CO 2 and CH 4 fluxes are decoupled from organic carbon loss in drying reservoir sediments

Reservoirs are a prominent feature of the current global hydrological landscape, and their sediments are the site of extensive organic carbon burial. Meanwhile, reservoirs frequently go dry due to drought and/or water management decisions. Nonetheless, the fate of organic carbon buried in reservoir sediments upon drying is largely unknown. Here, we conducted a 45-day-long laboratory incubation of sediment cores collected from a western Mediterranean reservoir to investigate carbon dynamics in drying sediment. Drying sediment cores emitted more CO2 over the course of the incubation than sediment cores 5 incubated with overlaying water (206.7 ± 47.9 vs. 69.2 ± 18.1 mmol CO2 m−2 day−1, mean ± SE). Organic carbon content at the end of the incubation was lower in drying cores, which suggests that this higher CO2 efflux was due to organic carbon mineralization. However, the apparent rate of organic C reduction in the drying sediments (568.6± 247.2 mmol C m−2 day−1, mean ± SE) was higher than C emission. Meanwhile, sediment cores collected from a reservoir area that had already been exposed for 2+ years displayed net CO2 influx from the atmosphere to the sediment (-136.0 ± 27.5 mmol CO2 m−2 day−1, 10 mean± SE) during the incubation period. Sediment mineralogy suggests that this CO2 influx was caused by a relative increase in calcium carbonate chemical weathering. Thus, we found that while organic carbon decomposition in newly dry reservoir sediment causes measurable organic carbon loss and carbon gas emissions to the atmosphere, other processes can offset these emissions on short time frames and compromise the use of carbon emissions as a proxy for organic carbon mineralization in drying sediments. 15


01.pdf
Figure 1.Methodological scheme for the collection, incubation, and analysis of sediment cores.Treatments with dark brown sediment represent cores collected from the "Wet" site, while treatments with light brown color represent cores collected from the "Dry" site.SEM: scanning electron microscopy; XRD: X-Ray diffraction.
a dry environment conducive to core drying, and to avoid CO 2 build-up inside the incubation chamber.Filtered water was periodically added to the "Incubation: Wet" cores throughout the incubation period to replace water lost through evaporation and maintain a constant water level.

Gas flux measurements
Core mass and fluxes of greenhouse gases were measured on the first day of the incubation and periodically thereafter.For the analysis of CO 2 flux to the overlying air, "Incubation: Dry" and "Incubation: Drying" cores were temporarily covered with air-tight, custom made caps.These caps created closed chambers with a surface area of 28.3 cm 2 and volume of 283 cm 3 .The chambers were connected to an environmental gas monitor (EGM-4, PP Systems, Massachusetts, U.S.A.) to directly measure CO 2 flux.CO 2 concentration within the chamber was analyzed every 4.8 seconds with an accuracy of 1%.Each chamber analysis lasted at least 300 seconds or until a 10 µatm change in CO 2 was recorded, whichever came first.CO 2 flux was measured approximately every other day.
"Incubation: Dry" and "Incubation: Drying" cores were capped as described above for 30 minutes for CH 4 flux analysis on days 0, 1, 5, 19, and 45 of the incubation.10 ml air samples were collected via syringe through a septa at the beginning and end of chamber deployment and transferred to pre-evacuated glass vials (Exetainers 339 W, Labco Lim.Lampeter, UK).
Samples were analyzed for CH 4 within 3 weeks of sampling using a gas chromatograph (7820A with a 77697 headspace sampler, Agilent, CA, U.S.A.).The gas chromatograph was calibrated at the beginning and end of each analysis session using a standard curve spanning 1-25 ppm CH 4 in a N 2 gas mixture, with a precision of 0.6 ppm CH 4 .
For "Incubation: Wet" cores, CO 2 and CH 4 fluxes were also measured on incubation days 0, 1, 5, 19, and 45.The overlaying water of all cores was covered with an airtight plastic seal for 20 minutes.Water was collected via syringe immediately before and after the 20 minutes and analyzed for DIC as a proxy for sediment CO 2 flux.DIC was analyzed via catalytic oxidation using a TOC-VCSH analyzer (Shimadzu, Japan).Additional water samples were allowed to equilibrate with headspace in enclosed syringes (10 ml water to 10 ml air).The CH 4 concentration of the equilibrated air was analyzed as described above for the other treatments.
Drying over the course of the incubation was periodically measured by placing each "Incubation: Dry" and "Incubation: Drying" core on a balance to track change in core mass.

