Google helps tackle plastic pollution in the Mekong River, aided by machine learning

UN Environment Programme will build and deploy a new machine learning model capable of uncovering a more detailed and accurate view of plastic pollution in the river.

The UN Environment Programme (UNEP) has partnered with Google to fight plastic pollution in the Mekong River, leveraging citizen science and machine learning to understand the magnitude of the problem impacting the longest river in Southeast Asia.

Aided by technical advisory support from Google, the CounterMEASURE project will build and deploy a new machine learning model capable of uncovering a more detailed and accurate view of plastic pollution in the river, which flows through China, Myanmar, Laos, Thailand, Cambodia and Vietnam.

“The plastic pollution problem demands creative solutions,” said Dechen Tsering, regional director of Asia Pacific at UNEP. “The CounterMEASURE project has already deployed an array of modern technologies to help map plastic pollution in rivers.

“With Google’s support, we are able to improve the detail and accuracy of this work, which will help UNEP develop guidance for local and national governments to effectively tackle plastic pollution in rivers.”

According to Tsering, one of the main challenges in fighting plastic pollution is determining how exactly it enters - otherwise known as “leaks” - into bodies of water.

Following work in the Mekong region between 2019 and 2020, the CounterMEASURE project - with the support of the Geoinformatics Center (GIC) at the Asian Institute of Technology - developed techniques which assessed plastic leakage into the Mekong River using geospatial data and images of plastic waste supplied by researchers and volunteers.

The new machine learning model developed by UNEP, Google and GIC will add to these efforts, creating a tool capable of generating a "more detailed and accurate view" of the plastic pollution problem in the Mekong and rivers beyond.

Specifically, citizen science will aim to strengthen the algorithm through community-sourced, annotated images with the machine learning model contributing to the development of a plastic leakage hotspot map. The map can then be used by local and national governments to determine how to target policies and resources to prevent plastic leaking into waterways.

“The plastic pollution crisis needs high quality, scalable solutions that can be used in areas that run the highest risk of leaking plastics into our oceans,” added Emmanuel Sauquet, vice president of Google. “Technology is critical to enabling these solutions.

"We are excited to support UNEP in creating this open source machine learning model that will help detect plastic pollution in streets and river banks. UNEP's influence with local governments will allow effective action to be taken to stop plastic leakage, and scale this solution globally."

Estimates show that rivers transport millions of tons of plastic into the oceans every year - some 95 per cent of that discharge comes from only 10 rivers, eight of which are in Asia Amongst those eight are the Mekong and the Ganges rivers, viewed as the lifeblood for hundreds of millions of people in Southeast Asia and India.

Going forward, plans are also in place to help expand the CounterMEASURE approach beyond the Mekong and Ganges. The collaboration will also contribute to the Global Partnership on Marine Litter.