Serial crime investigations usually involve more information management than singular incidents, said Inspector Glenn Woods, of the Royal Canadian Mounted Police (RCMP). "If a crime is committed for financial gain, there is a motive. If it is a violent domestic crime, there is a connection, a focus for the investigation. In serial crimes, however, the offender and the victim are strangers; it is the random crossing of the two that contributes to the death of the victim."
Linking Analysis
Linkage analysis systems are used by the FBI and large state agencies, RCMP and by Europol. The Homicide Investigation and Tracking System (HITS) used by the Washington State Attorney General's Office (WSAGO) is an example of this technology. HITS was developed in 1987 by Dr. Robert Keppel and Robert LaMoria, of WSAGO, and Dr. Joseph Weis, of the University of Washington. The system continues to evolve with advancements in technology and techniques -- with the addition of crime-related databases and links to vice and gang files, sex-offender registries, corrections and parole records, and department of motor vehicles databases. Although a single system, HITS works in conjunction with all of these by being able to simultaneously scan multiple databases, check and cross-check files, create datasets and timelines, delete redundancies and check itself.
The effectiveness of the system, said HITS Manager Robert LaMoria, largely depends on timely and accurate input of cases from the different agencies. Citing a 1994-95 serial rape case, he said, "When we received the information, the agencies in whose jurisdictions the rapes were occurring were reporting six attacks. We were able to link four more to the same offender, for a total of 10. After running the data through linkage analysis, we had over 1100 suspects, all within a 10-mile radius of the crime sites. We narrowed those down to five, profiled the first three and came up with a priority suspect. The investigators agreed. The agencies pulled in their proactive people and started a surveillance on the individual. Two nights later, they caught him attempting another rape. He was convicted on all charges -- some he confessed to, others he was linked to through evidence."
To make HITS even more effective, engineers are currently incorporating two recently-declassified Artificial Neural Networks (ANN). These smart networks are spin-offs from cold-war technology developed by the CIA, Boeing and scientists at the Hanford Nuclear Research Laboratory. ANN uses algorithms and statistical data and has up to several thousand interconnected processors. It functions much like neurons do in the human brain; it is capable of learning from mistakes, adapting, being trained and even training itself. To evaluate the effectiveness of the networks, engineers are making them analyze hundreds of case files for connections and matches in an effort to determine which is smarter. The results will be compared with the intuitiveness of the best criminal investigators in the field. The incorporation of ANN will likely eliminate all or most labor-intensive research by automatically gathering and analyzing data and providing profiles.
Will HITS replace investigators? LaMoria said no. "I don't think you are ever going to be able to take the human element out of this process, but investigations are going to become much faster, much less labor intensive."
LaMoria emphasized that HITS in its present form is far too expensive for law enforcement. "Washington had the good fortune to inherit the original mainframe and processors from discontinued federal operations and the neural networks through a grant from the National Institute of Justice," he said. To make the system cost effective for smaller agencies, it is being rewritten to function as a client/server system. The new version is expected to be operational by the end of 1998.
Geographic Profiling
Geographic profiling, a relatively new development in the field of environmental criminology, analyzes the spatial patterns of serial crimes and maps out the most probable location of the suspect's home. The original program was developed in 1990 by Detective Inspector Kim Rossmo of the Vancouver Police Department (VPD) during his doctoral research at Simon Fraser University (SFU) in British Colombia, Canada. Earlier studies by SFU criminologists Paul and Patricia Brantingham found that most offenders operate fairly close to home, and that "crime trips" decay with distance; the offender prefers to commit crimes closer to home. Their work produced a model based on where the suspect lived, approximating the most likely areas where he would operate.
Rossmo reasoned that if the serial criminal operates close to where he lives, it may be possible to approximate the location of his home by analyzing spatial patterns of the attacks. To test the theory, he wrote a program using proprietary algorithms and statistical analyses of serial crime data collected over several years. The program also incorporated a hunting typology of serial criminals based on the research. When later reviewing George B. Schaller's authoritative work, The Serengeti Lion, Rossmo found startling similarities in the hunting patterns of African lions and city-dwelling predators.
"Schaller's typology of how lions hunt matched almost perfectly with the hunting typology of the serial killer. Lions look for an animal that exhibits some indication of weakness -- the old, the very young, the infirm, the vulnerable. They will go to a watering hole and hang out because they know it is a draw for their potential targets. We see that all the time with criminal offenders; they go to target-rich environments to do their hunting. Spatial patterns are produced by serial killers as they search and attack. The system analyzes the geography of these, the victim encounter, the attack, the murder and body dumpsites."
Rossmo tested the research program by analyzing hundreds of serial and multiple crimes in Canada, the United States, the United Kingdom and Europe. The locations of offender residences produced by the program corresponded well with the actual addresses. By 1990, investigators in Canada, the United States and Europe began asking Rossmo for help on operational cases.
