The Age of Buried Carbon Changes the Greenhouse Gas Budget of a Dam

Dams are a globally relevant source of greenhouse gases (GHG), which impairs their function as a source of green energy. High burial rates of organic carbon (OC) in dam sediments may partly or fully offset the emissions. We argue that only the burial of carbon fixed in the timespan of dam operation changes the GHG balance. Here, we took sediment cores from a temperate dam. We analyzed radiocarbon age and OC molecular composition by laser desorption ionization mass spectrometry in the bulk OC and in four chemical extract fractions. The bulk samples contained modern OC, fixed after 1950. However, the extracted OC was of different ages (modern to 1900 years BP). Fractions with OC measured as old (>960 years BP) accounted for 57% of total sediment OC. Correlations of molecular composition with extract age suggest that these older fractions contained insignificant amounts of modern OC. We conclude that a substantial proportion of buried carbon did not originate from the contemporary atmosphere and cannot be offset against recent GHG emissions.


Introduction
For two decades, the greenhouse gas (GHG) balance of inland waters, and of dams in particular, has been controversially debated.The savings of fossil carbon for the generation of green electricity are potentially offset by high CH 4 and CO 2 emissions to the extent of about 60 Tg CO 2 -C yr 1 globally (Keller et al., 2021).Dams are important sedimentation sites, favored by high suspended matter loads, especially in mountainous catchments, and by high internal organic carbon (OC) production with sufficient residence time and availability of mineral nutrients (Clow et al., 2015;Lewis et al., 2013;Sobek et al., 2012).Estimates of OC burial in dams span a wide range and remain uncertain.Currently, we can assume a burial of 60 (20-110) Tg C yr 1 , which together with lakes (90 Tg C yr 1 ) is as high as the burial in oceans (Maavara et al., 2017;Mendonça et al., 2017).In dams, the OC burial is offset against the emissions in the GHG balance (Knoll et al., 2013;Li et al., 2015).
There has long been discussion as to whether OC transported from the continents to the oceans and permanently stored in ocean basins is recent in origin or aged.Only the transport of recently fixed carbon to the coastal ocean and its permanent sequestration lowers the CO 2 concentration of the contemporary atmosphere (Gomez et al., 2010;Hilton et al., 2012).Surprisingly, the issue of OC age has not yet been discussed in relation to the GHG budgets of dams, complicated by the fact that only a few age determinations of sedimentary OC from dams exist.It is therefore unclear whether OC burial can be considered a sink of current atmospheric CO 2 and offset against GHG emissions.
To be included in the carbon budget of a dam, the buried carbon must have been fixed during the operating lifetime of the dam, that is, it must carry a modern radiocarbon signal.Modern samples contain 14 C released by nuclear tests after 1950 A. D. and have percent modern carbon (PMC) values >100.Dam-internal (autochthonous) OC production can decrease the CO 2 level in the recent atmosphere by fixing modern inorganic carbon (IC) that invaded from the atmosphere into the water.Autochthonous photosynthesis could also refix modern carbon previously released by land plant decomposition in the soil and transported to the dam by groundwater or surface water.In contrast, fixation of IC from 14 C-aged sources (<100 PMC) such as carbonate rock or older soil OC in the dam represents a recent process, but it is based on carbon removed from the atmosphere prior to operation and is therefore not relevant to the GHG budget.Burial of terrestrial-fixed OC can improve the GHG budget if it carries a modern 14 C signal.
We studied the age of buried OC in a small temperate dam.Since the sedimentary OC can originate from several sources (e.g., leaves from the shoreline, soil, OC fixed in the dam, sedimentary rocks), one must assume that it does not have a uniform age but consists of different age fractions.Their composition and 14 C signal determine what portion of the total OC can be considered modern and be included in the GHG balance.In order to gain information about the age distribution and sources of the sedimentary OC, we devised a four-step sequential extraction based on previously developed single extractions and analyzed radiocarbon in the total and extracted fractions of OC.We combined radiocarbon age with molecular composition measurements by ultrahigh resolution mass spectrometry after laser desorption ionization (LDI-FT-ICR-MS) to estimate the proportion of modern OC.

