[1] Polar motion excitation involves mass redistributions and motions of the Earth system relative to the mantle, as well as the frequency-dependent rheology of the Earth, where the latter has recently been modeled in the form of frequency-dependent Love numbers and polar motion transfer functions. At seasonal and intraseasonal time scales, polar motions are dominated by angular momentum fluctuations due to mass redistributions and relative motions in the atmosphere, oceans, and continental water, snow, and ice. In this study, we compare the geophysical excitations derived from various global atmospheric, oceanic, and hydrological models (NCEP, ECCO, ERA40, ERAinterim, and ECMWF operational products), and construct two model sets LDC1 and LDC2 by combining the above models with a least difference method. Comparisons between the geodetic excitation (derived from the polar motion series IERS EOP 08 C04) and the geophysical excitations (based on those meteorological models) imply that the atmospheric models are the most reliable while the hydrological ones are the most inaccurate; that the ERAinterim is, in general, the best model set among the original ones, but the combined models LDC1 and LDC2 are much better than ERAinterim; and that applying the frequency-dependent transfer functions to LDC1 and LDC2 improves their agreements with the geodetic excitation. Thus, we conclude that the combined models LDC1 and LDC2 are reliable, and the frequency-dependent Love numbers and polar motion transfer functions are well modeled.