Emerging investigator series: the effect of wildfire on streamwater mercury and organic carbon in a forested watershed in the southeastern United States

Allison M. Jensen , Todd M. Scanlon and Ami L. Riscassi *
Department of Environmental Sciences, University of Virginia, 291 McCormick Road, Charlottesville, VA 22904, USA. E-mail: alr8m@virginia.edu

Received 9th September 2017 , Accepted 31st October 2017

First published on 3rd November 2017


Wildfires alter forested ecosystems, which include large stores of mercury (Hg) and organic carbon, two compounds that are closely linked in vegetation, soils, and streamwater. Studies have shown that wildfires release elevated levels of mercury to the atmosphere which can be locally redeposited and leave charred organic material (vegetation and litter) on the soil surface. Both can contribute to the elevated mobilization of Hg into lakes and streams. However, no studies have conducted a detailed examination of hydrological transport of Hg following a wildfire. This study investigates the coupled transport of mercury and carbon at Twomile Run, a headwater stream located in the forested mountains of Shenandoah National Park, in the year following a low-severity wildfire. Weekly baseflow samples and bi-hourly high-flow storm samples were analyzed for dissolved and particulate mercury (HgD and HgP, respectively), dissolved organic carbon (DOC), UV absorbance at 254 nm (UV254, surrogate for DOC quantity and character), and total suspended solids (TSS), and were compared with identical measurements taken from a nearby unburned watershed. For all flow conditions sampled at the burned site (which did not include the 2 months following the fire), streamwater HgD and DOC concentrations, and corresponding UV254, were similar to the unburned system. TSS concentrations varied between sites but overall differences were relatively small in magnitude and likely attributable to site differences rather than fire effects. Notably, the HgP per unit of TSS at the burned site was an order of magnitude higher than the unburned site (2.66 and 0.13 ng HgP per mg TSS, respectively) for 8 months following the fire, resulting in elevated HgP concentrations for the range of flow conditions, after which there was a rapid return to non-disturbed conditions. Streamwater total Hg fluxes roughly doubled (0.55 to 1.04 μg m−2 yr−1) as a consequence of the fire, indicating that in addition to changing atmospheric and terrestrial Hg cycling, fires can rapidly and significantly alter the streamwater Hg which has implication for downstream ecosystems. These findings are particularly relevant as the occurrence and severity of wildfires are expected to increase in the mid-latitudes in response to climate change.



Environmental significance

A low-severity fire in a southeastern U.S. forested watershed increased the streamwater particulate mercury per unit total suspended solids (TSS) an order of magnitude over non-fire impacted conditions, but had limited or no impacts on dissolved Hg, dissolved organic carbon and TSS concentrations over a wide range of discharge conditions. While Hg concentrations rapidly returned to non-disturbed conditions after 8 months, these results indicate that Hg released from terrestrial stores during forest fire can contribute to a brief increase in streamwater Hg flux, which has implications for downstream systems.

1. Introduction

Terrestrial landscapes, specifically forested uplands, are potential long-term repositories of mercury (Hg) due to their ability to store atmospherically deposited Hg,1,2 which has been enhanced from decades of fossil fuel combustion.3,4 Removal of Hg from watershed storage is typically gradual, via the processes of soil emissions5 and streamwater transport6. Soils account for the majority (average 91%) of Hg sequestered in U.S. forested ecosystems,7 with the highest Hg concentrations observed in surface litter and the upper organic layer of soils.8 Therefore, any surface disturbance to these landscapes has the potential to significantly alter atmospheric, terrestrial, aquatic, and ultimately biological Hg cycling, both on a global and local scale. Wildfire is an increasingly common landscape disturbance in upland forests that has the ability to immediately release large amounts of stored Hg to the atmosphere via biomass combustion and heat transfer to the subsurface.9–11 While atmospheric Hg fluxes from fires contribute significantly to the atmospheric Hg load and are considered in the global emissions inventory,12 the local impacts of wildfire on Hg cycling, specifically the potential immediate change in the concentration and flux of Hg transported downstream, has yet to be the focus of study. Alterations to the aqueous transport of Hg is of direct concern to ecological and human health, as increased amounts of Hg in an aquatic system can be delivered to downstream locations where it may be converted to its organic form, methylmercury (MeHg), by microorganisms. MeHg is a potent neurotoxin that bioaccumulates and biomagnifies in aquatic ecosystems, causing higher-order predators like fish to accumulate Hg in their body tissues.13 In this study, we examine the impact of wildfire on local stream inorganic Hg transport in a forested upland watershed located within Shenandoah National Park (SHEN), Virginia.

Mercury and carbon cycling in undisturbed forested systems have been well studied and the processes governing aqueous transport are well characterized. In forested watersheds, Hg is retained in association with organic carbon in the upper soil profile and in vegetation. In particular, the hydrophobic organic acid fraction (HPOA) of organic matter has been shown to account for the majority of Hg binding due to its higher reduced sulfur content14 and the presence of aromatic compounds.15 Mercury is transported to the aqueous environment in association with organic carbon primarily during high-flow periods when the upper soil profile becomes hydrologically connected to the stream. Both total mercury (HgT) and dissolved mercury (HgD) concentrations have been shown to increase with discharge and are consistently positively correlated with dissolved organic carbon (DOC)14,16–27 as well as UV absorbance at 254 nanometers (UV254 is frequently used as a surrogate measure of the HPOA of DOC).19,27,28 Likewise, particulate mercury (HgP) and particulate organic carbon (POC, or an alternative metric volatile solids, VS) or total suspended solids (TSS) concentrations sharply increase during high flow events and have a strong positive relationship over a range of discharge conditions.18,19,23,29–32 As a result of increased concentrations of both streamwater HgP and HgD during brief high-flow events, the majority of the annual stream Hg flux has been found to occur during these short-lived periods in a variety of systems.18,19,21,23,30 This includes three watersheds within SHEN, in which approximately 80% of the annual Hg flux was found to occur within <1% of the time.6 The amount of total HgT (measured as unfiltered Hg or calculated as HgD + HgP) exported annually represents between 5–30% of estimated atmospheric inputs,6,19,21,22,24,30,33 implying that in the absence of disturbance, the majority of deposited Hg may be retained within the terrestrial environment in these systems; non-fire volatilization also likely accounts for some portion of the difference.

The large flux of Hg volatilized from soils and vegetation during fires in the U.S. has been shown to be a significant source of Hg to the atmosphere, equivalent to 13–42% of the annual U.S. anthropogenic emissions.9 During a fire, up to 99% of the Hg held in vegetation and soil pools is volatilized to the atmosphere,34–37 although this amount depends on the severity of the wildfire.9,38 The majority of this Hg is emitted as gaseous elemental Hg in low-moisture fuels with smaller percentages (<1–14%) in particulate form;34 substantial portions of particulate emissions (up to 50%) are possible in fresh fuels.39 Particulate Hg can be locally redeposited to soils and streams or it can adhere to the forest canopy, making it susceptible to local redeposition via throughfall.40 While Hg is initially volatilized from the soil pool, burned soil, which is altered in chemical properties, has been shown to accumulate elevated levels of Hg (from the atmosphere) during the months following a fire.37 In addition, the small amount of Hg that is not volatilized during a fire is concentrated in charred organic material left on the ground surface. These fire-altered sources of Hg and carbon have the potential to mobilize into the surrounding environment41 and ultimately be transported downstream where they may accumulate in bottom sediments.42

Total Hg levels in lakes have been shown to increase when storm events follow a wildfire43,44 implying stream transport of Hg from the burned landscape. The Southern Rockies Watershed project, initiated in 2003 to quantify the impact of wildfires on water resources in the Rocky Mountain headwaters, noted that “high concentrations of mercury were evident” in streamwater when sampling began, the first year after a large (210 km2) severe wildfire, although no details were provided on the fraction (dissolved, particulate or total) of Hg that increased or the flow conditions during which high concentrations were measured.45 In another study, trace elements, including Hg, were monitored exclusively during high flow periods in several different streams in the three months after a large (680 km2), mostly moderate-severity fire in the sage scrub and chaparral forest outside of Los Angeles, California. Storm sampling indicated elevated concentrations in both filtered and unfiltered Hg within the burn area in comparison to those outside the burn area.46 The long-term impact of wildfire on streamwater Hg was assessed for a range of flow conditions in a forested watershed in Acadia National Park. Nelson et al.22 determined the annual streamwater Hg flux was significantly lower in a watershed that had been burned 50 years prior compared to an unburned watershed. Potential reasons for reduced Hg export included reduced soil Hg pools, greater retention of Hg in the soil and lower concentrations of Hg in post-fire throughfall. Overall, knowledge of stream Hg transport in the year following a fire in forested mountain watersheds is rather limited and a detailed understanding of changes to individual fractions (particulate and dissolved) in the context of hydrological variability is absent. A comprehensive report on the effect of wildfire on water quality in Eastern North America notes that “while mercury is a contaminant of major concern, few [studies] have addressed aquatic mercury dynamics following fire”.47

