Chapter 15 Crime Mapping Crime Mapping Methods Police departments are never at a loss for data. To use crime mapping is to take data from myriad sources and make the data appear on the computer screen or a published map. The following section explains one of the most popular techniques in crime mapping for converting tabular data files into mapped data: address matching. Assigning Spatial Locations to Events (Address Matching or Geocoding) We use a GIS to visualize data, specifically spatial data. GIS software has tools to turn tabular information with some form of spatial reference into data on a map. Once that data is on the map, we can commence data retrieval, manipulation, and analysis. Geocoding by definition is the process of converting analog data into a computerreadable format. In law enforcement, most references to geocoding refer to one type of geocoding: address matching. There are other methods for converting analog or real world data into a format to be read by computer software. For example, most GPS receivers have the ability to record and store locations in a file that can be downloaded to the computer and converted into mapped data. Did you ever wonder how all the map data you see on the Internet or available for sale on CD-ROM came into existence? GIS technicians have spent countless hours digitizing paper maps using a precision mouse and a digitizing table. A digitizing table is like an electronic drafting table where the surface is underlain with an electronic grid that enables the person doing the digitizing to trace map features on a map taped down to the table top with a precision cursor. Locations are identified, converted into numbers including coordinates, and then transferred to the computer. Address matching is the ability to match an address component to its geographic location on the ground. 1 At a minimum two things are required: a street database and a database containing addresses such as crime records in the records management system. In a perfect world we would have a feature on a map corresponding to every possible address. The reality is that most of us record address ranges associated with street segments. The high and low address range is linked to each street segment. GIS software then locates specific addresses in our crime data approximately along a street segment. If your incident address was 150 Main Street and the street segment represented the 100 block with addresses ranging from 100 to 199, the GIS software would locate the address of the crime approximately 50 percent of the way down the block. Please note that this is not a precise location. The illustration below comes from TIGER file technical documentation provided by the United States Census Bureau. Figure 15-8: How address matching works. Source: U.S. Census Bureau, 1997. 1 Glen Montgomery and Harold Schuch, The GIS Data Conversion Handbook (Fort Collins, CO: GIS World, 1993). 325
Part 3 Tools of Crime Analysis Address Matching Issues The crime mapper must understand how address matching works and the inherent weaknesses associated with the technique. 1. Street map data quality: Imagine the difficulty associated with keeping street networks up to date in a GIS. New streets must be added to the street database while vacated streets must be deleted from the streets database. Street name changes are common as well. In older, more stable areas, the task may be minimal. In high-growth areas the task is large. 2. Accurate address assignment: In addition to keeping streets in the database, the address ranges must be correctly assigned. This task sounds simple, but is not. Cities and counties go to great expense to put crews in the field to verify correct address assignments along street segments. 3. Street aliases: Streets often have alternative names or local names. For example, interstate highways have an official numeric designation, but may also have a local name such as the Blue Star Memorial Highway. Abbreviations may be common such as MLK rather that spelling out Martin Luther King Boulevard. 4. Apartment buildings: Apartment buildings generally have a single street address, but there are many units associated with that one address. 5. Street type abbreviations: Over time many abbreviations have been used to abbreviate street types. For example, the abbreviation for avenue may have been recorded as AV or AVE. 6. Police data: How well are addresses recorded in your police data? Are there typographical errors? Are there spelling errors? Is there missing data? Are addresses listed approximately such as in the 100 block of? Data integrity is paramount in crime mapping. For that reason, an entire chapter in this text has been devoted to the topic of data integrity. Please see Chapter 7 for additional information about the importance of data integrity in law enforcement. Visualizing Events on a Map Visualization, for the purpose of this text, will be adapted from MacEachren and defined as the use of maps, whether on paper or computer, to make spatial contexts and problems visible to the map reader. 1 In crime mapping, the first objective is to get events on the map. In previous sections we ve discussed some of the considerations for the geographic base and the crime data. Once data has been address-matched, or geocoded, the crime analyst is confronted with points or dots on a map. The initial results may be overwhelming or they may be revealing. Certainly, having too many dots on the map leads to information overload and there is no reasonable expectation that an analyst can discern patterns from the information. Another issue that arises is when dots on the maps representing events in the database are co-located. In other words, two or more events have taken place at the same location. In GIS software, the dots lie on top of each other masking the fact that there are two or more features there. 1 Alan M. MacEachren, How Maps Work: Representation, Visualization, and Design (New York: Guilford Press, 1995) 326
Chapter 15 Crime Mapping Spatial Queries Early on in this chapter we defined a GIS as having the ability to retrieve information. Retrieving data via a map is known as a spatial query, and there are several different types of spatial queries. GIS software includes a database management system, usually a relational database management system (RDBMS), to store attributes of the spatial data. GIS software also incorporates the capability of importing data from other database sources or accessing other data in database systems via links or connections such as Open Database Connectivity (ODBC). Database management systems support the concept of query to isolate a subset of data. Find is the simplest of attribute searches, usually with the intention of locating a single record. A restrict query enables the crime mapper to retrieve a subset of data by placing a restriction on the attribute of the data. For example, one could query a crime database by searching for all crime in a given month. In a database, we have the ability to find or restrict our queries, not solely based on descriptive attributes (such as date, or type of crime), but rather on geographic features. Geographic