Harnessing Artificial Intelligence to Tackle Plastic Pollution: Lessons learnt from the River Eye project

Rivers are the key to solving the increasingly urgent issue of ocean plastic pollution. Fortunately, incredible progress in the field of AI has the potential to revolutionise the field of riverine waste monitoring, increase the scalability of the projects and reduce costs. Banner: Blue Eco Line
Harnessing Artificial Intelligence to Tackle Plastic Pollution: Lessons learnt from the River Eye project
The Forum Network is a space for experts and thought leaders—from around the world and all parts of society— to discuss and develop solutions now and for the future. Aiming to foster the fruitful exchange of expertise and perspectives across fields to help us rise to this critical challenge, opinions expressed do not necessarily represent the views of the OECD.


The global use of plastic has been growing continuously, and so has the amount of micro- and macro-plastic leakages into the aquatic environment. The main route through which plastics enter the ocean are rivers, which work as highways for plastic to flow from land to sea. It has been indeed estimated that roughly 80% of all plastic waste in the ocean comes from land-based sources, a striking figure which highlights the extent of the problem of riverine waste. This estimate, at the same time, suggests that rivers are the key to solving the increasingly urgent issue of ocean plastic pollution.

William Kelvin once said that what is not measured cannot be improved and that what is not improved is always degraded. This quote sums up well the current problem that we are facing with respect to riverine waste and -more generally- with ocean pollution. Although some general estimates on the plastic input from rivers exist, we lack more precise cross-sectional measurement on the individual contribution of the most important water-ways, as well as data on seasonal and yearly trends.

Richer and larger datasets of these kind could help identifying the most critical settings, thus providing a sound basis to promote cost-effective interventions. Obtaining more data on this phenomenon would also help promoting scientific research in this field, would increase the saliency of this issue in the political arena, while helping raising awareness among the general public and thus possibly nudging individual behavior towards more sustainable choices. The need for more data on riverine waste has been recognized in the academic world as well as at the institutional level. The RIMMEL (Riverine and Marine floating macro litter Monitoring and Modelling of Environmental Loading) project, for example, represented a clear attempt to obtain large scale quantification of loads of floating litter entering European seas. Despite the general consensus on this research agenda, the collection of empirical data is however still lagging behind. Policy-makers still lack the necessary information for the implementation of well-designed targeted policies.

Read more on the Forum Network: Preparing to Protect the Environment: Strategies to Reduce Plastic Pollution During and Beyond COVID-19 by Tony R. Walker, PhD, Associate Professor, Dalhousie University

Read more on the Forum Network: Preparing to Protect the Environment: Strategies to Reduce Plastic Pollution During and Beyond COVID-19 by Tony R. Walker, PhD, Associate Professor, Dalhousie University

Why is empirical research on this topic progressing relatively slowly? Part of the reason is that collecting data on flow rather than stock variables tends to be inherently more difficult, as it requires acquiring information over time instead of simply sampling cross-sectional data points. What makes matter worse in the case of riverine litter is that floating plastic can take many different shapes and forms -especially at different stages of the degradation process. Moreover, river surfaces are likely to differ substantially across location and depend on the prevailing weather and geological conditions.

To simplify the matter, we can say that current estimates on riverine litter are obtained via two main methodologies, both of which suffer from important drawbacks. The first methodology relies on “manual” data collection, i.e. visual observation -for a minimum period of usually 30 minutes- of floating macro litter from an elevated position. These visual monitoring sessions are carried out by individuals who have to detect and classify objects in order to fill in short surveys. This method is extremely time-consuming, not necessarily reliable and clearly ill-suited for any large-scale continuous monitoring process. The second methodology builds instead upon theoretical models which combine geospatial and socio-economic information to obtain predicted estimates of plastic waste concentration in waterways worldwide. This approach is better suited to obtain a more comprehensive understanding of the phenomenon. However it often requires making numerous assumptions, and requires empirical data to calibrate the parameters of interest.

A third possibility has been receiving growing attention, however: the idea of harnessing recent technological advancements in the field of artificial intelligence. In recent decades, incredible progress has been made in artificial vision, object detection and object recognition algorithms. For example, even low-range smartphones are now able to perform accurate google searches based on pictures and images. Such advancements have the potential to revolutionize the field of riverine waste monitoring, increase the scalability of the projects, and reduce costs at the very same time.

At our start-up Blue Eco Line, which was founded in 2018 with the intent of tackling riverine plastic waste, we firmly believe that investing in these technologies represents the best way forward to close the information gap on this phenomenon. Since 2021, we have therefore started working on our River Eye project, which complements our River Cleaner system designed to intercept floating microplastic litter in waterways.

River Eye is an integrated system able to automatically and continuously acquire images of river surfaces, process them scanning for objects, and classify detected objects in different categories (e.g. organic waste, plastic…). The output of this two-stage (detection + classification) process - which relies heavily on neural networks - is an almost real-time estimate of the quantity and quality of floating river waste. As images are processed, River Eye also enriches a database of detected objects that can be used to progressively fine-tune the algorithms and thereby increase the accuracy of the detections.  

Exploring the conditions in which the system can operate, streamlining the acquisition and processing of images, as well as training the algorithms are all challenging tasks which will require time and effort. We have already installed our first pilot system on the Po River in Italy, and we are planning to install a few more on the Arno River in the upcoming months. Embracing the thought of Galileo Galilei, our main goal is not only to measure what can be measured but also to “make measurable what cannot be measured”. We believe in the potential of River Eye as a tool and concept to promote widespread reliable monitoring of riverine waste, and we hope that the information which we will soon be able to acquire will provide a substantial contribution towards solving the urgent and dramatic problem of ocean pollution. More generally, we believe that integrating digital and technological innovations into modern solutions for the monitoring and management of plastic pollution will be of crucial importance in the decades to come.





To learn more about plastic pollution, read also the OECD's recently released Global Plastics Outlook which aims to inform and support policy efforts to combat plastic leakage. And see what else the OECD is doing to prevent plastic pollution on this page compiling its  resources related to plastics.

Read the OECD's recently released Global Plastics Outlook which aims to inform and support policy efforts to combat plastic leakage.

Related Topics: 

Artificial Intelligence Climate SDGs Entrepreneurship

Comments

Go to the profile of Oscar Rivas
2 months ago

En México tenemos grandes problemas con la contaminación de los ríos. Esta noticia se vuelve muy importante para ser considerada en la elaboración de las políticas publicas