This thesis concerns the observation and modeling of solute transport processes, focusing in particular on the theoretical formulation by means of travel and residence time distributions. The work reports and analyzes data collected from a controlled experiment in a large lysimeter supplied with an underground measurement chamber. The system, protected from external rainfall input, is forced by pre-determined water injections according to a given Poissonian process. The presence of two willow trees induces high evapotranspiration fluxes, which, likewise discharge fluxes, are continuously monitored by means of load cells and bottom flow measurements. Five water inputs have been marked with neutral chemical compounds, i.e. different species of fluorinated benzoic acids. The tracers' concentration is evaluated at the bottom of the system, enabling travel time distributions to be directly calculated from experimental data. At a later stage, several models characterized by various degrees of complexity and relying on different mixing hypothesis are proposed and applied, and their ability to reproduce the observed concentration patterns is analyzed. Finally, travel time distributions inferred from the models applied are presented and discussed. The experiment shows that travel time distributions are eminently non-stationary, thereby supporting earlier theoretical conjectures. Modeling reveals both reliability and limits of the random sampling assumption (according to which ages are randomly sampled by the output fluxes in the same proportion as they are stored in the control volume) and provides appropriate elements for the understanding of the tracers' behaviour in soils, indicating new paths for the design of more complex and efficient transport models.