Environ. Sci. Technol. 2006, 40, 2910-2916 Segmented Organic Nitrogen Isotopes in Freshwater Wetlands Record Long-Term Changes in Watershed Nitrogen Source and Land Use EMILY M. ELLIOTT* AND GRACE S. BRUSH Department of Geography & Environmental Engineering, Johns Hopkins University, 3400 North Charles Street, Baltimore, Maryland 21218 Although historic land use is widely recognized as an important determinant of watershed N cycling, efforts to examine land use legacy effects are limited by incomplete historical data. This research evaluates N isotopes of sedimented organic matter (<315Norg), in a palynological context, as a long-term proxy of changes in N source to wetland biota. N and S isotope measurements of organic sediments, fossil plant fragments, and living plants are used to explore isotope stratigraphies of wetland sediment cores. Processes potentially contributing to isotope stratigraphies are investigated including the following: a change in N source, diagenesis, and denitrification. We document the <515Norg stratigraphy of a core from the Smithsonian Environmental Research Center, MD, U.S.A. spans approximately 350 years, during which time (5,5Norg increases from +2%> to +7%. Reconstructed population density and wastewater inputs to the watershed suggest that the increase in (5,5N reflects changing land use from forested conditions to increasing nutrient inputs from human waste. Our results illustrate the importance of hydrologic connectivity in delivering waste-derived N in a watershed characterized by relatively low human population density. These results also demonstrate how this approach can expand the temporal horizon over which we can assess human impacts to watershed N dynamics. Introduction Land use is an important determinant of nitrogen (N) biogeochemistry, whether in terms of N cycling {1,2), N fluxes (3, 4), or N sources to watersheds (5, 6). Isotopic studies comparing land uses in small watersheds have shown that forested ecosystems generally have (315N values <+3%o in soils, groundwater, and surface water, whereas watersheds with significant waste contributions from pasturage, feedlots, or septic systems generally have (515N values >+8%o (e.g., refs 7 and 8). In particular, Mayer et al. report that large watersheds with >80% forest cover generally have low <515NN03- (+3.5 to +5.5%o), whereas watersheds with >15% agricultural or urban land use have higher <515NN03- (+6 to +9%o) (5). Such studies provide valuable insight into contemporary dynamics between land use, human activities, and N sources to ecosystems. 'Corresponding author phone: (650)329-4649; fax: (650)329-5590; e-mail: eelliott@usgs.gov. Given the importance of contemporary land use on nutrient sources to watersheds, it is also important to consider that land use has changed dramatically in most Eastern U.S. watersheds since European settlement. For example, during the peak period of agricultural land clearance in the mid- to late-1880s, it is estimated that only 20% of the mid-Atlantic area was forested (9). The historic N dynamics resulting from such changes are unknown; however, it has been shown that historic land use affects contemporary N cycling (2). Attempts to understand N dynamics in watersheds prior to major anthropogenic N additions are generally limited to modeling exercises or examination of "pristine" watersheds. These approaches are limited by the lack of historical data for validating modeled results and the accuracy of comparing "pristine" watersheds given ubiquitous atmospheric ni- trate deposition, as well as the myriad factors that affect N cycling in individual watersheds (e.g., vegetation type, climate, etc.). In addition to 400 years of land use change in the eastern U.S., N sources to watersheds have also changed as a result of human activities. In the absence of human additions to the N cycle, ecosystems obtain N via biological fixation and, to a lesser extent, from lightning and forest fires. Of N-fixing organisms, symbiotic plant-bacteria associations provide the largest amount of fixed N in upland and wetland ecosystems {10). (315N derived from biological fixation generally varies over a small range (0 to +2%o) due to