Geographic Information Analysis: From GIS to GIA

The Science of the System

Applying “Animal Movement” Analysis Techniques to the Study of Crime

Sean Bair, Program Manager

Crime Mapping & Analysis Program

Introduction:
This is the first in a series of articles titled “From GIS to GIA.” My position is that, just as the development of GIS
(Geographic Information System) from the less powerful paradigm of simple mapping represents a necessary evolution in
technique, so must we advance beyond GIS into the next level of capability: Geographic Information Analysis, or GIA.
Most analysts acknowledge that GIS represents a fundamental paradigm shift from simple mapping, and that the power
brought to law enforcement analysis operations by a GIS is essential to modern police work. Now, however, we must face
the fact that our use of GIS is limited by our map-oriented perspective. To develop beyond this perspective, we must evolve
toward a more analytical construct - GIA.

GIS, GIA and Just Plain Maps:

In 1997, the Crime Mapping Research Center (CMRC) selected the term “Crime Mapping” to define - and popularize - the
use of GIS in Crime Analysis. This choice was attractive for a variety of reasons, and continues to be highly successful.
However, the term itself imposes a limitation upon the development of the technology by its emphasis on the word “Map.”
A map is simply a graphical representation of spatial data; a “picture” of information. The power of GIS transcends this
approach. Although the limitations of a "map" as compared to the power of a GIS may seem somewhat obvious to
experienced spatial analysts, it can be difficult to make clear to the inexperienced. After all, isn't a GIS just a faster way to
make a map? Not at all. Not only can a GIS construct powerful and elaborate models in three dimensions, it can now
perform four-dimensional, dynamic modeling, elevating the accuracy of the analytical processes significantly. It is
unfortunate that we consider such weak and two-dimensional artifacts as “Pinmaps” to be analytical products, when clearly
they are not. They are merely pictures, and not very good pictures at that.

So, how can a crime analyst, burdened by demand from patrol officers and administrators, laboring under deadlines, and
afflicted by multiple crime series, elevate his or her use of “Crime Mapping” or GIS into GIA? By recognizing that a map
is a useful product in the same way that a crime report is a useful product, or a radio transmission, or a mugshot - but it is
not, in itself, a finished product. The map does not, and should not, stand alone.

Many powerful and intuitive tools are available to the crime analyst willing to go beyond “mapping” and into the broader
fields of Geographic Information Analysis. Many other scientific disciplines have already met this challenge having created
tools and techniques which, I believe, can be adapted to the field of crime analysis. Meteorology, Vulcanology, Glaciology,
Oceanography, Medicine, Astronomy, Cosmology, Economics, and Biology each have developed spatial models and
methods for analyzing problems that can be applied to police operations.

Animal Movements:
One of the best sets of tools available to the modern crime analyst was developed for the science of Ecology; specifically,
the United States Geological Survey's homegrown ArcView extension, Animal Movements. Developed to study migration
and movement patterns, but containing many tools for general movement study, this professional extension was practically
unknown outside the science of Ecology. The potential of this extension to radically advance the field of tactical crime
analysis was realized in 1997. Dan Helms, Crime Analyst for the Las Vegas Metropolitan Police Department, unsatisfied
with existing methods for studying and predicting spatial behaviors, turned to game tracking experts in Southern Africa for
help in analyzing the movements of a serial sexual predator on the hypothesis that the seemingly disparate disciplines might

-1-
share some important truths.

What Helms discovered was a wealth of tactics, techniques and science created over decades and centuries to observe,
analyze and even predict animal movements. These techniques could make use of scattered observations to deduce such
factors as “Home Range,” “Hunting Grounds,” migratory patterns, and much more. Helms’ contacts stated that many of
their techniques have been incorporated into an ArcView script, Animal Movements. Helms downloaded the free extension
from the United States Geological Survey web page http://www.absc.usgs.gov/glba/gistools.

The Animal Movement Analysis ArcView Extension developed by the Alaskan Biological Science Center (USGS) is a free
add-on to ESRI's ArcView desktop GIS software application that provides ArcView users with numerous functions to
analyze and predict spatial information.

The extension was developed by Hooge and Eichenlaub in 1997. Dr. Hooge is a Research Population Ecologist and Mr.
Eichenlaub is the GIS Coordinator for the Glacier Bay National Park. The Animal Movement extension is a collection of
more than 40 functions to aid in the analysis of animal movement data. The extension fills a definite void in the availability
of affordable, powerful tools to incorporate into one's GIS. Let's take a look at just a few of the functions provided by this
extension and review their law enforcement applicability with real-world examples.