Settings
The studied dam is situated in the lower Harz mountains, Germany, at an altitude of 439 m a.s.l.(Table S1 in Supporting Information S1).It has a small surface area of 0.22 km 2 , a maximum depth of 17 m and a relatively low residence time of 52 days (Figure 1).It is part of the larger Rappbode reservoir system and referred to as Rappbode predam, intended to reduce the sediment and phosphorus load into the larger Rappbode main reservoir downstream used for drinking water production.The water discharges over the top of the dam, thereby keeping water level constant.There is only one major inflow and no dam upstream.Groundwater inflows can be neglected (Barth et al., 2017).The catchment area is 48 km 2 , 72% of which is covered by forest (Norway spruce), and 24% by agriculture (predominantly grassland) (Friese et al., 2014;Rinke et al., 2013).The bedrock within the catchment area consists mainly of sedimentary rocks such as clay-mineral rich mudstones (clay shale, greywacke, Silurian to Lower Carboniferous), which potentially contain 14 C-dead petrogenic OC (OC petro ).For a map of the geology of the catchment see (Friese et al., 2014).Operation started in 1961.The dam is moderately euthrophied (TP 23 μg L 1 ); it stratifies in summer, developing an anoxic layer in the deeper parts near the dam below ∼11 m depth.High rates of FeIII reduction in the sediment support the accumulation of dissolved Fe in the hypolimnion during stratification (Wendt-Potthoff et al., 2014).The surface layer is almost always supersaturated with CO 2 compared to the atmosphere and a source of CO 2 (Halbedel & Koschorreck, 2013).

Sampling
Four sediment cores were taken in October 2021, two cores from a station near the dam at maximum depth of 17 m (cores D1, D2) and two cores from a central station at 10 m water depth (C1, C2, Figure 1).A gravity corer (UWITEC, Mondsee, Austria) was used, equipped with clear PVC tubes of 6 cm diameter.The sediment cores were 31-39 cm long.There were no horizontal layers and no soil of the original valley bottom recognizable.The sediment thickness averaged at 45 cm (see below).Our cores, therefore, represent the largest part of the sediment that had accumulated after impoundment.Hence, they also include OC from after 1960, when particularly high amounts of 14 C-enriched CO 2 were present in the atmosphere due to nuclear weapons testing (Hua et al., 2022).
To estimate the amount of OC stored in the sediment, we measured the thickness of the sediment at 16 locations covering the reservoir area (Figure 1).We used a 2 m long metal lance (threaded rod 10 mm), which was weighted in the upper quarter with a 5 kg weight plate.The lower part of the lance was covered with white, self-adhesive Velcro tape, from which we could read the depth of penetration into the sediment.After each measurement, the sediment could be rinsed off.To test that the chosen weighting was sufficient to penetrate to the original valley floor, we had previously used a higher weighting of 10 kg, which gave similar results.With the metal lance we hit the partially stony ground, even in the places with thicker sediment.The depth of the sediment ranged between 21 and 77 cm and averaged 45 cm.There were no systematic differences in the sediment layer thickness between the near-inlet and the deeper near-outflow areas.OC accumulation was calculated based on the sediment depth, water content (78%-84%) and OC content (67-76 mg C g 1 ) measured in the four sampled cores (C1, C2, D1, D2) and at five additional sites that together cover the longitudinal extent of the dam (Figure 1, Table S2 in Supporting Information S1).

Sample Processing
In the laboratory, all core sediment was gently and thoroughly homogenized using a slow-moving stirrer.Subsamples of ca.400 g fresh weight were freeze dried, pestled and stored in a desiccator.The sequential extraction procedure developed for this study was applied to the samples of the second cores (C2, D2).The protocol is described below.Briefly, five sediment OC fractions were operationally defined: loosely adsorbed and exchangeable OC mobilized by ionic strength increase (IST) in a 1 M MgCl 2 solution, OC bound to reactive Fe and released by Fe reduction (RED) with sodium dithionite, the remaining acid hydrolyzable low-molecular weight fulvic acids (ACD) removed by 1% HCl, the remaining fulvic and humic acids and Al-bound OC dissolved in alkaline solution (ALK) of 1% NaOH, and the residual OC (RES) largely consisting of humic compounds.Using comparatively low concentrated HCl and NaOH, we aimed to characterize the more labile OC in contrast to the chemically stable and presumably microbially less available residual OC (Lukkari et al., 2007).The procedure was carried out in triplicates.
Blanks were separately processed and analyzed after combustion of bulk sediment from cores C2 and D2 for 4 hr at 550°C.We measured remaining OC concentrations of 0.6 mg g 1 in the bulk sediment and of 0.1-0.2mg g 1 in the sediment after the IST, ACD and ALK extraction (Table S3 in Supporting Information S1).We attribute the presence of small amounts of OC not to contamination but to the incomplete removal of OC in our blanks, which were combusted at a lower temperature than in the later analysis (550°C vs. 950°C).To derive our blanks, we have chosen a temperature that is not too high in order to minimize changes to the sediment matrix.The blanks measured after the RED step indicate that 0.2-0.5 mg g 1 of OC remained from citrate, which was added during extraction and not completely removed during the three subsequent washes.The residual citrate represented 1.4%-2.3% of the total OC extracted during the RED step and was considered not significant.Our blank values were comparable to those measured by Faust et al. (2023) who estimated OC values of 0.1-0.2mg g 1 after dithionite extraction with carbon-free silica sand.
The OC extracted by IST was quantitatively small.Due to the low OC and high salt contents in the extracts, we could only determine the age of this fraction for core D2 and with larger uncertainty.The carbon isotope composition of the RED fraction was estimated by a mass balance (see below).In the mass balance of core C2, we assumed the same mean value for the IST fraction as in core D2 and a large error range (Table S3 in Supporting Information S1).To test an extreme scenario, we varied the missing 14 C-OC value of the C2 IST fraction over a theoretically possible but unlikely wide range from 28 PMC (corresponding to OC produced after last glaciation 10 kyrs BP, that is before 1950) to 160 PMC (atmospheric 14 C-CO 2 surpassed only in years [1963][1964][1965][1966][1967].We calculated corresponding 14 C values of the C2 RED fraction between 100 and 128 PMC, representing modern OC.

