Geoprofiling Moves Into The Limelight

circletimessquare writes "Interesting and timely. A short piece at CNN talks about the software helping to track down the sniper currently terrorizing the Washington DC area. It was the doctoral thesis of a cop, Kim Rossmo, who developed it while walking the beat in Vancouver and reading about the hunting patterns of African lions. Googling, I found an older but deeper piece which mentions more of the tech behind the software, called Rigel. That led me to the website of ECRI, the company that makes Rigel. More good tech there."

Comments

saw this on TLC

[twiggy]

There was a great special on this software on TLC not that long ago. Basically, they were able to calculate the odds of the suspect living and/or acting in a certain area based on where the crimes were, etc.

They ended up catching the killer, and he was a cop!

From discussions I'm seeing about these shootings, it may very well be a cop or someone in the armed forces. The ballistics of the gun/ammo being used just don't fit right since people are saying they don't hear the shots, or don't hear very loud shots, so people are theorizing that there's special subsonic rounds being used to minimize noise - not easy to find with these types of bullets, from what I gather.. But I dont' know a lot about guns, so.. yeah...

Anyhow if I remember the name of the TLC special I'll post it here, it was on recently enough that it will probably be on again soon.

Re:saw this on TLC

[Anonymous Coward]

Subsonic bullets would negate the entire reason for the .223, namely the extra powder charge. The point of the rifle bullet is to fly very fast. Most likely they didn't hear loud shots because the urban environment is actually very noisy.

Re:saw this on TLC

[spoonist]

It is possible to put a suppressor on a .223 rifle. This would make it much harder to identify, by sound, from where the shot originated. BTW, you don't have to "find" subsonic .223 rounds, you can just reload.

Re:saw this on TLC

[NeoSkandranon]

There is no such thing as a subsonic .223 round, but not because of the reason you gave. The long .223 case means that if you lessen the charge any (which is what is done to produce subsonic ammo) then the powder can move around in the case, thus causing problems iwth ignition and firing. There is a solution, which uses a necked down .338 case which is shorter, but that's further offtopic. Also, making a subsonic .223 round does NOT negate its purpose, as then you have a subsonic projectile that still has good armor piercing ballistics (as opposed to a subsonic 9mm or .45 round, which would literally bounce off body armor)

Re:saw this on TLC

[Happy go Lucky]

From discussions I'm seeing about these shootings, it may very well be a cop or someone in the armed forces. The ballistics of the gun/ammo being used just don't fit right since people are saying they don't hear the shots, or don't hear very loud shots, so people are theorizing that there's special subsonic rounds being used to minimize noise - not easy to find with these types of bullets, from what I gather..

Easy enough to make, though. It's not uncommon for hunters or competitive shooters to load their own ammunition at home. To make a slower bullet, just use less powder. (Okay, it's a tiny bit more complex than that, but you see the general idea.)

Also, it's not hard to mistake the sound of a gunshot for something else, and especially not in an urban area. A month or so ago, I took a complaint of a guy whose truck had been shot. With some sort of .30-caliber solid-construction bullet, original weight above 200 grains, and probably faster than 2700 feet per second from the muzzle based upon the deformation. If you don't know what that means, that's a damn loud round. I try and shoot an elk with a round like that every year, and thank god for Peltor earmuffs. Anyway, almost nobody in the neighborhood remembers a gunshot. However, everybody remembers a car backfiring fairly loudly. Coincidence? Maybe, but I don't believe in them.

So, you see where I'm going with this? It's easy to mistake the sound of a gunshot for something else if you don't know much about them. I'm going to take a stab in the dark and guess that people in an area where private firearm ownership is almost nonexistant (like much of the DC area) may not know what they did or didn't hear.

If only life was like UT2003...

[davidstrauss]

we'd have a trail from every shot of the sniper.

Ruining the Model

[Anonymous Coward]

Doesn't awareness of the geoprofiling model by the suspect make the model less accurate, or is there something built into it that takes this into account?

Re:Ruining the Model

[bjohnson]

No. This is based on the theory that most criminals (this technique is most often used with property crimes like burglary or armed robbery) like to work in familiar territory.

Forcing them into unfamiliar territory to screw up the profiling them loses them the advantages of commiting crimes on known ground and makes it more likely they'll be seen/caught.

media and the software

[SeanWithoutPants]

I'd be somewhat curious if letting the media know that they're currently using this technique to catch the sniper is a good thing. It seems like this guy (or gal) loves the media attention and would certainly hear of this-although given how many times he has shot people in the last few days, I'd imagine it would still be very helpful. Would "security through obscurity" be a good thing here?

