The Washington, DC region is great >> and it can be greater.


More riders, less rain increase bicycling on Custis Trail

Since the fall of 2009, Arlington County has been automatically collecting data from dedicated bike and pedestrian counters. Traffic on the Custis Trail increased from 2009 to 2010, though more rain in 2009 makes it difficult to pin down the exact amount definitively.

Photo by M.V. Jantzen on Flickr.

Arlington has had two counters installed for more than a year. One is on the Custis Trail near Rosslyn. Another is located on the Four Mile Run trail near Shirlington. The county is continuing to add counters, which will allow for very rich and useful data.

The counter on the Custis Trail at the top of the Rosslyn hill near mile post 3.5 (map) has collected the most data, tracking at 15 minute intervals continuously for more than 15 months.

Bike Arlington reported on the CommuterPage blog that data from these counters will be available online and the site will include graphing tools. Allowing citizens to do their own analysis is a great attribute.

Three different analyses comparing 2009 to 2010 all show an increase in ridership at this counter. However, the magnitude of the increase ranges from 1.3% to 11.2% depending on the analysis.

Click for larger image.

This graph shows ridership for 5 weeks starting at the first of November through the first week of December. The days of the week have been aligned, so 2010 actually starts on Sunday, October 31. Total ridership over the 5-week period increased from 24,015 in 2009 to 26,714 in 2010. That's an 11.2% increase.

That's a pretty good bump for one year. However, it turns out that 2009 had a lot more rain.

The chart above shows days with rain as darker bars. 2009 had 13 days of significant rain and 3 days with tiny rain. 2010, however, had only 5 days with significant rain and another 7 with a tiny bit of rain. Whether the timing of the rain on each day was a factor for cyclists is unknown.

Let's look at two approaches to removing rain as a factor.

The above chart shows ridership for the five weeks starting in mid-October through the third week of November. For each day of the week (Monday, Tuesday, etc.) the highest ridership for that particular day is shown. That is, the Monday with the highest ridership in 2009 and the Monday with the highest ridership in 2010 regardless of whether it was from the same week.

Five of the seven days of the week showed higher ridership in 2010 than 2009. The total for all seven high ridership days for each year are 9,826 in 2009 and 10,617 (+ 8%) in 2010.

It should be noted that the highest ridership day of all was actually Saturday, Oct 30, 2010 with over 1,600 cyclists. This number was clearly an anomaly; ridership was skewed by the Stewart/Colbert rally on the mall, so I disregarded it for the sake of this analysis.

For just the five weekdays the totals are 6,085 in 2009 and 6,590 (+ 8%) in 2010.

This analysis, which eliminates rain as a factor, shows an 8% gain in ridership but not as much as the first chart above. This comparison is probably closest to capturing the "unadulterated" increase in ridership, because it most well illustrates that there were, in fact, more riders (maximum ridership was up). However, let's look at an alternative analysis.

To eliminate rain as a factor, this approach takes the seven weeks from mid-October through the first week of December and disregards all days that had any rain in either year, matching up the days of the week. There were 20 days that were completely dry in both years, and they are shown in this chart:

This approach essentially attempts to create an apples-to-apples comparison from year to year, comparing the same days with each other and ignoring days with rain. Total ridership for these 20 comparable days were 15,970 in 2009 and 16,180 (+1.3%) in 2010.

In contrast to the other two, this last approach shows only a very small gain in ridership from 2009 to 2010. However, other factors, such as temperature for instance, may have played a role as well. As the automated counters continue to click each fifteen minutes, 24/7, we'll keep gaining a clearer and clearer perspective on use of these trails.

Steve Offutt has been working at the confluence of business and environment for almost 20 years, with experience in climate change solutions, green building, business-government partnerships, transportation demand management, and more. He lives in Arlington with his wife and two children and is a cyclist, pedestrian, transit rider and driver. 


Add a comment »

Steve, this is great analysis, thanks for it. Have you tried simply running a regression on all days, with dummy variables representing rain and seasonality? Then, the coefficient on "rain" should isolate the effect of rain alone, and show its magnitude and statistical significance. This might be an easier way to pinpoint the effect of rain.

Also, I'm a little confused by the title of the post. "More riders... increase bicycling" seems self-evident - did you mean something else?

by Just161 on Feb 7, 2011 1:07 pm • linkreport

Actually, you might add "inches of rain" to the regression as a scalar variable, and "degrees below 60" or some such as well, in order to analyze the whole year and capture the effect (if any) of temperature on bicycling. WashCycle's earlier post suggests that high summer heat doesn't have much effect.

Is this source data posted anywhere?

by Just161 on Feb 7, 2011 1:19 pm • linkreport

I wouldn't exclude the Stewart/Colbert rally - big events on the Mall are the norm, not an anomaly.

by TheGreenMiles on Feb 7, 2011 3:25 pm • linkreport

I've said this before, the counter location in Rosslyn is NOT optimal. I suspect it overcounts bikes as many pedestrians don't use that part of the trail.

In terms of measuring increase in bike traffic, efforts to include/exclude snow, rain and Colbert rallies seems rather pointless. That is what bike traffic is -- weather dependent. If we ran out of gasoline for a week (unlikely) I'm sure bike traffic would spike too.

Bike traffic probably is dependent on some degree to gasoline prices and general economy (people can't afford a car.) North Arlington isn't hurting that badly though, so this data might not point to that direction.

by charlie on Feb 7, 2011 4:20 pm • linkreport

Great comments; thanks.

The regression analysis idea sounds good. If only I hadn't taken statistics in the 80's I might remember how to do that.

Over a full year, I would agree that events, etc. shouldn't be discounted, but since I was trying to compare two short periods between two years to get a sense of how ridership had changed, I felt an event like that would overwhelm the other factors.

Lastly, Arlington is adding a number of new counters in several new locations, which will allow for more in-depth and rich data about ped and bike usage throughout the county.

by Steve O on Feb 8, 2011 9:08 am • linkreport

Add a Comment

We are switching over to our new website. Commenting on the old site is now disabled. Thank you for your patience and pardon our dust!

Support Us