Labelling using the stable 15 isotope of nitrogen allows a close monitoring of the fate of the fertilizer applied to arable crops. Because 15N data give direct estimates of nitrogen transformation rates, they also provide more stringent tests for N models than those based on bulk inorganic N dynamics. They may therefore point at flaws in models that had previously gone unnoticed, especially if 15N was monitored on short time steps which capture even rapid processes like nitrification.Here we tested the simple, process-based model SUNDIAL on two such data sets obtained in Northern France under winter wheat and winter rape crops receiving various doses and forms of fertilizer N. In both experiments, microplots (≈ 1 m2 in size) within larger blocks were dressed with 2·0 atom-% enriched labelled 15N, as urea, or ammonium-nitrate as NH+4–15NO3 or 15NH+4–H–NO3. Replicate micro-plots were subsequently sampled on four occasions after fertilizer application, and 15N enrichment was monitored in plant roots and tops, and at several depths in the soil in inorganic and organic forms.Comparison between observed and simulated data showed that, shortly after application, SUNDIAL either underestimated (rapeseed) or overestimated (wheat) the rates of crop uptake. Also, the gradual incorporation of 15N into soil organic matter was too quick in autumn and too slow in spring under the rapeseed crop. The simulation of the rapid depletion of the labelled soil inorganic N pool was correct under wheat, whereas under rape, SUNDIAL predicted an accumulation of nitrate which was not observed.After a longer time interval (1–2 months), the simulated and observed amounts of fertilizer-derived N in the crop and in the soil became more comparable. However, SUNDIAL only accounted for part of the unrecovered labelled N. Additional measurements indicated that denitrification and ammonia volatilization were responsible for most of the losses, with discrepancies occurring because SUNDIAL failed to volatilize ammonia after fertilizer spreading. The other major source of error lay in the simulation of crop demand for nitrogen, which may be improved based on sounder eco-physiological concepts, such as that of a regulation of plant uptake by shoot biomass.