The geomagnetic field 1900-1995, including the large-scale field from magnetospheric sources, and the NASA candidate models for the 1995 revision of the IGRF

This paper reports on a continuous representation of the main geomagnetic field of degree 13 for the 1990-1995 time period, including a degree 1 representation of the field of external origin, designated GSFC(S95). The model employs a cubic B-spline basis with equi-spaced knots for the temporal variation in the secular variation of the internal field. Hence, the temporal variation of the spherical harmonic coefficients is represented by integrals of cubic B-splines. In the derivation, a suite of different forms is utilized for representation of the external field: (1) GSFC(S95-a), in which the external terms are proportional to the annual averages of the aa index, (2) GSFC(S95-s), in which the external terms are represented by unconstrained cubic B-splines, (3) GSFC(S95-sc), in which the cubic B-spline representation of the external field is constrained to be near the GSFC(S95-a) model for years prior to about 1940, and, (4) GSFC(S95-nx), in which there is no external field representation. The NASA candidate models for the 1995 revision of the IGRF are extracted from GSFC(S95-sc). Data sources include the Magsat and POGO satellites, observatory annual means, decimated land survey, marine total-field, aeromagnetic, and repeat data. Random data uncertainties are assigned by statistical binning procedures, while systematic error is accounted for via the correlated weight matrix procedure of Langel et al. (1989). The data are not sufficient to resolve all model parameters, and thus, regularization via quadratic penalty functions is employed. For the internal field this included minimizing the average of the square of the radial field secular variation and acceleration over the core-mantle boundary and through time. Comparison of the GSFC(S95-sc) model with the ufm1 model of Bloxham and Jackson (1992) for their common time span shows good general agreement, especially with respect to secular variation coefficient signatures and overall data statistics. The major differences are manifested in a better fitting of Magsat and POGO data by GSFC(S95-sc), but better fitting of early survey data by ufm1 This is expected and reflects the relative influence of the data and penalty function in the particular model. The external field of GSFC(S95-sc) exhibits a very prominent solar-cycle variation in the q(1)(0) coefficient, though with about a 2-year time lag. The internal field exhibits a well behaved R-n spectrum throughout the time span indicating sufficient constraints being applied to the poorly observed, high-degree secular variation. Finally, the westward drift synthesized for degrees 2 to 5 shows a 0.76 linear correlation with the length-of-day (led) variations of Jordi et al. (1994) when the latter leads the former by 11 years.