Mercury's sodium exosphere has been observed only few times with high spectral resolution from ground based observatories enabling the analysis of the emission spectra. These observations highlighted the energetic state of the sodium exospheric atoms relative to the surface temperature. More recently, the Doppler shift of the exospheric Na atoms was measured and interpreted as consistent with an exosphere moving outwards from the subsolar point (Potter et al. 2009). Using THEMIS Solar telescope, we observed Mercury's sodium exosphere with very high spectral resolution at two opposite positions of its orbit. Using this very high spectral resolution and the scanning capabilities of THEMIS, we were able to reconstruct the 2D spatial distributions of the Doppler shifts and widths of the sodium atomic Na D2 and D1 lines. These observations revealed surprisingly large Doppler shift as well as spectral width consistent with previous observations. Starting from our 3D model of Mercury Na exosphere (Mercury Exosphere Global Circulation Model, Leblanc and Johnson 2010), we coupled this model with a 3D radiative transfer model described in a companion paper (Chaufray and Leblanc 2012) which allows us to properly treat the non-maxwellian state of the simulated sodium exospheric population. Comparisons between THEMIS observations and simulations suggest that the previously observed energetic state of the Na exosphere might be essentially explained by a state of the Na exospheric atoms far from thermal equilibrium along with the Doppler shift dispersion of the Na atoms induced by the solar radiation pressure. However, the Doppler shift of the spectral lines cannot be explained by our modelling, suggesting either an exosphere spatially structured very differently than in our model or the inaccuracy of the spectral calibration when deriving the Doppler shift.