The Mw6.8 Tottori earthquake, Japan, does not exhibit any surface trace but was particularly well instrumented. Strong motion displacement records and GPS coseismic data are used to constrain the evolution of the slip on the fault plane in time and space. We adopt in this study a two-plane fault geometry based on aftershock distributions and analysis of close station records. In a first step, our inversion allowed surface slip. The model obtained has a significant surface slip, which contradicts the absence of clear surface slip reported by geologists. In a second step, models with no slip at the surface (buried faults), compatible with geological observations, have been tested. The tests with different fault depths show that when slip is allowed to occur close to the surface, the fit to seismological and geodetic data is increased. These tests confirm that slip actually occurred at shallow depth. Despite the nonuniqueness of the solution, all the inverted source models show (1) a large slip amplitude patch at a depth of about 4--5 km and (2) relatively small slip in the hypocentral area. The rupture velocity is about 2750 m/s in the asperity region. The total rupture duration is about 8 s. The slip distribution seems to be controlled by the variation of fault properties with depth. Another feature that could control the rupture of this earthquake is a fault plane almost perpendicular to the main fault NW of the epicenter, which apparently inhibits further rupture propagation.