EDITOR’S NOTE: This is the second in a four-part series of posts this week by statistician Mike Logsdon. Mike bikes across the Fremont Bridge every day, and put together this analysis of the bike counter data “mostly for the amusement of friends and coworkers.” I hope you enjoy it as much as I do.
Seattleites talk a tough game with the rain. We proudly eschew umbrellas and hoods, and claim not to be bothered by showers and gloom. But how true is that? Do we really treat the rain like the bishop in Caddyshack golfing the round of his life? Or do rainy days send Seattleites running for the relative comfort of cars and busses?
To answer this question, we can merge the bicycle counter data with weather data. I used NOAA Quality Controlled Local Climatological Data from Boeing Field, to record temperature and precipitation corresponding to the observed times in the bicycle data. Being from Boeing Field, the weather data probably aren’t a perfect match for Fremont, but for the purposes of this exploration should be plenty similar. (The website at seattle.gov has more detailed weather data from the University of Washington plugged into a Tableau interface. I had QCLCD data readily available, so that’s what I used. The UW weather recorded more elaborate measurements, like Solar Watts/m2 and wind speed, but I assume that season, temperature, and rain encapsulate most of the variability explained by the more nuanced weather information. If somebody was paying me for this I would spend the time to secure the more detailed data.)
To investigate the effect of weather, we can first view the somewhat complicated in appearance Figure 3. This graphic shows daily total bicycle trips as a function of daily average temperature, with the points sized according to daily precipitation. The data were separated according to weekend and weekday, with overlaid trend lines with respect to temperature. Most obvious is the strong trend of ridership with respect to temperature, a no-brainer. Notice also that the largest bubbles, which denote the rainiest days, almost all occur beneath the respective trend lines. That suggests, in at least an exploratory fashion, that rain indeed deters bicycle trips, even when adjusting for temperature1. The rainy days nearly all fall below the counts suggested by temperature alone.
But we’d like to make it official. One obvious way of estimating the effect of rain on bicycle trips is to fit a regression model. It was already clear from Figure 3 (and common sense) that temperature strongly correlates with volume of bicycle trips. However, daylight and seasonality also surely play a role. I know that I personally take fewer bicycle trips in the dark, regardless of weather. Similarly, potential riders may also experience a wintertime mental block, where bicycling to work on a nice day in February never even occurs to them, because it’s February. As such, I fit a regression model to the daily total ridership as a function of daily average temperature, daily total precipitation, and a sine wave to encompass seasonality and daylight (again, with the more detailed UW weather data I could have also considered incident solar and wind speed, but I assume that the terms I included account for most of the variability explained by weather).
Let Yi denote the number of bicycle crossings for a particular day i. Similarly, let tempi denote the average temperature, precipi denote the total precipitation, and datei denote the date, scaled by 2π/365 to convert to a sine wave with period one year. The form of the regression model was as follows:
EYi = β0 + β1tempi + β2precipi + β3sin(datei) + β4cos(datei)
Fitting this model, which is to say adjusting for season and temperature, one additional tenth of an inch of rain was associated with 92 fewer bicycle crossings per day, with 95% confidence interval (68, 115). We may safely assert that Seattleites do indeed hang up the bike shoes and messenger bags when the rain starts falling.
In the next part, we will look at the effect Bike Month has on ridership levels.
1In Seattle, precipitation is confounded by season. A plot of bridge crossings against precipitation will surely show declining ridership on wet days, but we wish to disentangle whether that decline is related to the rain itself, or the fact that the rain typically falls during colder, darker winter months.
Comments
22 responses to “A statistical analysis of biking on the Fremont Bridge, Part 2: Rain”
My personal version of this is that I am significantly less likely to bike on a rainy day if it is raining in the morning. I don’t like arriving at work soaked and cold and it is hard to overcome the initial trepidation of stepping out into the rain from a warm house. If I’m already at work and it starts raining… oh well, I’m just going to bike home in it anyway, no big deal.
My bail out days tend to be those when it is forecast to rain hard all day and it is raining medium to heavy when I wake up. Not very many days meet both those criteria.
I am not affected by temperature or darkness, with the exception of cold days with significant ice on the roads.
I love the rainy days, fewer clueless riders out there. Especially those days that are rainy in the morning and clear up at midday.
I’m with you on this. It’s also why I like riding through the cold Winter.
It’s really brilliant how the inexperienced riders can provide us with self worth all year long. In the summer we can feel indignant about how we need to use our superior skills to navigate safely in their inept vicinity. In the winter we can feel smug that we, the ones who suffer wind, hail, and snow, are the only ones true enough to be out in the dark drizzle.
Funnily enough, from time to time I experience a similar behavior with a certain subset of drivers, the sort who claim that all drivers from city x don’t know how to drive.
Some similar work using Vancouver data has been done with older data:
http://canadianveggie.wordpress.com/2011/02/20/dunsmuir-bike-lane-by-the-numbers/
One question about rain that I have had for a while is if there are echo effects after rainy days. In other words, do people ride less the day after a rainy day? Are people really checking the weather every day and considering the likelihood of rain rationally or reacting to the fact that they recently got soaked and the sky is gray again? It seems like winter here features very long stretches of dark gray but not necessarily a ton of rain. Perhaps our weather has a stronger deterring effect on cycling than temperature or cumulative rain would suggest, because the clouds are just so relentless.
