Natural colloid separation at different pH using dialysis membrane allows quantitative evaluation of trace element complexation. Display omitted ► Dialysis assessed the distribution of trace elements (TEs) between colloidal and dissolved (<1kDa) phases. ► Proportion of colloidal forms was measured as a function of solution pH (from 3 to 8). ► Two groups of elements can be distinguished according to their behaviour during dialysis. ► Results allow prediction of colloidal form proportion in organic- and Fe-rich boreal waters. A dialysis procedure was used to assess the distribution coefficients of ∼50 major and trace elements (TEs) between colloidal (1kDa–0.22μm) and truly dissolved (<1kDa) phases in Fe- and organic-rich boreal surface waters. These measurements allowed quantification of both TE partitioning coefficients and the proportion of colloidal forms as a function of solution pH (from 3 to 8). Two groups of elements can be distinguished according to their behaviour during dialysis: (i) elements which are strongly associated with colloids and exhibit significant increases of relative proportion of colloidal forms with pH increase (Al, Ba, Cd, Co, Cr, Cu, Fe, Ga, Hf, Mn, Ni, Pb, rare earth elements (REEs), Sr, Th, U, Y, Zn, Zr and dissolved organic C) and (ii) elements that are weakly associated with colloids and whose distribution coefficients between colloidal and truly dissolved phases are not significantly affected by solution pH (As, B, Ca, Cs, Ge, K, Li, Mg, Mo, Na, Nb, Rb, Sb, Si, Sn, Ti, V). Element speciation was assessed using the Visual MINTEQ computer code with an implemented NICA-Donnan humic ion binding model and database. The model reproduces quantitatively the pH-dependence of colloidal form proportion for alkaline-earth (Ba, Ca, Mg, Sr) and most divalent metals (Co, Cd, Mn, Ni, Pb, Zn) implying that the complexation of these metals with low molecular weight organic matter (<1kDa fraction) is negligible. In contrast, model prediction of colloidal proportion (fraction of 1kDa–0.22μm) of Cu2+ and all trivalent and tetravalent metals is much higher than that measured in the experiment. This difference may be explained by (i) the presence of strong metal-binding organic ligands in the <1kDa fraction whose stability constants are several orders of magnitude higher than those of colloidal humic and fulvic acids and/or (ii) coprecipitation of TE with Fe(Al) oxy(hydr)oxides in the colloidal fraction, whose dissolution and aggregation controls the pH-dependent pattern of TE partitioning. Quantitative modeling of metal – organic ligand complexation and empirical distribution coefficients corroborate the existence of two colloidal pools, formerly reported in boreal surface waters: “classic” fulvic or humic acids binding divalent transition metals and alkaline-earth elements and large-size organo-ferric colloids transporting insoluble trivalent and tetravalent elements.