Sampling of shallow alluvial groundwaters and shallow wetland ponds occurred periodically from June 2016 to September 2018 at the Swamp White Oak (SWO) Nature Preserve, a Nature Conservancy-owned restored savanna oak forest dominated by Quercus bicolor (Dettman and Mabry, 2008), occupying the floodplain of the lower Cedar River in east-central Iowa near Conesville, Iowa (lat. 41°24′36″, long. 91°17′06″, see map in Figure 1). Samples were also collected nearby from the main channel of the Cedar River at the site of a United States Geological Survey (USGS) gaging station (USGS station number 05465000), for which continuous discharge measurements have been available since 1940. The Cedar River is a major tributary of the Upper Mississippi River, with a watershed area of approximately 20,200 km² at the sampling station. Corn and soybean cultivation comprise the major land cover (72.5% of land area) with smaller areas of grasslands (14.9%, including pasture), forest (7%) and impervious surface (2.7%). The climate of the region is temperate, with hot, humid summers and cold, dry winters. Mean annual precipitation in nearby Muscatine, Iowa is 810 mm/year, with May, June, and July typically the wettest months of the year (NCEI Climatology Database). Long-term mean annual discharge of the Cedar River at Conesville, Iowa is 164 m³ s⁻¹ (5,800 ft³ s⁻¹).

A monitoring well network encompassing microtopographic highs (sandy ridges) and lows (silty swales) was installed in 2005 as part of studies of riparian hydrology, nutrient processing, and carbon dynamics within the SWO preserve (Schilling and Jacobson, 2009; 2012; 2015). Monitoring wells were installed to a depth of approximately 2 m with a 1.5 m long polyvinyl chloride (PVC) well-screen and 1.5 m long PVC riser (Schilling and Jacobson, 2009). The water table in these floodplain monitoring wells responds rapidly to Cedar River stage indicating a high level of hydrologic connectivity between the floodplain and the river (Schilling and Jacobson, 2009). Significant differences in groundwater quality, with respect to nitrate, DO, and DOC, exist between ridges and swales, with wells in microtopographic depressions exhibiting lower DO and nitrate and higher DOC (Schilling and Jacobson, 2012). Schilling and Jacobson (2009)A subset of three topographically “high” wells (>180.5 m land surface elevation; Center, West, and South) and three “low” wells (<180.5 m land surface elevation; North, East, and Swale) were selected for our study of Hg biogeochemistry.

All samples were collected utilizing a clean hands/dirty hands approach (Fitzgerald, 1999) to minimize sample contamination. Monitoring wells were sampled using a peristaltic pump with acid-washed Teflon tubing and a short piece of silicone tubing on the pump head. Prior to sampling, depth to water table was measured to the nearest cm using an electronic water level tape (Solinst Model 101). Wells were flushed for 2–3 min and allowed to recharge for a minimum of 1 h prior to sampling. Samples from the two ponds were collected by wading and filling a sample bottle at maximal reach from the area of sediment suspension caused by wading. Samples from the Cedar River and the SWO outlet stream were collected by lowering a virgin polyethylene terephthalate glycol (PETG) bottle or acid-washed Teflon bottle to a depth of about 30 cm below the water surface. All samples were stored in PETG bottles with zero headspace on ice until return to laboratory to minimize sample oxidation. DO, pH, specific conductance, and water temperature were measured immediately on a separate subsample using a multimeter (YSI Pro DSS or PASCO Xplorer GLX) with optical DO sensor. Samples for dissolved Fe(II) and sulfide were immediately filtered in the field through a 0.2 µm nylon membrane filter and preserved in ferrozine colorimetric reagent (1 mM ferrozine, 10 mM HEPES buffer) for Fe(II) analysis (Stookey, 1970) and sulfide antioxidant buffer (SAOB) for sulfide analysis (Clesceri et al., 2000).

Samples were returned to the laboratory, refrigerated, and processed within 24 h of collection. Samples were filtered through 0.45 µm glass fiber filters (baked for 4 h at 550^OC in a muffle furnace and stored at room temperature in a double-bagged sealed plastic container) using an acid-washed PFA filtration assembly (Savillex; Eden Prairie, MN) into virgin PETG bottles. A known volume (generally about 150 mL) of filtrate was collected for both THg and MeHg and analyses and preserved with 1% trace metal grade HCl. Filters corresponding to particulate THg and MeHg were retrieved and stored frozen until analysis. All MeHg analyses were completed within 60 days, within the acceptable holding time (180 days) per EPA Method 1630 (US EPA, 1998). For THg, the majority of samples were analyzed with 90 days, the EPA recommended holding time (US EPA, 2002). 25% of samples (corresponding to THg samples from 7/7/17, 8/17/17, 7/19/18, and 9/4/18) were held acid-preserved and refrigerated up to 10–18 months prior to analysis. Prior work indicates that preserved THg samples are stable for at least 10 months (Parker and Bloom, 2005). Filtered THg samples for the Cedar River on 6/3/16 and 7/17/17 and particulate MeHg on 6/3/16 and 9/8/16 were lost during processing. A separate subsample of the filtrate was preserved with 2% HNO₃ for other major and trace element analysis, carried out within 6 months of sample collection. Samples for anions and DOC were collected in 50 mL polypropylene tubes and stored frozen until analysis, typically within 30 days of sample collection.