Re:media and the software

[jnik]

Would "security through obscurity" be a good thing here?
Questionable. As with any technique designed to discover patterns in human nature, the focus already is on the aggregate, on how people tend to behave. There's going to be deviation no matter what. A good model accounts for this. A good detective understands the nature of the model and its limitations.

Take a look at the second article: Rossmo puts emphasis on certain locations based on his psychological assumptions about the quarry. At the same time, he discards or discounts other locations that he believes might skew his findings. This is just one tool in his arsenal: an important one, but other tools feed data in and yet others interpret what comes out. Sounds like the way to go.

Re:This is scary

[bjohnson]

Actually this isn't new, nore is it all that freaking scary.

It's a technique that's been in use for a long long time by police departments, only with a less quantitative aspect. But they dind't call it 'geographic profiling' they called it 'sticking pushpins into the map'...

They're not *tracking* people, they're entering crime data into a GIS.

The pattern stops here...

[Anonymous Coward]

It all sounds good, but when i read this "Currently, Rigel runs only on a Sun Microsystems UltraSparc workstation. But ECRI is reprogramming it for use on Windows NT workstations and servers".

It all lost it's beauty.

You're running low on virtual memory, pick a smaller town, fewer crimes or reboot yur machine

A Computerized Profiling technology I'll Support

[doc_brown]

There has been alot of talk and show these days about all those new computerized profiling technologies. (Face recoginition, et al.)

Finally, here is one that I think is right on the money.

Here is one that makes the computer just another tool in the policeman's tool box. This is in sharp contrast to present trends. For now the computer is helping solve the crimes and prevent future crimes, but it's not laying the blame on people who have yet to commit a crime.

I know this is mostly due to how the creator uses his experience, but (IMHO) that's what makes this soo nice.

More links

[Anonymous Coward]

• Geographic Profiling: talk by Dr. Kim Rossmo
• cached copy of the APBNews article
• Discussion at NIJ (PDF file)

Not subsonic with a .223

[typical geek]

That's a real small bullet, about as big as a .22 we all grew up with. In order for it to be lethal, it needs to be shot at about 1000 mps (Mach 3 ish).

It's probably a disaffected, over intellectual loner in high school or college with an M-14 or a bolt action .223 hunting rifle with a scope, who's taking out his feeling of inadequacy and powerlessness against random people. Needless to say, he's never been laid, either.

Hmm, I just described half of Slashdot. I hope you have your alibis

Re:Not subsonic with a .223

[Lord Omlette]

Nonono, this is clearly the work of some godless hippie liberal who hates America. Clearly the only solution is to invade Iraq!

But what if the sniper has the software too?

[Anonymous Coward]

Could he profile himself and then know where NOT to go to find his next victims?

Hmmm, maybe it was seeing Red Dragon

[sielwolf]

But I've been thinking of Lecter's advice to Clarice in Silence of the Lambs: this looks a little too random.

Ever since I first saw the movie I've always wondered how often that is the case: serial criminals who commit the first crime locally, realize it, and then make a point of trying to be "random".

This entire scenario it doesn't look like the case: the first and fifth shooting were very close together and the entire field of action seems to be very localized. But still these sort of things always make me think of that quote. Guess because it was so imporant in the movie.

sniper anagrams

[recalci]

The DC sniper left a "Death" tarot card with the words:

"Dear Policeman, I am God"

written on it.

Some anagrams for "dear policeman i am god"

go and implode America
Laden doom pig America
impaled good American
magic doomed airplane
megalomaniac drop die
an imperial dogma code
good, an epidemic alarm

Some anagrams for "dear policeman i am god death"

imperial hated and good came
imperial death and good came
I'm a degraded emotional chap
I'm delegated macho paranoid
homicidal dead eager top man
Peter, a homicidal dead man go
dead homicidal game not rape

programs response

[UniverseIsADoughnut]

It's Mrs. White in the library with the revolver

The "Deeper" Piece Seems to be Slashdotted:

[Ribo99]

GEO-PROFILING: POTENT NEW POLICE TECHNIQUE
Cracking the Toughest Serial Criminal Cases
Dec. 31, 1998

By Jim Krane

SAN DIEGO (APBnews.com) -- Picture a small city in eastern Canada whose residents were rarely touched by violent crime. Then, startlingly, a serial rapist began attacking women, injecting a dose of fear into a normally tranquil community.

By the time the assailant sexually assaulted his 11th victim, police were desperate. They compiled a list of 300 possible suspects and prepared to conduct expensive, laborious DNA tests on each one, hoping to match DNA residue taken from victims.

