Environ. Sci. Technol. 2010, 44, 935–940
Protocol to Reconstruct Historical Contaminant Loading to Large Lakes: The Lake Michigan Sediment Record of Mercury RONALD ROSSMANN* U.S. Environmental Protection Agency, Office of Research and Development, National Health and Environmental Effects Research Laboratory, Mid-Continent Ecology Division, Large Lakes and Rivers Forecasting Research Branch, Large Lakes Research Station, 9311 Groh Road, Grosse Ile, Michigan 48138
Received July 30, 2009. Revised manuscript received November 19, 2009. Accepted December 8, 2009.
Samples of opportunity from Pb-210 dated sediment cores collected from Lake Michigan between 1994 and 1996 were analyzed for mercury. The storage of anthropogenic and total (post1850) mercury in the lake is calculated to be 186 and 228 t, respectively. By setting the sum of mercury stored in a representative core equal to the mercury storage within the entire lake, the time variation of annual mercury loading to the lake is calculated. The modern (1980-2002) mercury flux to the lake represented by the surface of the core at the time of collection in 1994 was 21.4 µg/m2/y. The preindustrial flux (e1850) was 3.09 µg/m2/y, and the peak flux in 1946 was 53.3 µg/ m2/y. These yield modern and peak enrichment factors of 6.92 and 17.2, respectively. Modern fluxes exceed published atmospheric deposition estimates and, therefore, include terrestrial point sources, atmospheric deposition to watersheds, and atmospheric deposition to the lake. The modern net mercury load to the lake’s sediments was 1157 kg/y in 1994. The atmosphere is estimated to contribute 91% of this load directly to the lake.
Introduction Accurate, complete historic records of contamination are sparse. In many cases, measurements of contamination do not begin until after identification of problems associated with the contaminant. Sediment cores can contain a record of contamination for a region. However, the record can be influenced by local and long-range transport, global background signals, and local sediment redistribution processes integrated into the record. For mercury concentration, trends have been investigated in Scotland, the Mediterranean Sea, Norway, the United States, the United Kingdom, Finland, and Canada to mention just a few (1–7). Information on the time variation of mercury fluxes can be obtained from dated sediment cores. Some of the locations for which this has been done include Finland, Sweden, Russia, the United Kingdom, Greenland, Denmark, Norway, Ireland, Brazil, and Czech Republic (8–17). Locations for which mercury fluxes have been reported for North America include Alberta, Minnesota, Vermont, New Hampshire, Alaska, Florida, Ontario, Yukon, Northwest Territories, Hudson Bay, California, * Corresponding author e-mail:
[email protected]; phone: 734-692-7612; fax: 734-692-7603. 10.1021/es902307c
Not subject to U.S. Copyright. Publ. 2010 Am. Chem. Soc.
Published on Web 12/23/2009
Colorado, Bay of Fundy, Vermont, New Hampshire, New York, Lake Superior, and Manitoba (18–31). The application of dated sediment cores for documenting historic mercury fluxes is discussed by Fitzgerald et al. (32). For those contaminants that are not subject to significant redistribution of mass by sediment diagenesis and diffusion, dated sediment cores can be used to reconstruct the time variation of a contaminant’s flux to a region represented by the cores. Most of these reconstructions have been done for relatively small lakes. If the goal is to reconstruct atmospheric deposition, the use of small lakes is advantageous because many such lakes occur in small undeveloped watersheds where sediment redistribution is less complex and mercury sources are primarily from the atmosphere. For these lakes, a core that is subjected to a minimal amount of postdepositional redistribution can often be obtained from deep areas of the lake where the signal-to-noise ratio is high. The task becomes much more complicated when trying to reconstruct historical loadings to a large basin such as Lake Michigan. Large basins receive loads from tributaries that are not spatially evenly distributed and that may not have the same loading histories. For a large lake, sediment deposition not only varies in space and time but also in response to the postdepositional processes; including biotic and physical mixing, sediment resuspension by waves and currents, and redistribution by currents. All these processes can act on lake sediment to varying degrees and lead to erroneous interpretations. Though sediments of a large lake contain a history of persistent contaminant loads, the loads are those to the site acted upon by local conditions. Thus cores from a large lake may provide histories with a high variance when compared to one another. The basic problem is which, if any, of the cores represent the lake’s contaminant load history. This problem, though challenging, must be addressed for the reconstruction of contaminant histories to large lakes. Protocols must be developed to address the identified issues that make interpretation very complex. Defensible analysis of sediment cores from large lakes provides an independent means for estimating modern and historic contaminant loads for use in mass balance models. Development of loading history of a contaminant to a lake that is representative of contaminant loads from all sources to the lake as a whole requires that the loading history described by a core is consistent with all other cores analyzed. This paper uses mercury data to present a protocol for understanding and integrating multiple stratigraphic profiles of a persistent contaminant from the large basin. The protocol can be applied elsewhere if dated sediment profiles are available, the investigator has an understanding of the depositional zones in the basin, and contaminant storage in a basin is known or can be estimated.