In June and July of 2018, sediment cores were retrieved from the Center and South ponds at SWO via hand-coring with 5 cm diameter, 45 cm long polycarbonate tubes, predrilled at 1 cm intervals with openings sealed with silicone. Sediment cores were 12–16 cm in length and filled with overlying water leaving no headspace, capped with plastic seals wrapped in vinyl tape, and transported to the laboratory at ambient temperature. Upon return to the laboratory, cores were stored overnight at 4°C, then transferred the next day to incubators at field temperature (22.5°C–23.7°C) for in situ methylation potential determination or sectioned at 2 cm intervals for ambient total Hg (THg) and MeHg analyses. Sub-sectioned sediment core samples were homogenized and stored frozen until analysis.

Core injections for methylation potential (k _m) determination followed the method of Mitchell and Gilmour (2008). Sediment cores were injected at 1 cm intervals through the silicone septa with 100 µL of 4 μg/mL enriched ²⁰¹HgCl₂ (pre-equilibrated for 1 hour with 0.2 µm-filtered overlying water from the respective ponds). Injections were performed while moving the syringe needle in a back-and-forth motion while withdrawing the needle to try to evenly inject the ²⁰¹Hg spike across the core’s entire cross-sectional area at each depth. After injection, the cores were incubated at site temperature in the dark for 3 h. Cores were then sectioned at 2 cm intervals and samples frozen for analysis.

Samples for MeHg (either filtered water, filters with particulate matter, or sediments) were analyzed by steam distillation, ethylation with sodium tetraethylborate, and isotope dilution gas chromatography inductively coupled plasma mass spectrometry (Hintelmann and Evans, 1997) using enriched Me¹⁹⁹Hg as the internal standard (Oak Ridge National Laboratories, 91.95% ¹⁹⁹Hg). All MeHg analyses were carried out with a manual purge and trap GC system as described in Kanzler et al. (2018) interfaced to an Agilent 7500ce ICP-MS. Quality assurance/quality control (QA/QC) measures included analysis of field blanks (filter blanks), distillation blanks, a certified reference material (NIST 1566b oyster tissue), and duplicate sample analyses. QA/QC data for MeHg analyses is summarized in Supplementary Table S1.

Sample digestion and analysis protocols for THg analyses differed depending on the subsample type. Filtered water samples for total Hg analysis were digested overnight in 1% BrCl. Filters (for particulate THg) and sediments were digested in 7:4 v/v HNO₃:H₂SO₄ at 190°C for 2 h, diluted with deionized water (resistivity of 18.2 MΩ-cm), then preserved with 1% BrCl. For filtered water samples and filters, THg was determined by isotope dilution purge and trap dual gold trap Hg amalgamation coupled to ICP-MS. Briefly, a 50–100 mL sample was added to a bubbler with 100 µL of 20% w/v hydroxylamine hydrochloride (to reduce excess BrCl), 300 µL of 60 g/L SnCl₂ reductant, and a known amount of enriched ²⁰¹HgCl₂ (Oak Ridge National Laboratories, 98.11% ²⁰¹Hg) and purged for 15 min onto a sample trap containing glass beads coated with gold (Brooks Rand Instruments, Seattle WA). Gold sample traps were heated via a nichrome wire to release amalgamated Hg in a stream of argon gas onto a second gold trap (the analytical trap), which was heated to release Hg directly into the Agilent 7500ce ICP-MS. A small subset of filtered THg samples were analyzed by an automated total mercury analysis system (Brooks Rand Instruments MERX-T using SnCl₂ reduction and dual gold trap amalgamation) interfaced to a Perkin Elmer Nexion 2000 ICP-MS. Sediment samples which had higher THg concentrations were analyzed by ICP-MS with flow injection reduction of Hg(II) to Hg(0) using 60 g/L SnCl₂ in 1% HCl (Graham et al., 2012). QA/QC measures for THg analyses were identical to those for MeHg (except that NIST 2709a San Joaquin soil was used as certified reference material) and are summarized in Supplementary Table S1.

