Findings support other crime-mapping sur-veys by several governmental crime-research agencies Crime Mapping Research Center, 1999; Police Foundation Crime Mapping Laboratory, 2000, where
Trang 1Crime Map Analyst: A GIS to Support
Local-Area Crime Reduction
Paul Brindley, Max Craglia, Robert P Haining,
and Young-Hoon Kim
CONTENTS
6.1 Introduction 113
6.2 Current Crime Pattern Analysis 115
6.2.1 GIS Crime Systems in the United States 115
6.2.2 Past Crime Mapping and Analysis Research 115
6.2.3 Background to Crime Map Analyst 118
6.3 Overview of Crime Map Analyst 118
6.3.1 Density Maps 119
6.3.2 Repeat Victimization Identification 121
6.3.3 Temporal Analysis 122
6.3.4 Area Profiles 124
6.3.5 Origin=Destination Analysis 125
6.3.6 Ancillary Tools 126
6.4 Conclusions 127
Acknowledgments 129
References 129
The importance of geographical information systems (GIS) for crime analy-sis, and strategic and tactical deployment of forces, has been increasingly recognized in both the United States and the United Kingdom This was forcefully endorsed by former New York mayor, Rudolph Giuliani, during his visit to London in February 2002
Trang 2Senior police officers are keen to learn from the New York experience while
Mr Giuliani visits London In the eight years he was mayor of New York crime plunged
The success was credited to CompStat, the computerised system which keeps track of week-by-week crime figures for each precinct, the basic division of the city’s police department
(The Guardian, 14th February 2002) CompStat is of course only part of a wider strategy of crime reduction, but
it makes the point that the regular analysis of crime for small geographical areas is crucial for the effective deployment of resources, monitoring and evaluating impacts, and sharing intelligence The increased emphasis by the Home Secretary on increasing detection rates by concentrating police resources into selected hot spot areas goes in the same direction
The ability to visualize and analyze the data geographically is at the heart
of GIS These types of systems are already widely used in the United Kingdom, but there are significant variations among the forces in extent and purpose of use (Weir and Bangs, 2007) There are therefore opportun-ities for using GIS more and better, with stronger integration to crime-reduction strategies both in the forces themselves and as part of the wider crime-and-disorder partnerships
GIS can add value to the data already held by police forces and become a more integrated tool in crime-reduction strategies There are two essential preconditions to make this happen
1 Geo-coded data
GIS can only operate effectively if the data to be analyzed have accurate and consistent geographical locations attached to it Although this may sound a purely technical matter, it is in fact a largely organizational one It must become a routine to report the location of crime events as accurately as possible, and against a standard gazetteer of locations Not all forces have adopted such practice and the assigning of coordinates to past crime data pro-vides context to the analysis
2 Awareness and training
Training staff in the use of GIS, or any new system, is of course time consuming and expensive The advantage however of adopt-ing off-the-shelf and widely used software is that there are already well-developed courses, training packages, and learning resources, and that there is a support network of millions of users on which to build This minimizes training costs and makes the most effective use of the investment made Perhaps, more crucial is ensuring that the necessary awareness exists among senior managers of the value of such investment, and that adequate support is provided
Trang 36.2 Current Crime Pattern Analysis
6.2.1 GIS Crime Syste ms in the United States
The importan ce given in the United Stat es to GIS for crime analys is is mos t clearl y demo nstrated by the work of the Map ping & Analy sis for Publi c Safet y (MAP S) program (formerly the Crim e Mapp ing Rese arch Center) at the Nationa l Institute of Justi ce (http: == ww w.ojp usdoj.go v =nij=maps ) This center was established in 1997 to promote, research, evaluate, develop, and disseminate GIS technology and the spatial analysis of crime The lessons learned out of the U.S situation are valuable in the continued development
