Geographic Profiling: Psychology, Spatial Behavior, and Investigative Use
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Geographic profiling is an investigative method that uses the locations of a potentially linked series of crimes to estimate areas where an offender may have a meaningful base of activity. Its central question is spatial: given the locations of events attributed to one offender, which areas deserve further investigation as possible home bases, workplaces, or other regularly used anchor points? The answer is usually a ranked geographic search surface, not a person's identity, address, psychological diagnosis, or finding of guilt.
The method draws on environmental criminology, research on everyday mobility, offender decision-making, geographic information systems (GIS), and statistical modeling. It is particularly relevant when investigators have multiple reliable event locations and a defensible basis for treating the events as related. The U.S. National Institute of Justice (NIJ) describes its purpose as providing information about a possible offender base of operations and emphasizes evaluating software against solved-case data and simple control methods Rich and Shively, 2004.
A geographic profile is an inference about where investigative attention might be allocated. It is not evidence that a resident of a high-scoring neighborhood committed a crime. Its value depends on the accuracy of the event data, the assumptions connecting place to behavior, the quality of independent evidence, and the way analysts communicate uncertainty.
What Does Geographic Profiling Mean?
In forensic and investigative psychology, geographic profiling refers to the analysis of crime-site geography to generate hypotheses about an offender's anchor point. An anchor point is a location around which a person's routine activity is organized; it may be a residence, job, relative's home, lodging, or another repeatedly visited place. The term “profiling” can be misleading: the output is usually a spatial prioritization rather than a description of personality.
Crime mapping answers where incidents occurred and how locations cluster. Geographic profiling attempts the harder inverse question: where might the person responsible for a linked series be based? General crime-hotspot analysis examines concentrations of events, sometimes from many unrelated offenders. Geographic profiling generally begins with an offender-specific series. Broader reverse spatial-pattern models can estimate likely offender locations at the neighborhood level, but that is a distinct modeling objective Hipp, Boggess, and Chamberlain, 2022.
The distinction matters because a map may be mathematically precise while the question behind it remains uncertain. A crime hotspot, a search-priority surface, a suspected residence, and a forensic conclusion refer to different kinds of information.
The Psychology of Spatial Behavior
Activity spaces and environmental familiarity
People encounter opportunities through their activities, including employment, commuting, shopping, recreation, and social relationships. The places they know, the roads they use, and the time of day they travel shape what they can notice and reach. Geographic profiling borrows this idea of activity space: behavior is often spatially organized because exposure to environments is spatially organized. Canter's account of investigative psychology places geographic behavior among several behavioral inferences that require research-based testing rather than intuition alone Canter, 2004.
An offender's routine geography can include several separate nodes rather than one fixed center. A person might leave home for work, encounter a potential target during another errand, and travel onward before an offense. The journey therefore need not begin at the residential address, nor end there. The meaning of a crime-scene location also changes when the recorded point is an initial encounter, the offense itself, a vehicle location, or a body-disposal site.
Distance decay: a population pattern, not a rule for every person
Distance decay describes a tendency for travel-related behavior to become less frequent as distance or travel cost increases. In crime research, some offense types and samples show more crime trips near familiar anchor points than far away. Researchers can estimate a journey-to-crime function from previously solved cases and use it to inform an unknown series. But the function is dependent on how distance was measured, which offenders were sampled, available target opportunities, transportation, and the definition of a trip.
A study of theft-related journeys reconstructed travel for 98 offenders and 449 offenses and identified biases in assuming a simple home–crime–home path; it reported substantial variation by offense and limitations in police location data Michaud and Proulx, 2023. A generic distance-decay curve should therefore be treated as a starting model whose fit to a particular setting needs examination.
Buffer zones: an especially uncertain assumption
A buffer zone is a hypothesized region around an anchor point in which offending is less frequent, perhaps because a person wishes to avoid recognition. Opportunity geography offers another explanation: there may simply be fewer relevant targets close to a base. Establishing either explanation requires information about where targets could have been encountered and how the offender actually moved.
