Detection and Monitoring of Floating Plastic Debris on Inland Waters From Sentinel-2 Time Series

Abstract

Floating plastic debris on water surfaces imposes both short- and long-term burdens on nature. Hence, identifying and monitoring plastic is important to document the location and scale of the phenomenon. Evaluating the opportunities provided by multitemporal Earth observation data, this article proposes a framework to detect and monitor plastic debris floating on inland waters using optical satellite image time series. First, the detection of plastic candidates is conducted with a rule-based approach relying on variations in signal intensity, temporal patterns, spectral features, and information fusion. Second, identified sensitive areas can be monitored over time, and the extent of plastic cover at subpixel level estimated using spectral unmixing. The method needs as main input parameters only a temporal time frame and an area of interest, therefore, no specific image must be selected nor region of interest manually outlined. Several examples are reported in which the same workflow, developed as a Google Earth Engine application, is successfully applied to identify critical affected areas in full Sentinel-2 scenes, across different continents and contexts, and exhibiting floating plastic debris varying both in type and dynamics.

Keywords

• Image time series
• inland waters
• plastic debris
• sentinel-2
• Google Earth Engine (GEE)
• spectral unmixing (SU)