Methods Cores were collected in 1994-1996 using a Soutar-type box corer which penetrated between 10 and 69 cm of sediment (Figure 1). Stations were spaced on a roughly 25-km grid; however, cores could only be obtained at 52 of the stations. Regions without representation by cores were nondepositional and could not be cored due to sand, rock, or glacial clays. After sectioning the cores in the field, samples were immediately frozen. In the laboratory, samples were freezedried and homogenized. Samples of the freeze-dried sediment were received as samples of opportunity once all dating was done elsewhere. Details of the collection and dating are VOL. 44, NO. 3, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 935
mercury storage were calculated. Annual mass fluxes to a 1-cm2 area of the lakes sediment were calculated. To calculate anthropogenic mercury loading for each core (eq 1), the annual mass flux (AF) of a 1-cm-thick interval in ng/cm2/y was calculated by first subtracting the preindustrial background (mean of all intervals in a core dated prior to 1850) mercury concentration in ng/g (HgB) from the measured concentration in ng/g (HgC) and then multiplying the result by the mass sedimentation rate in g/cm2/y (R) (Table S1) (36). AF ) (HgC-HgB) · R
(1)
The annual mass flux was then multiplied by the number of years (y) in the core interval to provide the total mercury mass within the interval (TAHg) in ng/cm2 (eq 2). TAHg ) AF · y
(2)
The anthropogenic mercury masses of all intervals (n) of a core were then summed in eq 3 to provide the total anthropogenic mercury mass per unit area (ng/cm2) within the core (THgAMASS). This was done for every core. n
THgAMASS )
∑ TAHg
i
(3)
i)1
FIGURE 1. Locations where box cores were successfully collected in Lake Michigan between 1994 and 1996. Regions of the lake lacking indicated locations could not be box cored due to a lake bottom consisting of sand, rock, or stiff glacial tills. Thus regions not having cores recovered do not have significant mercury storage. reported by Robbins, Edgington, and others (33–36). Pb-210 dating results were reported in 1999 by Robbins et al. (36). A total of 1580 core sections (1-cm intervals) from 52 cores were weighed and then analyzed by cold vapor atomic absorption spectrophotometry on a Milestone DM-80 analyzer (37) much in the same way as that described by Bennett et al. (38). Samples were combusted in the instrument with collection of mercury on a gold amalgam. Once collected, the amalgam was heated to volatize the mercury and flush it to the detector. Calibration was checked daily with check standards. All analyses were completed in 2003. The mean recoveries for SRM 1571 (orchard leaves) and Mess 3 (marine sediment) standard reference materials was 90 and 106%, respectively. Data were accepted only for SRM recoveries within the range of 80 to 120%. Samples failing these criteria were analyzed again. Blanks averaged 1.9 ng/g. The mean method detection limit based on the standard deviation of the blanks was 5.1 ng/g. All sample results were above the method detection limit. The mean percent relative difference for replicate samples was 4.5%.
Results and Discussion Throughout this section, the terms flux and load are used. Flux refers to the deposition of mercury to the lake’s water or sediment surface unit area per unit of time. Load refers to the mass of mercury delivered to Lake Michigan, excluding Green Bay, per unit time. The area of the lake used is 54,120 km2. This was estimated using the relative weights of Lake Michigan and Lake Michigan minus Green Bay cutouts proportioned to the reported Lake Michigan area of 58,016 km2 (39). Mercury Storage. Mercury storage was calculated based upon all cores analyzed. Both total and anthropogenic 936
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 3, 2009
Of all the cores, only six did not penetrate deeper than 1850. For these cores, the concentration in the deepest interval was used. These represented concentrations as late as 1870. A similar procedure, without background correction, was used to calculate the total mass of total mercury within a core for the period of 1850 to the date of core collection. Both anthropogenic and total mercury masses per unit area results along with x and y location data were converted to storage using the program Surfer (40) version 8.06.39. A boundary of zero ng/cm2 was used for the shoreline. With this boundary condition, the resulting distribution of total mercury storage in the lake’s sediments was consistent with respect to the location of depositional and nondepositional (sand, rock, glacial clay) areas based on sediment type (Figure 2). The data were interpolated using the Surfer software kriging defaults. The results from Surfer in ng/cm2 (Hgs) were converted to metric tons/km2 using the conversion factor of 10-5. The metric tons/km2 were multiplied by the area (A) in km2 of the lake to yield a mercury mass Hgm in metric tons stored in the lake (eq 4). Green Bay was not included for any of these calculations. Hgm ) Hgs · A
(4)
The bay was not represented by the current sampling. With a mean mercury flux to the surficial sediments of 200 µg/m2/y, the bay receives large loads of mercury (41). If mercury behaves like PCBs, which are also associated with particles delivered to the bay and redistributed by resuspension and currents, most of the load is trapped within the bay (42). Not including Green Bay, the post-1850 total mercury storage is 228 t, and anthropogenic mercury storage is 186 t. Normalization of Cores to Mercury Storage. The apparent mass stored in each of the cores was normalized to the actual total or anthropogenic storage previously calculated using Surfer. For the normalization, each core was assumed to be representative of the total or anthropogenic mercury stored within the lake. Using anthropogenic mercury as an example, the THgAMASS from eq 3 was multiplied by the lake’s surface area (A) to yield the apparent mass stored within the lake (ATHgM) as represented by the core (eq 5).
