Mount St. Helens is a stratovolcano of the Cascade volcanic arc and is one of the most active in the United States. The present morphology is mainly due to the destructive eruption of 18 May 1980, triggered by a five-magnitude earthquake that caused the collapse of a vast sector of the northern part of the volcano. Coda waves are an essential tool to monitor eruption dynamics and model volcanic structures jointly and independantly from velocity anomalies: it is thus fundamental to test their spatial sensitivity to seismic path effects. The thesis reports new research aiming to find a link between seismic imaging of the volcano and its actual geomorphology and surficial geology by qualitative and quantitative approaches. As a first step for coda attenuation imaging, space weighting functions and kernels were used to model the spatial sensitivity of coda waves. While the diffusive approximation is valid for an active seismic dataset to model coda waveforms, it is not valid for the passive seismic dataset of Mount St Helens and, therefore, the multiple scattering was introduce. To measure the relative importance of scattering attenuation vs absorption at Mount St Helens, the Multiple Lapse Time Window Analysis (MLTWA) was applied. The quantities obtained with this technique are the Seismic Albedo B0 (ratio between scattering and total coda attenuation, between 0 and 1) and the Extinction Length (Le, reciprocal of the total attenuation). Results show the typical dominance of scattering attenuation in volcanoes at lower frequencies (3 - 6 Hz), while absorption is the primary attenuation mechanism at 12 Hz and 18 Hz. Still, the seismic albedo is anomalously high (0.95) at 3 Hz. The scattering and absorption parameters derived from MLTWA are used as inputs to construct 2D frequency-dependant bulk sensitivity kernels for the S-wave coda in the multiple-scattering regime (using the Energy Transport Equations - ETE) and to validate the use of analytical diffusive kernels (AD, independant of MLTWA results). At 12 Hz, high coda-attenuation anomalies characterise the eastern side of the volcano using both kernels, in spatial correlation with low-velocity anomalies from literature, generally associated with the deep plumbing system of the volcano. The hypotheses underlying both MLTWA and sensitivity kernels are tested with radiative-transfer forward model of farand near-field envelopes in a half space using RADIATIVE3D. Here, a layered model was implemented with a shallower layer with increased scattering properties to model late coda envelopes. While the broadening of late coda phases improves, this model cannot explain the phases of the intermediate coda with higher amplitude than the direct waves. This behaviour is due to strong near-receiver scattering, which enhance anomalous phases in the intermediate and late coda across the 1980 debris avalanche and central crater. Only at frequencies higher than 3 Hz and in the far-field, diffusion onsets at late lapse times. At 3 Hz, the anomalous albedo, the forward modelling and the results of the tomographic imaging confirm that shallow heterogeneity beneath the extended 1980 debris-avalanche and crater enhance anomalous intermediate and late coda phases, mapping shallow geological contrasts. To better investigate the morphologies biasing the seismic wave path, a geomorphological map was necessary. Literature review pointed out the lack of this information for Mount St. Helens. So, a new geomorphology and surficial geology map has been created, based on a LiDAR acquired in 2003 with 3 m resolution. The effect of the 1980's blast on the entire area is evident, mostly on the northern sector of the volcano, which is covered by the debris avalanche and pyroclastic flow deposits. The volcano completely lost its typical conic shape, in favour of a horseshoe-shape crater rim, with an open side on the NS direction. The map is accompanied by a second one focusing on the geomorphological changes caused by the 2004 eruption, based on a LiDAR of 2009 (1 m resolution). Here, the main morphological variations affect the shape of the main glacier and the dome, both grown in the crater. A qualitative comparison between coda-attenuation imaging and the landforms of Mount St Helens has been done, using the first-ever geomorphological map of the volcano. The map highlighted extremely rough landforms (hummocky structures) in the complex debris avalanche deposit. The comparison with the tomography reveals several matches, not only with the debris avalanche itself but also with other areas of the south-western flank of Mount St Helens, like the Butte Camp dome, where a contrast in attenuation is possibly defining a fracture or a fault connected to this intrusive body. To better understand the influence of key morphologies, forward modelling of the seismic waveforms was fundamental. Starting from the velocity fluctuations obtained from the radiative transfer model, an forward model of the wave equation was introduced. The anisotropic model, following the qualitative observation, shows how the geomorphology is controlling the travel times of SH waves, in particular in two directions: WNW-ESE, following the trend of a buried fault along the South Fork Toutle Valley, and NS, consistent with the main morphological difference between south (mostly untouched by the 1980 eruption) and north (collapsed in 1980) of the volcano. A first analysis of the waveforms of a shallow event of 2005 recorded by three-components array shows a loss in high frequency content for the stations located on the debris avalanche of 1980, unlike other stations laying on older volcanic products. The isotropic model is well reproducing the arrival of the horizontal polarised S waves (SH) at higher frequencies (10 Hz), but not at lower ones (1 Hz and 6 Hz). The introduction of an anisotropic model at 1 Hz and 6 Hz is necessary to explain the arrivals for the stations in the north-northwestern side of the volcano.