Sequential OC Extraction Protocol
This describes a chemical extraction of OC from sediment by modification of ionic strength, reducing force and pH.Four known extraction procedures were combined to a sequential protocol.The operationally defined fractions include (a) loosely adsorbed and exchangeable OC, removed by an increase of ionic strength with 1 M MgCl 2 as used to extract adsorbed Fe (Poulton & Canfield, 2005).We refer to this OC as the IST fraction.(b) OC associated to redox-sensitive hydrated oxides, mainly Fe (oxihydr)oxides referred to as reactive iron (Fe R ) that are removed by the citrate-bicarbonate-dithionite extraction (RED) (Lalonde et al., 2012).Compared to earlier studies, we use a higher amount of dithionite (2 mg per mg dried sediment) and repeat the extraction in order to increase the extraction efficiency (Fisher et al., 2021).A control experiment, typically used to quantify the amount of OC released by the ionic strength of the added chemicals, is not carried out.This OC was already removed in the first step of our protocol.(c) remaining acid-soluble OC including remaining Fe R associated OC dissolved by HCl (1%) hydrolysis (ACD) and mainly consisting of low molecular weight fulvic acids (Abbott & Stafford, 1996;Grootes et al., 2004).(d) Remaining fulvic acids and humic acids, dissolved in alkaline solution of 1% NaOH (ALK) (Abbott & Stafford, 1996;Grootes et al., 2004).The residual sediment OC (RES) largely consists of humic compounds (Abbott & Stafford, 1996).The amounts of extracted OC and Fe are estimated from dissolved phases, except the RED step.Here, OC is calculated from sediment contents before and after extraction (see below) as a high amount of OC is added to the dissolved phase with the citrate.
The procedure starts with an inspection of the sediment.Worms, wood and needles are removed.The sediment is gently and thoroughly homogenized, freeze-dried, grinded to a fine powder and stored in a desiccator.Triplicate subsamples for total OC, N, and Fe as well as single samples for carbon isotope measurement ( 13 C, 14 C) are derived.
IST: Three replicates of 0.5 g sediment (read three decimal places) are transferred into 50 mL centrifuge vials and extracted with 30 mL 1 M MgCl 2 .The pH is adjusted to 7 with 5 N NaOH and checked again after 15 min.The samples are shaken overhead for two hours and centrifuged (15 min, 4,750 rpm).The extracts are weighed to be able to accurately calculate their volumes and then filtered (Whatman GF/F, ∼0.6 μm pore size, pre-combusted 450°C, 4 hr) to remove residual particles before analysis.Subsamples for OC and Fe quantification were collected from parallels and immediately stabilized with HCl (37%, 20 μL/mL) or HNO 3 (65%, 20 μL/mL), respectively.We note that the IST extracts are hygroscopic and difficult to freeze-dry because of their high salt content.The low OC contents after extraction and subsequent preparation for isotopic analysis resulted in high error ranges of the 14 C results of the IST fraction (see Table S3 in Supporting Information S1).The sediment residues are washed three times by centrifugation with high-purity demineralized water.

RED:
The sediment is subjected to a dithionite extraction.The residues from the previous step are mixed with 30 mL 0.27 M trisodium citrate plus 0.11 M sodium bicarbonate and heated to 80°C.1.0 g sodium dithionite is added.The dithionite and the samples during extraction must be well protected from light.The temperature is maintained for 15 min, shaking at least every 5 min.After centrifugation, the extracts are filtered (GF/F) and stabilized (HNO 3 ) to prevent Fe precipitation.With the sediment residues, the dithionite extraction is repeated.Before, the sediment must be thoroughly detached from the vial and homogenized using a vortex mixer and by repeated suction with a pipette.The extracts of both steps are now combined, weighed and stored for Fe quantification.To ensure that added citrate did not contaminate the samples (see Section 2.3), the sediment residues are washed three times.After resuspension in a few milliliters of demineralized water, about 35% of each residue is removed (hereafter referred to as abstracted sediment), dried (60°C overnight), weighed and stored in a desiccator for analysis of OC and C isotopes.