its derivation from atmospheric N (isotopically defined as 0%o). However, since 1890, human activities (e.g., fossil fuel combustion, fertilizer production) have increased fixed N contributions worldwide 9-fold (11). As fixed N, these inputs are readily available for biotic uptake. In addition, these anthropogenic inputs have relatively distinct ranges of d 15N values. For example, fertilizer (315N generally ranges from ?2%o to +4%o (12), whereas wet atmospheric deposition of nitrate in the U.S. ranges from ? ll%o to +3%o (13) and nitrate derived from wastewater, septic effluent, or manure generally has (315N values ranging from +8%o to +22%o (14, 15). ' This research investigates the use of freshwater wetlands as archives of land use change, human disturbance, and the associated impacts to the N cycle over long temporal scales. In particular, we examine the use of (515N of sedimented organic N (<515Norg) from freshwater wetlands as a means for reconstructing long-term changes in watershed N sources to wetland biota. As wetland biota assimilate inorganic N, the isotopic composition of the biomass acquires the (315N signature of the N source with minimal fractionation (< 2%o) (16). Ultimately, this organic N pool returns to the wetland surface and is buried by sediment, recording changes in N source to biota over time. Previous sediment core studies in lakes and coastal waters have demonstrated the utility of using isotopes to distinguish multiple N sources to aquatic biota, including atmospheri- cally derived N (17), biologically fixed N (18), wastewater effluent N (19), and salmon-derived N (20). Other work has shown that sedimented organic N isotopes record eutrophi- cation of lakes and coastal waters (21, 22). However, water column N cycling and trophic interactions in lakes and coastal waters can complicate the transfer of (315N from primary producers into the sedimentary record. In comparison, as the main primary producer in wetlands, wetland plants provide a direct link to evaluating nutrient sources without the complications of trophic transfer. In addition, freshwater wetlands regulate water, nutrient, and sediment fluxes at the terrestrial?aquatic interface, much closer to potential nutrient sources. 2910" ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 9, 2006 10.1021/es051587q CCC: $33.50 ? 2006 American Chemical Society Published on Web 03/30/2006 Experimental Section This research was conducted in a suburbanizing Coastal Plain watershed (1192 ha) at the Smithsonian Environmental Research Center, MD, U.S.A. (38?51 N, 76?32 W) (see Supporting Information). Two sediment cores were taken from Mill Swamp, a 59 ha, freshwater, nontidal, herbaceous wetland that flows into Muddy Creek, a tributary to Rhode River and the Chesapeake Bay. The two cores were taken approximately 100 m apart, with the SERC2 core upstream of the SERC 1 core. The sediment cores were collected using a vibrocorer and refrigerated upon return to the lab. After extrusion, sediments were divided into 1-cm increments for all subsequent analyses. Organic N and C isotopes were determined on increments from 1 to 10 cm and every 8?10 cm downcore (w=24). Organic sulfur was extracted at 5 sediment intervals (1-2 cm, 40-41 cm, 69-70 cm, 78-79 cm, 93-94 cm) for isotopic analysis. Plant fragments were extracted from sediments at intervals adjacent to those analyzed for isotopes and pollen (w=8). Pollen was extracted from sediments using methods documents by Faegri and Iverson {23), counted, and identified in increments from every 8 to 10 cm (?=19). Sedimentation rates were reconstructed following Brush (24), wherein changes in pollen assemblages are matched with historic records of land use change (see Supporting Information). Sedimentation rates were reconstructed for each 1-cm sediment interval between dated horizons by adjusting the average sedimentation rate in proportion to pollen concen- trations (24). Because pollen counts were conducted every 8?10 cm, interpolation was used to calculate sedimentation rates in intervals where pollen counts were not available. The pollen biomarkers used for dating sediments include the ratio of Quercus to Ambrosia (initial land