Dissolved Fe(II) was measured colorimetrically at 562 nm following complexation with ferrozine reagent (Stookey, 1970). Dissolved sulfide concentrations were measured using a sulfide ion selective electrode calibrated against a Pb-titrated sulfide standard (Clesceri et al., 2000). Major anion concentrations (chloride, nitrate, and sulfate) were determined using isocratic ion chromatography (Latchat 8000 series) with a carbonate eluent (2.6 mM Na₂CO₃ and 2.0 M NaHCO₃, pH = 10.44) with conductivity detection and ion suppression. Major and trace element concentrations other than Hg were determined by ICP-MS using either an Agilent 7500 ce or Perkin Elmer Nexion 2000. DOC concentrations were determined using a Shimadzu TOC-V_CSH/CSN analyzer. UV-Vis absorbance spectra were measured in quartz cuvettes on a Cary 60 UV-Vis spectrophotometer following addition of 0.2% w/v hydroxylamine hydrochloride to the sample to minimize interference due to formation of light scattering Fe(III) (oxy)hydroxides (Doane and Horwáth, 2010) upon sample storage after filtering. QA/QC measures for DOC, anions, and trace elements are provided in Supplementary Table S2.

Three methods were utilized for computing fluxes of THg and MeHg in the Cedar River. First, fluxes were estimated using the program LOADEST (Runkel et al., 2004) as implemented in R (R Core Team, 2019). LOADEST has the advantage of automated selection from multiple regression models using the Akaike Information Criterion, with computation of model coefficients using an adjusted maximum likelihood estimation (AMLE). A downside is the risk of overfitting given the relatively small number of observations (n = 16 across 3 years).

In a second approach, we calculated daily fluxes based on daily discharge and median THg or MeHg concentration. A third approach relied on using high resolution turbidity (5 min time interval) and discharge (at 15 min intervals) to estimate instantaneous fluxes at 15 min intervals. Turbidity data were available for a portion of 2018 (late February through mid-November), measured at the Conesville station at 5 min intervals (data retrieved from the Iowa Water Quality Information System; IWQIS: https://iwqis.iowawis.org/app/). The turbidity-based approach is discussed further in the results and discussion below. Fluxes were converted to yields by dividing the flux by the watershed area.

A hydrograph for the Cedar River during the study period is shown in Figure 2, and shows that discharge exceeded that required for overbank flow on several occasions during 2016–2018. The Fall 2016 event corresponds to the second largest discharge on record. Total unfiltered Hg concentrations in the Cedar River are also shown in Figure 2 and ranged from 1.6 to 14.6 ng/L (mean = 8.6 ± 3.4 ng/L, where uncertainty represents one standard deviation); median = 7.1 ng/L, n = 16), the bulk (nearly 80% on average) of which was particulate THg (mean = 6.7 ± 3.4 ng/L; median = 6.7 ng/L). No obvious seasonal trend is apparent, though few samples were collected during low flow winter months due to ice cover. Measured unfiltered THg concentrations are within the range reported by others during the 1990s and early 2000s for UMRB rivers at varying discharge scales (Hurley et al., 1995; Balogh et al., 1998; 2005). Particulate THg concentrations were strongly correlated to measured total suspended solids (TSS; r ² = 0.71, p = 0.0005), as shown in Supporting Information Supplementary Figure S2. Unfiltered THg concentrations were well correlated with discharge (r ² = 0.46, p = 0.004), likely due to a positive relationship between discharge and TSS.

Unfiltered MeHg concentrations in the Cedar River ranged from 0.05 to 0.82 ng/L (mean = 0.30 ± 0.21 ng/L; median = 0.25 ng/L, n = 14), with the particulate fraction accounting for about 64% of total MeHg. Total unfiltered MeHg was not strongly correlated with unfiltered THg (Supplementary Figure S2). The lower fraction of particulate MeHg compared to THg reflects stronger solid-solution partitioning for inorganic Hg relative to MeHg–calculated log K_d’s based on TSS (distribution coefficients with units of L kg^-1) for THg averaged 4.89 vs 4.21 for MeHg. Measured log K_d’s are similar to that reported for Wisconsin rivers by Babiarz et al. (1998). The mean unfiltered MeHg concentration for the Cedar River is similar to the 0.20 ng/L reported for the Minnesota River (Balogh et al., 2005) but higher than that reported for agriculture-intensive watersheds in Wisconsin (mean MeHg = <0.03–0.085 ng/L) (Hurley et al., 1995). Approximately 4.8% of unfiltered THg was MeHg, also higher than that reported for predominantly agricultural watersheds in Wisconsin rivers (Hurley et al., 1995). The highest filtered MeHg concentrations were observed in October 2016 May 2018 and September 2018 following overbank flow events, consistent with the idea that the floodplain is a significant MeHg source area.