Rossmo and Wheeler's 2024 simulation and case-study analysis found that identifying the shape of an individual offending-distance distribution, including a buffer-zone pattern, can require 50 or more crime trips. This is a requirement for reliable inference about that distribution's shape, not a universal threshold that a police case must meet before any spatial analysis can begin Rossmo and Wheeler, 2024. A visual ring on a map is insufficient proof that a person deliberately avoided nearby locations.
Context makes a difference. In a study of 30 South African commuter serial sexual offenders, Glass and Herbig reported weak support in that particular sample for several conventional distance-decay and buffer-zone assumptions. Those findings challenge universal application of models developed in other geographic and offender settings; they do not establish a new universal pattern for all sexual offenses Glass and Herbig, 2025.
Opportunity, transportation, and barriers
A short straight-line distance may involve a long journey if a river, highway, rail corridor, border, or limited crossing restricts movement. Traffic, public transportation, working hours, neighborhood land use, and where potential targets gather may influence observed clusters. Models that incorporate housing densities, barriers, and directional preferences illustrate why geography should be understood as an environment of unequal opportunities rather than a featureless plane Mohler and Short, 2012.
Population density can also alter the meaning of travel distance. Analysis of 1,186 stranger sexual-assault cases and 124 stranger homicide cases investigated differences associated with population density and with which crime-site location was used for measurement Almond and colleagues, 2022. Treating an urban result as a rural default, or vice versa, would confuse differing opportunity structures with supposedly stable personal traits.
How Geographic Profiling Works
1. Establish the analytic question
Analysts first identify the operational purpose. They may be trying to prioritize neighborhoods for reviewing existing leads, locate a probable base of operations, compare plausible anchor-point hypotheses, or decide where additional information collection could be most productive. These are investigative purposes. They should be stated separately from any subsequent decision about surveillance, a search warrant, arrest, or courtroom evidence.
2. Review whether incidents belong to one series
A map can be generated from any set of coordinates, even when the incidents are unrelated. The key prerequisite is a scientifically and factually defensible crime linkage hypothesis, supported by available investigative information. Behavioral similarities, time patterns, forensic traces, victim or target selection, and case information may contribute to linkage, but none should be assumed decisive. If unrelated events are mixed into the dataset, the geographic estimate may point toward an artificial center.
The literature on offender profiling shows why these problems deserve separate analysis. A large review found more developed statistical work on behavioral case linkage than on many forms of traditional offender-characteristic prediction; those findings cannot simply be transferred to geographic profiling Fox and Farrington, 2018. Geographic profiling begins after, or at least alongside, a contested linkage decision; it does not validate that decision by itself.
3. Check the location and time information
Investigators then distinguish crime sites from reporting locations and record how each point was established. Exact coordinates may be uncertain, a witness may recall a location imprecisely, incidents may span a route, and the important location may differ from the location used in a police database. Location precision, site type, temporal order, missing events, and data provenance should be documented before modeling.
For example, the place a victim was last seen, the place an assault occurred, and the place evidence was recovered are separate observations. Combining them without a defensible measurement rule risks creating a misleading geographic pattern.
4. Select an interpretable spatial model
Simple techniques include the mean center, median center, center of minimum distance, and geometric approaches based on the spread of incidents. More elaborate methods apply distance-decay functions, modeled buffer zones, probability density assumptions, travel networks, or environmental covariates. Multiple methods can be compared; the most visually complicated is not necessarily the most accurate.
A classic study of 70 U.S. serial-homicide cases examined many variants of a spatial decision-support system and reported search-cost results for that selected dataset Canter, Coffey, Huntley, and Missen, 2000. Its performance figures describe that study's defined search area, offender sample, and modeling choices. They are not an all-crimes probability of locating an offender.
Paulsen's comparison using solved serial-crime cases reported that simpler geographic strategies and human judgments performed similarly to more complex computer-based techniques under the tested conditions Paulsen, 2006. The practical implication is to benchmark sophisticated algorithms against transparent baselines before claiming added value.