clearance) and the abundance of Ambrosia (peak land clearance) and Castanea (extinction of American chestnut) pollen. The U.S. Department of Agriculture, Agricultural Research Service Lab (Beltsville, MD) measured 137Cs concentrations in sediments from 1 to 10 cm and identified the 1964 surface as the sediment interval where the peak 137Cs content occurred. Modern plants near the core site were identified, freeze- dried, and ground. Fossil plant fragments were extracted from core sediments using density separation {25) and analyzed for c515N and <534S. For organic N analyses, inorganic soluble N03~ and exchangeable NH4+ were extracted from bulk sediment subsamples with 2 M KC1. Residual organic N was dried and homogenized. Inorganic C was removed from separate bulk sediment subsamples using HC1 vaporization. <534Sorg was determined on separate bulk sediment subsamples after sequential extraction of acid volatile sulfur, chromium reducible sulfur, and elemental sulfur {26). N isotopes were analyzed at UC-Davis using a Europa Scientific Hydra 20/20 mass spectrometer. S isotopes were analyzed at the USGS in Menlo Park, CA using a Micromass Optima EA-IRMS. S isotopes of organic S and modern plants were analyzed in duplicate or triplicate. The limited plant fragment material recovered during density separation generally prohibited replication. The precision of isotopic measurements is ?0.13%o for (315N, and ?0.5%o for (334S. Isotope values are reported as parts per thousand (%o or permil) relative to standards of known composition using eq 1 <5 [%o] = (i?x/i?s-l)*1000 (1) where Rx and Rs are the ratios of the heavy to light isotope (e.g., 15N/ 14N, 34S/32S) in the sample and standard, respectively. N and S isotopes are reported relative to atmospheric N2 (Air) and Canon Diablo Troilite (COT), respectively. Local changes in human population densities were reconstructed from 1675 to 1993 using multiple sources. Local changes in housing density in the watershed were recon- structed using historic atlases (2 7), historic topographic maps {28, 29), and aerial photography {30). Population densities were estimated using census data for household size in Anne Arundel County from 1850 to 2000 {31). Prior to available census data, Earle's {32) reconstructed colonial land use history (1650-1783) of All Hallow's Parish, which encom- passes watershed 121, is used to approximate human and animal population densities in watershed 121. Although All Hallow's Parish (20 719 ha) is larger than watershed 121 (1192 ha), these data are unusually detailed and allow comparisons of land use change and sediment core characteristics prior to 1850. To estimate general trends in animal populations after 1783, census data for Anne Arundel County (1880,1900, 1940, and 1992) were used to calculate animal unit (AU) densities based on ref 33. Results Core Chronology and Sedimentation History. The 94 cm core (SERC2) taken from Mill Swamp spans approximately 350 years and, based on Munsell color, pollen counts, and texture, shows no indication of discontinuity. A total of 2116 pollen grains representing 36 plant types (genus, family, or species) were identified and counted. The pollen biomarkers used to date intervals and reconstruct sedimentation rates are shown in Figure 1 (see Supporting Information). Initial land clearance, indicated by the sharp decrease in the ratio of Quercus (oak) to Ambrosia (ragweed) pollen, is dated 1650 based on regional settlement history {32). Peak land clearance, indicated by the largest proportion of Ambrosia pollen, relative to other herbaceous and arboreal species, is dated as 1880 based on the highest percentage of improved farmland in Anne Arundel County reported by the U.S. Census Bureau. The decline of Castanea due to the American chestnut Blight is dated 1910 and is indicated by a sharp decrease in the concentration of Castanea pollen. The 1964 horizon is identified in the sediments at 7?8 cm where peak 137Cs concentrations occur, corresponding with peak atmospheric fallout in the northern hemisphere. Below 11 cm, 137Cs concentrations are below detection limits. The top of the core is dated as the year collected (2000). Sediment history