of crime mapping in the United Kingdom
Personal contact using the CRIMEMAP e-mailing group (crimemap@lists aspensys.com) was undertaken to disseminate crime mapping and analysis e-mails to all subscribers to discover the main GIS used within U.S crime mapping The survey was conducted during 11–31 October 2000, and a total
of 93 contacts were collected Findings support other crime-mapping sur-veys by several governmental crime-research agencies (Crime Mapping Research Center, 1999; Police Foundation Crime Mapping Laboratory, 2000), whereby over 50% of crime analysis and mapping in the United States was undertaken using just two software applications—ArcView and MapInfo
6.2.2 Past Crime Mapping and Analysis Research
Since the 1990s, the extensive usage of GIS has enabled police forces to map and analyze crime data efficiently, facilitating crime data analysis (Hirschfield et al., 1995) Computerized mapping technology has broad application areas in various police fields including operational, analytical, and strategic policing (Craglia et al., 2000)
GIS functionality has become widely used in many areas within crime data analysis, such as crime hot spot mapping and cluster detection, repeat victimization, temporal pattern analysis of crime incidents, and police pol-icy making for crime reduction and prevention For hot spot analysis, Ratcliffe and McCullagh (1999) developed a methodology for detecting various hot spots using a kernel estimate function on the basis of a local spatial autocorrelation statistic (Local Indicators of Spatial Association, LISA) to identify statistical hot spot variation Crime cluster detection has been carried out within several current crime mapping tools such as STAC and CrimeStat (Bowers and Hirschfield, 1999; Levine, 1999; Craglia et al., 2000) Farrell and Pease (1993) recognized the issue of repeat victimization
as a main criminological problem and suggested an implementation strat-egy for preventing crime repeats Anderson et al (1995) also provided strategic guidance for police forces to tackle repeat victimization Johnson
et al (1997) demonstrated the relationship between repeat victimization
Trang 4and other soc ioeconom ic factors, and exp lored analyti cal me thods to iden tify the re lationship Howeve r, geo-co ding pro blems have a profou nd impa ct upon the reliabi lity of spat ial repeat vic timizatio n iden tificatio n (Ratcli ffe and McCu llagh, 1998a)
The tempo ral as pect has been iden tified as a crucial factor to mo nitor crime incide nt change Instea d of using general discrete methods (for examp le, usin g mi d-point of bet ween from -time and to-time interva l), Ratcl iffe and McCu llagh (1998b) introdu ced a prob abilistic rate techn ique bas ed on aori stic rul es to estim ate a truer rate of crime incide nts A fuller desc ription of this method ology will be discusse d in Secti on 6.3.3 Practi cal use of this method was und ertaken to explor e differen t tempo ral pattern s of crimes with in a nu mber of ho tpots (Ratcliffe , 2002) Spa tial st atistics have been increasi ngly appli ed to crime data analys is in ord er to enhan ce novel capa bilities of GIS-bas ed analy sis of crime, suc h as local spatial statisti cs for crime pat tern a nalysis (Craglia et al., 2000), urban crime exami nation (Mur ray et al., 2001), or det ecting tempo ral chang es of crime (Ro gerson and Sun, 2001) Anselin et al (2000) introduced the extensive discussi on of spat ial analytical techni ques and poten tial of GIS for crime analysi s In add ition, for exp loring the relations hip with socioeco nomic area pro file and crime incide nts, Bowe rs and Hirsch field (1999) demo nstrated an examp le of GIS applicati ons in crime pat tern analy sis, a nd Craglia et al (2001) re ported the strengt hs of GIS -based spatia l analysi s with censu s data for mo deling hi gh-intensi ty urb an crime areas Hirsch field and Bowe rs (2001) sum marized extensive research contr ibutions of GIS and their prac-tical poten tial in crime data mapp ing and analy sis
