Sensitivity experiments were integrated for 1000 years. A difference between the models is that REcoM assumes the ligands to have a uniform conditional stability constant with pKFe′L=11pKFe′L=11, while the stability constant in PISCES is set to vary with temperature, following pKFe′L=17.27−1565.7/TKpKFe′L=17.27−1565.7/TK where TK   is absolute temperature. This leads to a pKFe′LpKFe′L of 11.5 at 0 °C and 11.9 at 20 °C. The net effect of this temperature dependency is that iron is scavenged slightly

more easily in colder waters than in warmer waters. To be able to evaluate the model, we compiled a data set of in-situ observations from the published literature. In doing so, we did not take into account that measurement methods for ligands still differ strongly in their methodology, e.g. through the click here application of different competing ligands in the electrochemical titrations, and consequently different analytical windows for ligand stability constants. We certainly do not see our data compilation as the last word on a ligand database, rather

as a first attempt to obtain at least a semi-quantitative data set for evaluation. The complete list of papers and data sets that we included can be found in the supplement to this paper. The compilation of in-situ observations (shown as filled circles in Fig. 1 and Fig. 2) shows that ATM/ATR inhibitor drugs a uniform constant value evidently is not supported by the data, and that, moreover, a constant value of 0.6 or 1 nmol L− 1 is an underestimate of the true ligand concentration. The distribution of ligands as it is produced by PISCES (model run LIGA, Fig. 1) clearly does a better job

than the assumption of a constant value. A few characteristic features are: Surface concentrations are highest in upwelling regions and over some shelves, somewhat elevated in the subpolar regions, and decrease towards higher latitudes and the centers of subtropical gyres. Below the euphotic zone, concentrations are more homogenous, but still present the same general pattern. In the mesopelagic, values become generally lower, but remain the highest below ocean regions that Methane monooxygenase are characterized by a stronger biological carbon pump. In the deep ocean the influence of lateral advection becomes apparent in elevated concentrations around Antarctica and the North Atlantic, while concentrations in the oldest water masses in the North Pacific are significantly lower. One may argue that several of the model predictions, such as the lower surface ligand concentrations in the Southern Ocean and the higher ligand concentrations there in the deep, are also seen in our collection of in-situ observations. There are other features where model and observations do not match so well, e.g. in the equatorial upwelling in the Pacific (albeit against one data point), where the model overestimates ligands, or the deep Atlantic, where modeled ligand concentrations are slightly too low.