, 1999), in general, ligand-bound iron can be taken up (e.g. Maldonado and Price, 1999), using a range of different uptake mechanisms (Maldonado and Price, 2001, Shaked et al., 2005 and Boukhalfa and Crumbliss, 2002). Several of these mechanisms are likely to result in a net loss of complexing capacity. In the model we thus describe the loss of ligands through uptake as Rupt = puptRFe, where pupt is a probability that iron uptake destroys a ligand molecule and RFe is the uptake of iron by phytoplankton.

Finally, part of the ligands is certainly colloidal (Cullen et al., 2006) and can aggregate with sinking particles. In the model this process is described as Rcol = pcolλL, ABT-888 in vitro where pcol is the fraction of ligands that undergoes aggregation and L is the total ligand concentration. λ is an aggregation rate, which we calculate from the concentrations

of dissolved and particulate organic carbon and aggregation kernels for shear and Brownian motion ( Jackson and Burd, 1998). At the moment, we assume that aggregated ligand is lost from the system completely, unlike for iron, where PISCES allows for re-dissolution of particulate iron. The ligand model as described above contains several parameters that must be chosen, namely rL:C, kphot, τmax, τmin, pupt and pcol. While direct measurements of each are unavailable at present, we can make first order approximations of their likely range from find more previous work (the sensitivity to each will be explored in additional model experiments). Concerning first the ratio of ligand to carbon rL:C, the seasonal variations in

ligand and DOC concentrations at the DYFAMED site in the Mediterranean by Wagener et al. (2008) show a good ligand:DOC correlation with a slope Aurora Kinase of ≈ 10− 4 mol L mol− 1 C. A second constraint comes from a linear correlation between iron solubility (a proxy for organic ligands) and regenerated phosphate in the Mauritanian upwelling ( Schlosser and Croot, 2009) with a slope of ≈ 10− 3 mol L mol− 1 P. Using the Redfield ratio of 106 mol mol− 1 for C:P this translates into a ligand:C range 10− 4 < rL : C < 10− 5 mol mol− 1. The shipboard incubation experiments with particles sampled in the water column at a polar and a subantarctic site south of Australia by Boyd et al. (2010) found a release of ligands and of iron in a ratio of ≈ 5 mol mol− 1. Assuming a typical Fe:C ratio in biogenic particles of ≈ 5 − 20 ⋅ 10− 6 mol mol− 1, this translates into a ligand:carbon ratio of 2.5 − 10 ⋅ 10− 5 mol mol− 1, within the range estimated above. Hansell et al. (2012) gives a range of degradation time-scales for dissolved organic carbon from 1.5 years for semi-labile DOC to 16,000 years for refractory DOC. We assume that the ligands that we are modeling are part of the continuum between semi-labile and more refractory DOC with a minimum degradation time-scale τmin of one year and a maximum time-scale τmax of 1000 years (at a reference temperature of 0 °C).