Evaluating Riverine Litter Monitoring Methods: A Comparative Study of Visual and Camera-Based Approaches

Riverine litter pollution poses substantial environmental challenges, necessitating effective monitoring techniques to assess and mitigate this environmental impact. Existing methods for monitoring riverine litter vary widely in quality, cost, ease of implementation and performance. The difference of these factors for different monitoring techniques remains underexplored, limiting the ability to effectively monitor floating litter flux over long time periods.

This study addresses this gap by evaluating four methods for riverine waste monitoring: (1) visual observations by human observers, (2) manual counting from camera images, (3) manual counting of AI-filtered camera images, and (4) fully automated AI-based counting of camera images. The evaluation focuses on two key objectives: assessing how well each method's recovery rate aligns with ground truth data and comparing plastic flux estimates derived from each method.

To this end, experiments are conducted in a semi-controlled waterway (lock). During these experiments, plastic litter is released in the water at random intervals to simulate natural litter transport. Human observers located on a bridge over the water count the floating litter and record data using the JRC Floating Litter Monitoring app. Simultaneously, high-resolution cameras capture images of the floating litter for the three camera-based methods. The flux estimates, as well as the implementation and the scalability of the different methods will be compared, to assess their overall effectiveness in monitoring. The study will provide insights into the strengths and limitations of each monitoring method, offering a basis for selecting the most suitable approach for various scenarios. This comparative evaluation will bridge a critical research gap, contributing to the development of more efficient monitoring strategies for addressing plastic pollution in waterways.