Vancouver Police Detective Kim Rossmo
That's when Det. Kim Rossmo got a call.

Rossmo, a detective inspector with the Vancouver Police Department, developed an investigative technique called geographic profiling. Using geo-profiling, police try to trace a serial criminal to his home or workplace by computing distances with geographic clues he's left -- such as dead bodies, sites of attacks and other known locations the lawbreaker visited.

Rossmo explained geographic profiling to attendees at the International Association of Crime Analysts here recently, giving criminal analysts a window into one of law enforcement's newest and least-known investigative techniques.

Rossmo's methodology would come in handy on the serial rapist case and many others.

Valuable search tool

As part of his doctoral research at British Columbia's Simon Fraser University, Rossmo developed an algorithm -- a mathematical model of repeated calculations -- that targets serial criminals by the spatial patterns they produce.

Since then, Rossmo's algorithm has been computerized, allowing it to make hundreds of thousands of calculations that pinpoint a criminal's hideout within a fraction of the crime site area.

Priority: danger

Rossmo most often gets a call when a serial criminal is on the loose. Since many agencies -- in Canada, the United States and Europe -- seek his services simultaneously, Rossmo said he gauges which community is most at risk.

In the eastern Canadian sexual assault case -- Rossmo didn't want to divulge the location -- his geographic profile turned out to be remarkably accurate. With 300 suspects on their hands, the local police could only look forward to a lengthy period of laboratory testing.

The red peaks in this image identify the probable location of an offender's residence in Vancouver, British Columbia.
But Rossmo's geo-profiling technique helped the police get their man much more quickly. The Vancouver detective visited crime scenes, read reports, and talked to victims and investigators. He analyzed the data using his computerized algorithm and found a neighborhood hot spot to focus on.

Seventh time's a charm

Instead of hauling suspects in alphabetically by last name, police matched suspects' addresses against Rossmo's findings and tested those who lived nearest the hot spot's peak. The seventh suspect lawmen tested was a positive DNA match. Police arrested the man and cracked the case.

"If they didn't have geographic profile prioritization, they might've started with Archer and ended with Young," Rossmo said.

Lazy to a fault

Despite its complicated mathematical calculations, geographic profiling is based on a simple theory. Criminologists say most humans -- criminals included -- are inherently lazy. Just as a person will shop in the grocery store nearest his or her home, a predatory criminal usually picks his victims in familiar areas -- except for a small buffer zone around his home, says Rossmo.

Thus, when an arsonist sets a series of fires, police can estimate his whereabouts (usually a residence) by dumping the addresses of buildings burned into the computer and calculating the location most central to the crime scenes.

Crime as topography

In reality, Rossmo's crime-busting technique is more complex. He walks through crime scenes, conducts interviews and reads police reports. With years of investigative experience under his belt, Rossmo puts emphasis on certain locations based on his psychological assumptions about the quarry. At the same time, he discards or discounts other locations that he believes might skew his findings.

Rossmo then keys his data into the computer. The machine converts street addresses into latitudinal and longitudinal coordinates and creates a three-dimensional "jeopardy surface" or topographical model of the data. The jeopardy surface looks like a mountain range, with colored bands of peaks and valleys that show where the addresses converge -- the peaks -- and where they don't -- the valleys.

When Rossmo superimposes the jeopardy surface onto a street grid, the result isn't an exact map to the killer's house, but it's something close to it.

Method used in 80 cases

Since 1990, Rossmo has used his geo-profiling technique in more than 80 cases, representing 1,800 crime locations. He believes his work helped crack about half of those cases.

But Rossmo doesn't measure his success only by cases cleared. He's interested in geographic accuracy.

In cases where an arrest has been made, Rossmo's been able to estimate the location of the offender's home within the top five percent of the search area. That means, if police believe the offender lives somewhere within a 10-square-mile area, Rossmo can tell investigators which half-square-mile section to search.

In some cases, he's more accurate. In the Canadian rapist investigation described above, Rossmo's suspect lived within the first 2.2 percent of the area searched.

The more a criminal strikes, the more clues Rossmo can enter into his computer. Theoretically, that makes his predictions more accurate. But Rossmo's computer doesn't spit out a name and address. After the computer does its thing, Rossmo writes a report suggesting strategies for capture.

"It's the investigator that solves the case. Our role is to support him or her," Rossmo said.

Cops, meet Rigel

Rossmo's algorithm has been incorporated into a software program called Rigel, manufactured by the Vancouver firm Environmental Criminology Research Inc. (ECRI). Rossmo is a member of ECRI's board of directors and acts as the company's chief scientist.

