Collecting data on plastic pollution has been an important focus for partners working on the Preventing Plastic Pollution project, helping to better understand sources and movement of plastic pollution through a river catchment. The majority of macro plastic data has been gathered using litter surveys compiled during clean-up events and activities, an approach that empowers volunteers to contribute to the research aims of the project. There are some limitations of this approach: manual data takes time to compile, and there are remote areas that humans can’t reach easily to clean up.
Occupying an area of approximately 36 km2 Poole Harbour is one of the largest natural harbours in the world, a haven for wildlife including Ospreys and White-Tailed Eagles and a site of special scientific interest (SSSI). The harbour also borders residential settlements, industrial areas, and numerous marinas housing commercial and leisure vessels.
With large parts of the harbour inaccessible due to tidal mud flats and salt marsh habitats, the research team at QMUL wanted to consider the extent of macro plastic pollution across potential inaccessible and accessible hotspots and have a data set that could complement the microplastic sampling of the four rivers that flow into the harbour.
The team at QMUL partnered with Ellipsis Earth to deliver a detailed baseline survey of litter across the harbour. Ellipsis have developed a technology platform that uses image recognition and machine learning to deliver automated litter identification with over 90% accuracy.

The Ellipsis team used a drone and smartphones to conduct the survey across seven sites, which were selected to be representative of the land use across the harbour and a potential repository for litter transported from the rivers Frome, Piddle, Corfe, and Sherford . Each site comprised 1000m x 50m lateral coverage along the coastline adjacent to the water’s edge.
Within each site, Ellipsis Earth geo-fenced areas of specific land use such as footpaths, reed beds, beaches, green spaces and the water’s edge, enabling comparison by land use between the sites.
Following the survey the image data was uploaded and the software then automatically detected, labelled and mapped 7,691 items across 47 OSPAR categories of litter. The total weight of waste was 136kg, calculated using average weight per item type, adjusted for level of degradation.

Analysis of the dataset provides a range of valuable insights, for example:
- Litter breakdowns by proportion, change, weight and environmental impact
- Distance/density calculations
- Geographic comparisons by region and land use
- Performance of bins
- Brand specific data

Some key findings from this baseline survey were:
Whilst the overall litter quantities found across Poole Harbour are extremely low in number, the survey sites’ locations so close to tidal water creates a high environmental risk: 1.71 million cumulative years. Presence of disproportionately high numbers of small fragments of degradation-resistant plastic further compound this risk due to their higher likelihood of being dislodged by water, wind or wildlife, their ease of transport once in the water system, and their ease of ingestion by marine life.
The results are heavily dominated by the disproportionately high concentrations of litter at one site (Holes Bay East – 72% of all litter, an item every 1.2m) where the litter profile is characteristically ‘roadside’ in nature (lower quantities of cigarette butts and larger single use plastic items, and a larger majority of industrial fragments, metal and rubber). Accessible and frequently visited sites were also high in littering across a wide range of categories, with a variety of litter items and high densities of littering in areas of highest footfall, such as carpark, beaches, and footpaths. . These land uses where not protected by reeds, contained 18% and 32% of all litter, respectively, with all sites being unmanaged/not frequently cleaned; the water’s edge sites posing a high environmental risk as mentioned above.
Conversely, reed beds were seen to have extremely low litter accumulation levels (an item every 1.65km on average, <1% of all litter, little to no recent or active deposition), which is a positive finding considering this land use makes up a significant majority of Poole Harbour’s coastline. Given the above-mentioned relationship between direct access and littering, water-borne/tidal deposition, high residence times and inaccessibility probably contribute to low levels of rubbish in reed beds.
Brand analysis highlighted the extremely low fractional contribution of branded litter, with only 13 branded items detected across the entire survey region, making up 0.2% of all litter and 0.9% of food and drinks litter. Composition of branded litter varied, from coffee cups, food containers, cartons and sauce packets, to water bottles and lids and aluminium drinks cans. Brands can certainly play a role in reducing littering, but prevention strategies need to address a range of behaviours and litter types to be effective.
Testing a different approach to data collection has been a valuable experience. Not only did this allow us to gather detailed data from remote areas but survey larger areas in a shorter period of time than if performing a standard litter pick. This baseline dataset provides a level of insight on the scale of plastic pollution across Poole Harbour, however, additional surveys will be necessary to provide sufficient evidence to draw conclusions.