In both models it is possible to produce a ligand distribution that is ALK activation closer to observations than a uniform low value. Moreover, the models reproduce some observed features well, such as a decrease along the conveyor belt circulation (e.g., Thuróczy et al., 2011 and Mohamed et al., 2011) a general decrease of ligand concentrations from the mesopelagic towards the deep ocean (e.g., Ibisanmi et al., 2011), and a horizontally and temporally variable concentration of ligands near the surface, with higher ligands e.g. near the European shelf seas. Both models also make strong predictions regarding the
gradient in ligand concentrations between regions of high and low productivity (e.g. between upwelling regions and the subtropical gyres) that can hopefully be tested in future fieldwork. In the model at least, this gradient is strongly dependent on the assumed photochemical degradation rate. Ultimately, the predictions of the model are regulated by the sources and sinks associated with each specific process (Table 2). In this regard, selleck compound process studies such as FeCycle that document
the time evolution of iron–ligand dynamics (Boyd et al., 2012) can provide important information for modeling efforts. For example, the maximum rates of ligand production from organic matter remineralization reach 0.25 and 0.05 nmol L− 1 d− 1 in PISCES and REcoM, respectively, of similar order, but towards the low end of Cell press the two estimates of 0.3 and 1.3 nmol L− 1 d− 1 from Boyd et al. (2010). Further such experiments that normalize the rate of ligand production to carbon solubilization would prove invaluable. Equally so, experimental constraints on the bacterial, photochemical and aggregation losses of ligands
would allow tighter constraints to be placed on these parameters. Modeling the ligand distribution dynamically instead of assuming a uniform and low constant concentration brings the average vertical profile of iron closer to the observed nutrient-like profile with a maximum near the oxygen minimum in the mesopelagic. However, as the ligand concentrations are now greater than those used previously, this raises the iron concentrations in the non-iron limited regions of the ocean such as the Atlantic and Indian oceans. A useful way to evaluate this effect is by looking at the excess ligand, denoted as L⁎ (e.g. Boyd and Tagliabue, 2014-in this issue), which is defined as: ligand minus dissolved iron. Our two models clearly overestimate the prevalence of negative L⁎ regions relative to that observed ( Fig. 8). The distribution of negative L⁎ in the models reflects external inputs of dissolved iron and highlights too low scavenging rates of uncomplexed iron. In REcoM negative L⁎ regions are restricted to the dust deposition regions, while in PISCES the large sedimentary iron fluxes that are absent in REcoM are also important ( Fig. 7).