The balance between SREM and Rcol in PISCES means that the subsurface ligand concentrations are only slightly higher in PISCES, relative to REcoM ( Fig. 1 and Fig. 2). The concentration of ligands affects the equilibrium distribution of iron between inorganic forms (mainly hydroxides for ferrous iron) and iron bound to the ligands. The high particle reactivity of the inorganic forms drives scavenging, a main loss process for dissolved iron. It is therefore expected that a spatio-temporal variation of ligand concentration has consequences for the distribution of iron in the ocean. This is indeed what is found: A comparison of the (globally averaged) vertical profiles (Fig. 3) of iron in model runs with variable

Ceritinib organic ligands with runs where the ligand concentration was kept fixed at a constant value throughout the ocean shows a general tendency of iron concentrations to increase in the upper 5 FU part of the ocean and to decrease somewhat in the deep part. A notable feature is that both models now show a more nutrient-like profile for dissolved iron than with constant ligand, with an intermediate maximum around 500 m depth, near the depth of the oxygen minimum. This is closer to observations than in the case with constant ligands, where deep iron tends to be too homogeneous compared to observations

(Tagliabue et al., 2012). It is interesting to note that, both with prognostic and with constant ligands, the average iron profile differs in several respects between the two models: PISCES has a local maximum near the

surface, and for constant ligand has a slight secondary maximum at 3000 m depth. Both features are absent in REcoM. This can be traced back to a different treatment of iron sources in the two models: PISCES has a comparatively strong sedimentary source of iron which is strongest on shallow shelves, and includes hydrothermal inputs of iron (Tagliabue et al., 2014), while REcoM has only a weak sediment source and neglects hydrothermalism altogether. Given this difference, it is encouraging that in both models, qualitatively, the introduction of prognostic ligands leads to a more nutrient-like iron profile. Of direct importance for biological productivity are of course mainly the Farnesyltransferase changes in iron concentration through prognostic ligands in the euphotic zone. Fig. 4 shows how near-surface (0–50 m) iron changes in the model runs with prognostic ligands, compared to a model run with constant ligand. Although details of the patterns differ slightly between the two models, the general picture is robust, namely that dissolved iron increases most in the Atlantic and Indian Ocean, while only small changes are seen in the Southern Ocean and the Pacific. This pattern reflects the fact that in the latter regions, production in the models tends to be iron-limited, so that here biological uptake is the main loss process for iron, not scavenging. An increase in ligands therefore does not lead to an increased lifetime in the surface ocean here.