The first generation of open-ocean iron enrichments (1993 to 2005) have all had broadly the same design. Enrichment of patches of ocean was typically on a 10 km length-scale, and experiments were of a duration of weeks. These scales were dictated by what could conveniently be achieved from research vessels, using tracers to track Lagrangian patches. The extrapolation of experimental findings to the larger scales required for carbon sequestration by ocean iron fertilization (OIF) leaves many uncertainties, to answer which, longer duration (i.e. months) and larger scale observations (100 to 200 km length-scale) are required. However, to extrapolate to a timescale of decades and to the scale of ocean basins, such observations must be conducted in parallel (and where possible assimilated into) detailed models of the physics and biogeochemistry of the fertilized waters. Our present understanding suggests that any carbon sequestration will occur as the net result of changes in the air–sea flux integrated over millions km2 and many years, and can only realistically be assessed by modelling. A central role of the observational studies will be to make such models as accurate as possible in their simulations and predictions. We present a scheme for the design of a second generation of ocean iron-enrichments and discuss the challenges that are evident in linking the modelling and observational components of such studies.