Producers and consumers brainstorm the potential of geospatial data for crime analysis

On Tuesday 20th the GeoCrimeData Project ran a workshop attended by 16 crime analysis stakeholders who will ultimately form part of the main user group for the data that we produce. There was a healthy 50-50 split between people working in crime-reduction organisations and academic researchers interested in crime and the analysis of spatial data more generally. This eclectic group helped to ensure that the discussion about access to and use of geospatial data was stimulating, focused and embraced an unusually wide range of perspectives.

The core of the meeting consisted of a breakout session where the delegates were split into two groups (developers of geospatial data and potential users of data) to discuss:

1. who might use the data and how they could use it
2. what the barriers are to the uptake of geospatial data
3. how these barriers can be overcome by developers.

The feedback was extremely useful for the project and will directly impact upon the spatial analyses that we perform and the ways in which we release and publicise the data. We are in the process of writing a summary about the barriers and solutions which will be relevant to many groups who use geospatial data, not just those interested in crime analysis.

Mapping riot locations using publicly available data

Following the riots in England, the Guardian newspaper have mapped the locations of some of the people charged with related offences and have compared them to deprivation data (the Index of Multiple Deprivation). This is an interesting and extremely relevant use of publicly available data.

Project source code is available

We have just uploaded the Java source code for the GeoCrimeData project to a repository:

http://sourceforge.net/projects/geocrimedata/

At the moment the software is able to read in spatial data and calculate the mean path length for each road in a network. Mean path length is a measure of how easy it is to get from one road to every other road; roads with a low path length have been found to correlate with high vehicle and pedestrian traffic.

The image below shows the mean path length of roads using Open Street Map data for parts of Leeds, calculated using the software. It appears that 'major' roads (such as the motorway running through the centre of the map) have lower path lengths than the surrounding minor roads as we would expect (a low/green value indicates higher traffic volume). There are preliminary results and the software needs some tweaking, but hopefully a measure such as this should be a useful statistic for people who are interested in estimating how permeable/busy/accessible a street or neighbourhood is.

The next stages will be to improve the measure and run the software on UK data (rather than just a small area in Leeds).

Results of a Survey of Crime Analysts

We have recently conducted a survey of ~30 eminent crime analysts from the UK and USA (as well as one person in Sweden!). The survey presented six maps which demonstrated some of the potential uses of geospatial data for crime analysts and asked the experts to give us their opinion on what they found useful/interesting about each of the new data sources.

 As an example, the figure left shows estimates of road traffic volume that were produced by looking at the structure of the local road network (using OS MasterMap data) and comparing this to occurrences of residential burglary. Analysts were asked whether they were familiar with this type of analysis, whether it would be useful for them to be able to easily access this type of data (or not) and how we might improve it.

The responses were extremely comprehensive and will be very useful in helping us to make sure that the products we produce will actually be useful to practitioners.

Forthcoming conference presentations

We are going to a number of academic conferences in the coming months to present ongoing work on the project. The conferences that we'll be attending are:

• Environmental Criminology and Crime Analysis (ECCA) in South Africa
• European Society of Criminology (ESA) - 21st to 24th September 2011, Vilnius, Lithuania.
• European Colloquium on Quantitative and Theoretical Geography (ECQTG) - 2nd to 5th September, Harokopio University of Athens, Greece.
• Regional Science Association (RSAIBIS) - 6th to 8th September, Cardiff.

 For more information about the presentations, see the project homepage.

Crime and the social mosaic

The very existence of geospatial data enables us to ask much broader questions about the context in which crime occurs. For example, rather than just identifying some areas as having higher crime than others we are able to move the discussion forward to ask questions about what types of area have high crime. Are they socially disadvantaged? Do they have high concentrations of particular types of housing? Are they student areas?

But geospatial data sets also enables us to ask questions about what is next to a particular area and what effect that might have on crime within an area. For example, do poorer areas have higher crime if they are also  surrounded by poorer areas? Do they have lower crime if they are next to affluent areas ? Do  affluent areas surrounded by disadvantaged areas have higher crime than affluent areas bordering middle or high income areas?. Geospatial data gives us the opportunity for the first time to explore the extent to which an area’s crime rate is affected not just by the crime risks within the area (i.e .the housing, natural surveillance, offender population and community cohesion) but by the characteristics of its neighbouring areas. This truly is breaking new ground!

Initial data analysis

At this stage of the project we are investigating the types of spatial data that are available and how they might help to provide contextual information to supplement crime data. For example, the figure below shows some burglary rates in two areas that are very close to each other. We have used the OS MasterMap Topography layer and spatially analysed each of the buildings to try to work out whether they are terraced, semi-detached, flats or detached.

In general, more crimes are committed in the area labeled "B" than in "A". It also appears that the houses in area B are very similar to each other; most are terraced houses of similar design. Therefore it is possible that if someone is able to burgle one of the houses in area B, it is easier to burgle others as their design is so similar. The same is not true in area A which exhibits many different types of house design.

Of course there are a number of other explanations for the difference in the crime rates between the two neighbourhoods, but undoubtedly the physical environment will have an effect. Therefore this type of information might be extremely useful to crime analysts or other people who are interested in understanding the context behind crime rates. The GeoCrimeData project will continue to explore these types of data and perform analyses that might reveal interesting spatial information.

Radio Interview with Professor Hirschfield

Alex Hirschfield was recently involved in a discussion about the new police.uk crime mapping website. The discussion was broadcast live on the Andrew Edwards Drive Live Show on BBC Radio Leeds at 5.30pm on 1st February 2011 and is available here.

During the discussion it becomes clear that although crime data is an essential part of the crime analysis toolbox, we cannot understand crime just by looking at crime data - we need to understand the 'context' within which crime occurs. Is there something about land use or the population out there in the 'environment' that might explain why crime levels appear to be high or low?

This is at the heart of the GeoCrimeData project, we will analyse existing environmental data (such as road networks and land-use datasets) to add context to observed crime patterns.

Budget

The following table illustrates how the budget has been allocated

Item	Amount of budget allocated (%)
Staffing, estates etc.	93
Dissemination (conferences) and travel	7

Projected Timeline, Workplan & Overall Project Methodology

Work Packages

The project will progress according to the work carried out in the following work packages. Mark Birkin is responsible for the overall project methodology.

WP1: User geospatial awareness, needs assessment and initial data exploration

The initial stage in the project is to engage with end users to explore their awareness of geospatial data, current levels and types of use and future analytical needs. This will be carried out using an online questionnaire that will be circulated to academic institutions and practitioners (police, community safety partners etc). Concurrently, initial data exploration will take place in order to identify potential data sets that can be used in the project.  Andrew Newton and Nick Malleson are responsible for surveying users and collecting data respectively.

WP2: Data Access and the Creation of New High Resolution Data for EC and Crime Analysis

This stage will involve accessing relevant geospatial datasets and then analysing/modifying them to meet the needs identified in WP1. These tasks will require new methods to be developed in order to spatially analyse the input data sets and derive new data. Nick Malleson is largely responsible for WP2.

WP3: Case Study Pilot

WP3 will document a scientific case study using the data generated in WP2. The purpose of the case study will be to exhibit the new data sources to the community and demonstrate how they can be used to inform crime analysis. Andrew Newton is largely responsible for WP3.

WP4: Validation and Sustainability

The goal of WP4 is to evaluate the usability of the new datasets through interviewing stakeholders (identified in WP1). This is work will be carried out by Nick Malleson and Andrew Newton.

Work Package Timeline

Gantt chart depicting timescales of each work package