Ocean models require independent datasets to verify fore-cast accuracy. Glider data, within an appropriate reference frame, can satisfy this constraint. In the present paper, pro-files from the northwest Mediterranean Sea are re-gridded to allow evaluation of modelled deepwater formation episodes. Time-series analysis of temperature, salinity, mixed-layer depth and ocean heat content show that the simulated response to surface flux is consistent with observations and the evolution of convective events is well represented. How-ever, discrepancies in the distributions of the simulated Levantine Intermediate Water (LIW) and western Mediterra-nean deep water (WMDW) remain. A new 'sweep' methodol-ogy, developed in the present paper, indicates that the loca-tion and duration of the simulated convective events are consistent with that observed, but spatial variability is under-represented. Variogram analysis ascribes integral scales similar to those observed for the mixed-layer depth, but sug-gests that simulated scalar fields are too diffuse. The ability to maximise the separation of temporal and spatial variability, inherent in this new methodology, confirms that glider data is suitable for validating high-resolution ocean models.