or spatial features have special geographic attributes, such as coordinates and measures, length, area, and perimeter, that enable spatial queries. Adjacency: Adjacency is the topological property of having an edge or boundary in common. Features sharing a boundary are in immediate proximity. You may ask: What jurisdiction is adjacent to mine? What other beats border Beat 12? On what side of the city boundary did the crime occur? Are our two rival gangs adjacent to each other? Proximity: A buffer is a zone around a point, line, or polygon that is spatially related to the feature. Buffering is a popular technique in crime mapping. Creating buffers is just one technique for determining proximity. Points, lines, and polygons can all be buffered with GIS software. Creating a buffer merely entails determining what the radius, or proximity, should be around a specified feature. For example, if you wanted to determine a 1000-foot drug free zone around a school you select the school and use the buffer tool in your software. If your school is shown as a point feature in your map base, your result will be a circle with a radius of 1000 feet. If your school appears as a polygon more closely representing the school yard in your base map then your resulting buffer will be a polygon that resembles your school yard only much larger and with rounded corners. With buffering we can answer questions such as what is near by or what is the relationship between registered sex offenders and their proximity to schools? GIS gives us the ability to manipulate data. Buffering is a manipulation of data. Creating a buffer creates new data that can then be used to do further analysis. After creating a 1000-foot buffer around your school you now have a new polygon in which you may want to query how many drug dealer suspects live or work within the buffer polygon. Area: Area is the measure of a planar region. How big is, or what is the area of my police beat? How does the area of my beat compare with others? Length or Distance and Connectivity: What is the length of a road segment? How long will it take me to travel from one end of town to the other on this road based on its length? Can I get there from here? 327
Part 3 Tools of Crime Analysis Figure 15-9: A 1000-foot buffer around a school Levels of Measurement Levels of measurement are the degree of subjectivity associated with measurement. The attributes associated with spatial data can be classified according to their level of measurement. There are four levels: nominal, ordinal, interval, and ratio. 1. Nominal: the root of the word nominal is nom which simply means name. A nominal classification of data provides the location and the name of the feature. 2. Ordinal: the word ordinal comes from order and implies rank. For example, small city, medium-sized city, and big city would be an ordinal classification. You use small, medium, and large map symbols or dots to illustrate, but you cannot state how much bigger one city is than another based on symbology alone. 3. Interval: interval data provides numeric information for different classes of data. A light crime area may have 1 to 10 burglaries; a high crime area may have 11 to 20 burglaries. Interval data usually begins measurement from a datum or point of beginning; in this example, zero. 4. Ratio: ratio data is, as the name implies, a ratio of one value to another. For example we may classify data to look at the percent of crimes to total population for a reporting district. 328
Chapter 15 Crime Mapping Figure 15-10: Crime maps using different levels of measurement Data Classification The purpose of data classification is to simplify complex geographic distributions into a presentation of the data that is meaningful to the reader. Data classification involves organizing data into groups or clusters based on a numeric or statistical distribution. There are two main components in a classification scheme: the number of classes into which the data is to be organized and the method by which classes are assigned. The number of classes is dependent on the objective of the analysis. Too often in crime mapping crime analysts rely on the default settings in their software for data classification. The natural breaks is the typical data classification default. Data classification can be controlled by the crime analyst and therefore he or she should use a variety of data classification techniques. Different techniques will yield different views of the data. There are several common techniques available in most GIS software packages. Natural Breaks: Most GIS software packages use the natural breaks technique as the default data classification option. This is a good technique for determining natural groups or clusters of values in the data. The technique looks for the breaks between groups of data values. Feature classes are set where there are relatively big jumps in the data. This technique is limited for time-series analysis such as comparing like time periods from one year to another because the range 329
Part 3 Tools of Crime Analysis of data may change significantly between years, making the distribution of data and therefore the natural breaks inconsistent between time periods. Quantile: Quantiles throw an equal number of observations or data points into each class as determined by the user. If the user sets four classes of data (quartiles), 25 percent of the data or records will be in each class respectively. This technique can be very misleading depending on the distribution of the data. A quantile classification is wellsuited to linearly distributed data. Figure 15-11: A choropleth map using a natural breaks classification Equal Interval: The equal interval technique breaks the range of data from low to high into a number of equal subgroups or classes as dictated by the user. For example, the number of crimes per beat ranges from 0 to 250. To determine five classes, the GIS takes the range, 250, and divides it by five (50), and makes the categories each contain 50 (e.g., 0 50, 51 100, 101 150). This is an excellent technique for making apples to apples comparisons between data of different time periods, as long as the user sets up the same equal intervals in period one as in period two. Figure 15-12: A choropleth map using an equal interval classification Figure 15-13: A choropleth map using a quantile classification Standard Deviation: This technique should only be used when the data approximates a normal distribution. This technique calculates the mean of the data distribution and then maps one, two or three standards deviations above and below the mean. Crime Maps The following sections examine common crime mapping techniques to assist the new crime analyst in focusing on common and popular cartographic techniques that are useful in the law enforcement community for visualizing and interpreting, spatially, crime in their jurisdiction. Most cartographers will identify three types of maps. General maps are general in nature, providing locations of data with place names attached. The street map in the glove compartment of your car is a general map. A second type of map is a chart. Those of you 330