How COVID-19 Affects Parking Congestion

Today, according to recent news reports, 75% of Americans and 33% of the globe are either in quarantine, self-isolation or another form of lockdown as a result of the Coronavirus pandemic. People are not driving, taking public transport, or otherwise going anywhere of significance aside from health-urgent or food-necessary destinations. For those of you who do drive, coming back home or parking near your destination could be challenging due to so few cars leaving their parking spots.

Figure 1: Empty street in San Francisco, March 18th, 2020.

At Parknav (an Ai Incube brand) we are dedicated to helping people find parking. Therefore, we decided to go and run a number of surveys and analyses to see how these lockdowns are affecting people’s parking pressures. What we found was somewhat surprising and not immediately obvious.

When daily life is disrupted due to big events or crises, such as the Coronavirus pandemic, parking behavior naturally changes. This post describes some of the changes we have seen related to parking behavior due to the current world crisis. We used a combination of data we collected manually along with several other different data sources to estimate the probability of finding parking at a given street at a given moment in time.

In the early days of the pandemic, when people decided not to use public transport, there was a temporary surge of parking pressure, particularly around business areas such as grocery stores and pharmacies. However, our data is now showing that a new standard appears to be emerging.

Figure 2: Mission District in San Francisco before the COVID-19 outbreak (left), and after (right). Red streets represent 0–1 spots, yellow 1 to 3 spots, and green 3–9 spots on average during 2019.

In Figure 2, the left-hand side shows the average parking availability in San Francisco’s Mission District before the COVID-19 outbreak. The colors represent the average number of available parking spaces per street where:

• Red means that there is less than 1 parking space available on average through out the day.
• Yellow means that there are between 1 and 3 parking spaces available in average.
• Green means that there are more than 3 parking spaces available.

The right-hand side of the figure shows the same streets after the Coronavirus outbreak. Overall, we see that the proportion of “green” streets in the area before the outbreak was 36%, while after the outbreak, increased to 49% showing a general increase in parking availability.

In order to demonstrate this with a more concrete example, take the street segment of the 20th street between Mission St. and Capp St. in San Francisco. In Figure 3, we show the probability of finding parking throughout the day in that street before (blue line) and after (orange line) the outbreak.

Figure 3: Parking probability in 20th street between Mission St. and Capp St. The blue line represents the probability before the pandemic, the orange line represents the probability after.

Before the COVID-19 outbreak, the hardest time to find parking on that street was around 1pm. Post COVID-19, the chances of finding parking is now much higher overall with the most difficult time to find parking around 2pm (albeit, still much easier at peak times than prior to COVID-19).

These parking patterns can be verified not only by our service, which includes machine learning models for events like these, but also through other similar data sources such as satellite images, traffic data, and mobility data (e.g., GPS tracking data from cell).

On average, the predictions that we show in Figure 3 have a normalized precision of 0.82 with a recall of 0.37. For more information about what these numbers exactly represent, please refer to my previous post: Evaluation Metrics for Parking Predictions.

In conclusion, a side effect of the current COVID-19 pandemic is the change in parking pressure within cities, with places that used to have an abundance of parking at certain times no longer having available spaces, and places that used to have parking congestion now being relatively empty. As the situation develops and behavior patterns continue to evolve, more data and further analysis will be necessary.

For now, if you are searching for parking, consider how the parking pressure in a street might have changed: streets in residential areas may have heavy parking congestion, while streets in commercial areas may have available parking spaces at different times than before.

Written by Juan Mancilla-Caceres
Head of Ai and Data Science at Parknav