Geostationary Operational Environmental Satellite-R Series (GOES-R) with Advanced Baseline Imager (ABI) onboard is the key component of NOAA's program for future environmental monitoring. As a contribution to that effort, a sea surface temperature (SST) algorithm for ABI is being developed by the SST Team as a part of the GOES-R Algorithm Working Group (AWG). The ABI SST production is prototyped with the currently available Meteosat Second Generation (MSG) Spinning Enhanced Visible and IR Imager (SEVIRI) data. The development strategy is to set up an end-to-end system, including near-real time acquisition of 15-minute, full-disk (FD) SEVIRI Level 1 data and processing them into Level 2 SST and associated top-of-atmosphere (TOA) clear-sky brightness temperatures (BT) products. The existing AVHRR Clear-Sky Processor for Oceans (ACSPO), developed at NOAA/NESDIS and currently operational with NOAA-18 and MetOp data, was adopted as a first-cut processor for clear-sky radiances, SST, and aerosol, and was modified to work with the SEVIRI data. AVHRR-like, single-channel aerosol optical depths, derived from the three SEVIRI reflectance bands, have also been preserved from ACSPO for diagnostic purposes. An extensive course of near-real time QC, Cal/Val, error characterization and monitoring for long-term stability and cross-platform consistency, will be performed outside ACSPO by other subsystems developed in the SST Team.

The continuous inflow of SEVIRI 15-minute FD images has been established in close collaboration with the AWG Land Team. McIDAS area files are being downloaded from NOAA operational servers in near-real time, reformatted into hdf4.2 files (similar in their structure to the EUMETSAT L1.5 product), and saved on STAR SAN storage provided by the AWG for shared use between different teams within the AWG. As of the time of this abstract, hdf files from January 2008 and on have been saved on spinning disk, and three more years (2005 to 2007) are being added from the University of Wisconsin/CIMMS archives. Near-real time processing will be done on SST Team Linux computers.

ACSPO version 2 provides two estimates of SST: one based on statistical regression (currently, split-window) and another on the physical inversion method. For inversions, the Community Radiative Transfer Model (CRTM) is used in conjunction with Global Forecast System (GFS) atmospheric and Reynolds first-guess SST fields. The two methods have been implemented side-by-side for cross-evaluation and for potentially merging into a single hybrid algorithm. That hybrid would combine the flexibility of accounting for local atmospheric transmission variations, inherent to the inversion method, with a potential adjustment of some retrieval algorithm parameters against in-situ data, utilized in the regression methodology. Retrievals are accompanied with a comprehensive cloud mask and QC flag.

In this study, the structure of the algorithms and product performance is discussed and illustrated with MSG-2 SEVIRI data. The study also shows results of sensitivity studies aimed at optimization of the ACSPO parameters for cloud mask, and for regression and physical SST algorithms, specific to SEVIRI geostationary observations. Preliminary estimates of SST retrieval errors are provided as a function of environmental (ambient clouds, water vapor, air and surface temperatures, etc.) and observational (angular dependency of retrievals) conditions. In-situ (buoy) and daily and weekly analysis (Reynolds) SSTs serve as references. Also discussed is the consistency of SEVIRI TOA BTs with CRTM simulations. Some preliminary analyses of the diurnal cycles in derived parameters are also presented.

Once fully implemented and tested, the complete SEVIRI SST system will run in near-real time and will be extensively checked for robustness, accuracy, and stability. It will be iteratively improved and fine-tuned until required SST products are generated, and their performance will be carefully evaluated by the comprehensive QC/QA and Cal/Val. Analyses will then extend to understanding in-depth the diurnal cycle of the SSTs and associated BTs. Potential to fill-in cloud gaps using high temporal resolution geostationary data will be tested, and derivation of higher-level products (hourly, daily, monthly) explored. Information content of 15-minute measurements will be compressed to effectively represent the major degrees of freedom within a day, and these condensed parameters will be subsequently trended, long-term. Cross-consistency with AVHRR-derived products and additional information content of geostationary data will be evaluated.