The surface analysis model currently used by FLENUMOCEANCEN uses an analysis technique known as Fields by Information as follows:

FIRST GUESS. The first guess is an estimate of what an analysis will look like without consider-ing current data. It is normally a blend of (1) the previous analysis extrapolated forward to analysis time, (2) a prognostic chart verifying at analysis time, (3) persistence from the previous analysis, and (4) climatology.

The first-guess provides continuity in data-sparse areas and gives an estimate of the shape (gradients, curvature, etc.) of the data field. In data-sparse areas, the accuracy of a final analysis depends partly upon the first-guess accuracy. The first-guess field is also useful in keeping "impossible" observations from being used in the analysis.

ASSEMBLY OF NEW IN FORMATION. In this step, reports of the parameter being analyzed, that is, pressure, wind, etc., are placed at their proper geographic positions on the grid. These observations are then compared to the first-guess values. If an observed value differs from a first-guess value by a pre-set limit, the observed value is termed "impossible" and is thrown out.

There is an inherent problem with the assembly step. The majority of the oceans are data sparse. When observations in an area are non-existent or are termed impossible, the model bases the analysis of the area on the first-guess data field. If the first-guess data field in the area is in error, the analysis ends up in error. Such errors are especially evident when atmospheric changes take place in an explosive manner. For example, if the model does not have an SLP observation(s) in an area undergoing rapid deepening or if it discards a report or reports in an area as impossible, the first-guess values are used. If the first-guess values in the area do not reflect explosive deepening, the area will be incorrectly analyzed. Any incorrectly analyzed region should be brought to the attention of the FLENUMOCEANCEN duty officer in order that the analysis can be corrected. Such corrections often require the insertion of a bogus report(s) into the data field. These made-up reports are designed to correct the analysis in a region in question.

BLENDING. Blending is the model step that corresponds to the drawing of isolines by hand. To cover data-sparse regions, grid-point values are adjusted and spread to surrounding grid points using gradient knowledge and mathematical gradient formulation. Blending spreads the data from high reliability grid points (grid points with values based on observations) to those having lower reliability (grid points based on the first-guess analysis). The degree of spreading is increased with higher reliability in the gradient.

COMPUTING RELIABILITY FIELD OF THE BLENDED PARAMETER. In this step, the computer assigns weight factors to the blended grid-point values. The higher the weight factor, the higher the reliability of the value. For example, a grid-point value based on an observation(s) normally has a higher weight than a grid-point value based on an extension of a gradient. The reliability of all grid points is increased through the blending process.

REEVALUATION AND LATERAL RE-JECTION. In step 5, FIB uses the blended parameter field and the weighted values to reevaluate each piece of information entered into the analysis. Reevaluation is a quality control done on each observation. A statistical value is computed for each report and is compared to the actual value. The statistical value measures how accurate a report really is compared to its expected accuracy as given by its assigned weight factor. The lateral rejection check takes place when each grid-point value, with its weight, is removed individually from the grid and compared with what remains, or the "background." If the report is within its expected reliability range, no change is made to its weight. If the value is greater than the expected range but within some upper limit, its weight is reduced. If the value exceeds the range limits, the report is rejected (that is, its weight becomes zero) and it has no effect when the next assembly and blending is done.

REANALYSIS. After the grid-point values are reevaluated and new weights are assigned, the reanalysis step begins. This step is no more than a repeat of the assembly step, using the first-guess field and the reevaluated data. The new field may be reevaluated and reanalyzed two or three times before the computer accepts it and sends it to the output section, where it is stored for transmission and for input into other programs.

Privacy Statement - Press Release - Copyright Information. - Contact Us - Support Integrated Publishing