A new concept of road runoff management, based on diffuse infiltration over the shoulder, was tested in a real-scale experimentation site located near Grandson, Switzerland. This new concept consists in diffusely infiltrating the road runoff in the infiltration slope adjoining the road shoulder; where road contaminants are retained. This presupposed that the road shoulder was as tightened as possible to effectively drive the road runoff to the slope, and that the infiltration slope effectively retains all road pollutants. Those two main postulates needed strong scientific verification. For this purpose, five different shoulder designs were tested in Grandson to assess which one presented the best hydraulic performance, i.e. which shoulder was the most impervious. Those designs were made of gravel mixed with humus (SGL), gravel mixed with clay (SGC), gravel seeded with lawn (SGL), a prolonged HMF road base (SH), and a bentonitic geotextile tightening (SB). Also, to verify the retention of pollutant, two infiltration slopes adjoining SGL (LW) and SH (LH) were set as lysimeter (basal geomembrane collecting road effluents); the third infiltration slope (LB) had a direct connection with the aquifer which was closely monitored (6 piezometers located up- and downstream from the road; influence of the road runoff was thus emphasized). Finally, to ensure that this new concept do not lower the road bearing capacity, deflectometers were placed in the road structure; 3 campaigns of leveling assessed the road settlement. Hydraulic assessment of the shoulders, based on 112 natural precipitation as well as 3 artificial watering tests, showed strong discrepancies between all shoulders behaviours. They had to deal with potentially high amount of water. Lag times ranged from a few minutes to several hours. Results showed that lag times were only function of the precipitation mean intensity. The infiltration process could be assessed by a modified Green & Ampt equation. The moisturizing front was frank and linearly progressed downward. Shoulders vertical hydraulic conductivities ranged from 4.6·10-6 (SGC) to 2.2·10-5 m·s-1 (SGL). The exfiltration volume is function of the water volume available at the shoulder surface. The drought preceding the rain event has no influence on the stock variation or the exfiltration volume. Road runoff used two different infiltration paths: the macroporosity drives the water efficiently and rapidly, while the microporosity is less efficient and retains the moisture longer. Tracer tests (Brine) proved that the contamination first flush passed over the shoulders. This is also confirmed by the flux calculation. Runoff coefficients CR ranged from 0.3 (SGL) to 0.9 (SB). This coefficient decreased rapidly once the shoulder is wet. Only SB really fulfilled its task, letting water through only in harsh hydraulic conditions (100 mm; 40mm/h). It is highly recommended for further road development: for new roads as well as old roads being refurbished. The shoulder SH proved to be inefficient for structural and material reasons. Other shoulders are comparatively ineffective. SGC must be enhanced; in that case it could offer a valuable alternative. The bearing capacity of the road was very good. The road lifetime was evaluated to more than 20 years (1·109 residual standard axle load). The Grandson experimental site could not conclusively demonstrate a worsening of the geotechnical behaviour of the road. This is due to the particularly good material and road pavement thickness used for this road. Infiltrations in shoulders are high enough and would have doubtless caused a loss of bearing capacity in case the road would have been more modestly dimensioned. It is therefore clear that the shoulder must be tightened. Geochemical results concerned six families of contaminant: MTE, PAH, Cx, BTEX, MTBE, and PCB. Batch and column tests allowed calculating MTE and PAH distribution coefficient Kd in the Grandson soils, as well as in other typical Swiss soils. Results showed that mobile elements are B, Br, Mo, Ba, Sb, Zn, and V (0 < Kd < 90). Less mobile elements are Ni, Cd, Cr, and Rb (200 < Kd < 300). Elements with very low mobility are Pb, Mn, and Ti (1500 < Kd < 3500). PAH mobility was high for the light PAH (naphthalene, acenaphtene, fluorene, anthracene; Kd < 75). Heavy PAH had lower mobility (Kd > 175). Comparisons with other typical soils enlightened the predominant influence of the pH. Batch test performed with acidic soils (pH = 5.5) confirmed lower MTE Kd. MTE are thus more mobile in acidic soils. Other physicochemical parameters have a smaller influence. The column test performed with a synthetically loaded solution (MTE and PAH) infiltrating the Grandson soils showed that MTE are more mobile in