Annual THg and MeHg yields were estimated using the daily discharge data and measured concentrations (Table 1). Automated model selection for both THg and MeHg in LOADEST resulted in a model of the form: ln f l u x = α 0 + α 1 l n Q (1) where lnQ is the ln(observed daily discharge)—center of ln(daily discharge). In the case of THg, this model described 81.2% of the variability in THg flux (p < 0.0001), whereas the model for MeHg only explained 40.6% of the variability in MeHg flux (p = 0.0039). Yields calculated using LOADEST-derived fluxes ranged from 2.6 to 6.9 μg m^-2 y^-1 for THg and 0.09–0.18 μg m^-2 y^-1 for unfiltered MeHg. Using the median concentrations of unfiltered THg and MeHg, yields were estimated to be between 2.5 and 4.2 μg m^-2 y^-1 for THg and 0.09–0.15 μg m^-2 y^-1 for MeHg. The median concentration and LOADEST approaches agree closely for MeHg, but the median concentration approach results in lower yields for THg in wet years.

An alternative approach for estimating THg yields is to use turbidity as a proxy measure of particulate THg concentration. Turbidity is strongly linearly correlated with TSS (Supplementary Figure S3, r ² = 0.96, p < 0.001) and TSS is strongly correlated with particulate THg (Supplementary Figure S2). A potential advantage of this approach is that turbidity and discharge measures are only weakly correlated (r ² = 0.01), an observation that is likely driven by significant discharge-TSS hysteresis, so the high-resolution turbidity measurements provide a more accurate real-time estimate of sediment flux not captured by the regression models. For example, during 2018, high turbidity occurred on the rising limb of the hydrograph, but fell much more quickly than discharge (Supplementary Figure S4). For 2018, where high resolution turbidity data were available, the particulate THg yield was estimated at 3.1 μg m^-2 y^-1 for the period 3/1 through 11/14/18. Assuming that the particulate THg flux is 80% of the total Hg flux (supported by concentration measurements in Figure 2C), the total Hg yield for this period would be 3.9 μg m^-2 y^-1, within 10% of the yield predicted based on median THg concentration (3.6 μg m^-2 y^-1 for the same interval). The LOADEST estimate for this same period is significantly higher, about 6.1 μg m^-2 y^-1. High resolution turbidity monitoring may be a useful proxy measure for THg in some watersheds (Riscassi and Scanlon, 2013; Vermilyea et al., 2017), as also demonstrated for phosphorus (Schilling et al., 2017), providing more accurate estimates of Hg fluxes especially under circumstances where suspended sediment delivery shows complex relationships with discharge (see Supplementary Figure S4).

Total unfiltered Hg and MeHg yields for the Cedar River are toward the higher end of what has been reported for agriculturally-dominated watersheds elsewhere in the Upper Midwest United States and even for many watersheds with much larger proportions of forests and wetlands (Bishop et al., 2020). Unfiltered MeHg and THg yields are above the 75th percentile (3.12 μg m^-2 y^-1 for THg, n = 80 and 0.091 μg m^-2 y^-1 for MeHg, n = 67) for all watersheds reviewed by Bishop et al. (2020). Hurley et al. (1995) reported mean watershed yields of 0.05 (fall) to 3.2 (spring) μg m^-2 y^-1 for unfiltered THg in agricultural watersheds in Wisconsin in the early 1990s. Balogh et al. (2005) reported THg yields of 0.3–6.4 μg m^-2 y^-1 for the Minnesota River and its tributaries in the early 2000s. Reported unfiltered MeHg yields range from 0.002 to 0.05 μg m^-2 y^-1 in Wisconsin rivers (Hurley et al., 1995) to 0.03–0.09 μg m^-2 y^-1 in Minnesota Rivers (Balogh et al., 2005). Wet deposition for the Cedar River watershed is estimated at 10–14 μg m^-2 y^-1 for the period 2016 to 2018 (NADP, 2019). Dry deposition and litter fall inputs of Hg within this watershed are not known [but are often similar in magnitude to wet deposition (Bishop et al., 2020)]. We are not aware of any local point sources of Hg within the watershed. Relative to wet deposition, observed THg yields thus represent watershed transfer efficiencies of about 25% to nearly 70%.

High THg yields likely reflect high TSS loads in rivers of this region. For 2018, we estimated TSS yield to be about 60 tons/km², similar to long-term observations of sediment yields for Iowa rivers contributing to the Upper Mississippi River Basin (Heimann et al., 2011; Jones and Schilling, 2011). High TSS loads can drive high THg and MeHg yields even when mean Hg concentrations of soil particles are fairly low (median THg concentration of suspended sediments was 48 ng/g and median MeHg concentration was 1.5 ng/g). Changing climate (increased frequency of intense rainfall events (Mallakpour and Villarini, 2015)) and land management practices (e.g., bringing conservation reserve program (CRP) lands back into production) have potential to increase sediment yields, and hence Hg yields, in this region.