5. Produce a prioritized search surface
Geographic models often assign each map cell a relative search score. Higher-scoring cells are areas to examine earlier according to that model's assumptions. Analysts may express this as a ranked surface or a cumulative search area. The map is a hypothesis generator: it does not identify an address as belonging to an offender, and a numerical score is not automatically a calibrated probability.
A useful report explains the geographic boundary used, incident count, model, parameter settings, sensitivity to alternate assumptions, and the operational action for which the ranking is intended. A profile that cannot be reconstructed from recorded inputs is difficult to audit or challenge.
6. Test competing explanations and update
Analysts should test whether the output changes when disputed incidents are removed, the possible base is defined as a workplace rather than a home, different travel measures are used, or opportunity and land-use information is added. Independent investigative facts may strengthen or weaken a spatial hypothesis. Updating a map after new evidence is legitimate; presenting an updated map as though it had predicted that evidence in advance is not.
What Does a Geographic Profile Actually Predict?
The usual target is an anchor-point location, conditional on an assumed crime series, chosen spatial model, observed locations, and reference population. This is different from predicting a perpetrator's identity, a future crime, a motive, or the probability that any particular resident is guilty.
“Most likely area” is also incomplete without explaining relative to what. A grid cell may rank highest among all cells even while a wide range of other cells remains plausible. A color gradient is a visualization of model output, not a statement that people living in red zones present greater danger.
An uncertainty-aware interpretation says that the observed crime geography is comparatively more compatible with some anchor-point locations under the selected model. A stronger statement needs independent validation that the model's scores actually correspond to well-calibrated probabilities in the relevant population.
How Accurate Is Geographic Profiling?
Research supports the possibility of useful geographic prioritization, particularly for some serial-event datasets, but performance varies with crime type, number and distribution of incidents, mobility patterns, model specification, local geography, and which outcome counts as success. The evidence does not justify a single global accuracy rate.
Search cost and hit rate
Search cost is commonly the proportion of a defined area that would need to be examined, beginning with the highest-ranked cells, before reaching the known anchor point in a solved test case. Lower search cost is preferable within a comparable geographic boundary. “Hit rate” can describe the proportion of cases whose anchors fall inside an explicitly defined top search area. Both metrics depend on the boundary and denominator. Comparing a 5% search area in one city to 5% in a much smaller district may conceal major differences in real-world workload.
In their historical 70-case study, Canter and colleagues reported an optimal average search cost of 11% of the defined area and identified 51% of anchors in the first 5% for the model and sample they tested Canter and colleagues, 2000. These are retrospective case-study performance measures, not prospective odds of locating an offender in any investigation.
Discrimination and calibration
Discrimination asks whether actual anchor points rank more favorably than alternative locations or a comparison model. Calibration asks a different question: when a system labels locations with certain probability levels, do actual anchors occur at approximately those rates over repeated appropriate test cases? Some geographic systems deliver scores without a validated probabilistic interpretation. A ranking can have practical discriminatory value while remaining uncalibrated.
Investigators also need operational comparisons: does the map shorten the work of analysts who already know about transit routes, clusters, and leads, or merely produce a polished representation of what a simple center calculation would show? The NIJ evaluation methodology explicitly called for comparison with controls, empirical solved-case testing, user feedback, and feature analysis Rich and Shively, 2004.
False positives and false negatives
A false positive in operational use may be an area prioritized by the model that does not contain the target anchor point; a false negative may be a low-priority area that actually does. These terms require an explicit threshold, outcome, unit of analysis, and reference population. When analysts call an entire region “high risk,” they must specify whether that means cells, addresses, persons, or expected anchor-point locations, because those units cannot be interchanged.
Raising the threshold may reduce wasted geographic search at the cost of missing anchors that fall outside it. Lowering the threshold can capture more anchors while expanding search effort. Because the map is not a person-level guilt classifier, its area-ranking performance cannot establish false-positive rates for accusations, arrests, or convictions.