FIGURE 3. Comparison of anthropogenic mercury flux profiles from various locations in Lake Michigan. The number of years that each curve was offset were -22 years for station 120, 0 years for station 41, -8 years for station 34, -19 years for station 61, -14 years for station 101, -19 years for station 58, -5 years for station 13, and -38 years for station 83. See Figure 1 for core locations.
FIGURE 2. Mercury storage (ng/cm2) sediments for the period 1850 to 1996.
in
ATHgM ) THgAMASS · A
Lake
Michigan (5)
The normalization factor was equal to the ATHgM divided by the Surfer calculated storage (Hgm). The factor was different for each core. This factor is called the apparent mercury focusing factor (Hg FF) (Table S2). With a mean of 2.37, it ranges between 0.182 and 12.9, illustrating the spatial variability of local sources and redistribution of mercury associated with sediments by physical forces. These means and ranges are larger than the Pb-210 (mean 1.65 with a range of 0.428-3.36) or Cs-137 (mean 2.10 with a range of 0.503-5.603) focusing factors reported by Robbins et al. (Table S1) (36). Differences among the mercury, Pb-210, and Cs-137 focusing factors illustrate that mercury, unlike Pb210 and Cs-137, has a variety of sources that are not evenly distributed over the basin. Time Variation of Mercury Loads to Lake Michigan. The time variation of the mercury load to Lake Michigan was calculated by normalizing the amount of mercury in a representative high quality core to the amount of mercury stored within the lake. High quality cores were those having a high sedimentation rate, a sampling interval that represented as few years as possible, a mixed layer representing as few years as possible, a high degree of responsiveness to annual fluxes to the lake, and the smallest time delay for delivery of mercury loaded near the shoreline to offshore core locations by recurrent sediment resuspension, transport by currents, and sedimentation. Cores having the highest sedimentation rates (g0.1 g/cm2/y) are those from stations 9, 15, 22, 41, 55, 61, and 87. Of these, the sample interval and mixed layer thickness resulted in surface resolutions of 0.91, 0.68, 0.73, 1.1, 3.1, 8.8, and 5.2 years, respectively. Because of resuspension and transport processes within the lake, the delivery of a mercury load to any location can be delayed. The delay was estimated by seeking those cores that had the earliest occurring historic peak maximum flux and using that mean date (1944) as the absolute time of the peak. Of the
high resolution cores, cores 9, 22, 15, 41, and 55 had an earliest ultimate peak occurring in 1957, 1950, 1944, 1946, and 1942, respectively. Core 9 had a double peak with one occurring in 1939 and the other in 1957. Thus it was not used. Core 55 was also somewhat strange in that the onset of mercury contamination was quite late (post-1920), suggesting a history of loads quite different from all other cores (Figure S1). The core 22 date for peak mercury loads appeared 5 years too late, attributed to redistribution processes, and was not used. The mean date of 1945 for cores 15 and 41 was chosen to represent the date of maximum mercury flux to the lake. These cores were within the southeastern region of the lake where sediment is resuspended during the isothermal period (November to April) in response to wind events as recorded by NASA/GSFC, MODIS Rapid Response satellite imagery (Figure S2) and initially reported by Eadie et al. (43). Either core 15 or core 41 seemed ideal candidates to represent the time variation of loads to the lake. Core 41 was chosen to represent loadings to the lake with little or no time delay for delivery of annual loads to its location. A comparison of loads to the lake for the two cores can be found in Figure S3. The sediments at station 41 had a mixed depth within the 1 cm surface interval sampled at the time of collection. The sampling resolution at this location yielded intervals that represent from 1.1 year of sedimentation at the top (1994) to 3.8 years near the bottom (1845) of the core. Within Lake Michigan, sediment resuspension and transport processes result in the delivery of older sediment with associated mercury to various locations in the lake. This produces a time offset that was estimated using the ultimate peak flux year at each location. The offset was determined by subtraction of the mean peak flux year of 1945 from the peak year of each core. Cores from various locations within the lake were selected to illustrate this delay for various regions of the lake relative to core 41. The cores selected include those from stations 13, 34, 58, 61, 83, 101, and 120 (Figure 1). The offsets for these cores are -5, -8, -19, -19, -38, and -22 years, respectively (Figure 3). Because of the integration of fluxes in the mixed layer and the time delay associated with resuspension and transport processes, only cores from stations 15 and 41 illustrate conditions representative of mercury loads at the time of collection (Table S3). All other cores were from stations that represent loads that occurred from a minimum of 5 years earlier (station 13) to a maximum of 38 years earlier (station 83) and that integrate the mercury flux for periods ranging from 5 years (station 61) to 28 years (station 58) due to surface sediment mixing (mixed layer thickness). Thus the record at all other illustrated locations is delayed and integrated with earlier fluxes to the lake. VOL. 44, NO. 3, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