ACD:
The sediment residues from the previous step are suspended in 15 mL HCl (1%).The samples are heated up to 60°C and the temperature is maintained for 4 hr, shaking every 15 min.After 30 min, the pH is checked and adjusted to 0.5 with HCl (37%).The extracts are removed after centrifugation, weighed and filtered.Subsamples for OC quantification are collected and stored at pH ≤ 2.5.The remaining extracts are first neutralized (NaOH 5 N) before additional subsamples are taken, which are immediately re-acidified using HNO 3 to ensure a consistent chemical matrix of Fe measurements in all fractions.The rest of the extracts is combined, acidified (H 3 PO 4 85%, pH < 2.5) and freeze-dried for C isotope analysis.The reacted solid phases are washed at least three times until pH ≥ 4, giving one hour between washes.Shake several times.

ALK:
To release the alkali-extractable OC, the acid residues are subjected to a NaOH (1%, 15 mL) extraction by repeating the procedure described in the previous ACD step.The extracts should be analyzed timely in order to avoid precipitation.The sediment residues are washed until pH ≤ 9.The sediment that finally resisted the four steps is defined as the RES fraction and can be dried and stored for OC, Fe and C isotope analysis.
The extracted OC contents are calculated from concentrations in extracts and extract volumes, except the RED and RES fractions.A density correction of 1.06 can be applied to convert the weight of IST extracts into volumes, while no correction was necessary for other extractions.The OC content in the RES fraction is measured directly in the residual sediment.For the RED fraction, OC is estimated as the difference of the specific OC (mg OC g 1 ) of the bulk sample minus the specific OC of the IST fraction minus the specific OC after the reduction step as measured in the abstracted sediment (see above).The Fe contents are calculated from extract concentration, except the RES fraction.Note that the specific OC and Fe contents of the ACD and ALK fractions must be calculated with respect to a lower amount of extracted sediment (initial weight minus weight of the abstracted sediment).The carbon isotope composition of the RED fraction is estimated on the basis of a mass balance C BULK -specific OC (mg g 1 ) in start OC before extraction I BULK -isotope composition (PMC or δ 13 OC, ‰) of start OC C IST -specific OC (mg g 1 ) of IST fraction I IST -isotope composition (PMC or δ 13 OC, ‰) of IST fraction C RED -specific OC (mg g 1 ) of RED fraction I RED -isotope composition (PMC or δ 13 OC, ‰) of RED fraction C ABS -specific OC (mg g 1 ) after the RED step as measured in the abstracted sediment I ABS -isotope composition (PMC or δ 13 OC, ‰) after the RED step measured in the abstracted sediment
The  Poland).The radiocarbon results refer to the oxalic acid II standard and were corrected for process and instrument blanks and for fractionation (Stuiver & Polach, 1977).The water contents were calculated from the difference of the fresh weight and dry weight determined after drying at 60°C overnight and storage in a desiccator.
The error ranges of burial rates were calculated on the basis of t-statistics and the additive and multiplicative error propagation law (Table S2 in Supporting Information S1).

