While light-non-aqueous phase liquids (LNAPLs) are one of the most important sources of soil pollution worldwide, their remediation remains complex and not always successful. At contaminated sites, significant fluctuations of the soil bio/geochemical properties (metabolisms types, redox) and significant spatial-temporal variations of the pollutant saturation occur, depending on the present LNAPL phase (pure phase floating above the groundwater, dissolved plume, residual droplets). Hence, the real pollutant soil saturation often differs from the LNAPL thickness measured in monitoring wells, and gaps exist between results from modeling and site monitoring. Furthermore, the multiple imbibition/drainage cycles caused by seasonal or pump-induced groundwater level variations also affect the three-phase fluids distribution (air/water/LNAPL), favoring the LNAPL release to the air and water. The relation between the groundwater variation dynamic and the LNAPL mobilization mechanisms need, however, to be further understood, especially in this climate change context where increased groundwater fluctuations are expected (changes in precipitation and recharge patterns, increasing groundwater use). For this purpose, a new project based on an original experimental system was developed at the GISFI (Scientist Group on brownfield sites) station, Homécourt, France. This work is based on a multi-scale approach (laboratory and controlled in situ) with an original experimental device (2m3 lysimetric columns), combining indirect (electrical conductivity), in-situ (pH, temperature, RedOx, electrical permittivity), and direct measurements in monitoring wells. This device allows the assessment of the LNAPL release (pure, dissolved, and gaseous phases) under a ‘current’ (based on regional climate records) and ‘extreme’ (based on the most extreme IPCC climate predictions) precipitation and groundwater fluctuation patterns. The remobilized hydrocarbons will be collected via gas collection chambers (gaseous phase) and suction cups (dissolved phase) as the groundwater table fluctuates and will be regularly analyzed (GC-MS, FTIR). This work aims to introduce an intermediate scale between "simplified" laboratory experiments (centimetric columns, 2D tanks) and the real complexity of contaminated sites, to (i) evaluate and compare different monitoring methods (ii) allow a better characterization and prediction of the LNAPL dynamics in a three-phase system (LNAPL/water/gas), (iii) identify and combine key processes and integrate them into existing numerical models. This project will make it possible to strengthen the recommendations concerning the characterization and monitoring programs for hydrocarbons polluted sites. Preliminary results will be presented to illustrate the capacity of this new instrumental system.