Because Hg is strongly associated with organic carbon, it is instructive to consider how fire impacts organic carbon in forested watersheds. Up to 98% of the organic carbon in vegetation and up to 95% of the organic carbon in upper soils is emitted to the atmosphere during a fire,34,35,48 with the exact percentage depending on the fire severity level.38 Organic carbon that remains associated with burned material on the ground surface can mobilize from the soil into streams, especially during high-flow events.49 Precipitation following a wildfire has been found to increase stream concentrations of total organic carbon (TOC),42 DOC49–51 and TSS.50 However, post-fire mobilization of carbon to streams is not stable over time, and greater variability in TOC has been observed after a fire.51 A fire's direct impact on the organic fraction of TSS remains unclear, although one study that sampled soil found increased concentrations of organic carbon in the top layer of soil after prescribed burning,52 which suggests that particulate organic carbon could increase in the stream.

Since Hg and organic carbon are linked in the terrestrial and aqueous environments and both elements are impacted by fire, we seek to quantify post-fire concentrations of stream Hg and organic carbon, both dissolved and particulate components, over a range of flow conditions and calculate resultant streamwater fluxes. In spring 2016, the Rocky Mount Fire burned for two weeks in SHEN and grew into the second-largest fire in the Park's 80 years history. One of the watersheds burned by the fire, Twomile Run, is located near a watershed where Hg and organic carbon dynamics were evaluated from 2008 to 2010.6,27 This provides a unique opportunity to sample streamwater in a recently burned watershed (Twomile Run) and compare post-fire Hg and organic carbon dynamics to prior findings from a nearby undisturbed headwater system (Paine Run). Our first hypothesis is that streamwater dissolved and particulate Hg and carbon concentrations will be similar between the burned and unburned watersheds during baseflow conditions, but higher concentrations will be measured in the burned watershed during high-flow conditions due to the increased potential of burned materials (which include both carbon and Hg) to leach into the subsurface and be carried to the stream via overland and subsurface flow. Our second hypothesis is that the ratio of HgD[thin space (1/6-em)]:[thin space (1/6-em)]DOC and HgP[thin space (1/6-em)]:[thin space (1/6-em)]TSS will increase during high-flow periods at the burned watershed compared to the unburned watershed reflecting leaching from burned organic matter and HgP re-deposition. Addressing these hypotheses will help determine the impact that wildfires have on Hg and organic carbon co-transport in forested watersheds, a question that has remained largely unaddressed in post-fire studies.

2. Site description

The Rocky Mount Fire in SHEN began on April 16, 2016 and was declared out on May 3, 2016. The wildfire burned over 40 km2 of mountain laurel, pine, and oak forest including the Twomile Run watershed, which has been visited quarterly by the University of Virginia for water quality monitoring since 1987. According to a fire severity assessment determined by remote sensing Landsat Thematic Mapper (TM) satellite imagery (L. Gibbons, NPS Northeast Region, lead fire effects monitor, written communication, April 2016), the majority of the watershed was burned at low/low-moderate severity (37.7%) followed by unburned-low severity (30.7%), with smaller areas burned at moderate (23%) and high (8.5%) severity.

Twomile Run, the burned watershed, and Paine Run, the unburned control watershed, are both located in Shenandoah National Park within the Blue Ridge physiographic province (Fig. 1). Lower elevation areas of the region experience modified continental climate, with mild winters and warm, humid summers. The mean annual temperature in the lowland area at Luray, Virginia is 12 °C.53 Higher elevations in this region experience winters that are moderately cold and summers that are relatively cool. The mean annual temperature at the National Atmospheric Deposition Program (http://nadp.sws.uiuc.edu) monitoring station at Big Meadows (VA28), located on the mountain ridge in central SHEN, is 9 °C.53 Precipitation is fairly evenly distributed throughout the year, with brief intense thunderstorms occurring in the summer. Snow and ice are common in winter, but melt usually occurs quickly.


image file: c7em00419b-f1.tif
Fig. 1 Map showing the area burned by the Rocky Mount Fire and the burn severity levels, the boundary of the study watershed, Twomile Run, and the boundary of the control watershed, Paine Run.

Twomile Run watershed and Paine Run watershed are similar in size (5.6 and 12.4 km2, respectively), stream gradient, forest composition (primarily chestnut oak), and bedrock class (siliciclastic), resulting in similar stream chemical properties (Table 1). In addition to quarterly streamwater samples (1988–present) and hourly discharge (1992–present), an in-depth study on dissolved and particulate Hg and carbon and associated water quality parameters was conducted at Paine Run from 2008 to 2010.6,27

Table 1 Summary of watershed properties and basic water chemistry for the control watershed (Paine Run) and the burned watershed (Twomile Run), based on quarterly measurements from 1988–2013
Stream Area (km2) Primary lithology Primary vegetation ANC μeq. L−1 pH SO42− μeq. L−1 NO3 μeq. L−1
Paine Run 12.4 Siliciclastic Chestnut oak 13.2 5.8 110.9 4.32
Twomile Run 5.6 Siliciclastic Chestnut oak 22.2 5.9 94.9 3.04


3. Methods

3.1 Field methods

Streamwater samples were collected at Twomile Run from June 2016 through May 2017. A weekly sample was collected for HgD and HgP analysis following U.S. EPA method 1669 ‘clean’ techniques for trace metals, which requires a two-person sampling team.54 If only one person was available for sampling, a duplicate sample was taken. Materials, cleaning procedures, and collection and handling of Hg samples are described in detail in Riscassi et al.55 Weekly grab samples were also taken in three separate bottles for (1) DOC and UV254, (2) pH, anions, and cations, and (3) TSS/VS analysis. During precipitation events, a stage-actuated automated sampler (Teledyne ISCO® model 3700) retrofitted for Hg sampling55 was used to collect either one or two high-flow samples at bi-hourly intervals. One sample was used for HgD, HgP, DOC, UV254, and pH analysis, while the second sample, if taken, was used for TSS/VS analysis. Four pairs of concurrent grab and automated samples were collected during one storm event to verify the accuracy of the ISCO sampling method under high-flow conditions at Twomile Run. Grab and automated samples were transported from the field to the UVA laboratory in a cooler and refrigerated until subsequent filtering, preservation, and analysis.

Stream stage was measured continuously by a pressure transducer enclosed in a stilling well since August 11, 2016 and verified by staff readings during weekly site visits. Routine manual discharge measurements were made with a SonTek FlowTracker Acoustic Doppler Velocimeter. Channel geomorphology and slope were measured with a CST/Berger Electronic Digital Theodolite Transit to estimate high-flow discharge using the Manning's equation. Both measured discharge and estimated high-flow discharge were used to establish a rating curve for the calculation of continuous discharge over the study period.

The data collected from 2008–2010 at Paine Run, including HgD, HgP, DOC, UV254, and pH, are used for comparison with the data collected at Twomile Run. Sample collection at Paine Run used the same field procedures as those used at Twomile Run.6,27 To confirm that Hg and carbon relationships at Paine Run have not changed since the earlier study, we collected 5 baseflow samples and one high flow sample at Paine Run concurrent with the study at Twomile Run. All samples were analyzed for dissolved Hg and carbon parameters, with the high flow sample also analyzed for HgP and TSS/VS. Although TSS was not measured at Paine Run during the earlier study, turbidity was measured using a sonde (YSI EXO2, with 6136 turbidity sensor) from November 2009–October 2010 concurrent with 39 stream samples. To convert turbidity measurements to TSS concentrations for each sample, we adopted a turbidity–TSS relationship developed in a nearby forested headwater stream, Staunton River.56 Turbidity and TSS were also measured concurrently (n = 21) for a one-month period at Twomile Run and the relationship was statistically similar to the one measured at Staunton River, giving confidence to the transferability of these relationships to similar watersheds in the region.

