Continental-scale joint inversion of Alaska and Yukon gravity and magnetic data
The 3D inversion of potential field data constitutes an increasingly important method of interpretation of geophysical data. A generalized inversion method first discretizes the 3D earth models into cells of constant density, susceptibility, or magnetization vector. In the case of continental-scale geophysical data collected by a combination of land, airborne and satellite measurements, the survey area may cover thousands and even millions of square kilometres, which makes the size of the inversion domain and the number of inverse model parameters extremely large. It is well known that for potential field data the computational complexity increases linearly with the size of the problem. Even a small-sized 3D inversion of huge amounts of data to 3D earth models with hundreds of thousands of cells can exceed the memory available on a desktop computer. In the case of several millions of discretization cells, the memory requirements may exceed the capacity even of the PC clusters. The second obstacle is the amount of CPU time required to apply a huge, dense matrix of the forward modelling operator to the data and model vectors, even using parallel computing.