Recent studies have suggested that accurate predictions of particle export flux can be derived from satellite-based estimates of phytoplankton biomass and net primary production (NPP), combined with models of the food web. We evaluate the performance of this approach using the output of a high-resolution, basin-scale coupled physical-biogeochemical model. There is tight correlation between the annual mean export flux simulated by the biogeochemical model and that predicted by the satellite-based algorithm driven by NPP from the model. Although the satellite-based approach performs well on the annual average, there are significant departures during the course of the year, particularly in spring. NPP and export flux can also become decoupled at the mesoscale, when the dynamics of fronts and eddies cause export to be displaced in space and/or time from the productivity event generating the particulate material. These findings have significant implications for the design of field studies aimed at reducing uncertainties in estimates of export flux.