Magnetic susceptibility (MS) is a very important physical attribute of the Earth's soils and surficial sediments. Soil-containing deposits in the Loess Plateau of central China offer a unique opportunity to study the relationship between soil reflectance and MS. In this paper, data on MS and diffuse reflectance spectra obtained from seven Loess Plateau sections were used to establish quantitative relationships between MS and spectral parameters related to remote sensing. By integrating the diffuse reflectance values corresponding to thematic mapper (TM) channels, five TM bands were simulated and served as independent variables for a multiple linear regression relating the bands to MS. The resulting calibration equation provides good estimates of MS values and has excellent correlation coefficients, >0.90, for all the test samples. The application of the regression equation for mapping soil MS was performed using the enhanced TM image from northwest part of Loess Plateau. The MS values obtained from TM data are convincing and compare well with the limited field test data. The spatial MS variations on the TM image clearly show a decrease toward the northwest and are in a good agreement with the monsoon directions. Results of this study demonstrate that the surface MS signal in soil can be mapped with data from satellite-based remote sensors operating in the visible and near-infrared bands. The future possibilities of this method in other soil types and climates need to be explored.