<p>This thesis deals with the application of the Spherical Cap Harmonic Analysis (SCHA) modelling technique to obtain geomagnetic field models for Indonesia, which have better resolution and accuracy than the International Geomagnetic Reference Field (IGRF). B-splines basis function and autoregressive forecasting are applied to improve estimates of secular variation and its forecast over the Indonesian region. The modelling technique is applied to geomagnetic observation data compiled from 68 geomagnetic repeat stations in Indonesia covering the period 1985 - 2015 from BMKG (Badan Meteorologi Klimatologi dan Geofisika / Agency for Meteorology, Climatology, and Geophysics) Indonesia, definitive data from five BMKG geomagnetic observatories and 13 INTERMAGNET (The International Real-Time Magnetic Observatory Network) observatories. Synthetic cartesian X, Y, and Z components at sea level at 17 fixed locations, calculated from IGRF-13, are also used. The area covered by the models in this thesis is the Indonesian region with a spherical cap half-angle of 30° and with the coordinate of the spherical cap pole at 122°E and 3°S. From statistical analysis and comparison with the IGRF, the SCHA model with index k = 7 is considered as the best SCHA model, both in resolution and accuracy. Compared with the root mean square deviation (RMSD) of the IGRF model, the RMSD of the SCHA model with index k = 7 is lower by 28 nT, 11 nT, and 34 nT for X, Y, and Z components, respectively. A model from interpolation of the SCHA with index k = 7 using the B-splines basis function for the year 1985.5 – 2015.5 shows that the SCHA model gives better results than the IGRF. The forecasting calculation for the year 2015.5 – 2020.5 suggests that the autoregressive order 3 of the SCHA with index k = 7 gives better results than the forecasting of the IGRF model, especially in the X, Z, and F components. However, in the Y component, the IGRF is still better than the SCHA model. The RMSD of the forecasted SCHA model is 154.92 nT, 200.87 nT, 104.39 nT, and 135.81 nT for X, Y, Z, and F components, respectively, while the RMSD of the IGRF model is 172.62 nT, 95.52 nT, 117.55 nT, and 162.38 nT for X, Y, Z, and F components. Thus, the forecasted SCHA model is suitable for data reduction of geomagnetic surveys in the Indonesian region but not preferable for navigation.</p>