Abstract
Exploiting the potential of space-borne oceanic measurements to characterize the sub-surface structure of the ocean becomes critical in areas where deployment of in situ sensors might be difficult or expensive. Sea Surface Temperature (SST) observations potentially provide enormous amounts of information about the upper ocean variability. However, the assimilation of daytime SST retrievals, e.g., from infrared sensors into ocean prediction systems, requires a specific treatment of the diurnal cycle of skin SST, which is generally under-estimated in current ocean models due to poor vertical resolution at the air–sea interface and lack of proper parameterizations. To this end, a simple off-line bias correction scheme is proposed, where the bias predictors include, among others, the warm layer and cool skin warming/cooling deduced from a prognostic model. Furthermore, a localization procedure that limits the vertical penetration of the SST information in a hybrid variational-ensemble data assimilation system is formulated. These two novelties are implemented and assessed within a regional ocean prediction system in the Ligurian Sea for the assimilation of daytime SST data retrieved with hourly frequency from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the geostationary satellite Meteosat-10. Experiments are validated against independent measurements collected by gliders, moorings, and drifters during the Long-term Glider Missions for Environmental Characterization (LOGCMEC17) sea trial. Results suggest that the simple bias correction scheme is effective in improving both the sea surface and mixed layer accuracy, correctly thinning the mixed layer compared to the control experiment, outperforming experiments with night-only data assimilation, and improving the forecast skill scores. Localization further improves the prediction of the mixed layer depth. It is therefore recommended that sophisticated bias correction and localization procedures are adopted for fruitfully assimilating daytime SST data in operational oceanographic analysis systems.
Highlights
Within regional ocean prediction systems and high-resolution environmental characterizations, the optimal exploitation of remotely sensed data becomes increasingly important [1,2], due to the high costs that deployment and maintenance of high-resolution in situ measurements require [3]
sea surface temperature (SST) measurements from infrared sensors onboard polar-orbiting satellites have been available for a long time, little use of daytime data has been made in operational contexts due to the difficulty of modeling the surface layer diurnal variability in Ocean General Circulation Models
We explore a different approach, where the skin SST model is embedded in the Ocean General Circulation Models (OGCMs), but the prognosed skin SST is used as a predictor within a multi-linear bias correction scheme together with other physically relevant quantities
Summary
Within regional ocean prediction systems and high-resolution environmental characterizations, the optimal exploitation of remotely sensed data becomes increasingly important [1,2], due to the high costs that deployment and maintenance of high-resolution in situ measurements require [3]. SST measurements from infrared sensors onboard polar-orbiting satellites have been available for a long time, little use of daytime data has been made in operational contexts due to the difficulty of modeling the surface layer diurnal variability in Ocean General Circulation Models. 2019, 11, 2776 difficulty of modeling the surface layer diurnal variability in Ocean General Circulation Models (OGCMs) This has induced most prediction systems to assimilate only nighttime SST m(OeGasCuMresm). Sophisticated physical–statistical observation operators may be used, where the skin SST minus model SST is calculated from high vertical resolution training data [18,19] Both approaches conceptually consider an observation operator capable of projecting the ocean state from the OGCM to the skin SST state while relying either on analytical or statistical relationships, respectively.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
