How magnetization within crustal rocks varies from place to place is fundamental to a variety of magnetic data processing and interpretation methods.It also provides basic constraints on the mechanisms involved in crustal formation and its subsequent history (e.g., deformation, alteration).A wide range of geophysical processes and rock properties has been described in fractal or scaling terms.Well log susceptibilities, surface susceptibilities and aeromagnetic fields all tend to support a model for a 3-D crustal magnetization distribution having a radially-averaged power spectrum proportional to some power of the spatial frequency.This simple model of the scale-invariant behaviour of crustal magnetization and the magnetic fields it produces can be exploited by several applications which require information on such spatial variation.A more realistic power spectrum, and equivalently, covariance model for crustal magnetization offers many advantages over the geologically incorrect assumption of a white power spectrum (equivalent to an uncorrelated distribution).Estimating crystalline basement depths from power spectra of magnetic data presumes some knowledge of the basement magnetization behaviour.Replacing the common assumption of an uncorrelated magnetization with a fractal description leads to a reduction in these depth estimates, producing more realistic results.Gridding randomly spaced magnetic observations using kriging requires an estimate of the covariance of the data.Using fractal covariances for kriging produces gfidded estimates which more closely reflect the statistics of the underlying magnetization process, producing maps with a justifiable degree of smoothness.Inversions of magnetic data for geological structure can also exploit realistic information on the spatial behaviour of magnetization.Using a scaling model of a priori magnetization values results in a smoothing of the inverted solutions.Calculated models are consequently structurally simple and contain only those features well resolved and justified by the data.Fractal magnetization models can be used to generate synthetic magnetic fields which have the same character as a given data grid.These synthetic values can then be used to fill in holes in the grid when the data are processed in the frequency domain.They may also be used to optimize the separation of magnetic fields caused by sources at different depths.