Abstract We have employed a coupled one-dimensional mixed layer /ecosystem /carbon cycle model to simulate both the normal annual cycle and the iron-fertilization experiment in the subarctic NE Pacific Ocean near Ocean Station P (50°N, 145°W) during summer 2002. We considered two size classes of phytoplankton, the larger representing diatoms, where each size class has a different degree of iron limitation, and compartments for nitrate, ammonium, microzooplankton, sinking detritus, and a prescribed annual cycle in mesozooplankton. The base ecosystem model is formulated in terms of nitrogen, but is coupled to sub-models of silicon and carbon. Diatoms formed aggregates during blooms that sink rapidly from the surface ocean, and diatoms also can be grazed by microzooplankton, consistent with observations. Using the same parameter set as for the base ecosystem model, we reproduce the basic responses to fertilization: an initial bloom of small phytoplankton (including calcifying coccolithophorids), followed rapidly by an increase of microzooplankton biomass; a continuing increase in diatoms that peak as silicate becomes limiting; and a later rapid sinking event of both carbon and silica particulates. Generally this sequence proceeds more rapidly in simulations than in situ. Simulations of the fertilization response show little sensitivity to the assumed fraction of small phytoplankton that are calcifiers, but a strong sensitivity to the assumed diatom uptake ratio Si:N. With an uptake ratio of 2.5, silicate is rapidly exhausted after fertilization, and 8 months later the regional p CO 2 was 14 μatm higher than in the case with no fertilization (assuming no exchange with surrounding waters): for all other simulations the p CO 2 anomaly is negative (indicating increased CO 2 exchange from the atmosphere) but small, 2–5 μatm, suggesting a persistence for a single large-scale fertilization of less than 1 year. Simulated mixed layer Si:N drawdown ratios for different fixed diatom uptake ratios of Si:N illustrate the dangers of interpreting uptake ratios from drawdown ratios: (i) for a fixed uptake ratio the drawdown ratio varies with time as the ratio of small to large phytoplankton changes, and (ii) simulated drawdown ratios are always higher than uptake ratios because of the more rapid recycling (according to the model structure) of N relative to Si in the surface layer. Sensitivity simulations, with the diatom uptake ratio for Si:N varying as an inverse function of iron limitation and with a higher remineralization rate for detrital Si, delayed the onset of silica limitation after fertilization by several days. The magnitude and timing of the diatom peak was unchanged, indicating that in the model the (early) termination of the diatom bloom following fertilization resulted from formation and sinking of aggregates, and not Si limitation.