Molecular-Level Organic Carbon Analyses
Freeze-dried sediment samples were measured by laser desorption/ionization Fourier transform ion cyclotron resonance mass spectrometry (LDI-FT-ICR-MS; Giannopoulos et al., 2021;Solihat et al., 2018;Zuber et al., 2020).Suwannee River Fulvic Acid (SRFA-III, ID 3S101F) was obtained from the International Humic Substances Society (IHSS) and used as a reference sample to monitor instrument performance.More details are provided in Text S1 of Supporting Information S1.Sediment samples of cores C1, C2, D1 and D2, and remaining sediment from sequential extractions of core C2 and D2 were used as received (freeze-dried).An amount of 50 μg of sample was mixed with 250 μl of ultrapure water (MilliQ Integral 5, Merck, Darmstadt, Germany) in an acidwashed micro test tube (Th.Geyer, Renningen, Germany), yielding a suspension with 16.7% sediment dry weight.
The mixture was shaken thoroughly, vortexed for 15 s, and 5 μl of sample were immediately spotted onto a ground steel target (MTP 384 target plate ground steel BC, Bruker Daltonics Inc., Billerica, MA, USA) and left to air-dry under a fume hood.Each sample was spotted four times to allow replicate measurements.
We here used a dual source ESI/MALDI-FT-ICR mass spectrometer equipped with a dynamically harmonized analyzer cell (solariX XR, Bruker Daltonics Inc., Billerica, MA, USA) and a 12 T cooled actively shielded superconducting magnet (Bruker Biospin, Wissembourg, France).The instrument was initially calibrated using SRFA.Laser desorption mass spectra were acquired within 1 day in negative ionization, broadband mode (mass range 147-1,000 m/z, Text S1 in Supporting Information S1).We used Compass DataAnalysis 5.0 (Bruker Daltonics Inc., Billerica, MA, USA) for processing of raw mass spectra.After peak picking at a signal-to-noise ratio of 4, mass spectra were internally recalibrated with a list of commonly found signals in SRFA and soils that were originally recalibrated based on iron hydroxide clusters (i.e., measurements of pure goethite and goethite exposed to SRFA).Calibrant masses with errors >|0.2| ppm were removed, and mass accuracy after linear recalibration was <0.1 rmse.The recalibrated mass lists were then exported and further processed with an inhouse software.
After processing (Text S1 in Supporting Information S1), 23,083 MFs remained in the data set (10,005 in SRFA spectra).For comparison of compositional trends, each spectrum's intensity scale was base peak-normalized (highest peak = 100).Difference spectra of sediment samples before and after extraction steps were calculated using this base-peak-normalized, averaged data, following the general formula: Pearson correlations with PMC data were obtained using modern (IST, ALK) and 14 C-old endmembers (ACD, RES) from each core C2 and D2 (n = 4 for each endmember).Fraction IST of C2 was assumed as 100 PMC based on bulk 14 C measurements.We conducted two-sided t-test and extracted the formulas with significant negative or positive correlation (p ≤ 0.05), less significant correlation (0.05 > p ≤ 0.25), or no correlation (p > 0.25), and visualized them in van Krevelen plots (Hydrogen-to-Carbon ratio vs. Oxygen-to-Carbon ratio of each MF) against all signals found across all samples as well as their absolute number and share of ion abundance in each fraction also BULK and RED (which were excluded from the initial correlation analysis) (Figure S6 in Supporting Information S1).Overall similarity in molecular composition was analyzed by Principal Coordinates Analysis (Text S1 in Supporting Information S1; Osterholz et al., 2016).

Radiocarbon in Bulk Organic Carbon and Extract Fractions
The bulk OC of all four cores contained modern carbon (>100 PMC, Table 1).In the extracted fractions, however, the OC was of different ages.The reductive-soluble carbon (RED) and the alkali extractable fractions (ALK) included modern OC (Figure 2), However, the OC in the acid hydrolyzable (ACD) and in the residual fraction (RES) was aged (<100 PMC).The two aged OC fractions (ACD, RES) contained carbon that was fixed from the atmosphere >964 years BP.They comprised 59% and 56% (57 ± 1%, mean ± range) of total sediment OC in cores C2 and D2, respectively (Figure 2).It could be concluded that more than half of the sediment OC was of ancient origin and was fixed before the dam was constructed.However, even within each extract fraction, the age distribution can be inhomogeneous.It is conceivable that rock-derived OC petro that does not contain 14 C (0 PMC) or ancient OC from deeper soil layers conceals substantial amounts of modern OC in extract fractions with <100 PMC considered as "old."OC petro is typically significant in steep catchments whose erosive streams transport large amounts of  S3 in Supporting Information S1, age of total inorganic carbon see text.
suspended sediment (Hilton et al., 2010).This does not apply to the Rappbode catchment studied here, but its bedrock consists of sedimentary rocks which potentially contain OC petro (see Methods 2).Therefore, from 14 C measurements it cannot be ruled out that OC age distribution within the sediment extract fractions was inhomogeneous and that rock-derived OC petro or ancient soil OC concealed modern OC in the "old" ACD and RES fractions.The proportion of old carbon in the sediment OC would then be lower than estimated above based on the percentage of old extract fractions.

