[en] Seismic wave propagation in multiphasic porous media have various environmental (natural risks, geotechnics, groundwater pollutions...) and resources (aquifers, oil and gas, CO₂ storage...) issues. When seismic waves are crossing a given material, they are distorted and thus contain information on fluid and solid phases. This work focuses on the characteristics of seismic waves propagating in multiphasic media, from the physical complex description to the parameter characterisation by inversion, including 2D numerical modelling of the wave propagation. The first part consists in the description of the physics of multiphasic media (each phase and their interactions), using several up-scaling methods, in order to obtain an equivalent mesoscale medium defined by seven parameters. Thus, in simple porosity saturated media and in complex media (double porosity, patchy saturation, visco-poro-elasticity), I can compute seismic wave propagation without any approximation. Indeed, I use a frequency-space domain for the numerical method, which allows to consider all the frequency dependent terms. The spatial discretization employs a discontinuous finite elements method (discontinuous Galerkin), which allows to take into account complex interfaces.The computation of the seismic attributes (velocities and attenuations) of complex porous media shows strong variations in respect with the frequency. Waveforms, computed without approximation, are strongly different if we take into account the full description of the medium or an homogenisation by averages. The last part of this work deals with the poro-elastic parameters characterisation by inversion. For this, I develop a two-steps method: the first one consists in a classical inversion (tomography, full waveform inversion) of seismograms data to obtain macro-scale parameters (seismic attributes). The second step allows to recover, from the macro-scale parameters, the poro-elastic micro-scale properties. This down-scaling step uses a semi-global optimisation method (neighbourhood algorithm), which allows the sampling of the full model space (thanks to the low numerical cost of the analytic direct model). With the a-priori knowledge of some parameters, a sensibility analysis shows that I can invert precisely skeleton parameters or the saturating fluid type, from the velocities only. Nevertheless, to recover the fluid saturation, it is preferable to use the attenuations. This two-steps procedure is tested on two realistic applications (reservoir monitoring and subsurface hydro-geophysics) and show that we can recover some constitutive poro-elastic parameters. (author)