A new ultra high resolution sea surface temperature analysis from GOES-R ABI and NPOESS VIIRS

Sea surface temperature (SST) is designated an Essential Climate Variable by WMO GCOS and is one of only two key performance parameters for the NPOESS VIIRS mission. In order to meet the increasing scientific and end-user needs for higher resolution SST information (i.e. higher than the ~0.5° – 1° of traditional global products), NOAA/NESDIS developed a new daily SST analysis combining Geostationary (GOES, MT-SAT and Meteosat) SST and POES (AVHRR) data into a single high-resolution (0.1°×0.1°) product. This resolution was chosen to approximately match the Nyquist sampling criterion for the mid-latitude Rossby radius (~20 km) in order to ensure preservation of mesoscale oceanographic features such as eddies and frontal meanders. The methodology employs a rigorous multi-scale optimal interpolation which approximates the Kalman filter, together with a data-adaptive correlation length scale to ensure the balance between detail preservation and noise reduction. The new analysis has been a significant success, even when compared to other modern products which purport to be similar or higher resolution.

While the new analysis satisfies many needs of mesoscale oceanography, there are many applications particularly within the coastal zone (one of NOAA's primary areas of operational responsibility) that require even higher resolution (approaching ~1 km). There are, however, some significant challenges to be met in order to ensure that such a product is optimal.

One of the biggest challenges in the new NOAA analysis has been the bias correction of the different data sets. The error characteristics of the GOES and POES data are substantially different. A major focus has been to ensure that the benefits of each data source are maximized while minimizing their respective weaknesses. The GOES SST data, while being somewhat coarser resolution (~4-km at the sub-satellite point, and closer to ~6 km at mid-latitudes) and slightly poorer accuracy (~0.5 K vs ~0.4 K), sample the SST field at a temporal frequency which is an order of magnitude greater than that achievable from polar-orbiting instruments. This sampling density permits very good coverage of the underlying SST even in conditions of broken cloud. However, the geostationary data are more subject to regional bias errors (typically of order 0.5 – 1 kelvin) than the polar orbiting data. At the moment, bias corrections are derived in a statistical manner and are updated on a daily basis. However, such corrections have to be smoothed over correlation length scales much greater than the analysis resolution in order to avoid spurious correction. Such correction approaches are unlikely to be suitable for application in coastal waters where SST gradients are high (which is why people want to study them) and bias conditions can change on much shorter length scales.

The new instrument technologies that will be available in the GOES-R and NPOESS era should permit higher resolution, better accuracy and an opportunity to derive physically-based SST retrievals and bias-corrections within a common retrieval framework. This should allow statistical bias correction to be relegated to the role of residual adjustment.