Riparian groundwaters and wetlands may be important sources of MeHg to the Cedar River. In order to understand biogeochemical processes controlling MeHg production, we monitored geochemistry in groundwater and surface waters within the adjacent river floodplain. Groundwater geochemistry was monitored in six shallow monitoring wells, representing three microtopographically “high” wells and three “low” wells over the period June 2016 through July 2018. The south well has the highest land elevation, but geochemistry was more similar to low wells. Hereafter, we exclude the south well from our analysis when discussing differences between low and high wells. Aquifer sediments at high wells tend to be coarser grained and have greater depth to water table (DWT) than low wells. Mean DWT was about 0.6 m greater in high wells (1.10 m) compared to low wells (0.47 m; p < 0.001 for one-way ANOVA). The rapid response of the water table elevation to Cedar River discharge is indicative of high hydrologic connectivity between the floodplain and the river. Supplementary Figure S5 shows the strong linear relationship (r ² = 0.78) between 5-day prior discharge of Cedar River and DWT at the swale well. Our results are in close agreement with insights provided by continuous monitoring of the water table in the center well from 2005–2007, which suggested a lag of 1–6 days between river discharge and water table elevation (Schilling and Jacobson, 2009).

Low and high wells also showed contrasting geochemistry. High wells had significantly lower specific conductance, higher DO, and lower DOC compared to low wells averaged across all sampling events (Supplementary Figure S6). Specific conductance was about 60% higher (p < 0.001 for one-way ANOVA) in low wells (mean = 352 ± 62 μS/cm) vs high wells (mean = 218 ± 51 μS/cm). DOC was about double (p < 0.001 for one-way ANOVA) in low wells (mean = 4.07 ± 1.74 mg/L) compared to high wells (mean = 2.09 ± 1.25 mg/L). Groundwater pH was similar across all wells (mean = 6.60 ± 0.34). Analysis of major cations and anions suggested similar distribution of ions (see Piper diagram in Supplementary Figure S7), though absolute concentrations were significantly higher in low wells compared to high wells, consistent with conductivity data. Alkalinity, estimated for these relatively hard groundwaters as the difference between strong base cation equivalents and strong acid anion equivalents, was nearly double for low wells (mean alkalinity = 2.74 meq/L) vs high wells (mean alkalinity = 1.35 meq/L). We posit that higher alkalinity and dissolved solids for low wells reflect greater rates of carbon mineralization at these sites, resulting in higher P_CO2 in groundwater, thus driving increased dissolution of carbonate minerals. Increased dissolved inorganic carbon and carbonate mineral dissolution in shallow aquifers in response to increased soil respiration in summer months are well-documented in other similar systems (Tsypin and Macpherson, 2012). Alluvial sediments at the high and low sites contain similar organic carbon in surficial sediments, but low sites have 2–5 times more carbon at depths greater than or equal to 0.5 m (Schilling and Jacobson, unpublished data), supporting the idea that increased carbon availability leads to greater subsurface microbial activity in the low wells. This hypothesis is further supported by observations of DO in these shallow groundwaters, as an indicator of aerobic microbial activity. DO was about 60% higher (p = 0.001 for one-way ANOVA) across all high wells (mean = 3.5 ± 1.9 mg/L) compared to all low wells (mean = 2.2 ± 1.4 mg/L). No significant differences were observed between low and high wells for other indicators of terminal electron accepting processes, including nitrate, dissolved Fe(II), dissolved Mn, and sulfate (Supplementary Figure S6).

Filtered THg concentrations in shallow groundwater (mean = 0.87 ± 0.92 ng/L; median = 0.57 ng/L across all sampling sites and dates) were similar to that observed in the Cedar River (mean = 0.94 ± 0.83 ng/L). Filtered MeHg concentrations (mean = 0.16 ± 0.26 ng/L; median = 0.087 ng/L) were slightly higher than that observed for Cedar River (mean = 0.10 ± 0.08 ng/L). MeHg represented about 22% of filtered THg for all groundwater samples. These averages, however, may mask significant difference in Hg and MeHg concentrations in time and space within the alluvial aquifer. Figure 3 shows differences in THg and MeHg concentrations averaged across all sampling dates. Filtered THg concentrations generally ranged from 0.5 to 2 ng/L, with only a few exceptions, and were not significantly different between wells, though both the south well and swale well did have larger median concentrations. Greater spatial variability existed for MeHg concentrations, with MeHg concentrations ranging from below detection (0.029 ng/L) to nearly 1.5 ng/L (a 50-fold difference). MeHg represented as little as 1% of filtered THg to as much as 94%, with a mean of 22% across all sites and sampling dates. MeHg concentrations were greatest in the low-lying north and swale wells.

