Elastic full-waveform inversion (FWI) is a physics-based method that can recover quantita- tive subsurface models by fitting synthetic to observed seismograms. In the near surface, where depths of interest are in the order of tens of meters, Rayleigh waves are the dominant energy in seismic records and carry strong sensitivity to shear-wave velocity. Exploiting this information is particularly valuable for engineering. The objective of this thesis is to construct a two-dimensional elastic reference model of the Herrenknecht gravel test site in southern Germany. This site serves as the validation ground for the newly developed Urban Vibro Truck (UVT), a mobile seismic vibrator intended for urban geophysical surveys. A reliable near-surface model is essential both for characterizing the test site itself and for calibrating the emitted source signal. Seismic data were acquired along two perpendicular profiles using hammer blows at 4 m intervals and vertical-component geophones at 1 m spacing. Two alternative starting models were tested: a four-layer model obtained from dispersion-curve inversion and a smooth linear-gradient model. In both cases, elastic 2D FWI converged to consistent solutions. The results reveal a low-velocity zone of approximately 200 m/s within the upper 3–5 m, underlain by a rapid increase to 400–500 m/s and a gradual rise to 600–700 m/s at 16 m depth. These features are consistent across both profiles and at their intersection point, demonstrating the robustness of the inversion. Multi-parameter tests examined whether compressional velocity (Vp) and density (ρ) could be reconstruct in addition to shear-wave velocity (Vs). While joint inversions achieved similar reductions in data misfit, the Vp and ρmodels remained poorly resolved and showed strong parameter trade-offs. In contrast, the Vs-only inversions yielded stable, geologically plausible results. Overall, the study confirms that Rayleigh-wave-based elastic FWI provides a reliable and reproducible Vs model of the shallow subsurface. This reference model will support future validation of the UVT source and illustrates both the strengths and limitations of Rayleigh-wave FWI in near-surface geophysics.