3.2 Laboratory methods

Grab samples were refrigerated and typically filtered within one week of sample collection. Automated Hg samples were retrieved within one week of the storm event; an evaluation of effects from delayed retrieval of automated samples showed minimal differences in Hg concentrations.55 Samples were filtered through a pre-baked quartz fiber filter (QFF) with a particle retention size of 2.2 μm in a vacuum desiccator chamber under a Class-100 Clean Bench with an acid-leached Teflon filter apparatus. The filter, filter apparatus, and filtering procedures were identical to those used by the USGS Mercury Research Laboratory.57 The filtered sample water was preserved with a 100% bromium chloride (BrCl) solution (5 mL BrCl per 1 L sample) within an hour of filtration and stored at room temperature until analysis. The filter with captured filtrate was placed in a Teflon bottle and 100 mL of a 5% BrCl solution was added, submerging the filter. The bottle was tightly capped, double-bagged, and placed in a 50 °C oven for 5 days to leach all Hg on the filter into solution. The solution was stored at room temperature until analysis. HgD and HgP were analyzed using a Tekran cold-vapor atomic florescence spectrophotometer according to U.S. EPA method 1631, Revision E.58 The method detection limit (MDL) for HgD was determined to be 0.09 ng L−1. The MDL for HgP was determined based on the volume of sample filtered. The MDL for HgP ranged from 0.12 ng L−1 for grab sample bottles (∼1 L) to 0.90 ng L−1 for automated sample bottles (∼200 mL).

DOC samples were filtered through a Whatman® Nuclepore™ hydrophilic membrane filter with a 0.4 μm pore size, preserved with H3PO4, and refrigerated until laboratory analysis. DOC was measured by UV/persulfate digestion according to U.S. EPA method 415.3 (ref. 59) using a Teledyne Tekmar Phoenix 8000 TOC analyzer. The MDL was determined to be 0.12 mg L−1. pH was measured at room temperature using a Metrohm 809 titration system equipped with a Metrohm Aquatrode Plus pH electrode.60 UV absorbance measurements at 254 nm were made with a Shimadzu UV-visible spectrophotometer with a stated accuracy of 0.003 absorbance units (cm−1). Sample water was analyzed within 24 hours of filtering in a quartz cell with a 1 cm path length.

Specific ultraviolet absorbance (SUVA254) was calculated by dividing UV absorbance at 254 nm (cm−1) by the concentration of DOC (mg L−1) and is reported as liter per mg DOC per centimeter (L mg DOC−1 cm−1). SUVA254 is an “average” molar absorptivity and has been used as an indicator of DOC aromaticity.61,62

TSS samples were filtered through a pre-baked QFF with a particle retention size of 2.2 μm, dried, and weighed to obtain TSS according to U.S. EPA method 160.2.63 The filtrate was then heated at 500 °C for 2 hours to burn off any organic matter and weighed again to obtain the non-volatile solids (inorganic or mineral solids) according to U.S. EPA method 160.4.64 The difference between TSS and non-volatile solids is the VS, which is often used as a surrogate for particulate organics. The MDL for TSS and VS was determined based on the volume of sample water filtered. The MDL for TSS ranged from 0.3 mg L−1 for grab sample bottles (∼2 L) to 6.9 mg L−1 for automated sample bottles (∼250 mL). The MDL for VS ranged from 0.14 mg L−1 for grab sample bottles to 3.40 mg L−1 for automated sample bottles.

3.3 Quality assurance and quality control

Quality assurance (QA) for Hg sample collection was maintained in the field through duplicates and blanks. The mean absolute difference for HgD field duplicates (n = 52) was 0.05 (+/− 0.05) ng L−1 and the mean absolute difference for HgP field duplicates (n = 46) was 1.65 (+/− 4.36) ng L−1. The higher absolute differences in HgP compared to HgD in duplicate pairs is due to higher concentrations and the larger variability in samples taken during high-flow periods. Under these conditions, suspended sediment levels are rapidly changing and the amount of sediment, and associated HgP, captured in the sample bottle is likely to vary due to non-uniform mixing. Four of the field duplicate pairs were comprised of grab and automated samples, confirming that HgD and HgP measurements obtained with either method are comparable. Field blanks (n = 2), which simulated grab sample procedures with deionized (DI) water, had HgD concentrations of 0.03 and 0.21 ng L−1 and HgP concentrations of 0.24 and 1.29 ng L−1. The relatively high HgP concentration for one of the two field blanks (1.29 ng L−1) was likely attributable to contamination introduced during the filtration process or the filter itself. Laboratory filter blanks, which evaluate the filtration process, were close to detection overall, but 9% (5 of 53 blanks) had concentrations over 1 ng L−1.

Overall QA for Hg laboratory analysis was determined by participation in an inter-laboratory proficiency test administered by Environment Canada in June 2017. The proficiency sample concentrations ranged from 1.5–97.2 ng L−1 (n = 5), and the mean recovery for our lab was 96.2% (+/−7.9%). Between-run laboratory accuracy was verified by ongoing analysis of a certified reference material, and recoveries were within +/−5%. For each batch of Hg samples (∼10), matrix spikes, laboratory duplicates, ongoing precision and recovery (OPR) analysis, and blanks (system, method, and filter) were used to evaluate instrument performance. The accuracy and precision of the matrix spikes (n = 75) and laboratory duplicates (n = 70) were maintained within +/−7%. The accuracy and precision of the standards (n = 70) and OPR samples (n = 79) were maintained within +/−5%. For the system and method blanks (n = 72), 96% and 99% were below MDL, respectively. All filter blanks for HgD (n = 49) were below the MDL. The filter blanks for HgP (n = 53) had a mean of 0.33 (+/− 0.56) ng L−1. Reported quality control metrics are all within U.S. EPA guidelines.58

QA for DOC sample collection in the field and analysis in the laboratory was maintained through field duplicates and filter blanks. The mean absolute difference for field duplicates (n = 10) was 0.085 (+/− 0.11) mg L−1. The filter blanks (n = 5) were all below MDL.

For UV absorbance, an outside check standard (In-Spec UV Standard #3) was analyzed with each batch of samples to verify instrument accuracy. QA was maintained through field duplicates, field blanks, laboratory duplicates, and filter blanks. The mean absolute difference for field duplicates (n = 50) was 0.003 (+/− 0.005) cm−1. All field blanks (n = 2) had an absorbance below 0.012 cm−1. The mean absolute difference for laboratory duplicates (n = 71) was 0.008 (+/− 0.01) cm−1. All filter blanks (n = 53) had an absorbance below 0.01 cm−1.

QA for TSS sample collection and transport was maintained through duplicates. The mean absolute differences for TSS and VS field duplicates (n = 6) were 1.34 (+/− 1.48) mg L−1 and 0.42 (+/− 0.46) mg L−1, respectively. In the laboratory, QA was maintained through filter blanks that were analyzed with each batch of samples. The filter blanks for TSS and VS (n = 19) were all below MDL.

3.4 Hg flux calculations

Streamwater HgD and HgP fluxes at Twomile Run were calculated for the one-year period subsequent to the fire (May 2016 through April 2017) using the “composite” method65 within the R loadflex package.66 The “composite” method combines the regression model method and the period-weighted approach and has been used to calculate Hg fluxes in small watersheds where samples were collected at both a fixed interval and at high frequency during storm events.19 Calibration data for the regression model included all measured concentration data (June 27, 2016–May 5, 2017) and associated discharge. For HgD, the model was run for the complete one-year study period using all available data. For HgP, the model and associated calibration data were run for two separate periods (prior to and after January 25, 2017) based on HgP analysis indicating two significantly different concentration dynamics, as discussed in the following sections. Hourly discharge for the three months post-fire for which data were not yet available at Twomile Run (May–August 11, 2016) were estimated from Paine Run discharge using a linear relationship of log transformed data (r2 = 0.74) determined from the 9 months of concurrent hourly discharge at the two sites (August 2016–May 2017).

