Study region: The study uses 78 groundwater head time series across 10 European countries with various geological and hydrological settings. Study focus: The estimation of groundwater recharge using time series analysis and lumped modelling based on groundwater head time series is a low-cost and practical method. However, lumped recharge estimation models based on groundwater level variations are uncertain, and successful applications are known to depend on both climate and hydrogeological setting. Here, we assess the suitability of three different models to estimate recharge (Metran - Transfer Function-Noise model, AquiMod - groundwater level driven hydrological model, and GARDÉNIA - lumped catchment model). New hydrological insights: Results showed that while all three models generally did well during the modelling of groundwater heads, the resulting recharge estimations from the models were different. The analysis showed that the transfer-noise modelling of groundwater heads with recharge and evapotranspiration in Metran is not generally applicable for recharge estimation. The addition of physical information in AquiMod improved the recharge estimations, but the reliability was still limited without control of the water balance due to non-uniqueness. By adding discharge information to the modelling, GARDÉNIA can provide more reliable recharge values. Thus, recharge estimation from groundwater head time series without water balance information must be considered uncertain with low precision, but applicability can be improved when including knowledge of the local system.