We apply inversion methods to first arriving P waves from explosive source seismic data collected along line 1 of the Los Angeles Region Seismic Experiment (LARSE), extending northeastward from Seal Beach, California, to the Mojave Desert, in order to determine a seismic model of the upper crust along the profile. We use resolution information to quantify the extent of blurring in the LARSE images and to smooth a damped least squares (DLS) image by postinversion filtering (PIF). Most of the original data fit is preserved while minimizing model artifacts. We compare DLS, PIF, and smoothing constraint inversion images using both real and synthetic data. A preferred PIF image includes larger-scale features in the smoothing constraint inversion image and finer-scale features in the DLS inversion image that are consistent with geologic information. We interpret principal model features in terms of geology, including faulting. The maximum depth of low-velocity sedimentary and volcanic rocks in the Los Angeles basin is 8--9 km and in the San Gabriel Valley is 4.5--5 km. A horst-like uplift of basement rocks occurs between these basins. The northeastern boundary of the San Gabriel Valley is imaged as a tabular, moderately north dipping low-velocity zone that projects to the surface at the southernmost trace of the Sierra Madre fault system. In the central and southern San Gabriel Mountains, velocity-depth profiles are consistent with intermediate-velocity mylonites overlying lower-velocity Pelona Schist along a shallowly southwest dipping Vincent thrust fault. Tomography does not provide a definitive dip for the San Andreas fault but, combined with other LARSE results, is consistent with a vertical to steep northeast dip. ¿ 1999 American Geophysical Union