A Lagrangian spectral parameterization of gravity wave drag (GWD) is developed and implemented into the NCAR Whole Atmosphere Community Climate Model (WACCM). Based on ray theory, the Lagrangian parameterization explicitly calculates gravity wave (GW) propagation that has been treated too simply in column-based parameterizations. For direct comparison with column-based parameterization, a hydrostatic and Boussinesq version of the Lagrangian parameterization is used in the present study. GW packet trajectories demonstrate that the Lagrangian parameterization calculates reasonably the GW-packet propagation and that the horizontal extent of GW propagation can be as large as 20°¬ as GWs approach critical levels. Comparison with column-based parameterization through one-month simulations indicates that the overall structures of GWDC are similar to one another, but the magnitude is much increased in the lower stratosphere and equatorial troposphere in the model with the Lagrangian parameterization. This magnitude difference is due mainly to the vertical convergence of GW packets. In climate simulations for 6 years, it is found that the zonal-mean zonal wind in the equatorial stratosphere and along the axis of the polar night jet is improved through the Lagrangian parameterization. Also, interannual variability in the equatorial lower stratosphere is significantly enhanced. The parameterization is validated by off-line calculations using TRMM and ECMWF data, and resultant temperature variance is compared with the observed in satellites. It is found that the Lagrangian parameterization produces much more realistic distribution of gravity waves than the currently used column-based parameterization, although with weak temperature variance compared the observed.