Assessing ocean ensemble drift predictions by comparison with observed oil slicks

Abstract

Geophysical models are cornerstone pieces in marine forecasting of floating objects and pollution, such as marine surface oil slicks. Trajectory forecasts of oil spills inherit the uncertainties from the underlying geophysical forcing. In this work we compare the forecast capabilities of an ocean ensemble prediction system (EPS) to those from a higher resolution deterministic model on the representation of oil slick drift. As reference, we use produced water (PW) slicks detected and delineated from 41 C–band Sentinel-1A/B satellite synthetic aperture radar images between April and December, 2021. We found that the EPS provided at least equivalent member-wise results relative to simulations forced with the deterministic model. Ensemble verification through rank histograms and spread-error relationship showed that including the ocean fields is necessary to address model uncertainties. Whether considering the ocean field or not, the modeled slicks were counterclockwise rotated between 20° and 30° relative to the ones observed in the satellite images, and these were deflected about 45° to the right of the observed wind direction.