Non-stationarity of solute travel time distribution observed in a controlled hydrologic transport volume
Experimental data were collected over a year-long period in a transport experiment carried out within a controlled transport volume (represented by a 2m-deep, 1m-diameter lysimeter fitted with bottom drainage). The soil surface was shielded from natural rainfall, replaced by an artificial injection (Poisson process) at the daily timescale. Bottom drainage out-flows were continuously monitored with leakage tipping bucket and evapotranspiration (prompted by a willow tree growing within the system) was measured trough precision load cells, which also allow an accurate and continuous reading of the total water storage. Five artificial soluble tracers (species of fluorobenzoic acid, FBAs, mutually passive) were selected based on low-reactivity and low-retardation in our specific soil and used to individually mark five rainfall inputs of different amplitudes and occurring at various initial soil moisture conditions. Tracer discharge concentration and hydrologic fluxes measurements provide a direct method for the assessment of the bulk effects of transport on the (backward and forward) travel time distributions in the hydrological setting. The large discrepancies observed in terms of mass recovery in the discharge (supported by ex post FBAs quantification in the soil and in the vegetation) and tracer out-fluxes dynamics emphasized the dependence of the forward travel time on the various injection times and the stages experienced by the system during the migration of the pulse. Rescaling the measured travel time distribution by using the cumulative drainage volume as an independent variable instead of the time elapsed since the injection also fails to yield to stationary distributions, as it was argued by Niemi (1997). Our experimental results support earlier theoretical speculations centered on the key role of non-stationarity on the characterization of the properties of hydrologic flow and transport phenomena. A travel time based model, with all in- and out- hydrological fluxes imposed by the experimental measurements, could accurately reproduce the large divergences of the five tracers’ behavior and recovery, using adequate assumptions on the mixing processes occurring within the controlled volume, thus discrediting simple plug-flow (old-water first) and well-mixed processes which fail at this aim.