Home Range and Hunting Grounds:

The first function we'll examine is the Jennrich-Turner Bivariate Normal
Home Range, or Jennrich-Turner Ellipse. It allows one to calculate a
bivariate normal probability ellipse, the major and minor axis, and the
arithmetic mean from a sample or population of event points. It is primarily
geared toward developing simple home range-habitat models. Simplistic in » »
function and flawed in its dependency on a bivariate normal distribution, the »
tool allows the analyst to construct possible hunting grounds based on
probability ellipses. »
The results of the function are ellipses providing an area accounting for 95, 90, » »
and 50 percent distribution respectively, as well as a graphic representation of »»
the polarity of the cases and the center (centroid) of the cluster under study. The
polarity graphic will allow an
analyst to discern the statistical
direction of the cluster. Many
crime analysts will recognize
similarities with analytical methods already used in the profession, most
» » notably Spatial and Temporal Analysis of Crime, (S.T.A.C.) Unlike the
» original S.T.A.C. program, however, the sensitivity of the Jennrich-Turner
Ellipse in the Animal Movements extension is infinitely adjustable, and
 
» ellipses are polarized along a complete 180 degree sweep, not constrained to
a rectangular grid. The graphic quality is also more dynamic, allowing for the
» » conversion of the ellipse to a polygon theme. Given this new theme, the
analyst can then perform spatial queries for further analysis using the
»» Jennrich-Turner results.

The Ellipse is a simplistic and limited tool, but we discuss it here because it
is so intuitive and easy to use, and because of its similarity to accepted law
enforcement techniques.

A much more powerful and slightly more complicated tool is the Kernel Home Range Density feature.

This function relies on the ArcView Spatial Analyst extension to create a harmonic density grid from known event points.
Each cell in the grid is assigned a probability weighting which indicates the probability, based on a finite set of observation

-2-
 points, that an animal will pass through the cell.

Imagine you are a wildlife researcher attempting to capture a specimen of a

particular bear. You have available a list of sightings, locations where, over a
» » period of time, observers have actually seen the bear. Viewed on a “Pinmap,”

» these present you with a spread of dots which might convey meaning, but
probably appear completely chaotic. The Kernel Home Range Density grid can
» tell you, based on past observations, where you should position yourself to watch
for your subject. The applicability to law enforcement is obvious.

» » The Kernel Home Range Density provides similar results to the Dragnet program

»»
created by Professor Canter at the University of Liverpool, and software
developed by Kim Rossmo of the Royal Canadian Mounted Police which was
based on research by the Brangtinghams at Simon Frasier University. However,
each of these programs intends to derive the home location of a serial offender
from the locations of his attacks. The Animal Movements Kernel Home Range
function is slightly different; it predicts where the subject will be observed.

Sequential Analysis:
Another function that has proven to have applications in Crime Analysis is the Point-to-Polyline function. This function
will “connect the dots” from a point file which has been ordered chronologically and will provide distance measurements
from point-to-point. While this capability might at first seem trivial, in fact it represents one of the most crucial advances in
viewing the distribution of serial crimes, and exemplifies how simply changing our perspective can revolutionize our ability
to analyze crime.

There are three basic techniques to integrate temporal analysis with spatial analysis in a typical desktop GIS environment:
Animation, Substitution and Sequencing. Animation is usually accomplished by blending a succession of still images and
then playing them in order. Substitution replaces one dimension with another; typically, the “Z” coordinate symbolizing
height is replaced with a “T” coordinate symbolizing time. Sequencing merely involves the display and analysis of events in
the order in which they occurred.

The application in Crime Analysis, as demonstrated by Helms, includes

classifying the newly created line theme with the arrow line symbol to display
the sequence or direction of cases. By analyzing relationships between intervals
in time and distances in space, Helms was able to calculate statistics on the
development of patterns between cases and provide enhanced predictions. Helms
» »
has taken the basics of this information a step further by applying a Lag  »
Variogram to the cases under study. This technique allows the analyst to identify
and analyze repeating cyclic patterns within a crime series - in other words, the
“tempo” of a series. For instance, if a suspect were to strike in distance intervals
»
of 1 mile, then 3 miles, then 1 mile, then 3 miles etc., we could analyze the
cyclic pattern of this series to reveal that the next strike may occur 1 mile from » »
the last strike. If one were to simply take the mean distance between strikes, the
results would be less accurate by yielding an average strike distance of 2 miles 
»»

and a prediction that the suspect may strike 2 miles from the last strike location.
A Lag Variogram applied to the distance measurement results returned from the
Polyline from Point file function enables the analyst to more accurately predict
next or future occurrences based on past behaviors.