dynamic conditions. This is easily explained by the reduced contact time between the contaminant and the soil aggregates. PAH were not detected at the column outlet, thus showing a very good retention in the column soil. Soil layers analyses demonstrated that PAH were mostly retained in the surface layers. Also, MTE with low mobility show the same behaviour. Mobile elements showed no preferential retention layer. Concerning the infiltration slope, lag times were usually greater than those encountered for the shoulders. Lag time were function of the rain mean intensity. The influence of the preceding drought was significant. Exfiltration volumes EL were high in case of large event; it could be approximated by a simple equation function of the API and surface available water. Moisture redistribution processes occurred in the soils as long as there is a volumetric water content θ gradient. Fluxes strongly varied in magnitude and direction during a precipitation event: usually from about 1·10-8 to 1·10-5 m·s-1 and from up- to downward (evaporation to infiltration). The first flush effect was poor to medium. NaCl signal was highly buffered. The pH was also completely buffered by the soil high carbonate content. It thus shifted from 6.5 in the rainwater to 8 in the LW exfiltration water. To especially emphasize the geochemical behaviour, an artificial precipitation test was created to control every aspect of the rainfall. It was judged very representative of a natural storm. MTE and PAH contaminants had two different behaviours: mobile elements moved mainly in solute and have concentration correlated with EC. Elements with low mobility had higher concentrations correlated with the turbidity peaks. They were transported sorbed to particles. Comparison between the particular and solute form concentrations confirmed that mobile elements are measured in similar concentrations whether the sample was filtered at 0.45 µm, at 1 µm, or not filtered at all prior to acidification. On the contrary, less mobile elements are easily sorbed to particles. The MTE concentrations under sorbed form range from 5% (Rb, Sb) to 98% (Pb) of the total concentration. They had strong first flush effects. B, V, Cr, Mn, Ni, CU, Zn, Br, Mo, Cd, Sb, and Pb were identified as road tracers. Fe and Al represented 80% of the road runoff. Overall, the total MTE concentration in the lysimeter first water collected decreased to 1/30th of the road runoff first flush concentration. Mobile elements were less retained. The PAH total concentration in the lysimeter first collected water was about 1'000 times lower than in the road runoff first flush. The PAH concentration was mainly constituted by mobile PAH naphthalene and phenanthrene. Cx were well represented in the road runoff. However, only heavy species were commonly found. C32 was the most common specie. The triplet C16-C18-C20 was always present in high concentration in the lysimeter exfiltration water. This triple is a great road tracer. BTEX, while present in the road runoff, were not detected in the lysimeter exfiltration water. The alluvial aquifer was also monitored. It is confined between the Arnon (north) and a basement till which shallows up southward. Eastern and western limit are less known but might coincide with the river. The aquifer is clearly linked to the river. This is proved by the piezometric level, EC and T follow-up. EC and temperature also emphasized the higher activity in downstream piezometer. The road runoff infiltration could not be emphasized by the NaCl tracer test. The infiltrated water concentration was possibly buffered by the aquifer volume. Also, the brine was heavier than the aquifer water: it could have plunged to the bottom of the aquifer. All families of contaminant are presented in the aquifer. However, the provenance of those contaminants remains uncertain. The groundwater was indeed as concentrated upstream as downstream. Moreover, the river concentrations had similar, or even higher, values. The contamination provenance is thus either the aerial deposition (equally distributed up- and downstream), either the Arnon River infiltrating the aquifer. The contaminant behaviours in the aquifer could be described. It confirmed the contaminant mobility previously noted during the batch, column and test precipitation experiments. Mobile substances are seen in higher concentrations, whereas substances with low mobility are sparsely found. This project thus demonstrated that the new "over the shoulder diffuse infiltration" concept is clearly implementable. Aquifer concentrations were compared to Swiss legal concentration limits. Except for Cr, the aquifer water was of drinkable quality. This proved that the concept is robust, environmental-friendly and conclusive.