Does having more crime locations help?
Additional reliable incidents can improve spatial information, but raw quantity is not a simple guarantee of increased accuracy. In a 2024 study of 333 Japanese burglars separated by observed series size, some between-group accuracy comparisons worsened in groups with more recorded crimes. A within-offender comparison using 150 offenders with at least ten crimes found a modest improvement as included incidents increased from three to ten, particularly for probability-distance methods Hanayama, Haginoya, and Kuraishi, 2024. These differences illustrate selection effects and the importance of asking exactly which comparison a study makes.
Newer approaches and open methods
Published work has compared alternative statistical models and introduced reproducible open-source implementations. Spaulding and Morris described the rgeoprofile package, which provides transparent centrographic and distance-decay methods that investigators and researchers can inspect and benchmark Spaulding and Morris, 2023. Spaulding also evaluated a serial-perpetrator residence model that uses different information depending on available event sites Spaulding, 2023.
A 2026 case study proposed incorporating land-use and environmental layers into a classic geographic-targeting framework Russo and colleagues, 2026. Such advances may improve model relevance to specific landscapes, but a single retrospective case study should not be treated as validation across jurisdictions, crimes, or policing contexts.
Geographic Profiling Example: How to Interpret a Hypothetical Map
Imagine a linked series of burglaries across several neighborhoods. Analysts establish the addresses where the crimes occurred, verify incident dates and the likely relationship among cases, and examine how transit routes, housing, shopping areas, and road barriers affect travel. One method ranks a central region highly; another puts greater weight on road access and produces a somewhat different pattern.
The appropriate conclusion is that investigators have two alternative spatial hypotheses about a possible anchor point. They might use the results to review existing leads or allocate limited, lawful field inquiries, while seeking independent evidence. They cannot infer that residents of either region are burglars or that a particular person is suspicious because a model favors the area. If the linkage hypothesis later changes, both maps must be reassessed.
This example illustrates why the methods work as decision support. The map organizes inquiry, while identification depends on independent investigative evidence and lawful procedures.
Geographic Profiling vs. Criminal Profiling, Crime Mapping, and Case Linkage
Traditional offender profiling attempts to infer characteristics of an unknown offender from crime behavior. Geographic profiling estimates likely spatial anchor points from linked event locations. Crime mapping describes geographic distributions of recorded incidents. Crime linkage tests whether multiple events may belong to one perpetrator or series. Investigative psychology is the broader research field that evaluates these and other inferences, including interview evidence and investigative decision-making.
These distinctions keep scientific claims appropriately scoped. Evidence that crimes can be linked by certain behavioral features does not show that a geographic algorithm can identify a residence. Evidence that crimes exhibit distance decay does not prove any single offender intended to create a buffer zone. Evidence that a map improved an investigation does not demonstrate that its output would be admissible expert testimony.
The larger profiling literature remains methodologically uneven. Fox and Farrington's systematic review of 426 publications found improvements in research rigor but relatively few strong empirical evaluations of proposed offender profiles or their operational effects Fox and Farrington, 2018. Geographic profiling requires method-specific evaluation rather than inheriting credibility or limitations indiscriminately from unrelated approaches.
When Geographic Profiling Is Most and Least Useful
The method is potentially useful when a meaningful series of crimes is credibly attributed to the same source, location information is reliable, there are enough informative events for the selected model, and regional transport and target opportunities can be represented sensibly. Cases with multiple plausible anchors can still benefit if uncertainty is made visible rather than forced into a single point.
Its usefulness diminishes when crimes are wrongly linked, sites are recorded inaccurately, the series spans unrelated phases of life, the offender travels unusually long distances or moves residence, or the environment differs sharply from the model's validation population. A profile may also become uninformative if nearly everyone in a densely populated area occupies a high-ranked region.
Geographic profiling of a single isolated event generally offers far less offender-specific information than a genuine series. Some related spatial methods can analyze one incident using contextual data, but they should not be presented as the same validated serial-offense method.
