The world’s leading publication for data science, AI, and ML professionals.

Impact of Snow on Road Traffic Operations: Case Study -State Highway 285

Hello everyone,

An Empirical Analysis of Snowfall Events on Macroscopic Traffic Variables

I was recently working on collecting traffic Data for a state highway going through mountains and I decided to investigate the effects of snowfall on traffic operations. While looking for similar studies, I found this research study from China looking at the impact of rain on traffic.

Survey the impact of different rainfall intensities on urban road traffic operations using…

Photo by Brett Sayles from Pexels
Photo by Brett Sayles from Pexels

I thought I would follow a similar approach to understand snow impacts and create a simplified macroscopic fundamental diagram (MFD) to summarize Traffic operations. To achieve this objective, I collected traffic data for state highway (SH) 285 and climate data from a nearby weather station. The gathered data are summarized below:

  • ATR Traffic volume (per hour) – From 2015 to 2019
  • ATR Traffic speed (per hour) – From 2015 to 2019
  • ATR Vehicular classification (per hour)
  • Climate data (daily) – Global Historical Climatology Network (GHCN) – From 2015 to 2019

I am not going to talk about how I collected traffic data in this piece but I will show how I gathered Weather data for the study site.

Climate Data Online

This beautiful API provides web services that provide access to current and historical weather and climate data. An access token is required to use the API, so, don’t forget to get yours. Each token will be limited to five requests per second and 10,000 requests per day (wow). A python wrapper for this API can be found in this git repo (thank you Jeff Ely and Remi Rampin!):

Jwely/cdo-api-py

As usual, let’s first import everything we will need.

The next thing we will do is to define our client and configure its query variables. A bounding-box is needed to find weather stations. Once we get them, we need to define the timeframe and data types such as snow, snow depth, etc. to retrieve the data. Don’t forget to update your token in the code below.

When the configuration is completed, the rest is to query each station and store climate data into a pandas data frame.

Okay, now we have our climate data. Let’s take a quick look at the distribution of snowfall at SH285.

You can see that his location had a maximum of 10 inches of snow in a day. Therefore, I divided my dataset into two different sets including Snow < 5'' and Snow > 5 '' to see the difference between light and heavy snow conditions. Of course, one can use different intervals to define such conditions.
You can see that his location had a maximum of 10 inches of snow in a day. Therefore, I divided my dataset into two different sets including Snow < 5” and Snow > 5 ” to see the difference between light and heavy snow conditions. Of course, one can use different intervals to define such conditions.

After retrieving the climate data, I coupled this with traffic data including volume, speed, and classification to have a rough idea of what happens to traffic operations. I created the MFD below using speed and flow variables and fitted a polyline for each.

It is apparent that we have some changes in the MFD because of snow. However, when the snow depth is over 5 inches, its impact on speed is less than light snow. This was really interesting so I decided to see how vehicle classification and volume change for light and heavy snow conditions.
It is apparent that we have some changes in the MFD because of snow. However, when the snow depth is over 5 inches, its impact on speed is less than light snow. This was really interesting so I decided to see how vehicle classification and volume change for light and heavy snow conditions.

I calculated the average volume for each vehicle type during no snow, light snow, and heavy snow conditions to see how they change. The table below summarizes the results.

From the table below, you can see that during light snow conditions, truck traffic is impacted the most. Truck volume goes down about 22%. However, heavy snow has only additional 8% decrease in truck volume while it takes vehicular traffic 30% more! During heavy snow conditions, passenger vehicles do not go on roads (35% less traffic) which may explain why traffic speed is increased overall. I am also certain that there might be other potential reasons for this phenomenon.

Another interesting takeaway from the data is that vehicles tend to go much slower when the hourly volume is very high or very low during heavy snow conditions. The figure pretty much illustrates that if you are the only one on the road or if there are many vehicles traveling with you, you prefer to drive slower than usual. The data also hints that the capacity of the road is reduced since the breakdown of speed starts at a much lower traffic flow (~2,000 vph) compared to the recurrent conditions (~2,500 vph).

The data shows that average traffic speed is decreased by snowfall, yet, this decrease is not as significant as the drop in traffic volume. In some cases, average traffic speed is increased due to the low volumes experienced during heavy snow conditions. I would like to collect more data from locations that get snow in Colorado to see if a similar relationship can be observed at various locations.

Thank you for reading!


Related Articles