Environ. Sci. Technol. 2009, 43, 3514–3521
Mercury in a Boreal Forest Stream Role of Historical Mercury Pollution, TOC, Temperature, and Water Discharge OLOF REGNELL* Department of Chemical Ecology and Ecotoxicology, University of Lund, SE-223 62 Lund, Sweden CARL J WATRAS Wisconsin Department of Natural Resources and Center for Limnology, University of Wisconsin-Madison, Trout Lake Station, Boulder Junction, Wisconsin 54512 BO TROEDSSON Eman Catchment Management Association, Box 237, SE-574 23 Vetlanda, Sweden ´E ANDERS HELGE Hultsfreds Municipality, Environmental and Health Office, Box 500, SE-574 21, Sweden TOMMY HAMMAR Kalmar County Administration, Environmental Unit, SE-391 86 Kalmar, Sweden
Received October 23, 2008. Revised manuscript received April 2, 2009. Accepted April 7, 2009.
via runoff from the catchment. In exceptional cases, natural geological sources contribute significantly to Hg budgets of boreal surface waters (1). It can be difficult to determine whether atmospherically derived Hg or historical Hg contaminants dominates the flow of Hg through a water system. Measurements above and below the historical point source will provide insight into this matter, but strong temporal variation in the contribution of Hg from internal and external sources may render conclusions based on discrete measurements unreliable. Mobilization of historical Hg contaminants in a stream is likely to increase with Q but so is the transport of atmospherically derived Hg because of the coupling between heavy precipitation events, increased atmospheric Hg deposition and decreased retention of Hg in the catchment (2). In order to protect ecosystems and human health from Hg poisoning, it is important to identify the limiting factors in the microbial production and release of highly toxic and bioavailable MeHg. Depending on source and chemical speciation, Hg may be more or less available for methylation and may or may not reach methylation sites. Also, increased amounts of Hg may not result in a proportional increase in MeHg production because of saturation of the Hg methylation potential, or because of microbial demethylation of MeHg (3, 4). We measured the advective fluxes of HgT and MeHg in a boreal forest stream at a reference site and at a moderately Hg contaminated site. The main aim was to assess the contribution of the historical Hg to the annual HgT and MeHg export of the stream. Another aim was to study relationships between HgT, MeHg, water chemistry Q, and T.
Material and Methods Over a one-year study period (2003), we monitored total Hg (HgT) and methyl Hg (MeHg) at two sites in a Swedish forest stream located above (Siteref) and below a stretch of Hgcontaminated sediments (SiteHg). We also monitored HgT, MeHg, and ancillary water chemistry in peat water close to the stream and HgT in open field wet deposition. Despite the presence of historical Hg contaminants, direct atmospheric Hg deposition and transfer of Hg from the catchment explained more than half of the annual HgT load at SiteHg. The concentrations of both HgT and MeHg were sensitive to changes in water discharge (Q) and water temperature (T) at both sites, suggesting that the stream HgT and MeHg load can change dramatically in response to changing weather conditions. The 2003 data together with data from 1996 disclosed intersite differences and temporal variation in the relationships between HgT, MeHg, and TOC (total organic carbon), reflecting variable sources of HgT, MeHg, and TOC and temporal changes in factors affecting Hg speciation.