937
TABLE 1. Annual Mercury Loads to Lake Michigan during the Lake Michigan Mass Balance Project (1994-1995)a load type
whole lake load (kg/y)(44)
adjusted Lake Michigan (excluding Green Bay) load (kg/y)
lake (excluding Green Bay) loads based on this work (kg/y)
load from atmosphere evasion to atmosphere net load from atmosphere total tributary load total net load
+1173 -453 +720 +230 +950
+1094 -422 +672 +107 +779
+1709 -659 +1050 +107 +1157
a Adjusted values are derived from those of Landis and Keeler (44) adjusted for the relative areas of the main lake (whole lake without Green Bay) to whole lake and for tributaries to the main lake only. Lake loads are based on core results with the assumption all missed loads were from the atmosphere. The atmospheric load and evasion proportion in the adjusted column was maintained for the lake loads column.
FIGURE 4. Time variation of total, anthropogenic, and background (presettlement) mercury loads (kg/y) to Lake Michigan based on the core from station 41 normalized to main lake (Green Bay excluded) mercury storage. Annual Loads in 1994-1995. Landis and Keeler (44) found that 84% of the mercury load to the lake, or 76% of the net mercury load, was from the atmosphere for 1994-1995. Adjusting their loads to exclude Green Bay and merging these with this study (Table 1), the net atmospheric load to the main lake accounted for 91% of the mercury stored in the sediments (1157 kg). The Landis and Keeler (44) net load to the lake in 1994-1995 (779 kg/y), based on model output that utilized field measurements, is 33% less than that calculated from mercury storage in the sediments. Either a load was missed, more mercury was emitted than calculated by Landis and Keeler (44), tributary loads were considerably larger than calculated, or a combination of all three. I believe the discrepancy between the Landis and Keeler (44) results and those based on storage are primarily related to mercury loads from the atmosphere or emissions of mercury from
the water column. The calculated tributary loads based on measurements were only 107 kg/y or 9% of the load and are thus a minor contributor to the total load to the sediments. For tributaries to account for the missing load to the sediments, their load would have to be more than tripled. Only five stations of data (Figure S4) were available to Landis and Keeler (44) for calculation of the loads. One of the stations was outside of the Lake Michigan basin. Thus the entire Wisconsin shoreline from Milwaukee northward into Michigan was devoid of stations. Within this stretch of coastline, loads from Milwaukee, Sheboygan, Manitowoc, Green Bay, Menominee, Escanaba, and Manistique would have been missed. These urban center locations are upwind of the prevailing southwest summer winds and northwest winter winds. Thus loads to the lake from these locations would have been missed. Vette, Landis, and Keeler (44) noted high mercury in precipitation and high total gaseous mercury concentrations offshore of Gary and Chicago. If these urban regions are representative of urban regions in general, then mercury loads to the lake associated with urban locations from Milwaukee northward would have been missed. This could account for the differences between the Landis and Keeler (44) loads and those calculated based on sediment storage. To fully resolve this issue, additional work is required. Historic Load Function. The total and anthropogenic mercury load functions were calculated using the storage of each in the lake, excluding Green Bay. The summed mass of mercury in the core was normalized to the total and anthropogenic mercury masses stored in the lake. This yielded the time variation of total and anthropogenic mercury loads to the lake with time (Figure 4). Because core interval represented a number of years, the results were linearly extrapolated between core intervals to provide a load for each year between 1850 and 1994 (Table S4). The areas under
TABLE 2. Comparison of Mercury Fluxes (µg/m2/y) to Lake Michigan with Those Elsewhere within the Great Lakes Region preindustrial (P) year
