The study of large-scale field-aligned currents (LSFACs) has depended on visual inspection of data plots, which is often subjective and time-consuming, and therefore limits the efficiency of analyses. The present paper reports a new procedure to automatically identify the spatial structure of LSFACs from satellite magnetic field measurements. The procedure is based on the concept of the first-order B spline fitting with variable node positions, which may be envisioned as fitting line segments to a line plot. The fitting is made for the maximum variance component of magnetic variations in the plane perpendicular to the background magnetic field. If the distribution of LSFACs can be approximated as an infinite sheet, each slope of the plot corresponds to the crossing of a FAC sheet. The number of node points, which determines the number of FAC sheets, is one of the fitting parameters and is optimized for each orbit so that the Akaike information criterion (AIC) is minimized. Whereas other methods, such as a spherical harmonic fitting, seek to obtain two-dimensional distributions of FACs from assembled data, the present method is basically the automation of the way we visually examine a plot of satellite magnetic field data, and it can be applied even to a single satellite pass. Therefore the procedure can be implemented to real-time data processing and now-casting. The procedure should also provide a powerful tool for data mining. Magnetic field data from the Defense Meteorological Satellite Program--F7 (DMSP--F7) were used for demonstration, and the present paper reports the initial results. ¿ 2000 American Geophysical Union