Orientation of an impact source is estimated from multicomponent land data. The seismic wavefield produced by an arbitrarily oriented impact source is represented by the linear combination of the nine independent wave-field components (Green's functions) for a full vector source. Coefficients of this linear combination are expressed in terms of the parameters that define source orientation. The parameters are an azimuthal angle, a vertical angle, and source strength, which are the unknowns of the inverse problem. Once the Green's functions are computed, source parameters are estimated by full wavefield inversion using a grid search over the source parameters. This inverse procedure also provides information about the uniqueness and resolution of the solution. Information on the source orientation is dominated by the relative amplitudes between components, so the inversion can be performed even when the source-time wavelet is not well known. The approach is tested using a synthetic common-source gather. Results for a field data example from northeast Texas show deviations between 11° and 16° from the vertical for an expected effective vertical impact, and between 19° and 23° from the horizontal for an expected horizontal cross-line impact, depending on whether a geophone rotation is performed to reduce the uncertainty produced by geophone misalignment. Results show the potential to quantify and isolate the effects of source misorientation in multicomponent land data.