top of page

How Tip and Cue Satellites Can Support Plastic Waste Monitoring in Southeast Asia 

One way to understand the plastic waste problem is by using satellite technology and earth-observation strategies like Tip and Cue monitoring to assess the numerous variables of the situation. 

beach-g22cad46e2_1280.jpg

Plastic. It’s a problem that’s persisted for decades, despite the very apparent issues that come along with its production and disposal (or lack thereof). While no country is without blame, and indeed some have done a better job of mitigating the pollution issues that come with it, some areas of the world have been, and remain, the primary producers of plastic waste globally. 

 

According to data from World Bank, Southeast Asia is the number one hotspot in the world for plastic pollution and inadequate disposal. Estimates show that more than 11 million tons of plastic enter the oceans annually - and more than half of the top-polluting countries are in Southeast Asia. 

 

There are many reasons this part of the world has such a hard time dealing with plastic pollution; urbanization, the growth of the middle class, inadequate waste infrastructure - the list goes on. Regardless of the reasoning, without a change in waste management, our natural resources, oceans, and the life within them will continue to suffer, and someday, it may be too late to fix. 

 

To begin the reconciliation process, countries must first understand the root of the problem, which should be understood using factual data that provide genuine insights into the ‘causes and effects.’ 

 

One way to understand the plastic waste problem is by using satellite technology and earth-observation strategies like Tip and Cue monitoring to assess the numerous variables of the situation. 

 

Authorities and conservation leaders need to know which sites are the highest contributors to plastic pollution in the oceans, where they are located, what the influx of waste is, and how it’s exiting the waste disposal sites and entering the oceans. It’s a lot to take in, and without modern technology and space-based imaging, it's challenging - and nearly impossible - to achieve. 

 

This report examines a case study performed by researchers using Sentinel-2 satellite data to monitor and assess terrestrial plastic waste in Indonesia. The study aimed to compare waste sites identified via satellite with those in public databases and evaluate the factors that pose a risk to ocean pollution. 

 

The report will serve as an example of how Tip and Cue satellite technology can help mitigate plastic pollution in our oceans coming from terrestrial sites - something that can immediately reduce the harm to our oceans, marine ecosystems, and planetary health. 

Study Background 

 

Plastics are one of today’s most significant pollutants, especially in our oceans. When plastics reach our oceans, they wreak havoc in countless ways, often in ways determined by the plastic's size, type, and biodegradability. Plastics can entangle wildlife, become entangled in coral reefs and other benthic surfaces, accumulate in giant patches and threaten birds and fish, be ingested by animals mistaking it as food, break down into microplastics and end up on our tables as food, and the list goes on.
 

How does most of the plastic in our oceans end up there? Mismanaged waste.

It is estimated that nearly ¾ of the plastic waste in our oceans comes from terrestrial sources, and as much as 90% or more enters the oceans through rivers, streams, and tributaries in watersheds that lead to the sea. More than a thousand rivers worldwide transport plastic into the oceans daily, and most of the top 10 severe are in South and Southeast Asia. 

 

Since plastic pollution has become so severe over the past decades, researchers, policymakers, and governments are beginning to take action with updated plastic management strategies. They have found that to make positive change, they must fully understand local municipal waste sites, hotspots for littering, where people are illegally dumping trash, and other leakage-related issues at waste sites. However, since data and information have not been collected consistently over the past decades, they must innovate to gain genuine insights. 

 

One way of digging into the root of the issue is with remote sensing and data analysis - which can allow users to do things like measure waste distribution, monitor sites, and build strategies backed by data rather than simply using prediction modeling. 

 

Recent studies using satellites and earth observation strategies have been executed successfully, showing promise in mediating the world’s pollution problem with these types of technologies. Methodologies include:

 

  • Characterizing spectral signatures of floating debris.

  • Analyzing spectral diversity between plastic waste and land.

  • Monitoring land surface temperatures in landfills to determine management strategies.

  • Using spectral, temporal, and structural information to determine how much plastic is in a landfill. 

  • Monitoring tributaries like streams and rivers to assess the outflow of plastic waste to the oceans.

 

Using some or all of these methodologies, users can monitor the waste management at known land waste sites and even detect unknown or undocumented waste sites yet to be recorded. The system is scalable, repeatable, and more cost-effective than previous strategies and allows new pathways to mitigating plastic pollution on land and at sea. 

 

Using the data and analyses described above, Tip and Cue monitoring is a strategy that can be deployed for further exploration, which we discuss more in the sections below. 

