Principal differences between the new analysis technique and the Reynolds-Smith Analysis, in addition to the change from 1.0 to 0.5 degree (latitude, longitude) resolution, are the following.
Ø Only 24-hours of satellite SST retrievals and in situ data are used in each analysis, rather than ingesting 7-days data;
Ø Satellite retrieved SST values are averaged within 0.5 degree grid boxes: with day and night “superobs” created separately for each satellite;
Ø SST reports from ships and buoys are not averaged within grid boxes;
Ø The first-guess is the prior (un-smoothed) analysis with one-day’s climate adjustment added;
Ø Surface temperature is calculated for ice-covered water using salinity climatology in Millero’s (1978) formula, for those grid-boxes where the ice cover exceeds 50 percent;
Ø An inhomogeneous correlation-scale-parameter g, for the correlation function: exp(-d2/ g2) , is calculated from a climatological temperature gradient, as g=min ( 450, max(225/|DT|, 100)), with d and g in kilometers.
Bias calculation and removal, for satellite retrieved SST, and all data quality control criteria are essentially the same.
Evaluations of the analysis products have shown it to produce realistically tight gradients in the Gulf Stream regions of the Atlantic and the Kuroshio region of the Pacific, and to be in close agreement with SST reports from moored buoys in both oceans. In contrast to the NESDIS and Reynolds-Smith analyses, this new analysis properly depicts the colder shelf water -- a feature critical in getting an accurate model prediction for the 29-30 December 2000, east coast winter storm.
Results of comparative evaluations will be presented.