Satellite Validation of Citizen Science Marine Pollution Data: Multi-Site Correlation Analysis of Sentinel-2 Floating Debris Index and Eyesea Ground-Truth Reports

Abstract

Marius Suteu

We present the first multi-site correlation analysis between the Sentinel-2 Floating Debris Index (FDI) and community-reported marine pollution data from the Eyesea citizen science platform. Across three geographically diverse coastal sites—Vasai-Virar (India), Santa Cruz, Galápagos (Ecuador), and Puerto Montt (Chile)—8,123 geotagged beach pollution reports were analysed against 91 Sentinel-2 L2A scenes spanning January 2023 to March 2026. We employ four spatial correlation methods, three temporal alignment approaches, seasonal decomposition, and multi-band spectral feature importance analysis. Statistically significant temporal correlations were found at all three sites, with the strongest at Santa Cruz (ρ = +0.773, p < 0.001). After seasonal decomposition, the residual correlation at Vasai-Virar reached ρ = +0.861 (p = 0.001), demonstrating that anomalous debris events are captured by both satellite and citizen scientists independently. Equatorial oceanic sites produced the cleanest signals, while estuarine and fjord environments required site-adapted analysis. These results establish citizen science marine pollution data as a scientifically verifiable ground-truth source for satellite remote sensing validation.

PDF