Monday, November 30, 2009

Mapping crime statistics UK style

Just to complement my latest post on Australian mapshups with crime statistics, a quick look on how the topic is dealt with in the UK – on the official map of crimes and antisocial behaviour from National Policing Improvement Agency. The map was launched with a great fanfare by Home Office Minister on 20th October, 2009.

The application shows monthly and quarterly crime statistics for the last couple of years (by type) for policing regions in England and Wales. Crime levels in neighbourhoods, within the policing regions, are shown on a Bing Map as shaded thematic overlays. Users can compare statistics of selected area to other locations as well as download the data as csv files.

Due to high demand the site went down frequently within the first 24 hrs after the launch but now performance is quite reasonable. However, the developers did not put much effort into generating neighbourhood boundaries since they do not align well with each other. As well, there may be some underlying issues with the data collection methodology – my UK based colleague pointed out that an area in his neighborhood with no population at all (ie. local park) is shown on the map as having the highest crime rate. This is of some concern since it my potentially influence purchase decisions and the demand for properties in the area. What is traditionally seen as a very desirable feature (ie. parkland) is now indicated as a hive of criminal activities. It highlights the perils of using statistics indiscriminately.

I believe that Australian developers participating in the MashupAustralia tackled technical and presentation challenges for equivalent set of crime statistics pretty well, and to a very high standard.


Paul S said...

Hi there,

I'm one of the developers responsible for CrimeMapper here in the UK. We did a (very long!) blog post that covers some of the ups and downs of the project, including accuracy of boundary data, take a look here:


Arek said...

Thanks Paul for the link. Very interesting insights into the whole project!

I am researching various approaches to presenting dynamic boundary overlays in online mapping applications such as Google or Bing Map and hence I "picked on" that issue. I agree, it's not a small challenge and especially considering the size of the traffic you have to deal with. And there is the issue of generalisation of polygons - here not even high end GIS software is able to be of much assistance if topological consistency is required at 15 different zoom levels. Unless Java based application is an option, the only alternative is 100's of hours of "manual labor" to make those boundaries fit.