There are few studies of Hg and MeHg in alluvial groundwater with which to compare our findings. Vidon et al. (2013) measured THg and MeHg in alluvial groundwaters at three sites in central Indiana and observed lower filtered THg concentrations medians ranging from 0.28 to 1.05 ng/L) and filtered MeHg concentrations (medians ranging from 0.01 to 0.05 ng/L (Vidon et al., 2013). The highest MeHg concentrations in groundwater at these sites were observed during the summer months of July and August. Creswell et al. (2008) reported THg concentrations in hyporheic wetland sediments from northern Wisconsin of 2–10 ng/L and MeHg concentrations of <0.05 to as much as 3 ng/L, more similar to the range of MeHg and THg concentrations we observe in these alluvial groundwaters in Iowa. Bradley et al. (2010) reported filtered MeHg concentrations ranging from 0.1 to greater than 1.5 ng/L in riparian pools fed by groundwater in a coastal plain floodplain of a blackwater stream in South Carolina. Filtered THg concentrations were reported to range from 1 to 10 ng/L in groundwater in a follow-up study (Bradley et al., 2012). Of these studies, the work of Vidon et al. (2013) is most similar in terms of geomorphic setting–a river floodplain developed on glacial outwash/alluvium with significant agriculture in upland portions of the watershed. Vidon et al. (2013) concluded that riparian zones in regions with low groundwater DOC may not contribute appreciable MeHg. Here, we show higher MeHg in groundwaters, especially in response to rapid groundwater table rise (see section 3.6 below), and suggest that such hydrologically connected floodplains do constitute important sources of MeHg during critical moments. For the Cedar River system, the moments of highest MeHg in groundwaters immediately precede the highest observed filtered MeHg concentrations in the Cedar River (Figure 2).

Comparing wells sampled on the same date (see Supplementary Figure S8), MeHg concentrations were on average about 2.5–7.9 times higher in the microtopographically low north and swale wells compared to the microtopographically high center and west wells. The low east well and unclassified south well were on average 1.8–3.7 times higher compared to the center well, but similar to the west well. Just as microtopography can influence carbon and nutrient cycling (Schilling and Jacobson, 2012), Hg dynamics are influenced by microtopography over very small spatial scales in this alluvial aquifer. We hypothesize that these microtopographic differences in groundwater MeHg reflect underlying geomorphic control on biogeochemical processing of carbon and other redox active solutes, as illustrated in the geochemical data in Supplementary Figure S6. Our finding highlights the need for consideration of microtopography in sampling design when investigating Hg biogeochemistry in groundwaters where small-scale spatial heterogeneity is likely to be important (e.g., river floodplains).

To better understand the controls on Hg speciation in these groundwaters, we examined relationships between MeHg and other geochemical variables (Table 2). Significant positive correlations were observed between filtered MeHg (FMeHg) and THg, DOC, absorbance at 254 nm (A₂₅₄, a parameter that depends on both DOM concentration and DOM character, especially aromatic carbon content (Weishaar et al., 2003)), dissolved Fe(II), and temperature. The fraction of filtered THg as MeHg (%MeHg) positively correlated with DOC, A_254, dissolved Fe(II) and Mn. The strong link between MeHg and DOC (r = 0.603) or A₂₅₄ (r = 0.640) has been noted in numerous studies of surface waters and wetlands (Hall et al., 2008) and may relate to the role of DOM as a principal control of microbial community structure and function, including Hg methylation (Graham et al., 2018), the role of DOM in altering the bioavailability of inorganic Hg(II) for methylation (Graham et al., 2012), and/or increased aqueous partitioning of MeHg due to MeHg complexation by thiols in DOM (Skyllberg, 2008).

Interestingly, we did not observe any significant correlation of MeHg or %MeHg with sulfate concentrations. The mean sulfate concentration across all groundwater sites was 20.7 ± 20.9 mg/L, about 35% lower than mean sulfate in the Cedar River. Sulfate reduction was evidenced by minima in sulfate during the summer months, as reflected in the inverse correlation of groundwater sulfate with temperature shown in Table 2. Dissolved sulfide was rarely detected above 2 µM (mean = 0.5 ± 1.9 µM; Supplementary Figure S6) and showed no significant correlation with sulfate. Precipitation of Fe-S minerals (e.g., mackinawite, FeS(s) (Rickard, 2006)) via reaction of sulfide with dissolved Fe(II) produced from dissimilatory Fe reduction may explain the low sulfide concentrations in these groundwaters. The highest sulfate concentrations, occasionally exceeding 100 mg/L, occurred during early spring, when groundwaters were more oxygen rich, suggesting internal cycling of sulfur stores in aquifer sediments (e.g., oxidation of reduced Fe-S minerals accumulating from prior sulfate and iron reduction). Compared to many sulfate-limited freshwater wetlands where Hg methylation has been extensively studied, sulfate supply may not limit Hg methylation in these systems and/or Hg methylation may be driven by microorganisms other than SRB (e.g., Fe-reducers, methanogens, or fermentative bacteria (Gilmour et al., 2013)). Recent studies investigating the diversity of hgcA in a variety of freshwater environments have shown that while Deltaproteobacteria (which includes SRB and FeRB) are often well-represented amongst Hg-methylators, other organisms, with diverse metabolisms, including those belonging to Firmicutes, Spirochaetota, Chloroflexota, Nitrospirota, and Methanomicrobia, can comprise an important fraction of the Hg-methylator community (Schaefer et al., 2014; Christensen et al., 2019; Jones et al., 2020; Xu et al., 2021).