Currently, Rigel runs only on a Sun Microsystems UltraSparc workstation. But ECRI is reprogramming it for use on Windows NT workstations and servers.

The software isn't cheap -- ECRI president Barry Dalziel priced a copy at $70,000, which includes some training and help with installation.

Rigel, emphasized Dalziel, isn't perfect. For best results, it should be used by a police investigator or crime analyst who undergoes a year of training, some of it under Rossmo's personal tutelage.

"If it sends them off on a wild goose chase, police investigators aren't likely to use the system again," said Dalziel.

It's a Canadian thing

Besides Rossmo's Vancouver Police, two other agencies have been trained in geographic profiling with Rigel: The Royal Canadian Mounted Police, Canada's national police force, and the Ontario Provincial Police. Rossmo said the British National Crime Faculty, another national law enforcement agency, will be certified in 1999.

No U.S. law enforcement agencies are on Rossmo's training list -- even though he's been invited to help crack dozens of cases in the States.

The real Robocop

If its geo-profiling uses weren't enough, Dalziel said investigators will be able to use a new version of Rigel to predict a serial criminal's next crimes, including dates and crime locations.

And cops will be able to predict and monitor the likely "hunting grounds" of paroled sex offenders by plotting past crime data and behavioral traits into Rigel, said Dalziel.

"Say there were crimes in that area that matched [a parolee's] M.O., his name would pop up," Dalziel said.

Jim Krane is APB News staff writer (jimk@apbnews.com).

You mean Mini14

[Wee]

That's a real small bullet, about as big as a .22 we all grew up with. In order for it to be lethal, it needs to be shot at about 1000 mps (Mach 3 ish).

For anyone curious, the .223 is about the same diameter as a .22 LR, but there the similarity ends. The .223 weighs in between 50 and 64 grains and travels at 2700-3300 fps. I think the .223 NATO round is 55 grains and moves at like 3100 fps. A .22 LR is 40 grains and travels at around 1050 fps. I might be a little off in my numbers, so don't quote me. The two are night and day as far as lethality and ballistics go, however.

It's probably a disaffected, over intellectual loner in high school or college with an M-14 or a bolt action .223 hunting rifle with a scope, who's taking out his feeling of inadequacy and powerlessness against random people. Needless to say, he's never been laid, either.

The M14 is .308, not .223. You mean a Mini14.

But I get your point. Feet first into the mulcher is too good a fate for this ass clown. Shooting old men and children and women. In the back. I'm having a hard time coming up with suitable retribution...

-B

Re: You mean Mini14

[Black Parrot]

> Feet first into the mulcher is too good a fate for this ass clown. Shooting old men and children and women. In the back. I'm having a hard time coming up with suitable retribution...

I'd go for ordinary imprisonment. Sure, this and lots of other crimes merit worse, but unfortunately our "justice" system is actually a "conviction" system, and doesn't appear to be batting too high an average on hanging the right guy.

Re:You mean Mini14

[scott1853]

3100 fps *drool*

Need... new... video... card... 8P

Re:You mean Mini14

[kubrick]

Shooting old men and children and women.

Something I've wondered in this case... why is it worse to shoot "old men and women and children" than it is to shoot anyone else? Are 35-year-old men some sort of second-class citizen, not worthy of sorrow? Sure, they may be more able to defend themselves in hand-to-hand combat, but that's not going to do them a lot of good when a sniper shoots them in the middle of the suburbs...

Kim Rossmo

[Anonymous Coward]

Kim Rossmo was also one of the first to suggest vancouver had a serial killer [robert pickton,pig farmer],which the VPD dissmissed promptly.The VPD also drummed Rossmo out via the old boys network because of interdept politics/powerplays.

Re:Better CNN article

[Anonymous Coward]

Please provide details about how the system works, and how a criminal could circumvent the profile to throw police off the track. I need this information ASAP.

Ph.D. in Criminology

[mbrubeck]

just curious, but what did Kim Rossimo get his Phd in?

Kim Rossmo got his Ph.D. in criminology at SFU. The ideas in his thesis weren't just sudden inspiration -- they came from his many years as a police detective working on investigations, and from rigorous academic study and research.

Some background on this guy....

[puppetman]

I live in Vancouver, where Kim Rossmo got his Ph.d and started his geoprofiling.

He was very successful, and it led to his rapid advancement in the Vancouver Police Department. But like most police departments, it's still old-boys, and alot of them resented an educated individual rising through the ranks so quickly.

Finally, they told him they weren't extending his contract when he was promoted too far. He sued. During the trial, the senior VPD members were made to look like fools for lying under oath.