A varie ty of crime mapp ing syst ems, extensio ns, and softwa re pac kages for GIS have been develo ped at practica l level s, as summar ized in
Tab le 6.1 Some packag es were develop ed for pin pointin g crime even ts and creating thematic choropleth maps, whereas other software systems were developed for locating hot spots and exploring spatial relationship with other socioeconomic data For example, STAC was a frontier stand-alone hot spot and cluster analysis package developed in the 1980s and still is useful for crime cluster analysis (Craglia et al., 2000) In the 1990s, many mapping packages have been developed as extensions of main GIS commercial software systems using their customization languages such as SCAS, CrimeView and Crime Analysis for ArcView, and Hotspot Detective for MapInfo To improve their user interface, computer lan-guages and scripts have been integrated such as Visual Basic (SCAS, RCAGIS, and CrimeStat), and MapObject (RCAGIS and Community Policing Beating Book, and MaxResponder) As an alternative, crime-oriented stand-alone mapping software has been developed such as PROphecy and CrimeWatch However, there has been limited success to tackle crime data analysis for various levels of U.K police force require-ments Therefore, this chapter demonstrates some of the key functions of GIS crime analysis that can meet various operational, tactic, and strategic police performance in the U.K police force
Trang 5TABLE 6.1
Summary of Main Available Crime Mapping Desktop Software
Name Source Primary Functionality Spatial and
Temporal
Analysis of Crime
(STAC)
Illinois Criminal Justices, 1993 ( http:==
www.icjia.state.il.us=public=index.
cfm?metaSection ¼ Data&meta Page ¼ STACfacts , assessed 18th October 2007)
Hot spot analysis package
Spatial Crime
Analysis System
(SCAS)
CMRC, 1994 ( http:==www.usdoj.
gov =criminal , assessed 18th October 2007).
Query interface; analytical mapping and reporting; installation flexibility; minimum reprogramming (avenue)
Regional Crime
Analysis GIS
(RCAGIS)
CMRC, 1994 ( http: ==www.usdoj.
gov =criminal , assessed 18th October 2007).
Low cost ( <$100); analytical mapping and reporting; interface with CrimeStat and MapObject Community Policing
Beating Book
ESRI, 1997 ( http:==www.esri.
com=industries=
lawenforce =beatbook.html , accessed on 6th June, 2002)
Simple query; mapping and reporting functions; MapObject application Crime Analysis
Extension
ESRI, 1999 ( http:==www.esri.
com=industries=lawenforce=
crime_analysis.html , accessed on 6th June, 2002)
ArcView application using Spatial Analyst extension; Hot spot and cluster analysis
CrimeView OMEGA, 1999 ( http:==www.
theomegagroup.com=crimeview.
htm , accessed on 6th June, 2002)
Query; density mapping and simple analysis; reporting; integrated with ArcView MaxResponder ESRI, 1999 ( http: ==www.
maxresponder.com = , accessed on 6th June 2002)
Query and mapping function; mobile GIS mapping functions
Hot spot Detective Ratcliffe, 1999 ( http: ==athene.csu.edu.
au =jratclif=index.html , accessed
on 6th June, 2002)
Hot spots; aoristic temporal analysis
Repeat Location
Finder
Ratcliffe, 1999 ( http:==athene.csu.edu.
au=jratclif=index.html , accessed
on 6th June, 2002)
Spatial repeat victimization identification
ReCAP-SDE Virginia Institute for Justice
Information, 2000 ( http:==vijis.sys.
virginia.edu=home.htm , accessed
on 6th June, 2002)
A stand-alone package; data handling and chart reporting functions CrimeWatch Spatial Data Inc., 2000 ( http:==www.
spatial-data.com=pCrimeWatch.
htm , accessed on 6th June, 2002)
Database; geo-coding and reporting functions CrimeStat Levine, 2000 ( http:==www.icpsr.
umich.edu=NACJD=crimestat.
html , accessed on 6th June, 2002)
Well-suited stand-alone software; set of spatial statistical modules for crime analysis; hot spot and clustering functions; compatible of main GIS software packages PROphecy ABM, 2000 ( http:==www.abm-uk.