Scientific Limitations and Common Sources of Error
Selection, missing data, and retrospective success
Many published evaluations use solved cases, for which the anchor point is known. Solved offenders may differ from unsolved offenders in mobility, offense choice, apprehension likelihood, and documentation. Models can also appear more accurate if analysts select cases that fit their assumptions or tune parameters after seeing the correct address. Stronger designs preregister evaluation choices, use separate training and test samples, report exclusions, compare baseline methods, and prospectively evaluate operational use.
Modifiable geographic boundaries and measurement
Search cost changes when the search boundary changes. Euclidean, road-network, time-based, and public-transit distances may produce different results. Small coordinate errors can matter near boundaries or for compact crime clusters. Transparent research must say what geographic units were used and why.
Offense categories and cultural transfer
Offending patterns depend on local environments and data collection practices. Findings from U.S. serial homicide cases, Japanese burglary records, or South African commuter offenders cannot be generalized mechanically to another population. Age, socioeconomic circumstances, access to transport, work routines, and policing practices can shape observed distributions. These are potential contextual factors for measurement and testing, not legitimate shortcuts to identifying a person through stereotypes.
Feedback and confirmation bias
After viewing a map, investigators may attend disproportionately to leads in its brightest area and neglect disconfirming evidence. If subsequent data are collected preferentially there, a feedback loop can make the original output seem more persuasive than it was. One safeguard is to record the pre-profile hypothesis, benchmark the output against alternatives, and preserve results from areas the model ranked poorly.
Security, privacy, and fairness
Geographic models can guide attention toward neighborhoods where many innocent people live. The distribution of recorded offenses is influenced by reporting, enforcement, and surveillance as well as offending. A method that disproportionately directs attention to already-policed neighborhoods can compound unequal scrutiny if geographic scores are translated into person-level suspicion without independent grounds. Agencies should document potential adverse impacts, avoid unsupported demographic inferences, and review how data collection influences model output.
Does Geographic Profiling Count as Legal Evidence?
Geographic profiling is typically an investigative analytic aid. Whether an analyst's methods or opinions can be introduced in court is a separate jurisdiction-specific legal question. The United States, individual U.S. states, and other legal systems apply different rules and procedures regarding relevance, expert qualification, reliability, and admissibility. A mathematical output does not itself establish any element of a crime, probable cause, criminal responsibility, or a person's legal capacity.
A responsible practitioner should clearly distinguish observed incident coordinates, the choice of geographic model, assumptions about travel or behavior, test-set performance, the resulting investigative inference, and any subsequent legal finding. The APA Specialty Guidelines for Forensic Psychology emphasize scientific grounding, impartiality, competent application, alternative explanations, multiple information sources, and separating observations from inferences and conclusions. Their applicability depends on professional role; the document is a professional guideline, not a universal statute governing police software.
If a geographic profile is described in legal proceedings, the court and parties need to know how the source incidents were linked, what validation studies used comparable populations, what uncertainty and alternative models were examined, and whether the method has an established error rate for the claimed application. A map cannot bridge evidentiary gaps merely by looking objective.
Can Geographic Profiling Predict Future Crimes?
Some spatial prediction approaches estimate where incidents may occur, and crime forecasting has its own empirical literature. Classic offender-focused geographic profiling usually estimates an unknown anchor point from known events. These are different targets. A spatial pattern associated with past events is not automatically an individual forecast of future conduct.
Predicting where, when, and whether a future offense will occur would require additional information, explicit outcomes, a validated forecasting framework, and analysis of intervention effects. Converting a ranked probable-home-base surface into an offender dangerousness score is a category error.
Ethical Use and Investigative Safeguards
Responsible use begins with a defined purpose and a record of case inputs. Analysts should preserve uncertainty in crime linkage and coordinates, compare against simple baselines, consider at least one alternative spatial explanation, test sensitivity to model parameters, and document what information was available before new leads emerged.