Average annual atmospheric Hg deposition to the watershed was estimated to place the fluvial Hg export in the context of the overall watershed Hg budget. Atmospheric Hg deposition was calculated as the sum of measured wet and estimated dry deposition. Wet Hg deposition was estimated as the mean of all available annual data (2003–2015) reported at the Big Meadows Mercury Deposition Network Station (VA28) located approximately 50 km northeast of the Twomile Run watershed. Dry deposition was estimated as measured litterfall Hg Deposition (2008–2009) at VA28.67 Litterfall Hg deposition has been shown to be very similar to the modeled annual dry Hg deposition at NADP Atmospheric Mercury Network sites.67

3.5 Statistical analysis

All descriptive statistical calculations and analysis were performed using MATLAB software (version 9.1.0.441655, The MathWorks Inc., Natick, MA). MATLAB was used to compute linear regressions, calculate associated r2 values, and determine whether the relationships between the two variables were statistically significant (p < 0.05) using the analysis of variance (ANOVA). Regressions were performed on raw data and log-transformed data. For HgD, HgP, DOC, and TSS regressed against discharge, a logarithmic transformation improved the relationship for one or both sites. For these regressions, the logarithmic transformation is reported in the results. For all other regressions, no transformations were necessary. MATLAB was also used to determine statistically significant (p < 0.05) differences in the slopes and intercepts of regression lines.

4. Results

4.1 Hydrology and basic water chemistry

Sampling at Twomile Run did not begin until almost two months after the end of the fire and during the intervening time approximately 290 mm of cumulative rainfall was received over multiple high-intensity events (Fig. 3). All results, therefore, reflect conditions in the year following a fire, excluding the initial rainfall-runoff events, referred to as ‘first flush’. The range of flow conditions observed and sampled during the Twomile Run sampling period (June 27, 2016–May 5, 2017) was representative of the moderate and high flows estimated for the 24 year period of record at the site (as estimated from the long-term Paine Run discharge record) (Fig. 2). However, due to a particularly wet spring/summer in 2016, historically low flows were not observed and therefore not sampled. In comparison to discharge conditions sampled at the control site, Paine Run (2008–2010 study and additional 2016–2017 samples), the high and moderate discharges were similarly sampled, while the low flows were not represented during Twomile sampling. Discharge conditions sampled at Twomile Run ranged from 0.0013 to 8.82 mm h−1 with a mean of 0.31 mm h−1. In comparison, discharge sampled at Paine Run ranged from 0.0001 to 5.32 mm h−1 with a mean of 0.35 mm h−1.
image file: c7em00419b-f2.tif
Fig. 2 Flow exceedence curves for a 24 year period of record (1992–2016) at Twomile Run and Paine Run. Symbols on the line represent discharge when samples were collected for Hg analysis.

During the Twomile Run study, a total of 45 baseflow samples and 68 high-flow samples were collected for analysis of HgD, HgP, DOC, UV254 and pH (Fig. 3). Anions (SO42−, NO3) and base cations (Na+, K+, Ca2, Mg+) were only evaluated for baseflow samples due to sample volume restrictions for automated samplers. TSS and VS were measured on a subset of 40 baseflow and 50 high-flow samples, respectively, which represented the full range of discharge conditions sampled at the site. High-flow samples collected during five separate rainfall-runoff events captured the rising, peak, and falling hydrograph limbs, while high-flow samples collected during three other events captured various parts of the hydrograph (Fig. 3). Weekly baseflow concentrations of SO42−, NO3 and the sum of base cations (Na+ + K+ + Ca2+ + Mg+) were not significantly different from concentrations measured from weekly samples taken during the concurrent period at Paine Run as part of an ongoing long-term monitoring program.


image file: c7em00419b-f3.tif
Fig. 3 The discharge hydrograph at Twomile Run for one year following the fire (denoted by red lines). Discharge at Twomile Run from mid-April 2016 to mid-August 2016 was estimated using a relationship with Paine Run, 95% confidence intervals are shown as a grey shaded area. Grab and automated sample collection is denoted with solid and open circles, respectively.

4.2 Discharge relationships with stream chemistry

The HgD concentrations measured at Twomile Run ranged from 0.36 ng L−1 to 3.27 ng L−1 and log-transformed HgD was significantly, positively related to log-transformed discharge (p < 0.05, r2 = 0.29) (Fig. 4a). At Paine Run, a similar range in HgD was observed, from 0.21 ng L−1 to 3.28 ng L−1, and there was a significant positive relationship between log-transformed HgD and log-transformed discharge (p < 0.05, r2 = 0.45) (Fig. 4a). The slopes of the linear best fit lines for log-transformed HgD and discharge were not significantly different between the two sites (p = 0.06), but the intercepts were significantly different (p = 0.01), with Twomile Run being higher. For all discharge conditions, this results in higher concentrations of HgD at Twomile Run; at the mean discharge of 0.31 mm h−1, this represents about a 47% increase in HgD concentration at Twomile Run as compared to Paine Run (1.38 and 0.94 ng L−1, respectively).
image file: c7em00419b-f4.tif
Fig. 4 Dissolved Hg (a) and particulate Hg (b) versus specific discharge, and corresponding regression lines, at Twomile Run and Paine Run. The two sites have similar ranges in discharge and HgD, however, higher concentrations of HgP were measured at Twomile Run.

HgP concentrations measured at Twomile Run ranged from 0.08 ng L−1 to 56.7 ng L−1 and log-transformed HgP was significantly, positively related to log-transformed discharge (p < 0.05, r2 = 0.23) (Fig. 4b). At Paine Run, lower concentrations of HgP were measured, ranging from 0.03 ng L−1 to 21.3 ng L−1. Similar to Twomile Run, log-transformed HgP concentrations had a significant positive relationship with log-transformed discharge at Paine Run (p < 0.05, r2 = 0.41) (Fig. 4b). The slopes of the linear best fit lines for log-transformed HgP and discharge were not significantly different between the sites (p = 0.97), but the intercepts were significantly different (p = 0.001), with Twomile being higher (Fig. 4b). For all discharge conditions, this results in a higher concentration of HgP at Twomile; at the mean discharge of 0.31 mm h−1, this represents a 7-fold increase in HgP concentration (4.41 and 0.65 ng L−1 at Twomile Run and Paine Run, respectively).

DOC concentrations measured at Twomile Run ranged from 0.71 mg L−1 to 5.53 mg L−1 and log-transformed DOC was significantly, positively related to log-transformed discharge (p < 0.05, r2 = 0.20). At Paine Run similar concentrations of DOC were measured, ranging from 0.53 mg L−1 to 5.38 mg L−1, and log-transformed DOC had a significant positive relationship with log-transformed discharge (p < 0.05, r2 = 0.47). The slopes and intercepts of the linear best fit lines were significantly different between the sites (p < 0.05), although both p-values were close to the significance level (Fig. S1). For all discharge conditions, this results in a higher DOC concentration at Twomile Run; at the mean discharge of 0.31 mm h−1, this represents about a 38% increase in DOC concentration (2.59 and 1.88 mg L−1 at Twomile Run and Paine Run, respectively).

TSS measured at Twomile Run ranged from 0.15 mg L−1 to 202.66 mg L−1 and log-transformed TSS was significantly, positively related to log-transformed discharge (p < 0.05, r2 = 0.59). The TSS estimated for Paine Run ranged from 1.67 mg L−1 to 161.1 mg L−1, and log-transformed TSS had a significant positive relationship with log-transformed discharge (p < 0.05, r2 = 0.32). The slopes of the linear best fit relationships for TSS and discharge were significantly different between the sites (p < 0.05), but the intercepts were not (p = 0.09). This results in lower TSS concentrations at Twomile Run during lower discharge, and higher concentrations at higher discharge (Fig. S2); at the mean discharge of 0.31 mm h−1, this represents a 6% decrease in TSS concentration (6.60 and 7.01 mg L−1 at Twomile Run and Paine Run, respectively).

