Partitioning of total mercury and methylmercury to the colloidal phase in freshwaters.

Using tangential flow ultrafiltration, total mercury (HgT) and methylmercury (MeHg) concentrations in the colloidal phase (0.4 microm-10 kDa) were determined for 15 freshwaters located in the upper Midwest (Minnesota, Michigan, and Wisconsin) and the Southern United States (Georgia and Florida). Unfiltered concentrations were typical of those reported for freshwater and ranged from 0.9 to 27.1 ng L(-1) HgT and from 0.08 to 0.86 ng L(-1) MeHg. For some rivers, HgT and MeHg in the colloidal phase comprised up to 72% of the respective unfiltered concentration. On average, however, HgT and MeHg concentrations were evenly distributed between the particulate (>0.4 microm), colloidal, and dissolved (<10 kDa) phases. The pool of Hg in the colloidal phase decreased with increasing specific conductance. Results from experiments on freshwaters with artificially elevated specific conductance suggest that HgT and MeHg may partition to different subfractions of colloidal material. The colloidal-phase HgT correlation with filtered organic carbon (OC(F)) was generally poor (r2 < 0.14; p > 0.07), but the regression of MeHg with OC(F) was strong, especially in the upper Midwest (r2 = 0.78; p < 0.01). On a mass basis, colloidal-phase Hg concentrations were similar to those of unimpacted sediments in the Midwest. Mercury to carbon ratios averaged 352 pg of HgT/mg of C and 25 pg of MeHg/mg of C and were not correlated to ionic strength. The log of the partition coefficient (log K(D)) for HgT and MeHg ranged from 3.7 to 6.4 and was typical of freshwater values determined using a 0.4 microm cutoff between the particulate phase and the dissolved phase. Log K(D) calculated using the <10 kDa fraction as "dissolved" ranged from 4.3 to 6.6 and had a smaller standard deviation about the mean. In addition, our data support the "particle concentration effect" (PCE) hypothesis that the association of Hg with colloids in the filter-passing fraction can lower the observed log K(D). The similarity between colloidal and particulate-phase partition coefficients suggests that colloidal mass and not preferential colloidal partitioning drives the PCE.