Dissolved iron supply controls half of ocean primary productivity. Resupply by remineralization of sinking particles, and subsequent vertical mixing, largely sustains 2 this productivity. However, our understanding of the drivers of dissolved iron resupply, and their influence on its vertical distribution across the oceans, is still limited due to sparse observations. There is a lack of empirical evidence for what controls subsurface iron remineralization due to difficulties in studying mesopelagic biogeochemistry. Here, we present estimates of particulate transformations to dissolved iron, concurrent oxygen consumption and iron-binding ligand replenishment based on in situ mesopelagic experiments. Dissolved iron regeneration efficiencies (i.e., replenishment/oxygen consumption) were ten-to one hundred-fold higher in low-dust Subantarctic waters relative to higher-dust Mediterranean sites. Regeneration efficiencies are heavily influenced by particle composition. Their make-up dictates ligand release, controls scavenging, modulates ballasting, and may lead to differential remineralization of biogenic versus lithogenic iron. At high-dust sites these processes together increase the iron remineralization length-scale. Modelling reveals that in oceanic regions near deserts, enhanced lithogenic fluxes deepen the ferricline, which alter vertical patterns of dissolved iron replenishment, and set its redistribution at the global scale. Such wideranging regeneration efficiencies drive different vertical patterns in dissolved iron replenishment across oceanic provinces. Globally, the productivity of major phytoplankton groups, including diatoms and diazotrophs, is set by dissolved iron (DFe) supply 1. Twenty years of research has revealed diverse modes of DFe supply, from dust to hydrothermal vents, and their regional influences 2,3. Iron biogeochemistry is a rapidly evolving field, driving the development of global modelling initiatives 4. However, iron cycling in the oceans' interior, a fundamental component of iron biogeochemistry 5,6 , represents a major unknown, and critically is hindering model development 7 .