was reconstructed using methods documented by Brush {24) (see Supporting Information). Error associated with sedimentation rate reconstructions was estimated using first-order propagation of error {34). These calculations indicate that error associated with sedimentation rate reconstructions are <0.04 cm yr_1. Large variations in sedimentation rates are apparent over the timespan of this core and range from 0.05 to 1.16 cm yr_1. Sedimentation rates increased nearly 15-fold from the mid- 1600s to the mid- 1800s due to deforestation, agricultural development, and subsequent erosion. Since the mid- 1800s, sedimentation rates decreased but are still 7-fold higher than the rates prior to settlement. These trends in sedimentation rates are similar to those reported by Brush {35) for cores taken throughout the Chesapeake Bay estuary. Isotope and Concentration Stratigraphies. Figure 2 illustrates the isotope and elemental concentration strati- graphies in the SERC2 core. In 24 measurements of (515Norg, values ranged from +1.7%o to +6.9%o. 615Norg values generally increase from the bottom of the sediment profile to the top. Plant fragments extracted from core sediments exhibit a similar trend as (515Norg. The (515N of plant fragments are on average 1.3%o lower than 615Norg. 634S values of organic S and plant fragments are generally <0%o and generally decrease from the bottom of the core to the surface. The average isotopic composition of 5 species of living plants {Sagittaria latifolia, Cinna arundinacea, Cornus amomum, Salix nigra, VOL. 40, NO. 9, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY ? 2911 2000^ 0 , 1970- 5 - 1940- 10 - 15 - 1910- 20 - 25 - 30 - 35 - 1880- "l850- ??40 - ^45- ?-50- ?55- 60 - 1820- 65 - 70 - 1790- 75 - 1760- 80 - 85 - 1730- 1650- 90 - 0.0 0.3 0.6 FIGURE 1. Reconstructed sedimentation rates, 137Cs concentration, profiles of pollen biomarkers used for dating sediments, and relative abundance of pollen from potential N-fixing plants. Depth from the surface [cm] is shown on the left axis with corresponding chronology. 2000, 0 n 1980- 5 - 1950- 10 - 1920- 15 - 20 - 25- 30- 1890- 35 - E40 - ~45 - 1860- 8-so- O55: 1830- 60 - 65 - 70 - 1800- 75 - 1770- 80- 1740- 85 - 1710- 16503 90 - r0 -5 - 10 - 15 -20 -25 -30 -35 -40 -45 -50 -55 -60 -65 -70 -75 -80 -85 -90 Percent N Percent C or TOC:TON S15N [%o] S34S [%.] FIGURE 2. Isotope and elemental stratigraphies in the SERC2 core. Depth from the surface [cm] is shown on the left axis with corresponding dates. Organic sediment fractions are shown as solid lines with solid circles for isotopic values (d,5Narg, d34Sarg) and hollow circles for concentration (percent organic N, organic S, organic C). Plant fragments extracted from core sediments are shown as dashed lines with a hollow triangle for isotopic values and a solid triangle for elemental concentrations. The mean isotopic value of living plants collected near the core site is indicated for dKN and dMS (n=5). The molar ratio of T0C:T0N is shown with hollow squares. Pileapumila) from the core site vicinity is +7.9%o and ?6.9%o for (515Norg and (534Sorg, respectively. Intrawetland Comparison. The effect of wetland spatial heterogeneity on the 615Norg stratigraphy was assessed using a second sediment core taken approximately 100 yards downstream of the SERC2 core. Sediment history in this downstream core was reconstructed using the same tech- niques as in SERC2 (137Cs and pollen biomarkers) to allow for comparisons of (515Norg in sediments of similar ages between the two cores. The (515Norg values in sediments of similar age are not significantly different before about 1950 (paired f-test, p=0.10). After about 1950, dl5N values of the downstream core are significantly different (paired f-test, p<0.05) and are an average of 0.55%o higher than d 15N values for equivalent dates in the upstream core. However, the small differences between the isotope stratigraphies in the two cores suggest that the stratigraphy of (515Norg in wetland sediments is recording a systematic change from low (315Norg values around the time of European settlement to higher (315Norg values in more recent sediments in both cores. Population Density Reconstruction. Local changes in human population densities were reconstructed for water- shed 121 from 1675 to 1993 and are shown in Table 1. Over this time period, population density in watershed 121 increased nearly 16-fold from 0.03 to 0.47 person ha-1. From 1675 to 1783, human population density steadily increased by approximately 6-fold. During the 1800s, a village was established upstream of the present wetland and resulted in 2912 ? ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 9, 2006 TABLE 1. Reconstructed Housing Density, Population Density, and Waste water Inputs to Watershed 121 housing density year (houses ha 1675 1685 1695 1705 1715 1725 1735 1745 1755 1765 1776 1783 1876 0.04 1904 0.05 1957 0.09 1972 0.11 1991 0.15 1993 0.15 persons population total wastewater N wastewater N per density watershed 121 input to watershed 121 load household3 [person ha 1]? population [kg N yr ']c [kg N/ha] 0.03 41 349 0.29 0.05 57 488 0.41 0.06 75 642 0.54 0.07 86 740 0.62 0.08 92 790 0.66 0.09 108 931 0.78 0.10 121 1038 0.87 0.12 141 1212 1.02 0.13 151 1301 1.09 0.15 174 1496 1.26 0.16 196 1686 1.41 0.17 201 1726 1.45 8.12 0.32 382 3285 2.76 5.02 0.26 306 2632 2.21 4.23 0.37 440 2464 2.07 4.23 0.47 560 3136 2.63 3.16 0.46 546 3058 2.57 3.16 0.47 565 3164 2.65 s Derived from Anne Arundel County, MD Census data from 1860, 1910, 1950, and 1990 U.S. Censuses. b Population density from 1675 to 1783 reconstructed using Earle, 1975. c Assumes 8.6 kg N person ^ from 1675 to 1950 and 5.6 kg N person1 from 1950 to 1993. the population density increase evident in 1876. From 1876 to 1993, the housing density in watershed 121 increased nearly 4-fold, while the housing occupancy rate decreased. The net result was a 1.5-fold increase in population density during 1876?1993. From these estimates, we estimate that the total watershed population increased from approximately 41 in 1675 to about 565 people in 1993. Discussion Historic Evidence for Source Change. Prior to major human N inputs, ecosystems generally obtained N via biological N fixation; this N source is isotopically similar to atmospheric N2 (~0-2%o). Hu et al. (18) document that (315N and pollen stratigraphies can record the presence of historic biological N-fixation in sediments. Here, we examined the presence of plant taxa known to have symbiotic relations with N-fixing bacteria in the palynological record, including Alnus, Legu- minosae, Myricaceae, Cornus, andRosaceae (10,36). Potential N-fixing plant taxa were more abundant in older sedi- ments (Figure 1), suggesting that historically more N may have been derived from this source. The percent faculta- tive N-fixers and <515Norg are inversely related (i?2=0.67, p<0.01). This suggests that lower (515Norg values in older sediments reflect the contribution of this fixed N or the derivation of N from this source after recycling through soil organic matter. Reconstructed population density in watershed 121 suggests that the trend of increasing d 15Norg reflects a change in the N source associated with historic land use change, as forested conditions gave way to initial land clearance (1650), peak land clearance (1880), increased population densities, and increasing N inputs derived from human waste. As shown in Table 1, we estimate that the net population in watershed 121 has increased by an order of magnitude from 1675 to 1993. This population increase would have contributed an increasing load of high <515NNo3- to the system over time from wastewater effluent. However, the relative amount of waste- derived N contributed by human populations likely varied with time according to sanitation practices. For example, during the mid-1800s, privy vault-cesspool systems were commonly used for waste disposal {37). Privy vault-cesspool systems are generally comprised of unlined holes in the ground from which waste flowed directly into the subsoil. Although septic tanks were introduced in the late 1800s {37), discharge of effluent into subsurface drain fields did not become common practice until the mid-1900s (38). It is estimated that the direction of septic tank effluent into subsurface drains and leaching fields reduces N losses by 35% (39). Given changes