Molecular Composition and Age
To obtain compositional information on the extract fractions, we analyzed the particulate OC with LDI-FT-ICR-MS (Text S1 in Supporting Information S1, spectra and sample compositions shown in Figures S1-S3 of Supporting Information S1).In general, the LDI total signal intensity was highly linear and paralleled the amount of extracted OC, with RED and ALK mobilizing highest amounts of OC and corresponding highest signal intensity and molecular formula (MF) numbers, which was also apparent visually in mass spectra (Figures S1 and S5 in Supporting Information S1).Notably, OC removal by IST led to improved detection of molecular signatures.
RED extracts indicated a surprising diversity in terms of molecular formula (MF) numbers and bandwidth of compounds present (in line with e.g., Coward, Ohno & Plante et al., 2018;Coward, Ohno & Sparks et al., 2018).ALK was composed mainly of aromatics with substantial contribution of O-and N-containing MFs (Figure S5 in Supporting Information S1).Low extractability of old OC (ACD) coincided with lower desorption/ionization yield (RES), and low molecular weight and diversity in both fractions (MF number, Figure S5 in Supporting Information S1).Compositionally, older fractions showed highest abundances of MFs associated to pyrogenic C (CH, CHN 1,2 and CHO 1 groups, Figures S2, S3 in Supporting Information S1; Solihat et al., 2019Solihat et al., , 2022)), thereby providing an explanation for their apparent persistence (Glaser & Knorr, 2008).
Molecular compositions of extracted sediments indicated significant differences in molecular composition of modern and the two oldest OC fractions (Figure S1 in Supporting Information S1).MFs that correlated significantly (p ≤ 0.05) or weakly (p ≤ 0.25) with extract fraction age appeared at low H/C and low O/C ratios (Figure 3, examples of correlations in Figure S6 in Supporting Information S1).Consequently, MFs correlated with 14 C-old extract fractions (RES, ACD) had by tendency higher H/O ratios (>2.5) than those correlated with 14 C-modern extract fractions (IST, ALK, Figure S7 in Supporting Information S1).
The statistical assignment already implies that modern fractions (IST, ALK) comprised more modern-correlated MFs than old fractions (ACD, RES) and vice versa (Figure 3).Of particular interest, however, is the different homogeneity of the composition of the extract fractions, which was not imposed by the chosen statistics.We find that the two old fractions (ACD, RES) contained predominantly MFs that were less abundant in modern fractions and only marginal proportions of modern-correlated MFs, in terms of ion abundance and number.In contrast, the modern and near-modern fractions RED and IST were more inhomogeneous, same as the unfractionated sample (BULK); they also included old-correlated MFs.The modern ALK fraction, however, was surprisingly homogeneous, given that a major portion of younger OC was already extracted by the prior RED extraction step, and the high amount of OC removed through ALK itself.This suggests that the alkaline extraction targeted a very specific, mostly modern fraction of OC.It can be concluded that the two old fractions ACD and RES and the modern fraction ALK were homogeneously composed, while the two modern and near-modern fractions RED and IST represented mixture of young and old OC.

Proportion of Aged Carbon to Sedimentary OC
The LDI-FT-ICR-MS results provided no indication that the two old fractions contained significant proportions of modern OC and in contrast confirmed that the 14 C-young fractions likely included 14 C-old material.Based on the share of 14 C-negative fractions in the sedimentary OC (ACD, RES), it can be estimated conservatively that 57% ± 1% of the stored carbon was of ancient origin (mean ± range of extractions from two cores).We deem our estimate conservative because, the LDI-FT-ICR-MS results indicate contributions of old OC in extracts classified as "modern" (RED, IST).Nuclear weapons testing has enriched 14 C in the atmosphere and therefore, a small amount of modern carbon can mask the signal of aged carbon in these samples.We conclude that a significant proportion of the sedimentary OC was fixed before dam construction.Taking into account the inherent uncertainties due to the inhomogeneous age distribution, no more than 40% of sedimentary OC should be offset against GHG emissions.

Sources of Sediment OC
Autochthonous production represents a potentially significant source of sedimentary OC.Emergent macrophytes rely on atmospheric CO 2 but were not significant in the dam studied.Algae and cyanobacteria predominated the photosynthetic biomass, their 14 C signal is identical to that of the dam total inorganic carbon (TIC), since radiocarbon values are corrected for fractionation (Stuiver & Polach, 1977).In the studied dam, 14 C-TIC was extensively measured in 2012-2013.We derived an annual average of 97.8 ± 0.4 PMC for the TIC of the dam photosynthetic layer and a similar value of 97.3 ± 0.3 PMC for the TIC of stream inflows (Tittel et al., 2015(Tittel et al., , 2019)).These values correspond to 14 C ages between 180 and 220 years BP.The 14 C signal of autochthonous production cannot be specifically assigned to any extract fraction (Figure 2a).
Unlike algal OC, which carries a uniform 14 C signal, terrestrial OC consists of a mix of compounds from different sources and of different ages.The inflowing stream delivered particulate OC that was found aged in six occasions (87.6-99.6PMC) but modern in one occasion (105.0PMC) during the annual cycle of 2012-2013 (Tittel et al., 2015).Modern sedimentary OC, as found in RED and ALK extracts, thus likely originated from the catchment, not autochthonous production.The sources of this recent OC include woody material or grass-derived vegetation such as needles and leaves of trees along the shoreline and OC of the top soil litter layer transported by overland flow.A previous study in the same dam revealed a redox-dependent adsorption/desorption behavior of dissolved OC on sediment Fe minerals.Under oxic conditions, unsaturated compounds with low H/C and high O/ C ratios (i.e., low H/O) bind on freshly formed Fe hydroxides (Dadi et al., 2017).It is therefore possible that modern OC extracted in the RED fraction was not associated to minerals from catchment soils, but only later in the dam, and the low H/O ratios of modern OC support this (Figure 3a).
Aged OC, with a 14 C age of ≥960 years, was found in ACD and in the RES fractions.The acid-extractable OC (ACD) constituted the quantitatively smallest fraction.Low OC/Fe ratios in ACD extracts (1.1-1.7,Table S3 in Supporting Information S1) were consistent with binding of OC to Fe hydroxides, yet acid also extracted significant amounts of Al (∼60% of total extracted Al).The residual fraction (RES) still comprised 55% ± 1% of bulk sediment OC (mean ± range of two cores).If we had used more concentrated HCl or NaOH, we would have retained less residual OC (Liu et al., 2022).The age and composition of the ACD and RES fractions suggests that refractory organic compounds and possibly OC petro from the catchment soils were buried in the dam sediment.

