In Washington D.C., we’re lucky to get a new chunk of Capital Bikeshare trip history data every quarter. For the end of the year, I created a new tool for examining the data from 2013. How does the weather affect ridership? How similar are CaBi ridership patterns to regular cyclists in the area? How similar is CaBi to other cities’ bikesharing systems?
I collected the data in an interactive bar chart, inelegantly called the 2013 daily ridership statistics for Capital Bikeshare. It has the daily “bikeout” (the number of bikes checked out) totals for CaBi, for both subscribers (those with monthly or annual memberships) and casual members (using 1-day or 3-day passes).
You can also get bikeout stats for an individual station. For many stations, this number may be much different from the number of bikeins, creating an imbalance, but for this study, I used only bikeouts.
The above is a sample bar chart – make your own chart here.
I got my weather data for the Washington region from wunderground.com, and the number of daylight hours from the Astronomical Information Center. Wunderground included 21 different measurements for each day. I looked at all of them and included the five most-influential of each type of measurement.
What are the most influential environmental factors that correlate highly with daily CaBi bikeouts? According to my study, they are:
- High temperature
- Minimum dew point
- Hours of daylight
- Average visibility
- Precipitation
You can find exact numbers in the table below. The high temperature was the most highly-correlated variable for CaBi ridership, including the subsets of casual users and subscribers.
To see how CaBi ridership compared to “normal” bike traffic, I got bike-counter data from Arlington, which has automatic counters on trails and in the streets. I used data from 13 of Arlington’s counters and added them together for the daily totals. It turns out this data set is also the most highly correlated with the high temperature. No surprise, people prefer biking when it’s warm out.
The minimum dew point is related to the temperature, and is a factor for both night-time temperatures and rain clouds. Average visibility is affected by storms.
The amount of daylight had a weak correlation with CaBi usage, perhaps because city streets are lit. For the Arlington bike traffic, where much of the data comes from unlit trails, it had a high correlation, suggesting that biking can be increased in Arlington by improving the lighting.
I expected precipitation to have a stronger correlation, but it’s important to note that we are comparing data sets for the daily totals. While few people are willing to bike while it’s raining, it’s less risky to bike on a day on which it has rained or on which it will rain.
You can also look at total ridership for Metrorail. Not surprisingly, Metro ridership shows barely any correlation at all with environmental variables. We know it takes a significant weather event for Metro to be affected (via PlanItMetro, Influencing Factors on Weekend Metrorail Ridership).
One of the most easily-discernable differences between CaBi subscribers and casual users is how they use the bikes on weekdays versus weekends. Subscribers have higher riderships on weekdays, suggesting they use the bikes to commute to work. Casual ridership does the opposite, with ridership doubling on weekends. The Arlington bike-count data also is higher on weekends, though it’s not as pronounced.
I also added data for bikesharing systems from London (Barclays) and the Twin Cities (Nice Ride). Barclays gets higher ridership on weekdays, while Nice Ride get higher ridership on weekends. Looking at daily ridership numbers, CaBi has a higher correlation with Nice Ride, even though Nice Ride shuts down for the winter.
Can you use this tool to discover your own insights? I’ve put together a few more of my own conclusions at Looking Back at 2013 CaBi Data. We’d love to hear your feedback in the comments below.
Correlation of biking with environmental factors, using daily totals from 2013
CaBi bikeouts | Arlington bike counts |
|||
---|---|---|---|---|
all | subscribers | casual | ||
max temperature | 0.80 | 0.70 | 0.60 | 0.76 |
mean temperature | 0.77 | 0.67 | 0.58 | 0.73 |
min temperature | 0.72 | 0.62 | 0.55 | 0.68 |
min dew point | 0.64 | 0.56 | 0.48 | 0.59 |
max dew point | 0.63 | 0.55 | 0.46 | 0.58 |
mean dew point | 0.62 | 0.55 | 0.46 | 0.57 |
hours of daylight | 0.61 | 0.46 | 0.57 | 0.71 |
mean visibility | 0.44 | 0.40 | 0.30 | 0.39 |
min visibility | 0.37 | 0.32 | 0.29 | 0.35 |
precipitation | -0.31 | -0.29 | -0.23 | -0.25 |
max gust speed | -0.25 | -0.25 | -0.14 | -0.18 |
mean wind speed | -0.25 | -0.26 | -0.13 | -0.19 |
max wind speed | -0.24 | -0.22 | -0.15 | -0.19 |
max sea level pressure | -0.24 | -0.22 | -0.16 | -0.21 |
cloud cover | -0.24 | -0.18 | -0.23 | -0.23 |
min humidity | -0.15 | -0.10 | -0.15 | -0.18 |
max humidity | 0.08 | 0.09 | 0.03 | 0.05 |
mean sea level pressure | -0.07 | -0.07 | -0.05 | -0.06 |
max visibility | 0.06 | 0.07 | 0.03 | 0.03 |
mean humidity | -0.05 | -0.02 | -0.07 | -0.08 |
min sea level pressure | 0.05 | 0.04 | 0.04 | 0.06 |
high correlation |
medium correlation |
low correlation |
Photo by M.V. Jantzen