in sewage disposal practices and popula- tion, we reconstructed waste-derived N loads to watershed 121 from 1675 to 2000. For this analysis, wastewater N loads from 1950 to 1993 are assumed to be derived from septic tanks and contribute 5.6 kg N person ' yr-1. This loading rate was reported by Castro et al. (40) and was determined using wastewater treatment plant releases of N and associated populations served by sewers. The loading rate reported by Castro et al. falls within the range of directly measured septic tank effluent (3.6?7.8 kg N per capita annually) reported by the U.S. EPA (38). Prior to 1950, it is assumed that sanitation practices consisted of privy-vault cesspool systems with average N loadings that are 65% higher than those of septic tanks with drain fields (8.6 kg N person ' yr-1). Using these assumptions, we estimate that based on population growth in the watershed, wastewater N loads to watershed 121 increased by an order of magnitude between 1675 and 1993 from 0.29 to 2.65 kg N ha"1 yr"1. The temporal trends in reconstructed population density and wastewater N inputs are generally reflected in (515Norg. As shown in Figure 3A,B, population density and corre- sponding waste-derived N in the watershed increased steadily from 1650 to 1783. (515Norg values are ~ +2%o from 1650 to 1730 and begin to increase after about 1730 when population density exceeds 0.09 person ha '. Although accurate popula- tion densities are not available from the late 1700s to the mid- 1800s, a small village was established by 1876. The high population density in 1876, coupled with the higher oc- cupancy rate during this period (Table 1), results in the highest estimated loads of wastewater-derived N to watershed 121 as shown in Figure 3B. This period from 1730 to 1876 was accompanied by a steady increase in (315Norg from ~ +2 to ~ +5%o. The introduction of septic tanks (~1950) and a decrease in household size reduced estimated net inputs of wastewater N from 1910 to 1950. Post World War 2 subur- banization results in nearly a 30% increase in population density after 1957. After 1972, watershed population growth levels out. During the period from 1876 to 2000,615Norg values continue the generally increasing trend. VOL. 40, NO. 9, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY ? 2913 6 ? E 4 o 1600 1650 1700 1750 1800 1850 1900 1950 2000 FIGURE 3. A comparison of temporal changes in reconstructed population density, waste effluent N, approximate animal unit density, and i)KN0n) in watershed 121 from 1650 to 2000. (A) Human population density (solid triangles) and d15Narg (hollow circles). (B) Waste effluent N inputs (solid triangles) and <515NorB (hollow circles). (C) Animal unit densities for watershed 121 (solid triangles) and Anne Arundel County (solid squares). In addition to human waste inputs of N, livestock animals also likely contributed waste-derived N during the post- European settlement of watershed 121. As shown in Figure 3C, we estimate that during 1650?1783, animal unit (AU) densities in watershed 121 increased nearly 4-fold between 1660 and 1745 to 0.18 AU ha"1. U.S. Agricultural Census (1880?1992) data for Anne Arundel County show a similarly high AU density in 1880. After 1880, AU density in Anne Arundel County decreases to 0.04 AU ha \ which is similar to 1660 levels in watershed 121. While the large increase in AU density would have been accompanied by a correspond- ing increase in waste inputs, we find it interesting that (515Norg does not reflect this increase in AU density. The relatively diffuse distribution of animal waste inputs to the landscape (e.g., pasture or applied as fertilizer), coupled with above ground delivery, may account for the relative insensitivity of the (515Norg signal to AU density. This lack of correlation between (515Norg and AU density supports our argument that hydrologic connectivity is important in transporting human- derived waste effluent to downstream ecosystems. To compare population density and wastewater loads directly with (315Norg, we used a subset of (515Norg values that fell within 6 years of population density and wastewater estimates (?