Method for transforming electromagnetic survey data acquired from a subsurface region to a subsurface resistivity model indicative of hydrocarbon accumulations or lack thereof. In one embodiment, data are selected for two or more non-zero frequencies (100), and a structural model of the region is developed based on available geological or geophysical information. An initial resistivity model of the region is developed based on the structural model (101), and the selected data are inverted to update the resistivity model (106) by iterative forward modeling (103) and minimizing an objective function (105) including a term measuring mismatch between model synthesized data and measured survey data, and another term being a diffusive regularization term that smoothes the resistivity model (104). The regularization term can involve a structure or geology constraint, such as an anisotropic resistivity symmetry axis or a structure axis, determined from the a priori information (102).