Wednesday, May 28, 2008

A Better Way to Measure Traffic, Part One

In our previous two posts, we outlined the county’s reliance on Critical Lane Volume (CLV) to measure traffic and identify problem intersections. We identified two problems with this system. First, the county’s measurement of CLV at each intersection only once every four years or so guaranteed the proliferation of fluky measurements. Second, CLV has different meanings under different circumstances. A low CLV could mean few cars or it could mean that few cars are able to move through an intersection due to excessive congestion.

There is a better way to measure traffic than this.

Any longitudinal data system must accomplish three things. First, the users must have reality-based data that actually measure what the users intend to measure. Second, there must be large numbers of observations to build an adequate sample size. Third, identical procedures must be applied over and over again in different conditions to isolate out their effects. The Planning Department’s current traffic measurement system has none of these characteristics.

But there is hope for something better that is buried deep within the planners’ own documents. In the appendix to the 2008 Highway Mobility Report, the staff has added something new: actual test runs by cars equipped with GPS devices on major corridors in rush hour. On pages 69-82, the staff shows the results of rush hour test drives on Wisconsin Avenue, Rockville Pike, Frederick Road, Georgia Avenue, Norbeck Road, Colesville Road, Columbia Pike, Connecticut Avenue, Clopper Road and Great Seneca Highway. Yes, these drives were only taken on one day each in May 2005, May 2007 or June 2007. But the results are very informative – you can actually see which intersections cause low speeds, and which stretches of these corridors have free flow.

Here is the kind of simulation that reflects reality. People do not drive through problem intersections once every four years, as the CLVs measure. They drive through long, clogged corridors every day, often during rush hour. The challenge is how to collect more of this data so it is not subject to outlier results determined by bad weather, accidents or other unusual conditions.

The solution is to draw on the hundreds of thousands of real live drivers who navigate this county every day. The Planning Department should offer county residents temporary GPS units for their cars in return for a payment of $100 per month. (That is the equivalent of one to two weeks of free gas!) These units would record all driving information and store them in internal memory for download upon return to the staff. Planning could rotate the units among different residents, perhaps 50-100 different people every month. After one year, Planning would have an unrivaled database of tens of thousands of actual drives under every condition imaginable. Planning would be able to judge the performance of any major roadway in the county under any weather condition, on any day, at any time of the day or night. Additionally, matching this traffic data with the police department’s online traffic accident database would enable planners to see the impact of auto collisions on roadway performance for every major route in the county.

How much would all of this cost? GPS units rent for as little as $5 per day and Planning could negotiate a better deal for lots of them. If Planning allocated them to 50 different residents each month and paid them $100 per month, the total rental and payment cost would be $151,250 per year. If Planning allocated them to 100 residents, the cost would be $302,500. Add on the cost of one full-time employee to monitor the program and the cost would be roughly $300,000 to $450,000 per year. The Planning Department’s total budget is approximately $19 million and it is currently suffering from cuts.

But Planning does not necessarily have to request lots of extra money. Why not pay for this, at least in part, by cutting back on CLV measurement? As we have seen, critical lane volume estimates taken in isolation do not by themselves provide reliable measurements of traffic. Do we really need 422 fluky, seldom-updated CLVs, many of which are taken at tertiary intersections with low volume? Why not reduce the number of intersections measured by CLVs and use the money for reality-based GPS measures instead?

Until the Planning Department moves to this sort of system, its existing test drive data provides a tantalizing peek at how a real traffic measurement system could work. In Part Two, we will begin looking at some of the new data Planning has already gathered.