com=uk=index.asp , accessed on 6th June, 2002)
Hot spots; temporal analysis
Trang 66.2.3 Background to Crime Map Analyst
This work is based on research undertaken by the authors between 2000 and
2002 for the Home Office and South Yorkshire Police It illustrates the implementation of some key requirements for a crime data analysis pack-age, which is sensitive to the needs of U.K police forces and which adds value to the data the police collect themselves The requirements for the package related to both operational and strategic levels of policing and so the analytical functions are of use in both these policy contexts They build
on existing GIS facilities but also exploit crime analysis tools and data visualization functions, which are not addressed in the current GIS-based crime mapping systems The results were implemented as an extension to ArcView 3.2, as this was widely used at the time, and have since been ported to ArcGIS 9.2 Of course, these same functions can also be implemen-ted on other packages if required What is important is the methodology for analysis rather than the software platform on which it is implemented The next section describes the main functionalities developed in Crime Map Analyst
Crime Map Analyst (CMA) is an extension for ArcView that enables the user to undertake various functions useful for the analysis of crime data The functionality is outlined below (Figure 6.1) and is subsequently discussed in greater detail Particular importance has been attached to
(1) Density surfaces: hot and cold spots
(2) Repeat victimization identification
(3) Temporal analysis
(4) Area profiles
(6) Ancillary functions
FIGURE 6.1
The main CMA menu.
Trang 7displaying outputs from the different functional tools against maps that help the user to relate findings to ‘‘real’’ geography
6.3.1 Density Maps
Crime analysts are increasingly using kernel-smoothing techniques to visu-alize and interpret crime data (Williamson et al., 2001) The density function takes data that are represented as points on a map and creates a continuous smooth surface of intensity CMA has the option to construct simple, quad-rat densities as well as more sophisticated kernel density maps
The quadrat density method employed within CMA places a grid over the area of interest It then simply counts the total number of points that fall within the defined search radius from each grid cell centroid This method does not take into account the fact that crime incidents are not randomly distributed and association between point locations may exist Crime events (points) frequently occur at the same locations (repeats) and have a ten-dency to cluster at specific locations
In comparison, kernel densities have several practical benefits for creating density maps (Bailey and Gatrell, 1995; Anselin et al., 2000) Significantly, considering spatial repeats and crime clustering, the method takes into account autocorrelation so that points are weighted according to their distance from the grid cell centroid Also, they do not have to adhere to socially constructed boundaries (for example, police beats, enumeration districts, or wards) like thematic maps that are using administrative bound-aries They are free from such constraints of shape unlike administrative boundaries or hot spots derived using circles or ellipses Also, altering the number of grid cells and their dimensions can alter the spatial scale at which the data is being investigated to fit the needs for which analysis is required
A further benefit is that the levels of crime intensity are easily visible, unlike with pin mapping techniques where crimes at the same location obscure the intensity of crimes at the same location, frequently causing data overload (Block, 1998) However, kernel densities are dependant on appropriate parameters used In particular, the search radius is critical because it determines the level of smoothing applied (Anselin et al., 2000) All the parameters within CMA required before creating a kernel density are determined by the extent of the area of interest This makes the process very simple for a user with little or no understanding of the method to produce meaningful results The parameters can be accessed within CMA
so that more experienced users can adjust them according to their exact need The user simply defines the area of interest by drawing a grid over the desired location The grid cell sizes are set so that the smallest extent (either horizontal rows or vertical columns) is set at a minimum of 400 grid cells and the other, larger extent is then scaled in the same proportions The search radius is predefined as a percentage of the perimeter of the area defined (default 0.4% of the perimeter) The default percentage used is
by no means the perfect solution, and it was considered outside the scope of this project, especially as more experienced users can easily adjust the search
Trang 8radi us to impro ve resu lts for their specif ic data However , the default per-centage was fou nd to provide good res ult given a variety of spat ial scal es
By comp aring dens ity map s for dif fering time periods , it is possibl e to iden tify hot and cold spot crime locat ions as illustrat ed (Fi gure 6.2) Thi s func tion is also of value in providin g a syno ptic v iew of recent chan ges and hel ping to evaluat e the impact of differen t poli cing strategies
A nove l featu re of the syst em is that it enab les the analyst to look in grea ter detail at parti cular locat ions on the map The user iden tifies an area
of interest with the mouse, and the De nsity-Zoo m-in func tion create s a new grid with as mu ch detail as requ ired (Figur e 6 3) This dynam ic zoomin g capa bility is particul arly useful in exp loring the areas of high densit y with in larger scaled hot spots
Summary view: areas of significant increase (black) or decrease (light grey)
FIGURE 6.2
Construction of hot and cold spots for domestic burglary in South Yorkshire between 1998 and
2000 (From 1991 Census: Digitised Boundary Data (England and Wales).)
