Regional-residual separation is a fundamental processing step required before interpreting any magnetic anomaly data. Numerous methods have been devised to separate deep-seated long-wavelength (regional) anomalies from the near-surface high-frequency (residual) content. Such methods range in complexity from simple wavelength filtering to full 3D inversions, but most procedures rely on the assumption that all long-wavelength anomalies are associated with deep source bodies: an incorrect assumption in some geologic environments. We evaluated a new method for determining the contributions of near-surface magnetic sources using frequency-domain helicopter-borne electromagnetic (HFEM) data. We inverted the in-phase and quadrature components of the HFEM data to produce an estimate of the spatial variation of magnetic susceptibility. Using this susceptibility information along with known topography and original survey flight path data, we calculated a magnetic intensity grid by forward modeling. There are two immediate benefits to this approach. First, HFEM systems have a limited effective depth of penetration, within the first hundred meters from the surface, so any magnetic sources detected by this method must be located in the near surface. Second, the HFEM-derived susceptibility is completely independent of magnetic remanence. In contrast, apparent susceptibility computed from the original magnetic intensity data incorporates all magnetic signal sources in its derivation. Crossplotting of versus served to reveal areas where the observed magnetic field was dominated by magnetic remanence and provided an estimate of the polarity of the remanence contribution. We evaluated an example, and discussed the limitations of this method using data from an area in the Bathurst Mining Camp, New Brunswick. Though it is broadly successful, caution is needed when using this method because near-surface conductive bodies and anthropogenic sources can cause erroneous HFEM susceptibility values, which in turn produce invalid magnetic field estimates in the forward modeling exercise.