Significant positive correlations between %MeHg and dissolved Fe(II) and Mn concentrations suggest that dissimilatory metal reduction may be important to Hg methylation in this system. Strong correlations between MeHg concentrations and dissolved Mn have been noted in agricultural wetlands (i.e., rice fields; (Alpers et al., 2014); and in a eutrophic reservoir (Beutel et al., 2020). We note that Geobacter, which include a number of species capable of Hg methylation (Gionfriddo et al., 2019), can reduce both Fe(III) and Mn(III,IV) (hydr)oxides (Richter et al., 2012). Mn(III,IV) hydroxides may also be reduced by dissolved Fe(II) (Schaefer et al., 2017), complicating interpretation of relationships between dissolved Mn and MeHg.

Wetland ponds represent another possible source of MeHg to the Cedar River. Total Hg and MeHg concentrations in the shallow wetland ponds tended to be greater than in groundwaters (Figure 3). Across all three wetland ponds, mean filtered MeHg was 0.26 ± 0.27 ng/L (median = 0.17 ng/L, n = 45) and mean filtered THg was 1.24 ± 0.65 ng/L (median = 1.02 ng/L, n = 45). Due to a combination of sediment suspension and primary production within these shallow wetland ponds and streams, these waters contain appreciable suspended particulate matter (Figure 4; mean SPM = 54.5 ± 71.4 mg/L; median SPM = 21.6 mg/L). Whole, unfiltered THg and MeHg concentrations are thus appreciably higher than filtered concentrations, averaging 4.89 ± 4.25 ng/L for THg (median = 2.96 ng/L) and 0.50 ± 0.48 ng/L for MeHg (median = 0.32 ng/L) respectively. Filtered MeHg comprised 55% ± 21% (median = 53%) of the total MeHg; for total Hg, this fraction was slightly lower, at 38% ± 24% (median = 37%). Distribution coefficients (log K_D’s) for partitioning between the dissolved phase and suspended sediments averaged 4.5 ± 0.6 and 4.9 ± 0.5 for MeHg and THg respectively, similar to that observed for the Cedar River. In general, no significant differences in MeHg and THg concentrations or partitioning behavior were observed across the three wetland ponds, excepting higher THg in the south pond that had the highest suspended particulate matter.

Similar to groundwater samples, we examined a correlation matrix of Hg species concentrations and associated geochemical and physical parameters for these wetland ponds (Table 3). Total MeHg was significantly correlated with dissolved Mn and the absorbance at 254 nm (Abs₂₅₄), a parameter that reflects both DOM concentration and aromaticity (Weishaar et al., 2003). Total Hg was most strongly correlated with SPM and DOC. Several earlier studies have highlighted the link between DOM and MeHg in wetland systems. Hall et al. (2008) found that MeHg in wetlands in coastal Louisiana was highly correlated with the hydrophobic acid fraction of DOM, but weakly correlated with bulk DOC. Mitchell and Gilmour (2008) observed strong correlation of Hg methylation rates in wetland sediments with SUVA₂₈₀ and slope ratio (S_R) (Helms et al., 2008), spectroscopic measures mostly strongly related to DOM molecular weight and aromaticity. Our results suggest tight coupling of DOM and MeHg in inland wetlands within agriculturally-dominated regions of the UMRB. Recent advances in measuring individual low molecular weight (LMW) thiols (Liem-Nguyen et al., 2021) and sulfur-containing compounds in DOM by high-resolution mass spectrometry (HR-MS) (Mangal et al., 2022) indicate the importance of thiols to Hg methylation in freshwater wetlands, presumably by controlling Hg bioavailability to Hg-methylating organisms (Graham et al., 2012; 2017). Increased anthropogenic sulfur inputs in regions with intensive corn and soy cultivation appear likely as soils may become sulfur-depleted following acid deposition recovery (Hinckley et al., 2020). Such sulfur inputs could induce changes in microbial community structure (Jones et al., 2020) and/or Hg bioavailability for methylation due to complex linkages between sulfide, DOM, and Hg (Graham et al., 2017; Poulin et al., 2017) that alter MeHg production in terrestrial ecosystems.

Observations of Hg speciation in wetland waters were supplemented by measurements of Hg speciation and Hg methylation potentials in shallow sediment cores. Total Hg concentrations in wetland sediments were generally between 70 and 100 ng/g dry weight and generally decreased slightly with depth (Figure 5). MeHg concentrations were greatest in near surface sediments, reaching 1 ng/g dry weight, up to 1.6% of THg. Methylation potentials were determined in short term incubations with injection of enriched ²⁰¹Hg pre-equilibrated with filtered surface water from each site and were greatest (k _meth exceeding 0.15 d^-1) in surface sediments.