4.3 HgD and dissolved constituents

At Twomile Run and Paine Run, there were significant positive correlations between HgD and DOC (both sites p < 0.05, r2 = 0.95 and 0.89 at Twomile Run and Paine Run, respectively) (Fig. 5). The slopes and intercepts of the linear best fit lines were not significantly different between the sites (p = 0.21 for the slopes and p = 0.32 for the intercepts). The relationships between HgD and UV254 were also statistically similar between Twomile Run and Paine Run (p = 0.17 for the slopes and p = 0.34 for the intercepts, data not shown). For both sites, the relationship between HgD and UV254, which represents both the quantity and composition of DOC, explained less of the variability in HgD than the relationship with DOC quantity alone (both sites p < 0.05, r2 = 0.92 and r2 = 0.79 at Twomile Run and Paine Run, respectively, data not shown).
image file: c7em00419b-f5.tif
Fig. 5 Dissolved Hg versus dissolved organic carbon (DOC) at Twomile Run and Paine Run. The slopes and intercepts of the linear best fit lines (thin line for Twomile Run, thick line for Paine Run) were not significantly different between the sites.

SUVA254 ranged from 0.019 to 0.053 L mg DOC−1 cm−1 at Twomile Run with a mean of 0.031 L mg DOC−1 cm−1. At Paine Run, SUVA254 ranged from 0.000 to 0.053 L mg DOC−1 cm−1 with a mean of 0.030 L mg DOC−1 cm−1. The full range in SUVA254 was present in both base and storm flow conditions at both sites. At Twomile Run, pH ranged from 3.51 to 6.29 (with an outlier at 8.08) and a mean of 5.95. At Paine Run, pH ranged from 3.73 to 6.99 with a mean of 5.65. The large range in pH was primarily driven by low values (below 5.5) only observed during storm flow conditions at both sites; baseflow pH values were higher and less variable at both sites.

4.4 HgP and particulate parameters

At Twomile Run, HgP was significantly, positively related to TSS, although TSS explained only a small amount of the variability in HgP (p < 0.05, r2 = 0.08) as compared to Paine Run (r2 = 0.93). To examine if changes in the HgP to TSS relationship over time contributed to its weak explanatory power, HgP and TSS were divided temporally into four periods. The criteria used to define the periods were that the periods must (1) include a storm event with at least three samples and (2) include a TSS measurement greater than 10 mg L−1. These requirements ensured that each period had samples covering a range of discharge conditions and TSS concentrations. The best fit lines for the HgP–TSS relationship during the first and second periods were statistically similar and the best fit lines for the third and fourth periods were statistically similar. As a result, Twomile Run HgP–TSS relationships were evaluated in two distinct periods, the earlier period, hereafter referred to as “Flow Period 1,” that included samples from June 2016 to January 25, 2017 and the later period, hereafter referred to as “Flow Period 2,” that included samples after January 25, 2017 through May 2017.

There were significant positive correlations between HgP and TSS for both flow periods at Twomile Run and for the full record at Paine Run (p < 0.05 for all data sets, r2 = 0.74 and 0.98 for Twomile Run Flow Period 1 and Flow Period 2, respectively, and r2 = 0.93 for Paine Run) (Fig. 6). Intercepts of the linear best fit lines were not significantly different between Twomile Run Flow Periods 1 and 2 (p = 0.37), but the slopes were significantly different (p < 0.05) at 2.66 and 0.11 ng HgP per mg TSS for Flow Period 1 and 2, respectively. While the slope for Paine Run, 0.13 ng HgP per mg TSS, is similar to the slope for Twomile Run Flow Period 2, the slopes and intercepts of the linear best fit lines were significantly different (p < 0.05) from each other as well as from Twomile Run Flow Period 1 (Fig. 6).


image file: c7em00419b-f6.tif
Fig. 6 Particulate Hg (HgP) versus total suspended solids (TSS) for Flow Period 1 and 2 at Twomile Run and the entire study at Paine Run. The slopes of the linear best-fit lines were significantly different between Flow Period 1 and 2 at Twomile Run, with Flow Period 1 having a slope of 2.66 ng HgP per mg TSS and Flow Period 2 having a slope of 0.11 ng HgP per mg TSS. The slope of the linear best fit line for Paine Run was 0.13 ng Hg per mg TSS.

HgP also had a significant positive relationship with VS for both flow periods at Twomile Run (both periods p < 0.05) (data not shown). For Flow Period 1, VS explained less of the variability in HgP concentrations compared to TSS (r2 = 0.27 and 0.74 for VS and TSS, respectively). For Flow Period 2, VS explained a similar amount of the variability in HgP compared to TSS (r2 = 0.97 and 0.98 for VS and TSS, respectively). Over the study period, VS ranged from 6.25% to 100% of TSS with a mean of 53.69%; the percentage of VS had no relationship with discharge.

4.5 Streamwater Hg fluxes

The streamwater discharge at Twomile Run for the one-year post-fire period, May 2016 through April 2017, was 282 mm per year, corresponding to an annual flow exceedance of 42%, indicating a slightly higher than average flow year. The HgD and HgP flux at Twomile Run was 0.24 and 0.80 μg m−2 yr−1, respectively. To estimate streamwater fluxes in the absence of fire, we assumed the DOC and TSS concentrations would have remained similar to those at Paine Run and calculated the HgD and HgP concentrations using the ‘non disturbed’ relationships developed at Paine Run (all relevant equations for conversions are listed in Table 2). Estimates of HgD and HgP fluxes at Twomile Run in the absence of the fire was 0.24 and 0.31 μg m−2 yr−1, respectively. The total additional flux of Hg released downstream due to the fire at Twomile run was therefore 0.49 μg m−2 yr−1, all in particulate form, resulting in an approximate doubling of HgT export. The fire increased the particulate fraction of the total streamwater Hg flux from 56 to 77%. Streamwater Hg export not attributed to the fire represented 3.2% of mean annual atmospheric deposition (17.2 μg m−2 yr−1) while Hg export attributed to the fire represented an additional 2.9% of annual atmospheric deposition. The increased Hg flux due to the fire was confined to the 8 months post fire period; the fire had no impact on fluxes calculated for the 9–12 months post-fire period.
Table 2 Values for the least squares linear regressions of TSS and HgP, and DOC and HgD
Site Fire? Data period Y X Y = mX + b
Variable Unit Variable Unit m b r 2
a Additional samples taken in 2016 and 2017 for HgD (n = 5) and HgP (n = 2).
Twomile Run Yes June 2016–May 2017 HgD ng L−1 DOC mg L−1 0.53 −0.04 0.95
Paine Run No March 2009–Oct. 2010a HgD ng L−1 DOC mg L−1 0.56 −0.09 0.89
Twomile Run Yes June 2016–Jan. 25, 2017 HgP ng L−1 TSS mg L−1 2.66 −0.52 0.74
Twomile Run Yes Jan 25–May 2017 HgP ng L−1 TSS mg L−1 0.11 0.92 0.98
Paine Run No March 2009–Oct. 2010a HgP ng L−1 TSS mg L−1 0.13 −0.05 0.93


5. Discussion

5.1 Impacts of fire on dissolved Hg and carbon dynamics

Our study found strong, positive relationships between HgD and discharge, as well as HgD and DOC, at both Twomile Run (burned site) and the control site (Paine Run), demonstrating that the fundamental controls on streamwater dissolved Hg did not significantly change (post-first flush) in the year following a low-intensity wildfire at Twomile Run. Despite these overall similarities, both HgD and DOC concentrations did tend to be higher in Twomile Run over the entire range of flow conditions. The resulting HgD[thin space (1/6-em)]:[thin space (1/6-em)]DOC ratio remained similar between sites, suggesting that the slightly higher HgD per unit discharge at the burned site was a direct result of the higher DOC concentrations. This is likely the result of natural spatial variability between watersheds rather than fire disturbance, as previous research has documented similar small-scale variability in DOC–discharge relationships between nearby SHEN watersheds.27 Furthermore, the statistical similarity in SUVA254 at Twomile Run and Paine Run indicates that the DOC composition is similar between sites, whereas obvious differences in SUVA254 would be expected if fire had a significant impact on the DOC transported to the stream.49,68

We hypothesized that during baseflow conditions, concentrations of HgD and the HgD[thin space (1/6-em)]:[thin space (1/6-em)]DOC ratio would be relatively unaffected by the fire, but increases in these quantities would be encountered during high-flow conditions. However, no such increases during high-flow conditions (post-first flush) were observed, which may be attributable to the relatively low fire intensity at this particular setting. Burton et al.46 found a several-fold increase in streamwater HgD subsequent to the 2009 Station Fire in southern California, and the strength of the HgD–DOC relationship indicated each were likely mobilized from the same source, attributed to leaching from ash and burned soil. Distinctions in fire severity (low compared to moderate), vegetation (forest compared to sage/chaparral), as well as climate (humid/subtropical vs. Mediterranean) likely resulted in less ash and char available for leaching and subsequent transport to streamwater in our study watershed. Our results demonstrate that for a low/moderate severity fire in an Eastern U.S. forest, the leaching and transport of DOC and HgD (post-first flush) may be insignificant. This possibility is supported by the study of Burke et al.37 who found minimal Hg leaching from newly-burned soils in a laboratory assessment. Alternatively, immediate leaching of ash and burned soil could have caused a brief enhanced release of HgD in the first two months after the fire that was unmeasured in this study.