The echo effect was a great suggestion! I re-ran the regression model including variables for the precipitation in the previous day, and the total precipitation in the previous two days. The effect of the current day’s precip changed to 86 fewer riders per tenth of an inch, and the effect of the previous day’s precip was a whopping 65 fewer riders per tenth of an inch. The previous day’s precipitation was almost as important as the rain that was actually falling (or forecast to fall later in the day). The effect of the two-day running total was not significant. It looks like Seattleites both look at the weather forecast and remember what happened yesterday when choosing whether to bike, but don’t necessarily care about what happened the day before yesterday.
Or (tongue in cheek) some people aren’t riding on rainy days and different people aren’t riding if their shoes are still wet from riding in the rain the day before… (because that can influence my decision)
You are awesome! Thanks for checking that out. I am shocked that the previous day’s effect is that high though it definitely jives with the way I tend to feel the day after a heavy rain.
Great analysis! Definitely matches my anecdotal evidence. When I wake up surprised by the beautiful weather I always find that the counter is lower than if the day is in a streak of nice days. Always figured that people make their decision about riding before the go to bed and even if they wake up and its nice, they’ve already decided on the bus/car so they go with it.
I wonder if there’s a contingent of riders who ride in on rainy days (perhaps before the rain starts) and then take alternate transportation home, leaving their bike at work/school/destination. Then the next day, they’re forced to go back to pick up their bike before riding it home. Might not possible to pick that up in the data unless there’s a strong AM/PM disconnect on rain days and the days after.
Nice question, Drew! I suspect there’s a longer-term effect – at some point a certain number of people decide that “bike season is over” and don’t ride again until “bike season” starts again, even on a glorious sunny day in February. Is bike season determined by temperature? day length? probability of rain? I don’t know, since I ride all year!
I’m shuddering just remembering the winter here. I don’t mind night riding, in fact, I’m about to go on a pleasure cruise after this post. I can get a kick out of riding in an occasional down pour. What makes me nuts is waking up every morning before dark and getting on my bike with no idea about what I’m going to get. Something about the combination of day after day of wet dark riding makes me so hateful. Even when it’s just a light drizzle, the wet air means less evaporation and more sweat in the clothes. Having to climb Capitol hill to get to work just made the sweat soak all the worse. Once something mildews, it never really smells fresh again. Wouldn’t it be cool if we had trails with little Safeco retractable roof all along? Combine that with little bike tunnels running east west through the north south hills in the North End. Wouldn’t it be nice?
I’m just glad it’s summer.
I find that the old school wool bike clothing meets the Seattle rain and defeats it. Yes I wear a waterproof outer layer for heavy downpours but the misting stuff, Ibex wool Shak jersey and Ibex long or knickers length tights make the ride a lot more enjoyable. As do a second pair of gloves for the ride home.
I like to bike in the rain because it poses more technical challenges and therefore requires a lot of concentration. I ride my all-weather bike, slow down, and stay out of traffic more than usual. I also bring as many changes of clothes as might be necessary.
Once the city constructed the Westlake sidewalk (Chesiahud Loop), it made it possible for me to commute safely in rainy & dark conditions, and therefore start commuting year round. It’s my hope that there will be more fail-safe alternative routes added like that and that more women will try biking to work year round. More of us would feel cheerful at work, regardless of the weather.
I continue to be concerned about some behaviors I’ve noticed, however, that I think make bike commuting potentially quite dangerous: 1) most bicyclists do not make an audible signal when they pass other bicyclists; and 2) the number of bicyclists commuting while wearing headphones is increasing. One morning I noticed 20% that passed me in the Westlake parking lots were wearing them. This really concerns me- it endangers me and it endangers them. I don’t think it’s ever a good idea to close one’s senses off while getting around – you need all of them! I can always pick out someone biking or walking with headphones – they don’t look around as much.
If you want to try a commute in the rain vicariously, here is a recording I made riding from Fremont to downtown a few years ago (warning, a bit monotonous, but shows what a safe commute in the rain feels like): https://www.youtube.com/watch?v=RJ-BY7ye5hE
If you’re interested in verifying the Boeing-Field-is-close-enough-to-Fremont assumption, I’ve got a data set for you:
https://github.com/bobhall/Commuting-weather
This is a log of my commute from Crown Hill to downtown Seattle. For each morning and evening of my commute, I assign it a status of either Wet or Dry. Sadly, the first date in this log is only Sept 23, 2013.
It would be interested to see how well this correlates with the Boeing Field data.
[…] ← A statistical analysis of biking on the Fremont Bridge, Part 2: Rain […]
Mike, thanks for doing these analyses. I love thinking about what influences biking in Seattle. I’m particularly interested in the outliers on your plots, such as the really cold but non-rainy day that had about 2500 trips (basically the leftmost blue dot on the graph) and the 64 degree-ish day that only had about 1600 trips. Any thoughts on what might have caused those? For the cold day with lots of riders, maybe it was the one sunny day amidst a week of gray weather? For the warm day with less riders than expected, maybe it was a Thursday or Friday following a week of great weather?
I’m not sure I have a compelling story for any of those days. Even the unusual ones you described still fall within the bounds of normal daily variation when adjusting for season and weather (the residual trips of Part 3, which will show up again in Part 4). The counts from those days might have been caused by some pattern of nice weather and day of week, or they might have just been random.
[…] Seattle Bike Blog has some great analysis of biking trends using data from the Fremont Bridge. Yesterday’s rainfall is almost as important as today’s rainfall when estimating daily bicycle […]