Operational actions require a separate legal and ethical justification proportionate to their intrusiveness. A profile can assist resource allocation while remaining subordinate to corroborated investigative information. Investigators should communicate the map's limitations in language a nontechnical decision-maker can understand, and avoid labeling districts or residents dangerous because of a high spatial score.
Independent review and periodic testing in the relevant region are valuable. Performance reports should specify crime type, observed event count, search boundary, anchor definition, error metrics, subgroup or geographic differences where measurable, and the circumstances in which the tool performed poorly.
Myths and Misconceptions
Myth: Geographic profiling reveals an offender's home address. In practice, it usually ranks areas that may contain a home or other anchor point, subject to substantial uncertainty.
Myth: A computer-generated heat map is proof of scientific accuracy. Presentation quality is separate from validation, calibration, and operational value.
Myth: Offenders always avoid crime close to home. Buffer-zone effects vary with the case, opportunities, sample, and measurement method. The hypothesis must be tested rather than assumed.
Myth: Every linked serial offender stays within one geographic territory. People can commute, travel, move residence, maintain multiple anchors, and offend in different contexts.
Myth: A high-profile case in which a profile helped proves general validity. Strong general claims require well-designed evaluations across representative cases, compared with relevant baselines.
Myth: Geographic profiling determines whether someone is dangerous, mentally ill, or legally responsible. Those are distinct clinical, behavioral, and legal questions that a spatial-search model does not answer.
Frequently Asked Questions
What is geographic profiling in simple terms?
It is a way to use locations of plausibly connected crimes to rank geographic areas where an unknown offender may have a base of activity, such as a home, workplace, or frequently visited location.
Who developed geographic profiling?
The field developed through work in environmental criminology, investigative psychology, and spatial modeling. D. Kim Rossmo is widely associated with criminal geographic targeting, while David Canter and colleagues developed and studied related investigative geographical decision-support approaches. Different algorithms and research traditions should be credited specifically rather than treated as one system.
Does the FBI use geographic profiling?
Geographic and behavioral analysis have applications in law enforcement, including in the United States. That general fact does not verify adoption of a particular named program by an agency at a particular date. Operational methods, personnel, and software should be confirmed from the responsible agency's current official records.
How many crimes are needed for geographic profiling?
There is no scientifically universal minimum. More reliable, independent event locations can be useful, but suitability depends on the model, event distribution, target question, and case quality. The often-discussed threshold of 50 or more trips in recent research concerns estimating the shape of an individual's travel distribution, not a blanket rule for producing a preliminary map Rossmo and Wheeler, 2024.
Is geographic profiling accurate?
Some methods perform usefully in defined retrospective samples. Accuracy changes across offense categories, geography, event count, model assumptions, and evaluation design. There is no single validated universal success rate; performance should be reported in metrics such as comparable search cost and independently evaluated hit rate.
Is geographic profiling the same as a geographic information system?
No. GIS is technology for managing and analyzing geographic data. Geographic profiling is an investigative inference task that may use GIS tools and spatial models.
What is the difference between a marauder and a commuter offender?
These are descriptive spatial pattern categories used in some geographic-profiling literature. A marauder's known offenses tend to occur within an area organized around a relevant anchor; a commuter may travel from an anchor to offend in a different activity area. They are analytic simplifications, not clinical diagnoses or fixed personality types, and an individual can show mixed patterns.
Can a suspect be arrested because they live in a high-ranked area?
A location score alone provides no individual identification and cannot establish guilt. Rules for investigative detention, arrest, warrants, and use of expert evidence depend on the jurisdiction and the full evidence available.
Does a geographic profile explain criminal motivation?
No. The output estimates potential spatial relationships. Motives and psychological mechanisms require distinct behavioral evidence and cannot be read off an anchor-point map.
Can geographic profiling be used outside serial homicide?
Yes. Researchers have studied burglary, sexual offenses, and other linked incidents. Every application needs its own population-relevant validation rather than assuming findings from serial homicide transfer directly.
