In year of El Niño, new uses of big data can aid disaster management

UN Global Pulse integrates artificial intelligence and the Internet of Things to improve response to crises

An army officer helps to distribute masks to car passengers driving through the Indonesian city of Palangkaraya during the 2015 haze crisis. Aulia Erlangga, CIFOR
29 March 2019

A looming El Niño event could make 2019 the hottest year ever recorded, and this is concerning news for countries in Southeast Asia. The mega-fires caused by the severe 2015 El Niño ravaged more than 26,000 kilometers of land in the region, destroying peat forests and causing a massive haze crisis that spilled over to neighboring countries. Resulting premature deaths around the region numbered more than 100,000.

Data science has since stepped in to help authorities better manage fire and haze events, and UN Global Pulse is now exploring some of the latest advances in artificial intelligence and the Internet of Things to help authorities and citizens make timely, informed decisions in environmentally critical moments.

In Indonesia, the crisis analysis and visualization platform known as Haze Gazer combines official data with that of satellites and social media to map haze hotspots and the response of affected populations. Developed by Pulse Lab Jakarta, the tool has been adopted as part of the Indonesian Government’s early-warning system, and a prototype of it can also be freely accessed online.

Pulse Lab Jakarta has emerged as a leader in driving the use of big data and artificial intelligence for sustainable development and disaster management, a mission that entails both developing prototypes and making the case for their adoption worldwide.

“By layering new and traditional sources of data, the platform provides insights on the situation on the ground in near real time,” explains Pulse Lab Jakarta’s data scientist Anthony Mockler. “The analysis can drill right down to a city or a township level – we can analyze what people in that area are saying about the haze on social media.”

A river in Sebangau National Park in the Indonesian province of Central Kalimantan, one of the regions the 2015 fires affected most. The country’s peatland fires that year contributed to an economic loss of at least USD 16.1 billion. Aulia Erlangga, CIFOR

DEEP LEARNING, BACKYARD SENSING

The Haze Gazer resembles a tiered cake in that it can integrate layers upon layers of data in a single platform. Disaster management and government authorities can adapt the platform to its needs by selecting the sources of data most relevant to a region or country.

Developers at Pulse Lab Jakarta are continually improving the tool as well. Currently, they are working to add two new functionalities.

The first uses a set of machine learning methods to predict current air quality using photos that citizens post on social media. Using algorithms inspired by the human brain, this model learns from vast quantities of data to make its predictions – a subset of machine learning known as deep learning.

“Official data might deliver a daily reading for Jakarta at best, but this new model allows us to forecast air quality using social media images in conjunction with a few other conventional data sets,” says Mockler.

The lab is also exploring Internet of Things sensors. “Weather stations are expensive, but these sensors are affordable and can be installed in any backyard,” notes the data scientist. “These sensors can use Wifi or your own cell phone network to provide air quality information at a much greater level of detail.”

Ideally, a Lab partner would deploy the devices, and the Lab’s data analysts would then integrate the data into a dashboard where it can be visualized for further analysis. “By combining official air quality figures with estimates from our social media learning model and the sensors, we should be able to have amazing coverage,” he says.

IN THE NOW

Among other disaster-management tools, Haze Gazer offers governments the ability to see what’s happening on the ground in areas where disaster strikes nearly in real-time.

But it’s not just for high-level use; citizens and NGOs can also benefit from it, says communications specialist at Pulse Lab Jakarta Dwayne Carruthers. Humanitarian organizations, for example, can use the tool to see how fire and haze hotspots overlap with the location of vulnerable groups of people. This can help them respond more effectively and efficiently to emergencies.

Potentially, the platform could send information on immediate air quality via SMS to target groups, such as people with respiratory diseases and school administrators. “People could get notifications that the haze is really bad, and they’d better stay indoors or take the bus instead of the motorbike,” explains Mockler.

Carruthers notes that wider regional uptake of Haze Gazer depends on the will of governments; the negotiation of partnerships that grant access to a wider variety of data; and efforts to contextualize the platform to the needs of each country and the available data sources.

The good news is that more and more institutions and organizations are turning to Pulse Lab Jakarta to learn from their experience in building Haze Gazer. “Partners are looking around for the best solutions to avoid having to reinvent the wheel, so we can hopefully modify Haze Gazer and make it work in different contexts,” says Mockler.

The use of artificial intelligence and big data for disaster management is still an emerging field, but the experts at Pulse Lab Jakarta are determined to make it grow. “We want more partners and more data,” says Carruthers. “Anyone interested in talking about air quality, come speak to us!”

According to the Indonesian Forum for the EnvironmentMore than 2,200 fire hot spots were recorded across Indonesia in 2018, after relatively fire-free years in 2016 and 2017. UN Global Pulse

Leave a Reply