16:45–17:00 | ABSTRACT 1285
LONG-TERM REMOTE SENSING OF NATURAL OIL SLICK DYNAMICS IN THE SANTA BARBARA CHANNEL
Oil enters marine environments through anthropogenic and natural sources. Anthropogenic spills have been tracked using airborne and satellite remote sensing. Natural sources—e.g., seepage, which contributes ~47% of oil entering marine environments—are often continuous rather than acute, thereby producing lower surface marine oil (SMO) concentrations harder to remotely detect. Mapping hotspots of naturally occurring SMO would improve knowledge of how natural seep dynamics affect water-quality gradients and relate to the global hydrocarbon budget. Presently, spatiotemporal variability is poorly understood, including whether human activities such as extraction modify seep flux rates by altering well pressure. We investigated algorithms to map SMO features associated with natural seepage in the Santa Barbara Channel (SBC). We included simulated sensor waveband configurations from hyperspectral reflectance observations, e.g., AVIRIS, and considered spectral absorption, air-water interface effects, and spatial patterns. We found that classifying Landsat observations using convolutional neural networks supports long-term (1984-2026) SMO monitoring, with expert and automated classifications of oil-containing tiles producing highly correlated area estimates (R² = 0.84-0.92) with the lowest performance associated with nearshore elevations in turbidity. We present the longest, to our knowledge, ocean remote sensing satellite time series of SMO associated with a natural seepage hotspot. We test whether spatiotemporal variability in SMO constitutes a useful proxy for studying seepage rates and investigate whether the 42-year Landsat record indicates SMO trends in the SBC.
Camille Pawlak*, UCLA, [email protected], https://orcid.org/0000-0003-0740-3956
Logan Jewell, SUNY Brockport, [email protected], https://orcid.org/0009-0007-8535-9619
Henry Houskeeper, Woods Hole Oceanographic Institution, [email protected], https://orcid.org/0000-0001-5848-5440
