Knowledge Bank

A set of 38406 1°x 1° anomalies were compiled starting from a smaller set supplied by the Defense Mapping Agency Aerospace Center. These anomalies were used to determine a set of 1507 5° equal area anomalies using least squares prediction techniques and a direct mean approach. Several different covariance functions were used to examine their effect on the resultant prediction and accuracy. On the average the estimation is insensitive to the covariances used with the largest differences occuring when there was only a few known 1° blocks within the 5° block. The direct mean estimation process was not considered as good as the least squares estimation because the estimated anomaly field was too rough when compared through anomaly degree variance data derived from satellite data. A complete 1654 5° anomaly set was formed by estimating the 157 remaining blocks using the 10 closest previously predicted values. The determination of potential coefficients from this data was made considering a smoothing operator as well as the effects of the atmosphere; a spherical approximation; and the effect of the terrain. The smoothing operator is important in these studies above degree 10. The terrain correction effects amount to 10 to 25% of the low degree coefficients; however, the actual corrections can only be approximated with such approximations not yielding improved solutions at this time. Potential coefficients and their accuracy were computed and listed to degree 52, Tests indicate that it is imperative to use integrated kernal expressions in these computations above degree 5. Anomalies were computed from the potential coefficients for comparison to the original anomalies. Tests showed that the agreement improved considerably as the degree increased to the maximum considered. These tests shed some doubt on the rule of thumb that a block of size theta can be represented by a spherical harmonic expansion to 180° /theta.