=10, mean difference = 3 years). In this subset of data, (315Norg values are strongly correlated with population density (i?2=0.91, p<0.001) and wastewater N inputs (i?2=0.86, p<0.001) for equivalent periods (Figure 4). These land use history reconstructions suggest that increasing contributions from waste sources are responsible for the increasing dl5N values in the core. Other potential N sources to the system, including industrial fertilizer and atmospheric deposition are ruled out based on isotopic values for these sources, which would generally not result in increasing (315N values with increasing inputs. Other con- temporary studies have shown that (515N values in biota are correlated with wastewater N inputs (e.g., refs 6 and 41). In particular, Mayer et al. reported higher d15N values (> +6.5%o) in riverine nitrate from 16 large watersheds in the Eastern U.S. with wastewater contributions >1.5 kg ha ' (5). In watershed 121, this threshold (1.5 kg ha-1) of wastewater contributions was crossed between 1783 and 1876, although (315Norg values began increasing before 1783. In this same study, Mayer et al. also report correlations between popula- tion density and (515NNCB- where population densities ranged from 0.08 to 5.6 person ha ^ (5). In watershed 121, we observe this same relationship, with stronger correlations, at a much lower range of population densities (Figure 4). This suggests that examination of long-term changes in a single watershed document the contributions of wastewater N at lower population densities than possible when comparing a cross- section of watersheds. Other Factors that May Contribute to Stratigraphic Changes in ?15Norg. The strong correlations we observe between d 15Norg, human population density, and wastewater N inputs to watershed 121 suggest the stratigraphic trends we observe result from increasing input of waste-derived (315N characterized by high dl5N values. However, several additional factors must also be considered. As in any paleoecological study, it is important to recognize the potential for postdepositional or diagenetic alterations of sediments. Sediment and soil profiles affected by diagenetic processes generally have increasing dl5N values with depth and decreasing N concentrations and C:N ratios with depth {42, 43). In this study, we observed decreasing (515Norg and slight increases in the percent organic N, organic C, organic S, and TOC:TON with depth. Although we cannot rule out the influence of diagenetic alterations, the stratigraphic trends we observed in sediment cores for this study are generally 2914 ? ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 9, 2006 7 - A R2 = 0.91,p<0.001 y% B R2 = 0.86, p<0.001 % 6 * /* 10%, Quercusi'Ambrosia <1. 129 0.47 ? 0.024 1880-1877 48-49 111 0.88 1859-1857 50-51 225 0.43 1855-1851 60-61 84 1.16 1832-1830 68-69 125 0.78 1819-1816 70-71 260 0.37 1814-1809 78-79 264 0.37 1779-1774 80-81 243 0.40 1770-1766 88-89 311 0.31 1734-1729 90-91 1720 (1710-1740) Ambrosia <10%, Quercusi Ambrosia >1. 384 0.06 ? 0.002 1720-1699 91-92 331 0.09 1699-1686 93-94 477 0.06 1670-1650 94-95 1680 (1620-1680) Ambrosia < 1%, Quercusi Ambrosia >10. 438 0.08 ? 0.003 1650-1634 Elliott and Brush S2 Environmental Science and Technology Calculation of sedimentation rates: a,_,=?x]f (i) "i-j where /?,-._,- is the sedimentation rate for interval i toy; n is the average number of pollen grains per unit area in the interval; rit-j is the number of pollen grains per unit area in the interval i toy; R is the average sedimentation rate ( ? ) where d is the depth of the dated horizon and t is the time in years. Between horizons where pollen was counted, sedimentation rates were linearly interpolated. Between dated horizons, offsets in chronology from linear interpolation of sedimentation rates were normalized by the ratio of years between dated horizons and estimated years in between dated horizons reconstructed from sedimentation rates. Error associated with sedimentation rate calculations are estimated using first order propagation of error from Benjamin and Cornell (1970) (see reference #34 in main text). This calculation considers the ranges of dates for each horizon as indicated in the table. Elliott and Brush S3