10:00–10:15 | ABSTRACT 1156
A PROPOSED OPERATIONAL OCEAN COLOUR PRODUCT FOR GLOBAL SURFACE OCEAN DIMETHYLSULFIDE ESTIMATION
Dimethylsulfide (DMS) is the dominant biogenic sulphur gas emitted from the ocean surface and an essential driver of marine aerosol formation and cloud albedo. Despite its climatic significance, no operational satellite DMS product exists. Existing predictive models rely on nutrients or chlorophyll-a as inputs — quantities that are themselves derived from empirical or prognostic models or climatologies, propagating errors into the DMS prediction. We present DSMnR2, a Random Forest model that estimates global surface ocean DMS from satellite remote sensing reflectance (Rrs), sea surface temperature (SST), and photosynthetically available radiation (PAR). The model was trained on a quality-controlled and spatially balanced subset derived from a consolidated global database of 993,447 observations spanning the global ocean from 1998 onwards. Evaluated on a held-out 20% test set, DSMnR2 achieves R2 (natural log) = 0.872, median relative error of 18.9%. Model architecture is designed to reduce underestimation at DMS concentrations greater than 20 nM — critical for accurate sea-air flux estimation where episodic high-DMS events dominate. SHAP analysis confirms physical interpretability, with spectral reflectance encoding phytoplankton community state and solar irradiance governing the seasonal DMS production cycle. DSMnR2 generates a global daily, 5-day and monthly DMS product at up to 0.05-degree resolution, currently spanning 1998 to 2025 and continuously extendable as new satellite observations become available. This represents an important step toward an operational ocean colour DMS product.
Shrivardhan Hulswar, Institut de Ciències del Mar, [email protected], https://orcid.org/0000-0002-1346-9823
Martí Galí Tàpias, Institut de Ciències del Mar (ICM-CSIC), [email protected]
Rafel Simó, Institut de Ciències del Mar (ICM-CSIC), [email protected]