One of the interesting things that came out was that he suspected (back in June, 2001) that a serial killer was involved in the disappearance of 20 to 30 Vancouver women. Well, he was right. The Vancouver police are conducting a huge investigation at a pig farm in the Vancouver area, and Robert William Pickton is now Canada's most prolific known serial killer with 16 or so charges in the works, and more pending as they find more DNA at the farm.

I don't know much about the technology (or psychology) involved, but I do know that when he applied his software to some of Canada's other serial killers (Paul Bernardo, Cliffard Olsen, etc) his software picked a 4-block area which included the killer's home. It was also used to catch a killer in Abbotsford.

Thanks to a bunch of fat old men who's ego has extended past their intelligence, Vancouver has lost what appears to be a top-rate talent.

Reaching, aren't we?

[Inoshiro]

I know it's human nature to find patterns in random data, but this seems just a wee bit far fetched to me.

Chances are s/he just wanted to say that s/he's god, and because of that has power over life and death (with no way for them to stop s/he).

Sometimes a nutjob is just a nutjob.

Re:Reaching, aren't we?

[Ralph Wiggam]

Have you ever watched a movie? The bad guy always taunts the hero with an anagram.

-B

Re:Sound cool but

[ReconRich]

what if the guy your trying to track is a complete nutcase
Well, I'd say he's certainly some kind of nutcase. Examination of his method, however, reveals a few interesting things. First, he acts like a military sniper; that is he selects a target, takes 1 shot, and then (presumably) leaves his position, taking everything (like shell casings) with him. In other words, all the work is in preparing his site, and leaving it. The actual shooting doesn't take much time or effort. Second is his selection of target. It appears to be random, although shooting a kid going to school could be designed to cause fear. Its almost as if the target is secondary to the location; this nutball may be picking sites from which he thinks he can get someone, and then killing whoever shows up. If this is the case, this location software may be just the thing; on the other hand, his criteria for picking a site might not have anything to do with where he lives. Anyway, I hope they catch this bastard soon, and I really don't care how they do it.
-- Rich

Is it just me?

[PerryMason]

who developed it while walking the beat in Vancouver and reading about the hunting patterns of African lions

Is it just me or does anyone else think they might have more chance of catching the guy if cops dont walk the beat reading a book!!

I have worked on this

[SheldonYoung]

I was part of the team that implemented an early version of the Rigel software used by Kim Rossmo.

At least in the early version, the algorithm was very simple. It was so simple you would have though it would never be useful. The beauty is that the algorithm doesn't need to pinpoint the house, just the neighborhood. It was much better to have a simple and easily provable algorithm than get another half a block of accuracy.

The available databases to convert from street address to spatial locations sucked. To me a big part of the magic was converting addresses where a crime occured to a UTM coordinate.

Most importantly, the magic of Rigel and Kim Rossmo is not the geoprofiling algorithms, but the marketting and public relations.

Re:Shooter is accurate too

[ReconRich]

That is a good point, This guy hits reliably and doesn't freak out. My own take is that this is someone with military sniper training. What bothers me most about that is that sniping is hardly glamour killing; its boring, monotonous work. I have a very hard time understanding why someone who is killing people for some sick kind of emotional reward would do it this way. It is, however, probably the best way to get away with it. Then again, I worry about anybody who could get inside this guys head. He is very scary.

-- Rich

Happy that someone's listening to him

[Stephen Samuel]

Although Rossmo developed his software in Vancouver, it didn't get a very good reception here. Rossmo used his software to conclude that 50+ women missing in Vancouver's Downtown Eastside weren't a coincidence. Unfortunately for him, the missing women were mostly low-rent prostitutes, and the Downton Eastside includes the postal code with the lowest average rent in Canada. It wasn't the mayor's highest priority by far.

Vancouver's Mayor had more police manpower directed towards a high profile pot shop in the area than the case of the 50 missing women. Rossmo's thesis was pooh-poohed and he was demoted and effectively run off the force.

The missing prostitute case continued to be a willfully low priority of the Vancouver police department until it recieved some publicy (including, I believe, being featured on "America's Most wanted" -- "Vancouver's a great place to be a serial killer -- cops cry '50 missing and all's well!'"

A little over 2 years later, they've charged a guy with killing 15 of those missing women, and are searching for more remains on his pig farm.

From what I've been able to piece together, he abused them, killed them, ran their bodies through a meat grinder (or branch grinder) and buried the ground-up bits on his farm.

In the meantime, Downtown Eastside residents who were formerly unwilling to report mysterious disappearances of friends to the cops have now brought the number of missing women into the 60 person range.

More info on the missing women case can be found on the CBC website.