For broader context, see Investigative Psychology: What It Is and How It Differs From Criminal Profiling.
Related Articles
References
Almond, L., et al. (2022). Exploring the impact of population density on journey-to-crime in cases of stranger sexual assault and stranger homicide. Journal of Investigative Psychology and Offender Profiling. https://doi.org/10.1002/jip.1598
American Psychological Association. (2013). Specialty guidelines for forensic psychology. American Psychologist. https://www.apa.org/practice/guidelines/forensic-psychology.html
Canter, D. (2004). Offender profiling and investigative psychology. Journal of Investigative Psychology and Offender Profiling. https://doi.org/10.1002/jip.7
Canter, D., Coffey, T., Huntley, M., & Missen, C. (2000). Predicting serial killers' home base using a decision support system. Journal of Quantitative Criminology, 16, 457–478. https://doi.org/10.1023/A:1007551316253
Fox, B., & Farrington, D. P. (2018). What have we learned from offender profiling? A systematic review and meta-analysis of 40 years of research. Psychological Bulletin. https://doi.org/10.1037/bul0000170
Glass, D., & Herbig, F. (2025). Do theoretical assumptions of geographic profiling hold when scrutinising commuter serial rape offenders in South Africa? Journal of Investigative Psychology and Offender Profiling. https://doi.org/10.1002/jip.1645
Hanayama, A., Haginoya, S., & Kuraishi, H. (2024). Testing the accuracy of geographical profiling considering the effect of the number of crimes. Journal of Investigative Psychology and Offender Profiling. https://doi.org/10.1002/jip.1625
Hipp, J. R., Boggess, L., & Chamberlain, A. (2022). Locating offenders: Introducing the reverse spatial patterning approach. Computers, Environment and Urban Systems. https://doi.org/10.1016/j.compenvurbsys.2022.101888
Michaud, P., & Proulx, J. (2023). Deconstructing journey-to-crime's questionable validity in theft-related crimes. Journal of Criminal Justice. https://doi.org/10.1016/j.jcrimjus.2023.102052
Mohler, G. O., & Short, M. B. (2012). Geographic profiling from kinetic models of criminal behavior. SIAM Journal on Applied Mathematics. https://doi.org/10.1137/100794080
Paulsen, D. (2006). Human versus machine: A comparison of the accuracy of geographic profiling methods. Journal of Investigative Psychology and Offender Profiling, 3, 77–89. https://doi.org/10.1002/jip.46
Rich, T., & Shively, M. (2004). Methodology for evaluating geographic profiling software: Final report. National Institute of Justice. https://nij.ojp.gov/library/publications/methodology-evaluating-geographic-profiling-software-final-report
Rossmo, D. K., & Wheeler, A. P. (2024). The journey-to-crime buffer zone: Measurement issues and methodological challenges. Journal of Criminal Justice, 95, 102272. https://doi.org/10.1016/j.jcrimjus.2024.102272
Russo, S. M., Bottini, G., Quattrociocchi, D., & Leitner, M. (2026). Enhancing Rossmo’s criminal geographic targeting model through environmental and land-use spatial layers: A case study of the Atlanta homicides (1979–1981). Crime Science, 15, Article 18. https://doi.org/10.1186/s40163-026-00278-w
Santosuosso, U., & Papini, A. (2022). An analysis about the accuracy of geographic profiling in relation to the number of observations and the buffer zone. Journal of Geographical Systems, 24, 641–656. https://doi.org/10.1007/s10109-022-00379-5
Spaulding, J. S. (2023). Prediction of serial perpetrator residence: Part II—Evaluation of prediction model accuracy. Journal of Investigative Psychology and Offender Profiling. https://doi.org/10.1002/jip.1606
Spaulding, J. S., & Morris, K. B. (2023). An open-source implementation of geographic profiling methods for serial crime analysis. Journal of Geographical Systems. https://doi.org/10.1007/s10109-023-00417-w