Greenhouse Gas Budget
The CH 4 emissions of the dam are low compared to other dams with 2.2 ± 0.9 g CH 4 m 2 yr 1 (mean ± SD of annual averages of four sub-areas of the dam, Table S4 in Supporting Information S1).Assuming a global warming potential of methane of 27.9 (IPCC, 2022), the CH 4 emissions correspond to GHG emissions of 17 ± 7 g CO 2 -C m 2 yr 1 .Beyond that we could add the CO 2 emissions of the dam, which were measured at a rate of 88 g CO 2 -C m 2 yr 1 (Tittel et al., 2019).It must be qualified, however, that these emissions likely include CO 2 that the stream would emit if the dam were not in place.Hence, we only consider emissions in an amount that corresponds to the CO 2 release of the dam sediment that has formed on the bottom of the valley after impoundment (38 ± 4 g CO 2 -C m 2 yr 1 ) (Wendt-Potthoff et al., 2014).The flux was estimated from accumulation rates in the anoxic hypolimnion during summer.It is an integrative and robust method, the error results from the combination of the errors of the analytics and the hypolimnion volume as well as the sediment surface and is most probably ≤10% according to the authors' personal information.Coupled with the CH 4 emissions, a warming potential of 55 ± 11 g CO 2 -C m 2 yr 1 can be estimated.
To compare emissions with the OC-burial, we estimated OC accumulation in the sediment using additional sediment cores taken over the longitudinal extent of the dam (Figure 1).We calculated a burial rate of 98 ± 7 g C m 2 yr 1 (mean ± 95% error range, Table S2 in Supporting Information S1).In quantitative terms, burial exceeds GHG emissions (55 ± 11 g CO 2 -C m 2 yr 1 ), which would make the dam a net carbon sink of the atmosphere.However, assuming that only 43% of the buried OC (42 ± 6 g CO 2 -C m 2 yr 1 ) can be accounted for in the GHG balance according to the proportion of modern fractions, the amount of buried modern OC would not exceed the emissions, but would be approximately equal to them.Our data show that sediments can contain a considerable amount of OC, which was fixed in the catchment long before the dam was constructed and therefore cannot be included in the GHG balance.
Even if the burial rate or the emissions could be calculated more accurately, the final GHG balance would remain uncertain and difficult to constrain.It depends on how much GHG's would be released into the atmosphere if the dam were not present (Prairie et al., 2018).If OC were not reoxidized but alternatively stored in the catchment, burial in the dam could not be included in the GHG balance.We only considered the modern portion of the sedimentary OC in the GHG balance.Thus, we assumed that the modern OC would be mineralized back to CO 2 and emitted in downstream areas such as wetlands, estuaries or coastal oceans, while aged sedimentary OC would continue to resist oxidation.Earlier studies revealed a limited postdepositional degradation and millennial-scale storage times of terrestrial OC in river and lake sediments (Chmiel et al., 2015;Torres et al., 2020;Wan et al., 2019).This complexity is not taken into account in our analysis and remains a general source of uncertainty in the assessment of anthropogenic changes in greenhouse gas balances.