Development
The program attracted support from The National Research Council of Canada, Simon Fraser University, the RCMP and VPD. In 1995, a cooperative venture with Environmental Criminology Research Inc. and Facet Decision Systems Inc. produced the first commercial version of the program. An expanded version, Rigel Geographic Profiling, followed in September of this year. Three agencies currently have the Rigel system: VPD, RCMP and the Ontario Provincial Police. Several other agencies have requested that Rossmo assist them with geographic profiling, including Scotland Yard, the FBI, the Los Angeles County Sheriff's Office and the New York Police Department.
Rigel analyzes spatial and temporal patterns associated with a series of related crimes to determine the most probable location of the suspect's home. The system incorporates GIS technology, a graphical user interface and a series of open database connectivity modules that allow trained profilers to store and manipulate data from a wide range of sources. Each crime series is stored in a file that can be modified or updated at any time; the system has the speed to create different scenarios with nearly instant results. It can analyze the hunting behavior of the serial offender and graphically indicate if a series of crimes may have been committed by the same suspect, if there are connections between the spatial and temporal patterns of the offender, and if he might have moved to another location. Rigel works with map overlays, digital orthophotographs and geocoded or lat/long locations. It runs on the new Sun UltraSPARC line of workstations.
Geographic profiling begins with a map of the area under study. The profiler enters the temporal and spatial coordinates of the specific crime sites and other case-related data. The next step is the calculation of the geographic patterns using the criminal-hunting algorithm and the locations of the crime sites. Once the profile has been developed, the profiler searches linkage analysis datasets, or individual databases, for connecting data. The latter may include police records, sex-offender registries, DMV files, postal/ZIP codes and records of 911 calls -- all have spatial components. Geographic profiles are displayed as 2D or 3D surface maps -- referred to as jeopardy surfaces -- depending on the need for topographic perspective. Different colors indicate the statistical probabilities assigned to each section of the map. When an area is targeted, the police may mail a description of the suspect to the residents or move in directly to canvas the neighborhood.
Combining Systems
Rossmo and others agreed that geographic profiling is most effective in conjunction with psychological profiling and linkage analysis systems -- HITS, VICAP (FBI's Violent Criminal Apprehension Program) and ViCLAS RCMP's new Violent Crime Linkage Analysis System).
Data sharing is the key, said Woods. "If you have murders in Los Angeles, Reno, New York City and El Paso, unless these jurisdictions are entering the cases into a central system, they may never be recognized as being linked to the same offender. That is what linkage analysis does, and that is why geographic profiling works so well with it; not only can we see links between cases, we can often get characteristics and traits through the profile and focus on the individual's anchor point."
In a recent case involving a serial arsonist, RCMP asked Rossmo to do a geographic profile of the suspect in conjunction with the psychological profile that they had developed. The resulting analysis enabled investigators to focus on a very small geographic area where the suspect lived. An individual who fit both profiles was subsequently prioritized as the number one suspect. But before the police could move in to canvas the area, the arsonist was caught trying to set another fire. His actual home address turned out to be across the street from where the geographic profile had estimated.
Geographic profiling has also been effective in single crimes with multiple sites or events. The recent Abbottsford Killer case in British Columbia involved two teenage girls who were attacked together on the street. Shortly afterward, the killer began making taunting phone calls to the police, daring them to catch him. The locations of the calls were traced through the enhanced 911 system, but because the killer always hung up after a few seconds, the police could not reach the phone booths in time. Those locations, however, were being entered into the geographic profiling system. By the time the killer racked up 10 calls, Rossmo produced a map of the area where the caller lived. Although he was caught and convicted through a special police call-in arrangement -- a woman identified the caller's voice as that of her son -- geographic profiling estimated the location of the caller's home within three-fifths of a mile from where he actually lived.
Not A Panacea
Woods cautioned against the notion that these computerized systems are simply a matter of holding data, pressing a button and spitting out an address. "The fact is," he stated, "they are a way of managing information, particularly when you get into large investigations involving serial homicides or rapes. Geographic profiling enables you to focus the investigation in a small area of the community, rather than on the whole metropolitan area. It cuts the time required to do the initial part of the investigation, and it reduces the amount of resources required to get involved in cases of that magnitude." Woods estimates that one system may be sufficient to serve an entire state or region. To learn geographic profiling, individuals from law-enforcement agencies in Canada and the United States understudy with Rossmo in a rigorous 18-month training program.
Rossmo stressed that geographic profiling is an investigative tool designed to work together with other knowledge-based systems. "When they are applied together, the power of each becomes greater. I think it's going to be a very exciting time in the next few years as these systems begin working together."
Bill McGarigle is a freelance writer specializing in communication and information technology.
December Table of Contents