In the previous image, we see how Helms has applied an arrow line theme to the Point-to-Polyline routine to depict
movement. To increase the readability of the analysis, one might ramp shade the order of points or use graduated symbols to
depict chronology or change in time and space. Instead of trying to identify patterns based on the locations of points, the
analyst can study the changes between points, symbolized by the lines, to create a more accurate and refined forecast.

-3-
Geographic Centers:
It is often desirable to identify the geographic center of a group of points, such
as crime locations. This information has often been used successfully to deduce
the location of the offender's home, work, or school location, and for a variety » »
of other analytical techniques. The Spider Diagram function will create a »
Polyline file using the Arithmetic mean, Harmonic mean, or another Polygon »
theme to calculate the center of a series.

This technique has been used on numerous occasions to develop useful

» »
analytical products such as prioritized lead lists, offender home, work, and »»
school location estimates and refined hunting ground and home range
estimates. It can also be used as a spatial selection criterion to help identify
streets and roads possibly used by the offender.

Conclusion:
The tools described and illustrated in this article are only a sample of the comprehensive suite of utilities provided in the
Animal Movement extension for ArcView. There are several other tools in this extension that can enhance an analyst’s
ability to analyze, predict or forecast criminal activity.

The use of uncommon spatial statistical techniques employed by other disciplines can have an application in the study of
crime. Like predatory animals, criminals make decisions based upon their environment, needs, habits, addictions and
limitations. In the study of criminal behavior, there are no random spatial events. Criminal behavior can be spatially and
temporary forecasted or predicted given adequate study and understanding of the cases. However, although human behavior
is not random, it is certainly chaotic – that is, unpredictably complex interactions of small events can result in enormous and
unexpected influences. Suspects will miss the bus, get sick, their car will break down, and they sometimes get arrested in
another jurisdiction before giving us the pleasure of catching them using these techniques.

Helms’ innovative use of the Animal Movement script is only one example of how Crime Analysts can draw from other
disciplines and professions’ analytical practices. Future articles will demonstrate how relatively untapped disciplines and
sciences such as Signal Analysis, High Finance, Meteorology and Glaciology can be used to study spatial and temporal
criminal behavior.

I hope that I have provided you with an understanding of some of the additional tools that are available to use when
analyzing human behavior. Although these procedures developed by the USGS relate to animal behaviors and are based on
analysis using algorithms and processes developed from numerous studies of animal movements and behaviors, I believe
that much of the methodology can be applied to the study of human behavior. Analysts have only begun to scratch the
surface as to what GIA can provide us in our quest to fight crime.

-4-
To obtain the Animal Movement script, visit the USGS Internet Web Page at: http://www.absc.usgs.gov/glba/
gistools.

The techniques presented here will soon be taught in the Crime Mapping & Analysis Program’s (CMAP) Advanced Course
currently under development. Please contact Alisa Anthony, Program Coordinator for CMAP, for inquires about future
course dates and locations at 1-800-416-8086 or [email protected]. Visit our CMAP website at http://www.nlectc.
org/cmap to learn about upcoming CMAP classes and general information regarding the CMAP program.

Comments, questions or suggestions regarding this article should be directed toward Sean Bair at (303) 871-3851 or
[email protected].

Bibliography:
Levine, N. (2000), CrimeStat Manual. National Institute of Justice.

Cantor, David (1999), Criminal Shadows. Harper Collins Trade Publications.

Brantingham, P. L.; Brantingham, P. J. (1984), Patterns in Crime. MacMillan Press.

Brown, Donald (1999) A New Approach to Spatial-Temporal Criminal Event Prediction. 1999 ESRI Users’ Conference
XIX Proceedings. http://www.esri.com/library/userconf/proc99/proceed/abstracts/a889.htm

Helms, Dan, (1999) The Use of Dynamic Spatio-Temporal Analytical Techniques to Resolve Emergent Crime Series.
http://www.bairsoftware.com/dstapaper.pdf.

Hooge, P. N. W. Eichenlaub, and E. Solomon. (1999) The Animal Movement Program, USGS, Alaska Biological Science
Center. http://www.absc.usgs.gov/glba/gistools.

Silver, Steven (1998) Hydrologic Model Development and Water Resource Administration Within an ArcView GIS
Environment. 1998 ESRI Users’ Conference XVIII Proceedings.

-5-