A prognostic high-resolution model is established to provide an integrated view of the evolution of the spring bloom during the Programme Oc¿an Multidisciplinaire M¿so Echelle (POMME) experiments carried out at sea from February to May 2001 (16--22¿W and 38¿--45¿N). Data collected during the first survey were used for model initialization, and data from three other cruises were used for model validation. The model successfully predicts the time evolution of the main reservoirs and fluxes, except for a storm event during postbloom conditions, for which the biological impact is underestimated. The bloom is long in duration (2 months), has low intensity (1 mg Chl m-3), and is characterized by a small f-ratio (0.45) and a small e-ratio (0.05). Furthermore, the model reveals much stronger space and time variability than sampled in the data. This large variability results both from the synoptic atmospheric variability and from the stirring induced by oceanic mesoscale eddies. In particular, the bloom starts in specific submesoscale features that correspond to filaments of minimum mixed layer depth. On short timescales (2--3 days), space and time variability have the same order of magnitude. On the seasonal timescale, time variability is larger than space variability. Considering the transient state of the system, this modeling exercise is also used to quantify the nonsynopticity of the observations, which occur mostly during bloom conditions, a crucial point for the data interpretation.