Header image

Using AI/Big Data to Analyze Urban Mobility Patterns During the Pandemic

October 16, 2020 blog-post big-data geospatial partnership traffic traffic-waze-geospatial transport waze business development government health healthcare logistics transport

In the second part of our series of blogs, the ADB and Thinking Machines Data Science explore using Artificial intelligence and Big Data to craft solutions in this pandemic. This piece focuses on changing patterns in urban mobility and reflects on the opportunities and limitations of data science techniques and freely available open datasets.


In this blog, we use aggregated openly available anonymized transport data from several transit applications (e.g. Waze Connected Citizens Program, Google and Apple). All these apps have a symbiotic relationship with their users: users benefit from useful, real-time information (e.g. traffic patterns), in return for providing data points based on their own journey. This means that the more active users, the better the quality of the data and “the better the App”. In economics this is referred to as a “thick market externality” that captures the strong non-linearities in networks.

It is important to note that transit applications are not ubiquitous, and the data generated reflects only those individuals with access to such digital technologies. It therefore provides only a snapshot of the overall trends. With Google and Apple, our view is limited to only those with access to smartphones with location history switched on, while Waze Connected Citizens Program (CCP) data is limited to app users with access to private vehicles--accounting for only 20% of the larger population reliant on public transport. However, we see this blog as the commencement of the dialogue with real-time data providers to better understand changing patterns of mobility and support decision-makers to make data-driven, evidence-based plans.

As countries slowly recover from the pandemic, we are seeing a significant shift in transport patterns. Using publicly available transport data we found that in Southeast Asia ridership fell by over 90% in March compared to pre-pandemic patterns in January (Figure 1). This was amidst various degrees of lockdown and social distancing measures imposed by governments, including restrictions on public transport.

Kuala Lumpur had the biggest drop of 94.2% in late April. As public transport across cities slowly returned to full operation in recent months, ridership has partially recovered, led by Bangkok and Singapore.

As city lockdown measures were eased or removed, there was a rise in private car use. In late July, Apple mobility data recorded over 30% more activity from private vehicles in Bangkok and Jakarta than in January (Figure 2).

We examined what this means for the Philippines. We explored mobility data across three metropolitan areas in the Philippines--Metro Manila, Cebu City, and Davao City. Metro Manila is by far the largest, with a population of more than 12 million and encompassing 16 municipalities.

Throughout the pandemic, the Philippine national government implemented varying degrees of movement restrictions. In March, Metro Manila, Cebu City, and Davao City were under strict movement restrictions and traffic flow as measured by jams on Waze CCP dropped almost completely. According to Waze CCP data, there was a 94% drop in traffic flow from private vehicles across the three metropolitan areas during the strictest quarantine measures (Figure 3).

In July, Cebu City was 94% lower than the previous year and Davao City was 78% lower. During the same period, Metro Manila was normalizing with traffic flow from private vehicles at about half of the previous year’s level.

With even greater granularity, we also observed that for some of the most congested roads in Metro Manila, Waze CCP data on private vehicle traffic flow in July came close to the previous year’s traffic patterns, with some roads only 6% away from previous levels. At the road level, we see that although private vehicle traffic flow dropped during the strictest restrictions, it began to rise quickly as they were eased, similar to municipality- and city-level patterns in Metro Manila.

In light of COVID-19, mobility undoubtedly plays a significant role in both the spread and containment of the virus. In municipalities across Metro Manila, we observe a pattern between the number of reported traffic jams and COVID-19 cases (Figure 4). Indeed COVID-19 spikes occur around the same time that mobility is rising again, and the same trend is apparent even when breaking it down to the individual cities.

However it’s important to note that the rise in mobility is defined by the rise in traffic jams as recorded by Waze CCP, primarily from private vehicles. This reflects some degree of correlation but does not inform us of causality as there is not sufficient depth of data to trace human contact, frequency, origin or destination, or mode of travel. These limitations on data sources and COVID-19 cases mean we need to consider other factors outside mobility.

This is made apparent when we look at the data in Cebu City and Davao City. In Davao City, local government restrictions continue to keep mobility low as recorded by traffic jams from private vehicles on Waze CCP, and we do not see a rise in travel as we do across Metro Manila (Figure 5). Throughout July, citizens in Davao City were still under a 9 pm to 5 am curfew and trips outside for essential shopping were only allowed on certain days of the week and required the use of passes.

Cebu City saw spikes in COVID-19 cases in June despite the fact that the data shows that mobility remained extremely low. What else could be at play here in the rise of cases? This points to the need to consider other factors beyond mobility, including social distancing measures, sanitation and hygiene protocols, testing capacity, and access to healthcare services and facilities.

According to Google mobility data (Figure 6), Metro Manila residents visited groceries and pharmacies more than any other destination outside of their homes, followed by workplaces. As restrictions ease, businesses are reopening, and people are returning to their regular office hours. From June to July, the data also shows that residents are avoiding recreational destinations, namely parks, retail locations, and public transport stations.

Moving forward, we need big data, but we also need open data and open source initiatives from the public transport sector for a more complete analysis. Our view thus far has been formed by users with access to private vehicles using transit apps. In order to obtain this, we need responsible and ethical data sharing between the public and private sectors. We can build on existing regulations, such as the Philippines’ Data Privacy Act 2012, which requires consent and agreement between parties when sharing and collecting data.

Trends in the Philippines and around the region are pointing toward shifting behavior in transport. We need to build back better to address these changing demands by improving data systems in order to adapt, innovate and be more responsive.

Check out the abridged version of this blog over at the Asian Development Blog here.

This is the second blog in a series by the Asian Development Bank and Thinking Machines Data Science, Inc. where we explore ways we can use big data, artificial intelligence, and machine learning to craft development solutions during the pandemic.


AUTHORS

Stephanie Sy, Anica Araneta
Thinking Machines

Hanif Rahemtulla, Bruno Carrasco, Stella Balgos
Asian Development Bank

MORE STORIES

Kailan ka ikakasal? Four data-driven comebacks to use at your next family reunion

Actually, Tita, the number of Filipinos getting married has declined 15% in the past 10 years. It's part of a nationwide trend.

A second look at PNP crime statistics: Taking population into consideration

A look at the number of index crimes per 10,000 residents for 15 Philippine cities.

Makeover Your Charts with These 5 Design Hacks

Small tweaks can make a big difference when it comes to data design. Here are five case studies from our recent data storytelling workshop.