While only a few studies have evaluated changes in streamwater Hg the year following fire, a much larger number of studies have documented changes in streamwater DOC relative to unburned systems. The majority of these studies have been conducted in the western U.S. and reported DOC increases of relatively large magnitude (2–5 times higher) during high-flow periods.46,49–51 However, studies conducted in wetlands of the Southeastern U.S.69 and moorlands of Northern Ireland68 found little to no change and a decrease in DOC concentrations, respectively. As with HgD, the low burn severity and site characteristics may have contributed to the lack of increases in DOC in the current study. The delayed onset of sampling should also be taken into consideration, particularly in light of findings from other locations where elevated post-fire DOC concentrations were measured during high-intensity precipitation events. These DOC increases were relatively short lived,49–51 although in no cases as brief as two months.

5.2 Impacts of fire on particulate Hg dynamics

At Twomile Run, HgP and TSS had significant positive relationships with discharge, similar to the relationships observed at Paine Run and documented in other undisturbed forested watersheds.32,70,71 However, at Twomile Run, HgP concentrations per unit of flow were 7 times greater during the first 8 months post fire compared to concentrations at Paine Run while TSS remained consistent between sites. While some studies have shown increased mobilization of sediment to streams during precipitation events that followed a wildfire,46,50 others have found that low-severity prescribed fires do not burn far enough in the soil column to significantly increase suspended sediment in surface waters,72–74 which seems consistent with our findings.

A potential explanation for the increase in HgP is that local Hg redeposition during and after the fire increased the Hg bound to sediment on the ground surface, which was then mobilized to the stream. Rainfall following a fire has been shown to transport elevated concentrations of Hg, presumably from the smoke/ash, from the canopy to the ground surface via throughfall.40 This likely occurred in our system since (1) the majority of the canopy did not burn in the fire and, therefore, could act as a surface for deposition, and (2) substantial rainfall occurred during and immediately after the fire event providing the transport mechanism for Hg that had accumulated on the live foliage. Another potential source of new HgP is the layer of charred material with activated carbon that remains after a fire, which has a strong affinity for Hg and results in the sorption of atmospheric Hg (both dry and wet deposition) to the burned material.36,37,75,76 A combination of local redeposition of HgP and increased efficiency of Hg binding to burned organic matter might explain why HgP concentrations at Twomile Run were significantly greater than HgP concentrations at Paine Run while TSS concentrations remained at similar levels.

Our study found an elevated amount of HgP per unit of TSS for the first 8 months (excluding the first-flush period) after the fire (Twomile Run Flow Period 1), followed by a return to pre-fire HgP levels for the rest of the study period (Twomile Run Flow Period 2). Perhaps most surprising is the temporal nature of the change in HgP per unit of TSS: we observed a rapid step-change in the HgP to TSS ratio rather than a gradual decline from elevated HgP to control-level concentrations. This suggests that the post-fire pulse of easily mobilized HgP was effectively flushed out of the system after 8 months. This finding contrasts with the one other long-term post fire study that found elevated Hg concentrations for two years following a fire.45 However, the fire was characterized as ‘severe’ and no details were provided on discharge and as a result, the streamwater flux was not quantified. The relatively rapid return to pre-fire levels at Twomile Run could have been due to a lower severity fire producing smaller amounts of potentially mobile Hg11 and/or because the 810 mm of rainfall that occurred during the 8 month post-fire period was enough to facilitate the complete flushing of elevated HgP from the watershed.

5.3 Streamwater Hg flux and watershed export

The streamwater HgT flux estimated without the fire disturbance for Twomile Run was 0.55 μg m−2 yr−1, which is less than half of the HgT flux measured at Paine Run for the 2010 water year (1.3 μg m−2 yr−1) but within the range of Hg fluxes measured in undisturbed forested systems (0.25–6.5 μg m−2 yr−1 as reported in Table 3 in Riscassi and Scanlon6). Although low compared to the prior Paine Run estimate, the Twomile (non-disturbed) flux is reasonable considering the annual streamflow at Twomile Run for the study period was half of that measured during the Paine Run study period (282 mm per year and 533 mm per year, respectively).

The measured streamwater HgT flux for Twomile Run attributed to the fire represented an estimated increase of 0.49 μg m2 over non-disturbed conditions. Although the fire resulted in an almost doubling of HgT export and a large change in the relative physical fractioning (Hg exported as HgP increased from 56 to 77%), the total amount of Hg transported downstream due to the fire was small relative to mean annual atmospheric deposition (2.9%). Overall these findings imply that additional HgP mobilized by fire has an immediate (2–8 months post fire) and significant impact on streamwater Hg concentrations and fluxes, but the enhanced (post-first flush) hydrological fluxes are not responsible for releasing a large portion of the Hg stored in watershed soils. While no other study has quantified post-fire Hg fluxes in the year immediately after a fire, Nelson et al.22 determined the annual streamwater HgT flux 50 years post fire was approximately one third lower than the flux measured in a similar non-fire impacted system (0.4 and 1.3 μg m−2 yr−1, respectively) due to Hg depletion in vegetation and soil. These results, together with our study, suggest a temporally changing post-fire streamwater response, ranging from elevated fluxes in the months following a fire to reduced Hg export in the long-term.

5.4 Applications to other systems

The results of this study can be applied to other forested headwater systems in the Eastern U.S. that have experienced a low-severity natural fire or prescribed burn. Our study shows that low-severity fires do not significantly impact post-first flush stream mobilization of HgD, DOC, or TSS following the fire. In addition, our study shows that low-severity fires can significantly increase the amount of HgP per unit TSS in streamwater. Since we found no change in the amount of sediment mobilized in a burned watershed, our study points to important mechanisms of increased HgP: local redeposition of HgP and enhanced atmospheric Hg binding to charred organic material on the ground surface. In addition, the results show that post-fire increases in Hg can return to non-fire impacted levels within a year.

Site-specific factors such as canopy type, ambient weather, long-term climate, moisture content of fuel, and fire severity are all likely to impact the post-fire hydrological transport of Hg. After a fire, different types of tree canopies have been shown to impact the amount of Hg that is scavenged and brought to the ground surface via throughfall.40 In addition, the intensity and timing of precipitation events following the fire can affect the amount of carbon and associated Hg that is mobilized to streams.50 Fuel moisture has been shown to be the overall driving factor defining the amount and relative fraction of particulate Hg emitted during a fire.39 Fire severity has also been shown to have a significant impact on the amount of Hg released from the terrestrial environment11,38 and on the amount of TSS (and associated HgP) that is mobilized to the stream as a result of greater vegetation and litter loss and subsequent erosion of bank sediment.77

5.5 Summary and implications

Our study has implications for prescribed burning, a common forest management technique used across the U.S., and wildfire management, as it demonstrates how a watershed's Hg budget can change in the aftermath of a wildfire. Natural and prescribed fires release Hg that is typically held in vegetation and soils, which can lead to local HgP redeposition as well as subsequent atmospheric Hg binding to charred organic material. This research has shown that in the year following a low-severity fire, newly available Hg can be rapidly transported downstream during the post-first-flush period in the absence of significant changes in TSS, DOC, or HgD. The low-severity fire in this study essentially doubled the amount of Hg exported from the headwater system. Any increase in inorganic Hg downstream is significant since it has the potential to be methylated and can lead to enhanced Hg accumulation in fish.13 The results of this study are especially important because fire activity is expected to increase across North America with warming caused by climate change.78,79 As wildfires become larger, more frequent, and more severe, a greater number of forested watersheds will experience alterations in the local Hg cycling. This study demonstrates that low-severity wildfires can cause significant increases in the amount of Hg mobilized to headwater streams.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

Funding was provided by the National Science Foundation – Division of Environmental Biology grant (DEB-1644500). This research is a contribution to the Shenandoah Watershed Study and the Virginia Trout Stream Sensitivity Study. Susie Maben and Jeff Atkins of the Shenandoah Watershed Study provided assistance in the laboratory and field, respectively. Lauren Tabor provided laboratory and data entry support. The authors also acknowledge the anonymous reviewers for their detailed evaluations which improved the quality of the manuscript.

