The methodology behind Global Plastic Watch is explained in detail in arXiv, an open-access repository of pre-publication found here. In brief, waste sites are detected using artificial intelligence that analyse free and publicly available satellite imagery from the Sentinel-2 satellite program operated by the European Space Agency. The resolution of this dataset is too poor for the human eye to be able to identify plastic waste aggregations. However, neural networks have been trained to analyse the spatial, spectral, and temporal information in the satellite data to classify whether a given region contains plastic waste. After a site is confirmed, the platform pulls information from external data sources to generate site attributes, like the number of people living nearby, or distance of the site to a waterway.

Every waste site that the Global Plastic Watch system detects is validated by a trained reviewer using multiple sources of high-resolution satellite imagery and on-the-ground imagery where available. The system are unable to detect different types of waste.

Global Plastic Watch was designed to be updated monthly. Monthly updates will allow monitoring of changes in the sites over time, which will allow users to monitor the success of interventions to reduce plastic waste as well as identify sites where risk of leakage to the environment may be increasing.

This map identifies waste sites which existed between January, 2017 and May, 2021. For each of these locations, we monitored their boundaries every month from as early as December 2016.

Twenty five countries have so far been mapped by the Global Plastic Watch tool. They include all of South-East Asia, Australia, and the top 20 countries in annual plastic leakage into the oceans according to scientific publication Science Advances. The first edition of Global Plastic Watch was published in May 2022 and will continue to be updated regularly as more data is analysed.

The Global Plastic Watch tool makes no automated distinction between managed and unmanaged waste sites. The system predominantly detects and displays aggregations of waste containing plastics. The system can also detect different types of waste.

Each satellite image identified by our machine learning algorithm as a potential waste site is reviewed by a multidisciplinary team composed of data scientists and remote sensing imagery analysts. However, there are still waste aggregations that go undetected.

Examples of missed detections might include waste that is shielded from the satellite’s view underneath tree or cloud cover, waste aggregations smaller than 10 x 10 metres, or compositions of waste that are different from those that we trained our networks to detect.

Given the scope of this work, we do not detect buried waste, such as that buried in landfills. Future versions of the system will focus on detecting a wider variety of waste aggregations, with a particular focus on identifying smaller sites.

On Global Plastic Watch you can gain access to a variety of data for each identified waste site with plastic:

• Location of formal and informal land-based waste aggregations containing plastic waste, including micro dumpsites (as small as 25 m2).
• Validation of land-based waste aggregations through Google Street view or high-resolution satellite imagery.
• Temporal trends in the extent of the surface area of land-based waste aggregations, month-to-month as far back as December 2016 in some locations.
• Each site also displays a list of additional site attributes, queried from other publicly available datasets. The parameters include soil type information (clay and sand percentage, fine earth density, and soil great group identity from OpenLandMap), site elevation, drainage direction, slope (SRTM), landform type (Global ALOS Landforms), distance to nearest water bodies (OpenStreetMap), nearby population (WorldPop), and upstream drainage (MERIT Hydro).
• Individual sites can also be filtered by distance to water, population density, and waste site size.

The methodology behind Global Plastic Watch is explained in detail in arXiv, an open-access repository of pre-publication found here.

Data will be accessible through an API endpoint. You can register your interest here.

Minderoo Foundation, [year of update], Global Plastic Watch, electronic dataset access date: [month] [year],<URL>

• Plastic Waste Site: Aggregations of waste that the system has identified as present and likely to contain plastic.
• Waterways: We use data from OpenStreetMap to determine whether there is water near to a dump site. Some waterways or water bodies are missing from OpenStreetMap, but it is one of the most comprehensive datasets of this type that exist. These water types include natural water like rivers, streams, lakes and ponds, as well as man made water bodies like reservoirs and irrigation canals.
• Nearby Water: Nearby water is shown if the centre of a site is located within 250 metres of a known waterway or water body.
• Surface Area Change: We monitor the footprint of each site on a monthly basis. Because of the low resolution of the satellite data and dynamic nature of sites, this is challenging to do with pinpoint accuracy and precision. You may see that the waste site boundary does not align with the image on the map. This is because the map image is a snapshot from a single point in time. Waste aggregations are likely to have shifted before or after that image was captured.
• Site area: shown in the Site Attributes tab, refers to the median waste site footprint across all time points.
• Surface Area: as seen in the Change tab, is the area of the assessed waste site boundary at a given time point.

No, the population number is included to give an indication of the number of people living near a waste site. This helps with the prioritisation of interventions and helps quantify both the scale of the problem and potential impacts on local communities.

The tool detects plastic waste aggregations but can not differentiate between types of plastic. 

Email us at: gpw@minderoo.org

Email us at: gpw@minderoo.org

We follow the United Nations standards for country names and territories and express no opinion about the legal state of any country, territory or area nor concerning its delimitation, frontier or borders.