Trang 9It is also possible to create densities based on a population field This can
be used to weigh some points more heavily than others, or to allow one point to represent several observations For example, one address might represent a condominium with six units, or some crimes might be weighted more severely than others in determining overall crime levels
A specific function in the menu (contained within the ancillary functions) has been created to facilitate the display of density hot spots over raster images, without obscuring the background This helps the analyst to recog-nize areas and focus on patterns of events in terms of specific areas or neighborhoods
6.3.2 Repeat Victimization Identification
Repeat victimization is where the same offense occurs a number of times against the same victim (be this a person or entity, such as a house or vehicle) According to Farrell and Pease (1993), 10% of the victims
of crime account for up to 40% of crimes in a given year The rationale is
to reduce crime by targeting prevention at repeat victimization loca-tions However, defining repeat cases is notoriously difficult (Ratcliffe and McCullagh, 2001) CMA uses two different functions to identify repeats: spatially or textually defined repeats
When performing a search for spatial repeats, the program examines a set of points to find those with the same geo-coordinates This detects repeats at the same place There is also an option (radius) to search for points that are in a close proximity to others This may assist users if the geo-referencing of offenses is not entirely accurate Another use of the radius tool is to search for ‘‘clusters’’ where numerous offenses are committed over
a limited geographical space (e.g., within 50 or 100 m from the identified location)
To investigate repeats for the same person (such as a vehicle owner) or entity (such as a vehicle registration) a textual repeat finder can be employed, which searches for identical text strings within a field This is
Low intensity
High intensity
Motorway
A roads
S
E W
FIGURE 6.3
Dynamic zooming capability (From 1991 Census: Digitised Boundary Data (England and Wales).)
Trang 10of course a standard facility on any database package, but the advantage of this method within a GIS package is that the outcome is immediately localized on a map There are opportunities to extend this function to explore spatial patterns of offenses based, for example, on the description
of the offender This would be of particular value in the current priority to reduce street crime
A third function reveals if the events are within a certain time period of each other (e.g., within 6 months), adding time as a variable in the identi-fication of repeat victimization A comparison of the results using the diffe rent iden tification pro cedures is shown in Figures 6.4 and 6.5 The combination of spatial and textual repeat functions can also assist in assess-ing the quality of the data For example, it may highlight instances in which different addresses may have been given the same location or identical addresses appear ‘‘displaced.’’
6.3.3 Temporal Analysis
Police records attach a start time and an end time to each crime event This is
to characterize the uncertainty that exists over exactly when the offense occurred Most current analysis tends to employ some measure of the middle point between the start and end times A more accurate method— termed aoristic search—considers that the offense might have occurred at any instance between the start and end times (see Ratcliffe and McCullagh, 1998b for further details)
FIGURE 6.4
Spatially defined domestic burglary repeats (ß Crown Copyright=database right 2007 An Ordnance Survey =EDINA supplied service.)