Thursday, July 19, 2007

Not the only one with that thought

I participate in one and only one listserv. It’s the Crime Mapping list, operated by (take a deep breath) the Crime Mapping and Analysis Program of the Rocky Mountain Region of the National Law Enforcement and Corrections Technology Center at the University of Denver. Try fitting that on a business card. Basically, the list is a couple thousand GIS and research wonks—mostly academicians who post and police crime analysts who lurk.

I have something of an overblown and undeserved reputation in this field, and I keep an eye on this list, which has only episodic activity, so it takes little time. On rare occasion I will post something. This week, someone started a thread on something called kernel densities. I happened to catch one of the posts in the thread from Tony Berger, a detective sergeant at the Pierce County Sheriff's Office in Tacoma, Washington who was trying to figure out the best way of displaying 21,700 traffic accidents on a map of Pierce County. Since 21,700 dots will overwhelm a pin map, the usual way of doing this is with a continuous-surface grid, which ends up looking sort of like a weather map—with the hot spots where the greatest intensity of events occur, whether they are burglaries, traffic accidents, or whatever.

Unlike weather, though, which is be spread across the entire surface, traffic accidents only happen on roadways. The "weather map" approach covers every place. So I suggested making a map of roadways, and giving the individual street segments a cool-to-hot color based upon the number of accidents that happened along that stretch or at that intersection.

To illustrate my original thought, I made a quick map of Lincoln’s 5,464 traffic crashes so far this year using this method, and posted it on an obscure location on this blog for the crime mappers to take a look. It was rudimentary--just meant to be a quick visual aid to illuminate the concept. After doing so, I thought that some of the other readers of this blog might be interested in seeing what I’m talking about. Click the image for a larger view.

For those of you familiar with Lincoln, those darkest red segments along Nebraska Highway 2 eastbound at 27th Street and on down the road between N. 40th and N. 56th Streets are pretty obvious. You can also see the predictable hot spot on N. 48th Street between O and R Streets, as well as several segments along N. 27th Street on the approaches to Vine, Holdrege, Cornhusker, and Superior Streets.

Many types of crime are closely related to street activity, and this same methodology might be a useful way of examining other phenomenon. Apparently I wasn't the only one thinking about the potential of using streets as the thermometer to display police incidents. One of the other participants in this online discussion, Dr. Isaac Van Patten from Radford University posted an example, looking at street robberies in Roanoke, Virginia. Ian Oldfield, an acquaintance at the London Metropolitan Police posted some samples of the same kind of work he had done beginning several years ago. So much for my originality.

Gaston Pezzuchi from the Argentina's Federal Police in Buenos Aires, Dr. Martin Anderson from Simon-Fraser University in British Columbia, Dr. Lee Hunt at the High Point, NC police, Bryan Hill and the Glendale, AZ police, and a dozen or so other people from various U.S. police departments and Universities in from Florida to San Diego, Washington State to New Jersey chimed in. There had just been a short virtual conversation among police personnel and academicians thinking along similar lines (pun intended) and crossing the entire United States, Canada, South America, and the United Kingdom. That was rather phenomenal, when you think about it.


Anonymous said...

Is your soap box reinforced so it doesn't collapse. Just kidding interesting ideas but how do you turn all of these stats into data that your street officers can and DO use. Its all nice and neat but what for. It doesn't take a college degree to know 27th is where the majority of the traffic accidents occur

Tom Casady said...

Sorry, Anonymous 9:37, but you're wrong.

As even this rudimentary map shows, 27th and O Street is nowhere near the top hotspot for traffic accidents this year. Among Lincoln's street intersections, it ranks 22nd.

I would agree, though, that this map primarily demonstrates the obvious--if you're a Lincoln police officer. I think most of our officers would identify the traffic accident hotspots with pretty good accuracy with no need for a map. These are the spots where you've been hot, tired, cold and wet at often.

That wasn't the point. This was a short discussion about a method for displaying information, not an attempt to figure out where accidents happen in Lincoln.

Here's some examples of stuff officers really can use that you can find from crime analysis and problem analysis in our huge data stream:

1. Thefts from autos are commonly occurring at auto dealers and repair shops along and near Cornhusker Highway.

2. Car break-ins at recreation areas such as ball fields, fitness centers, trail heads, swimming pools are more common than you might think.

3. Open garages with spare refrigerators are shopped for beer with regularity in the evening hours, particularly in Lincoln south of Old Cheney Road and east of S. 27th Street.

4. That suspect we all know and love specializes in glass break burglaries at small commercial strips, favoring businesses that have cash receipts and no alarms, like nail salons and hair stylists.

I could go on and on. When you handle 400 incidents a day, it's easy for patterns and trends to get lost in the volume--especially for high-frequency events such as residential burglary or thefts from autos.

Police officers are spread across the calendar, the clock, and the landscape. You could easily have four separate officers who have investigated a flasher exposing himself at drive-through restaurant windows over a six-week period, yet none of the officers realizes that the cases are connected.

The purpose of crime analysis is to find actionable information that would otherwise be buried in the usual business. This might be an emerging modus operandi, a connection between cases not previously recognized, or a long-term trend amenable to a prevention effort or legislation--like thefts of catalytic converters, or shoplifing of Sudafed.

Joe Bob said...

All them purty 'puters is great there boss, but my geek alarm was goin off at max output while reading that one!

Magnum Opus (only a real geek will understand that)

Tom Casady said...

Geek alarm? Gee, captain obvious, how'd you ever notice that??

Anonymous said...

You had me at "wonk."

Obviously a tinge of influence seeping in from the Londoners.

Tony Berger said...

I'd like to say to the citizens of Lincoln that you are very fortunate to have the Chief that you do. I am the aforementioned Tony Berger from Pierce County, WA (yes, the guys who are on COPS all of the time) and as a cop of 21 years I am here to tell you that only a handful of police administrators understand where technology can take us in law enforcement these days, and even fewer embrace the technology of today to reach out to the citizens of their cities. Chief Casaday not only believes in using the vast amount of information that we gather today using computers but he personally puts these ideas into practice, like this blog of his. I know of no better way to explain the successes and challenges of police work today that this type of forum. You can accuse him of patting himself on the back or pontificating from a soapbox all you want -- at least he is taking his time to keep you all informed on a day to day level. (Try to find out what is happening in New York from the NYPD!) This just does not happen in very many places, trust me. His response to me regarding viewing our traffic collisons as they related to street line segments was just the method I was looking for. I was over-engineering the project and by following his lead he saved me and our office a bunch of time. The practical use for this, to the bravely identified "anonymous", is so that we can place our meager pro-active traffic division (of about four deputies) in areas of high collisions where they will have the highest likelihood of stopping the most wrecks. If, by using this method, we stop an injury accident, then I think the "purty" map will have done its job. (Thanks again, Chief!)