References

  1. V. L. St Louis, J. W. M. Rudd, C. A. Kelly, B. D. Hall, K. R. Rolfhus, K. J. Scott, S. E. Lindberg and W. Dong, Environ. Sci. Technol., 2001, 35, 3089–3098 CrossRef CAS PubMed.
  2. E. K. Miller, A. Vanarsdale, G. J. Keeler, A. Chalmers, L. Poissant, N. C. Kamman and R. Brulotte, Ecotoxicology, 2005, 14, 53–70 CrossRef CAS PubMed.
  3. P. Lorey and C. T. Driscoll, Environ. Sci. Technol., 1999, 33, 718–722 CrossRef CAS.
  4. P. F. Schuster, D. P. Krabbenhoft, D. L. Naftz, L. D. Cecil, M. L. Olson, J. F. Dewild, D. D. Susong, J. R. Green and M. L. Abbott, Environ. Sci. Technol., 2002, 36, 2303–2310 CrossRef CAS PubMed.
  5. N. V. Smith-Downey, E. M. Sunderland and D. J. Jacob, J. Geophys. Res.: Biogeosci., 2010, 115, G03008 Search PubMed.
  6. A. L. Riscassi and T. M. Scanlon, J. Hydrol., 2013, 501, 92–100 CrossRef CAS.
  7. D. Obrist, Environ. Sci. Technol., 2012, 46, 5921–5930 CrossRef CAS PubMed.
  8. D. Obrist, D. W. Johnson and R. L. Edmonds, J. Plant Nutr. Soil Sci., 2012, 175, 68–77 CrossRef CAS.
  9. A. Biswas, J. D. Blum, B. Klaue and G. J. Keeler, Global Biogeochem. Cycles, 2007, 21, GB1002 CrossRef.
  10. A. Biswas, J. D. Blum and G. J. Keeler, Sci. Total Environ., 2008, 404, 129–138 CrossRef CAS PubMed.
  11. J. P. Webster, T. J. Kane, D. Obrist, J. N. Ryan and G. R. Aiken, Sci. Total Environ., 2016, 568, 578–586 CrossRef CAS PubMed.
  12. United Nations Environment Program (UNEP), Global Mercury Assessment 2013: Sources, Emissions, Releases and Environmental Transport, 2013 Search PubMed.
  13. E. N. Kelly, D. W. Schindler, V. L. S. Louis, D. B. Donald and K. E. Vlaclicka, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 19380–19385 CrossRef CAS PubMed.
  14. M. Ravichandran, Chemosphere, 2004, 55, 319–331 CrossRef PubMed.
  15. J. Dilling and K. Kaiser, Water Res., 2002, 36, 5037–5044 CrossRef CAS PubMed.
  16. R. K. Kolka, D. F. Grigal, E. S. Verry and E. A. Nater, J. Environ. Qual., 1999, 28, 766–775 CrossRef CAS.
  17. D. F. Grigal, Environ. Rev., 2002, 10, 1–39 CrossRef CAS.
  18. J. B. Shanley, P. F. Schuster, M. M. Reddy, H. E. Taylor and G. R. Aiken, EOS, Transactions of the American Geophysical Union, 2002, 83, 45–48 CrossRef.
  19. J. B. Shanley, M. A. Mast, D. H. Campbell, G. R. Aiken, D. P. Krabbenhoft, R. J. Hunt, J. F. Walker, P. F. Schuster, A. Chalmers, B. T. Aulenback, N. E. Peters, M. Marvin-DiPasquale, D. W. Clow and M. M. Shafer, Environ. Pollut., 2008, 154, 143–154 CrossRef CAS PubMed.
  20. M. E. Galloway and B. A. Branfireun, Sci. Total Environ., 2004, 325, 239–254 CrossRef CAS PubMed.
  21. M. A. Mast, D. H. Campbell, D. P. Krabbenhoft and H. E. Taylor, Water, Air, Soil Pollut., 2005, 164, 21–42 CrossRef CAS.
  22. S. J. Nelson, K. B. Johnson, J. S. Kahl, T. A. Haines and I. J. Fernandez, Environ. Monit. Assess., 2007, 126, 69–80 CrossRef CAS PubMed.
  23. P. F. Schuster, J. B. Shanley, M. Marvin-Dipasquale, M. M. Reddy, G. R. Aiken, D. A. Roth, H. E. Taylor, D. P. Krabbenhoft and J. F. DeWild, Water, Air, Soil Pollut., 2008, 187, 89–108 CrossRef CAS.
  24. M. E. Brigham, D. A. Wentz, G. R. Aiken and D. P. Krabbenhoft, Environ. Sci. Technol., 2009, 43, 2720–2725 CrossRef CAS PubMed.
  25. J. A. Dittman, J. B. Shanley, C. T. Driscoll, G. R. Aiken, A. T. Chalmers, J. E. Towse and P. Selvendiran, Water Resour. Res., 2010, 46, W07522 CrossRef.
  26. J. D. Demers, C. T. Driscoll and J. B. Shanley, Water Resour. Res., 2010, 46, W01511 CrossRef.
  27. A. L. Riscassi and T. M. Scanlon, Water Resour. Res., 2011, 47, W12512 CrossRef.
  28. J. A. Dittman, J. B. Shanley, C. T. Driscoll, G. R. Aiken, A. T. Chalmers and J. E. Towse, Environ. Pollut., 2009, 157, 1953–1956 CrossRef CAS PubMed.
  29. R. P. Mason and K. A. Sullivan, Water Res., 1998, 32, 321–330 CrossRef CAS.
  30. T. Scherbatskoy, J. B. Shanley and G. J. Keeler, Water, Air, Soil Pollut., 1998, 105, 427–438 CrossRef CAS.
  31. R. K. Kolka, D. F. Grigal, E. A. Nater and E. S. Verry, Soil Sci. Soc. Am. J., 2001, 65, 897–905 CrossRef CAS.
  32. A. L. Riscassi, K. J. Hokanson and T. M. Scanlon, Water, Air, Soil Pollut., 2011, 220, 23–36 CrossRef CAS.
  33. C. J. Allan and A. Heyes, Water, Air, Soil Pollut., 1998, 105, 573–592 CrossRef CAS.
  34. H. R. Friedli, L. F. Radke, J. Y. Lu, C. M. Banic, W. R. Leaitch and J. I. MacPherson, Atmos. Environ., 2003, 37, 253–267 CrossRef CAS.
  35. M. Mailman and R. A. Bodaly, Environ. Pollut., 2005, 138, 161–166 CrossRef CAS PubMed.
  36. M. A. Engle, M. S. Gustin, D. W. Johnson, J. F. Murphy, W. W. Miller, R. F. Walker, J. Wright and M. Markee, Sci. Total Environ., 2006, 367, 222–233 CrossRef CAS PubMed.
  37. M. P. Burke, T. S. Hogue, M. Ferreira, C. B. Mendez, B. Navarro, S. Lopez and J. A. Jay, Water, Air, Soil Pollut., 2010, 212, 369–385 CrossRef CAS PubMed.
  38. R. K. Kolka, B. R. Sturtevant, J. R. Miesel, A. Singh, P. T. Wolter, S. Fraver, T. M. DeSutter and P. A. Townsend, Int. J. Wildland Fire, 2017, 26, 296–305 CrossRef CAS.
  39. D. Obrist, H. Moosmueller, R. Schuermann, L. W. A. Chen and S. M. Kreidenweis, Environ. Sci. Technol., 2008, 42, 721–727 CrossRef CAS PubMed.
  40. E. L. Witt, R. K. Kolka, E. A. Nater and T. R. Wickman, Environ. Sci. Technol., 2009, 43, 1776–1782 CrossRef CAS PubMed.
  41. I. Campos, C. Vale, N. Abrantes, J. J. Keizer and P. Pereira, Catena, 2015, 131, 149–159 CrossRef CAS.
  42. C. A. Caldwell, C. M. Canavan and N. S. Bloom, Sci. Total Environ., 2000, 260, 125–133 CrossRef CAS PubMed.