Questions for Future Research
To date, very little radiocarbon data is available from dam sediments.Significantly aged OC was found near a check dam at the Chinese loess plateau and in a dam in Taiwan (Zeng et al., 2020;Zheng et al., 2020).In a preliminary study of sediments from three German dams, we measured the 14 C of residual OC comparable with our RES fraction (Table S5 in Supporting Information S1).We found aged OC, fixed between 99 and 214 years BP, which was in line with the results from Rappbode reservoir.More data from sequential extracts could also help to better constrain the molecular imprints of aged or modern OC.As shown in our study, aspects of fraction homogeneity can be addressed by LDI-FT-ICR-MS, and may therefore help to reduce uncertainty.Ideally, LDI-FT-ICR-MS would be combined with other, complementary solid-state tools, such as FT-IR or NMR spectroscopy, to improve these estimates.Our data show that the sediment from dams can contain a significant amount of OC that was fixed before the dam was constructed and therefore cannot be offset against GHG emissions.
It is commonly assumed that reductively mobilizable (Fe-bound) OC represents long-term stable OC (Faust et al., 2021;Lalonde et al., 2012).However, we here found that in this fraction (RED) fresh carbon with modern 14 C signature was predominantly bound.In contrast, old carbon that had escaped remineralization for centuries was found in other, reductively stable, fractions.LDI-FT-ICR-MS revealed that within the RED fraction, the molecular composition was noticeably more heterogeneous compared to the other fractions (Figure S1 in Supporting Information S1).The origin, age, and fate of sedimentary OC bound to Fe and Al phases therefore remains an issue that should be explored in future research.
Compounds correlated with modern or ancient extract fractions differed by their H/O ratios, with a rough boundary of 2.5 between markers of modern and old OC.Correlation was more pronounced in compounds with low H/C and low O/C ratios, which may be owed to the particular analytical window of LDI mass spectrometry and could be compared with other ionization techniques or other complementary tools, such as FT-IR or NMR spectroscopy (Blackburn et al., 2017).The bandwidth of molecules detected was however very comparable to other studies (Figure S4 in Supporting Information S1; Solihat et al., 2018), and total spectrum magnitude indicated good fit with OC removal.Extract compositions would be challenging to assess with classical ESI-FT-ICR-MS due to the wide differences in solubilities and extractants used.LDI of remaining OC after extraction therefore represents an elegant and promising alternative which should be tested with other samples or extractions.

Figure 1 .
Figure 1.Bathymetric map and sampling sites of the studied Rappbode predam.Stations C and D (red symbols) represent the sites sampled for OC quality ( 14 C, LDI-FT-ICR-MS).Duplicate sediment cores were taken, from which the OC of one core at each site was subjected to sequential OC extraction.OC content and water content were analyzed at nine sites (red and yellow symbols).Sediment thickness was measured at 16 locations.
EX = I B -I A with I = intensity after base-peak scaling; EX = predicted intensity in extract; B = measured intensity in sediment sample before extraction step; and A = measured intensity in sediment sample after extraction step.Only those MF pairs were considered that followed the condition I B > I A (i.e., yielding positive I EX values).Nomenclature of sample names in FT-ICR-MS analysis is detailed in Text S1 of Supporting Information S1.Based on the stoichiometry of the MFs, we calculated intensity-weighted mass spectrum averages (i.e., intense peaks contributed more to the average) of molecular descriptors, compound groups and formula classes (Text S1 in Supporting Information S1).All spectra and formula distributions are shown in Figures S1, S2 and S3 of Supporting Information S1.Molecular descriptors, formula class and compound group distributions are shown in Figures S4 and S5 of Supporting Information S1.

Figure 2 .
Figure2.Chemical and isotope measurements in fractions of depth-integrating sediment samples.OC, organic carbon; TIC, total inorganic carbon; PMC, percent modern carbon; IST, OC extracted by increased ionic strength; RED, OC associated to redox-sensitive (Fe) oxides; ACD, acid hydrolyzable OC; ALK, alkaline extractable OC; RES, residual OC.Values represent means ± SD of triplicate extractions, except radiocarbon (percent modern carbon, single measurements ± analytical errors of combined three parallel extractions).The age of the ionic strength increase fraction was measured only in Core D2 (Methods).Data in TableS3in Supporting Information S1, age of total inorganic carbon see text.

Figure 3 .
Figure 3. Compositional trends in fractions of depth-integrating sediment samples.(a) Molecular formulas (MFs) correlated ( p ≤ 0.05) positively (blue) or negatively (red) with percent modern carbon, in van Krevelen space; gray dots are all signals.(b) Absolute numbers of MFs and fractions of PMC-correlated MFs.(c) Relative abundance (based on percent ion abundance) of PMC-correlated MFs across spectra.Values represent means and error bars SD of two extractions (cores C2, D2).
sediment OC and N was measured with a Vario EL analyzer(Elementar, Hanau, Germany).Dissolved OC in extracts was quantified by near-IR absorption in accordance with EN 1,484 (Dimatoc 2000 analyzer, Analysentechnik Essen).Fe concentrations were measured by inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 7300 DV, Perkin Elmer) according toDIN EN ISO 11885 (E22).A Delta V Advantage mass spectrometer (Thermo Fisher Scientific, Waltham, Massachusetts) was used for 13 C isotope analysis.The δ 13 C values were expressed relative to standard Vienna PeeDee Belemnite.Radiocarbon samples were combusted at 900°C with CuO as oxidant and measured by accelerator mass spectrometry at the Poznan Radiocarbon Laboratory (

Table 1
Bulk Sediment Chemical and Isotope Characterization Values represent mean ± SD of triplicate analyses of mixed sediment from each core, except isotopes (single measurements ± analytical errors).All samples contained modern carbon (>100 PMC).
TITTEL ET AL.