Study Methodology

 

A neural network system was developed to analyze different characteristics of satellite data from the Sentinel-2 constellation operated by the European Space Agency and identify terrestrial waste sites that collect a significant amount of plastic waste. 

 

The idea was that with enough data, the neural network could be trained to identify significant sites with enough waste to target for monitoring purposes. However, only a handful of known waste sites were available to start, and when used alone, spectral and/or spatial data returned unclear insight into the waste components for each site. 

 

Consequently, the neural networks that the team designed and used to assess the Sentinel-2 data aggregated and analyzed spectral, temporal, and spatial information - working together to cross-validate each set of individual data. 

 

The first stage operates on a per-pixel basis to maximize the number of data from a given site and reduce overfilling of spatial information for a waste site. The second component of the network is temporal, serving as a way to clarify some of the diluted data returned from only a spatial analysis. 

 

By analyzing changes in time and seasonality for different waste components, it becomes more clear which type of waste (i.e., plastic, organic, etc.) a waste site is composed of. The dataset used to train the data model is updated regularly so as to ensure the most accurate results possible. 

Data Sources

 

The primary data source for this study was the open-access dataset of the Sentinel-2 satellite program run by the European Space Agency. The dataset comprises data with moderate to high spatial resolution, a multi-spectral range, and a temporal revisit rate of five days. The log contains data collected since 2015, which is used alongside open-source information and imaging (i.e., Google Earth, etc.) for site validation. The data portal is available to the public and contains parameters like soil type, soil makeup, soil density, site elevation, landform types, distance to rivers and tributaries, and population estimates for nearby locations. 

Data Labeling

 

To start, the team used a set of 10 known waste sites located in Bali, Indonesia. Negative-class sites were chosen to capture the terrain distribution and bias sites with more waste present than natural vegetation. Bali is dominated by tropical forests outside of urban areas, so the models were trained to confirm or deny sites using this information. 

 

After training the dataset throughout 13 rounds of training, 213 locations were identified with significant waste, and 345 regions were identified without waste.

Validating Data Outputs

 

Once the system/neural network is trained and is able to identify sites with a growing or continuously changing waste profile, the observed analyses must be validated.

 

The team used high-resolution satellite imagery hosted on Google Earth and similar platforms like Planetscope to validate the data outputs. The high-resolution data from these sources are often available and clear enough to identify important characteristics within or around a landfill site. 

 

However, this approach has significant drawbacks, including inadequate or outdated imaging available to assess an ongoing threat or problem. Further, the data provided by sources like Google Earth and OpenStreetMap is limited. Of the 213 sites identified with significant plastic waste, only 126 had imagery available on Street View. The alternative would be to use imagery from Planetscope, but this imagery is often outdated and won’t provide the detail needed to assess the threat adequately. 

 

So, what can users do to overcome the challenges associated with data validation, as described above?

How Tip and Cue Methodologies Can Be Deployed to Overcome Data Validation Challenges 

 

Rather than rolling the dice on whether or not data is up-to-date and of high-enough quality to satisfy research demands, why not employ companies that operate small satellites in LEO to deliver high-resolution data and imagery in real time?

 

Companies like Spire Global, Planet Labs, and ICEYE all offer earth observation services with small satellites, which can be used with Tip and Cue monitoring strategies to save time, money, and resources. 

 

In the above study, researchers could have partnered with one of these companies to task high-resolution satellites to observe and collect data on all 213 waste sites in real time, eliminating any concerns of wasted effort. Doing so would allow for conclusive information that could drive genuine change and ensure future efforts do not fall by the wayside. 

Study Results

 

Identified waste sites in Indonesia were evaluated 9 different times between January 2019 and March 2021 - producing more than 160 million predictions at the pixel level and over 600 million patch classifications. 

 

The model detected 374 waste sites across Indonesia, manually confirmed through a review process. This number was more than twice those identified in public databases before the study - showing a massive void in awareness of the plastic pollution problem in the country. The nature of each site varied quite a lot, with some sites being government-run and others being informal local dump sites. 

 

Results showed that the system could detect approximately 3 previously unknown waste sites for every site it missed. While this isn’t perfect, it is a step in the right direction. 

 

However, as described in the sections above, employing companies operating small satellites with Tip and Cue capabilities could help turn the tide toward success. This strategy will be imperative if the country wishes to gain a better hold on the plastic pollution problem and develop better waste management strategies in the future.

bottom of page