Introduction In waters to which Hg has been released directly, typically by an industrial point source, the presence of Hg may be controlled both by internal mobilization of historical Hg contaminants and by influx of atmospherically derived Hg. The latter may reach the water both as direct deposition and * Corresponding author phone: +46 46 2223781; e-mail:
[email protected]. 3514
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 10, 2009
Study Site. We conducted our studies in Paulistro¨msån, a boreal forest stream in southeastern Sweden (Figure 1). Its catchment (219 km2) consists of dense coniferous forest (60%), water surfaces (14%), scattered coniferous forest (10%), farm land and pastures (10%), deciduous forest (3%), wetland (2%), rock ( 0.60). Hence, despite lower TOC concentration in 1996 than in 2003, the years appeared not to differ with respect to HgT concentration. As will be shown below, there are several possible reasons for a variable relationship between HgT and TOC. In 2003, HgD ( 0.05), but the Wilcoxon’s signed-rank test indicated a slight elevation (P < 0.05). In 1996, mean MeHg at SiteHg was 0.26 ng Hg/L (range: 0.07-1.2). LogMeHg did not differ significantly between 1996 and 2003 at SiteHg (P > 0.7). MeHgD ( 0.05). Only slightly better fit was obtained using the product Q × T instead of T as the independent variable (r2 ) 0.35, P < 0.005). Similar relationships were seen at SiteHg the same year (2003) (SI Figure S4). In 1996, logTOC was significantly correlated with Q (r2 ) 0.43, P < 0.005) but not with T (r2 ) 0.21, P < 0.05), and Q × T explained more variation than Q alone (r2 ) 0.55, P < 0.0005). When TOC varies with Q rather than with T, water flow path, erosion, and resuspension are likely controlling factors in the TOC transfer to streamwater, whereas strong correlation with T might arise when previous high flows have depleted mobile OM. Improved fit using Q × T as the explanatory variable can be understood as a combined effect of T-dependent production of mobile OM and Q-dependent flushing of soil and sediment. Also, microbial activity may be stimulated not only by high T but also by a high water table and high Q (20-22). Compared with logTOC, logHgT showed stronger correlation with Q at both stations and in both years (SI Figures S4 and S5). A multiple linear regression model with TOC (mg/L) (P < 0.0001), Q (m3/s) (P < 0.0001), and site (Siteref ) 0, SiteHg ) 1) (P < 0.01) as independent variables explained 86% of the variation of the pooled 2003 HgT (ng/L) data: HgT ) - 2.15 + 0.26(TOC) + 1.50(Q) + 0.47(Site)
(2)
The correlation between logMeHg and the product Q × T was highly significant (r2 > 0.45, P < 0.001) and Q × T explained more variation in logMeHg than either Q or T alone for both sites and both years (SI Figure S6), suggesting that high discharge during summer causes higher stream MeHg loads than high discharge occurring earlier or later in the year. The same conclusion was drawn from studies of MeHg in Minnesota streams (18). There must be time lags between Q × T and MeHg that depend on the rate of change in Q × T, and it should matter whether a change in Q × T is a result mainly of a change in T or in Q (see the following). Moreover, water level fluctuations may be conducive to MeHg production (23), further complicating the role of Q. In any case, if Q and T were strongly correlated, it would be statistically unsound to use Q × T as a predictor variable, and the relationship between Q × T and MeHg may drastically change when resuspension becomes important.
FIGURE 4. Concentrations of indicated elements, compounds and suspended solids, and absorbance of riparian peat water (PW1-PW4) during the spring flood of 2003. DO ) dissolved oxygen, together with streamwater discharge (a), FeD ) total Fe in filtered water (b), sulfate, together with SD ) total S in filtered water (c), methyl Hg in unfiltered water ) MeHg, and in filtered water ) MeHgD (d), total Hg in unfiltered water ) HgT and in filtered water ) HgD (e), TSS and LOI (f), absorption of 254 nm UV by filtered water (g). MeHg below detection was given the value 0.06 ng Hg/L. MeHg vs Water Chemistry. At both Siteref and SiteHg in DO than those close to the lower edge of each site (PW2 and 2003, logMeHg showed stronger correlation with logFe than PW4) (Figure 4a). In all wells, DO subsequently declined, tracking the decline in Q in three of the wells. In the well with Q × T, possibly because Hg methylation, release of MeHg showing the highest DO levels (PW3), DO declined even before to water and Fe oxide dissolution are processes that are Q did. strongly linked to sulfide formation. However, there was no All filtered samples from the wells were filtered directly significant correlation between logMeHg and logFe at SiteHg in 1996. Conditions may not have been reducing enough in the field. As expected, FeD (
