choose GeoJSON as your format and make sure you set the coordinate system to WGS 84. That is the global standard used on Open Street Map and Google Maps. On this map the coordinates are defined in the Swedish RT90 format, which you can see in the statusbar if you hover the map.
Step 3: Making the actual map
Congratulations you’ve now completed all the difficult parts. From here on it’s a walk in the park.
Go to CartoDB, login and click Create new table. Now upload your newly created GeoJSON file.
You can double-click the GeoJSON cells to see the embedded coordinates.
Move all the way to the right of the spreadsheet and locate the unemployment column. If it has been coded as a string you have to change it to number.
If you go to the Map view tab your map should now look something like this:
To colorize the map click the Style icon in the right-side panel, choose choropleth map and set unemployment as your variable. And, easy as that, we have a map showing the unemployment in Sweden:
I have played around with CartoDB for a couple of days now and really come to like it. It still feels a little buggy and small flaws quickly start to bug you (for exemple the lack of possibilities to customize the infowindows). And the biggest downside of CartoDB is obviously that it is not free of charge (after the first five maps). But apart from that it feels like a worthy competitor to Google Fusion Tables. Once you get a hold of the interface you can go from idea to map in literally a few minutes.