Abstract

Plastic pollution is a widespread environmental issue, with larger macroplastic objects resulting in harm to the environment, economies, and human health. Satellite remote sensing can detect plastic over widespread areas, but quantifying plastic on the spatial scale within sensor pixels (sub-pixel) in imagery poses challenges to medium-resolution sensors. Determining the factors affecting the ability of plastic to be detected at small scales within satellite imagery and testing little-explored spectral unmixing sub-pixel plastic quantification on these open-source satellite images can increase the value extracted from this data source. Spectral measurements of plastic packaging at changing abundances were measured using a hyperspectral sensor under laboratory conditions, identifying a non-linear trend of reflectance with surface density which makes its utility in plastic quantification unreliable. The experiment also identified High Density Polyethylene (HDPE) plastic had the largest measured absorption features at a surface coverage of 5% and showed that visible colour did not impact the depth of Near InfraRed (NIR) and Short Wave Infrared (SWIR) absorption features and impede plastic identification. Increased thickness of translucent Low Density Polyethylene (LDPE) targets produced a stronger trend between surface density and absorption feature depth, making the plastic quantity easier to identify through absorption peak depth. Resampling spectra to the resolution of the Sentinel-2 satellite sensor removed visibility of absorption peaks, but trends of reflectance against surface density remained. Expanding these findings to quantifying plastics within satellite imagery, Multiple Endmember Spectral Mixture Analysis (MESMA) was identified as an untested method with potential. Agricultural polytunnels were used as a testing target, and over Almeria, Spain, and Littywood Farm, UK, MESMA was able to calculate the plastic surface quantity to an accuracy of 87.56 ± 10.09 %, and 93.33 ± 5.08 % respectively. To test if MESMA was suitable for individual plastic pollution objects, a methodology for cataloguing floating marine plastic was provided to oceangoing vessels to match with available Sentinel-2 satellite imagery. 18 samples were logged and 5 matched to same-day Sentinel-2 imagery. Glint reflecting from the water surface restricted the effectiveness of MESMA over these sampled objects, as well as published spectral indices tested alongside. This thesis has shown the potential of an alternative, accessible approach to quantifying sub-pixel plastic surfaces with open-access satellite imagery, alongside factors affecting the accuracy of the quantification measurement.

Awarding Institution(s)

University of Plymouth

Supervisor

Matt Telfer, Richard Thompson

Document Type

Thesis

Publication Date

2026

Embargo Period

2026-04-23

Deposit Date

April 2026

Creative Commons License

Creative Commons Attribution-No Derivative Works 4.0 International License
This work is licensed under a Creative Commons Attribution-No Derivative Works 4.0 International License.

Download

Share

COinS