This study presents the first basin scale section of the partition coefficients (Kd) of 14 rare earth elements (REE) along the GEOVIDE (GEOTRACES GA01 cruise) transect, in the subpolar North Atlantic (May–June 2014, R/V Pourquoi Pas?). Although Kd data are very scarce in the literature, it is a key parameter of element cycle modeling, controlling the scavenging efficiency by particles. Because of this lack of data, Kd are generally adjusted in the models to obtain the best agreement between modeled and observed concentrations. This shortcoming likely contributes to make the modeling of the oceanic REE cycles challenging. This is well illustrated by the difficulty to satisfactorily simulate both Nd concentrations and isotopic compositions.

Here, we determined Kd(REE) at 10 stations, calculated from previously published dissolved and particulate REE concentrations, and from suspended particulate matter (SPM) concentrations. The data required for the calculation of SPM concentrations were obtained in the framework of the GEOVIDE cruise. Kd profiles displayed minimum values at the surface, variations in the upper 400 m, and an increase with depth below 400 m. The expression of the ratio of particulate to dissolved REE concentrations as a function of SPM concentrations showed that manganese oxides (MnO2) were the main driver of REE scavenging, followed by the lithogenic phase and iron hydroxides (Fe(OH)3). Our study also highlighted that REE scavenging preferentially depended on the relative proportion of these three phases rather than on their absolute concentrations. These conclusions were evaluated by a Pearson correlation test, with correlation coefficients of 0.83, 0.84, and 0.75 between Kd(Nd) and the MnO2, lithogenic, and Fe(OH)3 fractions, respectively.

Finally, we proposed two Kd(Nd) parameterizations as a function of the particulate phase fractions. The first parameterization included MnO2 and Fe(OH)3, phases that are currently not represented in Nd cycle models, and showed a good agreement between observed and calculated Kd(Nd) (R2 = 0.84, p < 0.05). The second parameterization did not include the MnO2 and Fe(OH)3 phases, as in existing Nd cycle models, leading to a lower correlation between observed and calculated Kd(Nd) (R2 = 0.71, p < 0.05). Given that scavenging is an important process in the water column, these results reveal a bias in Nd cycle (and more generally in REE cycles) with models. Determining these Kd values is promising and therefore recommended in the future for improving Nd (and REE) cycle models and our Nd (REE) cycle understanding.