Within recent years, percolation regimes of the Greenland Ice Sheet (GrIS) experienced significant changes. It remains unclear whether meltwater remains and refreezes within the firnpack and gradually fills up all pore space or near-surface refreezing causes the formation of impermeable layers, which result in lateral runoff. Both, homogeneous ice layers within perennial firn, as well as near-surface ice layers of several meters are observable in firn cores. However, the formation process of neither of them in real time has been observable before. Taking firn cores is a destructive sampling method and thus hampers monitoring. In addition, apart from surficial extent of melt over the GrIS, only very little is known about percolation depths, liquid water content and the amount of mass that is transferred into deeper layers during respective summer seasons. Temperature records in snow and firn may only indicate the depth of percolation of melt water generated at the surface. To overcome this deficit and provide data for model evaluations, we installed an upward-looking radar system (upGPR) 3.5 m below the snow surface in May 2016 close to Camp Raven (66.4779N/ 46.2856W) at 2120 m a.s.l. within the deep percolation zone of the GrIS. The radar is capable to monitor quasi-continuously changes in snow and firn stratigraphy, which occur above the antennas. For summer 2016, we observed four major melt events, which routed liquid water into various depths beneath the surface. The last event in mid-August resulted in the deepest percolation down to about 1 m above the antennas. Comparisons with results predicted by the regional climate model MAR are in very good agreement in terms of specific surface accumulation, while neither the timing of melt events nor the amounts of liquid water predicted by MAR correspond with upGPR data. Radar data and records of a nearby thermistor string, in contrast, matched very well, for both, timing and depth of temperature changes and observed water percolations. All four observed melt events transferred a cumulative mass of 56 kg/m^2 into firn beneath the previous year’s summer horizon of 2015. This contributes to the seasonal mass balance of the glaciological year 2015/16 at this point, which could be determined utilizing upGPR data as well. We find that the continuous observations of liquid water content, percolation depths and values for the seasonal mass balance are sufficiently accurate to provide valuable information to validate model approaches and help to develop a better understanding of liquid water retention and percolation in perennial firn. Furthermore, model approaches can be validated with observed temporal changes at exactly the same location, without having to account for spatial inhomogeneities.