Surficial sediments in the SWO ponds have methylation potentials toward the upper end of the range reported previously for other freshwater wetlands (Hoggarth et al., 2015), indicative of robust Hg methylation potential within wetlands of the Cedar River floodplain. Interestingly, the %MeHg observed in sediments (not exceeding 1.6%) is relatively low given the large values of k _meth observed. We hypothesize that MeHg degradation in SWO sediments may be comparatively fast and relatively invariant with depth, leading to the strong correlation between k _meth and MeHg in sediment (Figure 5G). Future studies at SWO will investigate biological and dark abiotic MeHg degradation processes, which, apart from MeHg degradation by merB containing organisms, remain poorly understood (Barkay and Gu, 2022).

Geochemistry in groundwaters and wetlands can change rapidly in riparian zones with high hydrologic connectivity. During the summer of 2018, we sampled at SWO just before and after overbank flow events of the Cedar River (Figure 6). We observed that filtered Hg and MeHg concentrations in groundwater increased markedly immediately following flood conditions. In the low-lying swale well (Figure 6B), groundwater MeHg concentrations increased 17 to 35-fold (from as little as 0.04 ng/L to 1.4 ng/L) following flood events. THg concentrations also showed significant increases, about 5-fold (from 0.5 to 2.7 ng/L) following flooding. During this period, we observed tight coupling with DOC concentrations. We hypothesize that water table rise into the OM-rich surface soils results in leaching of both DOM and THg and supports high MeHg production. A similar trend for MeHg and THg following flooding was observed within the center pond wetland, but DOC was actually highest during the June sampling with the lowest MeHg (Figure 6C). Apparent mismatch between peak DOC and peak MeHg and THg at this site may relate to shifting sources of DOM within the pond that have differential impact on Hg bioavailbity for methylation. DOM may be transported into the pond via both overland flow and groundwater, as well as being produced autochthonously via leakage from phytoplankton. Periods of groundwater table rise or overland flow are likely to increase the delivery of allocthonous DOM. During dry periods, groundwater recharge of the pond will be reduced, and DOM will be mostly derived from algal biomass produced within the pond. Laboratory experiments indicate that high molecular weight, more aromatic allocthonous DOM is especially effective at enhancing Hg methylation by Hg-methylating bacteria (Graham et al., 2013).

Importantly, we see no evidence that delivery of nitrate rich water suppresses Hg methylation during flooding. Nitrate concentrations in the Cedar River averaged 6.5 ± 1.7 mg N/L during the 2018 sampling dates but were markedly lower in wetlands (0.43 ± 1.0 mg N/L) and alluvial groundwaters (0.21 ± 0.51 mg N/L) over the same period. Nitrate inputs have potential to suppress other less-energetically favorable terminal electron acceptor processes. If the suppressed processes are the dominant metabolisms associated with Hg methylation, then nitrate addition may inhibit microbial Hg methylation. Nitrate addition to a sulfate-rich stratified freshwater lake resulted in lowered MeHg concentrations in hypolimnetic waters, likely related to both shifts in terminal electron acceptor processes and the formation/stability of iron-oxide layers at the sediment-water interface that adsorb and store MeHg (Todorova et al., 2009; Matthews et al., 2013). In this system, we see no buildup of nitrate in either alluvial groundwaters or wetland surface waters at or near flood conditions. The supply of labile organic carbon in this system is likely sufficient to allow for rapid denitrification and less-energetically favorable microbial metabolisms associated with Hg methylation (e.g., iron and sulfate reduction and methanogenesis).

Clearly, flood events result in significant increases in Hg and MeHg available for transport within the Cedar River floodplain. We posit that river floodplains are important “control points” (Bernhardt et al., 2017) for downstream delivery of Hg and MeHg. Hydrologic connectivity between floodplain and river likely plays an important role in the efficiency of both stimulation of Hg methylation and subsequent Hg and MeHg transport. Flood frequency has increased in the central United States over the last 50-plus years (Mallakpour and Villarini, 2015). Streamflow estimates derived from coupled downscaled precipitation projects and a hillslope link model for the Cedar River basin project a 50% increase in the Cedar River discharge during a flood with a 100-year recurrence interval by the end of the 21st century (Quintero et al., 2018). While increased streamflow will almost certainly lead to increased fluxes of Hg and MeHg (see section 3.1), our results suggest this effect may be amplified, especially for MeHg, by increased MeHg production within riparian zones during floods. Climate change thus has potential to partially offset reductions in export of Hg and MeHg from terrestrial ecosystems achieved by emissions reductions via implementation of the Minamata Convention. Long-term monitoring of Hg and MeHg in fluvial systems could serve an important role in measuring the effectiveness of the Minamata Convention and climate-change driven changes in Hg biogeochemistry and transport.