  43. R. K. Kolka, T. Wickman, E. Nater, M. Gabriel, L. Woodruff, W. Cannon, K. Gebhardt, J. Butrcher and E. Witt, Measurements of mercury mobilization and accumulation in fish in response to prescribed fire in a boreal forest ecosystem, USDA Forest Service, 2008 Search PubMed.
  44. S. E. Rothenberg, M. E. Kirby, B. W. Bird, M. B. DeRose, C.-C. Lin, X. Feng, R. F. Ambrose and J. A. Jay, Environ. Earth Sci., 2010, 60, 993–1005 CrossRef CAS.
  45. U. Silins, A. Anderson, K. D. Bladon, M. B. Emelko, M. Stone, S. A. Spencer, C. H. S. Williams, M. J. Wagner, A. M. Martens and K. Hawthorn, For. Chron., 2016, 92, 39–42 CrossRef.
  46. C. A. Burton, T. M. Hoefen, G. S. Plumlee, K. L. Baumberger, A. R. Backlin, E. Gallegos and R. N. Fisher, PLoS One, 2016, 11(5), e0153372 Search PubMed.
  47. R. K. Kolka, Effects of Fire and Fuels Management on Water Quality in Eastern North America, U.S., Department of Agriculture Forest Service, Southern Research Station, Asheville, NC, 2012 Search PubMed.
  48. B. T. Bormann, P. S. Homann, R. L. Darbyshire and B. A. Morrissette, Can. J. For. Res., 2008, 38, 2771–2783 CrossRef CAS.
  49. A. D. Revchuk and I. H. Suffet, Water Environ. Res., 2014, 86, 372–381 CrossRef CAS PubMed.
  50. S. F. Murphy, J. H. Writer, R. B. McCleskey and D. A. Martin, Environ. Res. Lett., 2015, 10, 084007 CrossRef.
  51. A. K. Hohner, K. Cawley, J. Oropeza, R. S. Summers and F. L. Rosario-Ortiz, Water Res., 2016, 105, 187–198 CrossRef CAS PubMed.
  52. M. Sofia Larroulet, E. Noemi Hepper, M. P. Alvarez Redondo, V. Belmonte and A. Maria Urioste, Arid Land Res. Manage., 2016, 30, 105–119 CrossRef.
  53. T. Sullivan, B. Cosby, J. Laurence, R. Dennis, K. Savig, J. Webb, A. Bulger, M. Scruggs, C. Gordon, J. Ray, H. Lee, W. Hogsett, H. Wayne, D. Miller and J. Kern, Assessment of Air Quality and Related values in Shenandoah National Park. Technical Report NPS/NERCHAL/NRTR-03/090, 2003 Search PubMed.
  54. U. S. EPA, Method 1669, Sampling Ambient Water for Trace Metals at EPA Water Quality Criteria Levels, 1996 Search PubMed.
  55. A. L. Riscassi, A. D. Converse, K. J. Hokanson and T. M. Scanlon, J. Environ. Monit., 2010, 12, 1833–1839 RSC.
  56. A. Riscassi, K. Hokanson and T. Scanlon, Water, Air, Soil Pollut., 2011, 220, 23–36 CrossRef CAS.
  57. M. E. Lewis and M. E. Brigham, Low-level mercury, in Processing of Water Samples. US Geological Survey Techniques of Water-Resources Investigations, Book 9, ed. F. D. Wilde, D. B. Radtke, J. Gibs and R. T. Iwatsubo, ch. A5, 2004 Search PubMed.
  58. U. S. EPA, Method 1631 Revision E: Mercury in Water by Oxidation, Purge and Trap, and Cold Vapor Atomic Fluorescence Spectrometry, 2002 Search PubMed.
  59. B. B. Potter and J. C. Wimsatt, Method 415.3, Revision 1.1. Determination of total organic carbon and specific UV absorbance at 254 nm in source water and drinking water, EPA Doc. EPA/600/R-05/055, 2005 Search PubMed.
  60. U. S. EPA, Determination of pH (Open System) in Handbook of Methods for Acid Deposition Studies, Laboratory Analysis for Surface Water Chemistry, EPA 600/4–87/026, 1987 Search PubMed.
  61. Y. P. Chin, G. Aiken and E. Oloughlin, Environ. Sci. Technol., 1994, 28, 1853–1858 CrossRef CAS PubMed.
  62. J. L. Weishaar, G. R. Aiken, B. A. Bergamaschi, M. S. Fram, R. Fujii and K. Mopper, Environ. Sci. Technol., 2003, 37, 4702–4708 CrossRef CAS PubMed.
  63. U. S. EPA, Method 415.3, Revision 1.1. Determination of total organic carbon and specific UV absorbance at 254 nm in source water and drinking water, EPA Doc. EPA/600/R-05/055, 1971 Search PubMed.
  64. U. S. EPA, Method 160.4. Residue, Volatile (Gravimetric, Ignition at 550°C) in Methods for the chemical analysis of water and wastes, EPA/600/4–79/020, 1971 Search PubMed.
  65. B. T. Aulenbach and R. P. Hoopert, Hydrol. Processes, 2006, 20, 3029–3047 CrossRef CAS.
  66. A. P. Appling, M. C. Leon and W. H. McDowell, Ecosphere, 2015, 6(12), 269 Search PubMed.
  67. M. R. Risch, J. F. DeWild, D. A. Gay, L. Zhang, E. W. Boyer and D. P. Krabbenhoft, Environ. Pollut., 2017, 228, 8–18 CrossRef CAS PubMed.
  68. C. D. Evans, I. A. Malcolm, E. M. Shilland, N. L. Rose, S. D. Turner, A. Crilly, D. Norris, G. Granath and D. T. Monteith, Ecosystems, 2017, 20, 813–829 CrossRef CAS.
  69. J. Battle and S. W. Golladay, Am. Midl. Nat., 2003, 150, 15–25 CrossRef.
  70. S. J. Balogh, Y. B. Huang, H. J. Offerman, M. L. Meyer and D. K. Johnson, Sci. Total Environ., 2003, 304, 305–313 CrossRef CAS PubMed.
  71. G. R. Wall, H. H. Ingleston and S. Litten, Water, Air, Soil Pollut., 2005, 165, 233–248 CrossRef CAS.
  72. L. W. Swift, K. J. Elliott, R. D. Ottmar and R. E. Vihnanek, Can. J. For. Res., 1993, 23, 2242–2254 CrossRef.
  73. P. R. Robichaud and T. A. Waldrop, Water Resour. Bull., 1994, 30, 27–34 CrossRef.
  74. K. J. Elliott and J. M. Vose, South. J. Appl. For., 2005, 29, 5–15 CAS.
  75. E. S. Olson, S. J. Miller, R. K. Sharma, G. E. Dunham and S. A. Benson, J. Hazard. Mater., 2000, 74, 61–79 CrossRef CAS PubMed.
  76. M. S. Gustin and K. Ladwig, J. Air Waste Manage. Assoc., 2004, 54, 320–330 CAS.
  77. U. Silins, M. Stone, M. B. Emelko and K. D. Bladon, Catena, 2009, 79, 189–197 CrossRef.
  78. M. D. Flannigan, B. J. Stocks and B. M. Wotton, Sci. Total Environ., 2000, 262, 221–229 CrossRef CAS PubMed.
  79. M. D. Flannigan, B. D. Amiro, K. A. Logan, B. J. Stocks and M. B. Wotton, Mitigation and Adaptation Strategies for Global Change, 2005, vol. 11, pp. 847–857 Search PubMed.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c7em00419b

This journal is © The Royal Society of Chemistry 2017