Sediment analyses
All cores were drained, sectioned and analyzed for water and organic matter content, either upon arrival to the lab for "Initial" or following the incubation for "Incubation" cores.Cores were sliced into 12 sections (0-1, 1-2, 2-3, 3-4, 4-5, 5-7, 7-9, 9-11, 11-13, and 13-15 cm in depth).Water content was determined by drying an aliquot of each section in a 70°C oven until a constant mass was reached (Fig. S2).This dry sediment was then combusted at 450°C for 4 hours and re-weighed to determine the percentage of weight lost on ignition (LOI), a proxy for sedimentary organic carbon content.We assumed that organic carbon content was equivalent to LOI/2 (Dean and Gorham, 1998).All remaining sediment was frozen at -20°C.Sediment from near the surface (1-2 cm in depth) for all cores was later defrosted and analyzed for pH and alkalinity using a Metrohm 848 Titrino Plus Titrator.Sediment was suspended in a 2:1 deionized water:sediment solution and filtered, and the pH of the resultant filtrate was analyzed.Frozen incubation treatment sediment was freeze-dried for additional mineralogical analyses.Freeze-dried sediment was ground, and its mineralogical composition was determined using a Siemens D500 automatic X-Ray diffractometer (working conditions: Cu K-alpha, 40 kV and 30 mA).The identification of mineralogical species was carried out using EVA software attached to the diffractometer and their quantification was done using the standard procedure (Chung, 1974).The uncertainties associated to the quantification method are 5% wt.Albite, calcite, clinochlorite, dolomite, gypsum, kaolinite, microcline, muscovite, and quartz were identified and quantified.To qualitatively assess differences between treatments, freeze-dried sediment from 0-1 cm (i.e.shallowest) and 13-15 cm (i.e.deepest) was coated with gold and imaged using a scanning electron microscope JEOL J-6510 equipped with an EDS detector at the Scientific and Technological Centers of the University of Barcelona.
The influence of biological activity in carbon dioxide flux was assessed using defrosted sediment from 1-2 cm in depth from a randomly selected "Incubation: Dry" core.This sediment sample was split into 3 replicate sections of 5 ml each, each of which was placed in a separate 100 ml glass beaker and covered with an airtight seal.Baseline CO 2 flux from each section was analyzed as described above for sediment cores.Samples were then sterilized by exposure to UV light under a laminar flow hood (AH-100, Telstar, Catalonia, Spain) for 45 minutes followed by microwaving at 700 Hz for 90 seconds in 30 second increments, each separated by 1 minute of shaking.Sediment was then allowed to return to room temperature, re-sealed, and analyzed again for CO 2 flux.

Data analysis
All statistical analyses were conducted in R (R Core Team, 2018).To test differences in CO 2 flux, CH 4 flux, core mass, and water content between treatments, one-way analyses of variance (ANOVA) were conducted on mixed effect models with treatment considered as a fixed effect and replicate core within treatment as a random effect using the lmer function of the package nlme (Pinheiro et al., 2018).Depth was considered an independent variable for water mixed effect models, and time was considered an independent variable for CO 2 and CH 4 flux and core mass mixed effect models.To analyze organic carbon content differences between treatments, we first identified different sediment core layers as defined by a clustering analysis of the organic carbon content profiles from all cores.Clustering analysis was performed using the chclust function from the R package rioja and constraining the result by sample depth (Juggins, 2017).This analysis was performed to identify the depth of the surface layer affected by organic carbon changes during the incubation, for we expected organic carbon changes to be unlikely beyond a surface layer of unknown depth a priori.After identifying the depth of the surface layer most affected by organic carbon changes, we assessed differences in surface layer organic carbon content between treatments using ANOVA.
Post-hoc Tukey tests were conducted to identify differences between treatments using the lsmeans package (Lenth, 2016).
Net CO 2 flux during the incubation was determined by performing trapezoidal integration under the curve of observed gas flux data points along time, using the trapz function in the package pracma (Borchers, 2018) Extreme outliers were removed from all data sets following examination of box plots, Cook's influential outlier tests, and Cleveland boxplots (Zuur et al., 2010).All plots were created using ggplot from the tidyverse package (Wickham and Team, 2017).

Incubation carbon gas fluxes
Gaseous carbon fluxes differed between each incubation treatment (F = 68.3,p < 0.001 for pairwise post hoc tests) (Table 1 and Fig. 2)."Incubation: Wet" incubation cores generally displayed positive CO 2 fluxes (i.e.out of the sediment) that declined in magnitude over time (Fig. 2).Meanwhile, "Incubation: Dry" cores displayed positive fluxes for the first week but then consistently negative (i.e.into the sediment) fluxes for the remainder of the incubation (Fig. 2).Post-incubation analysis showed that this CO 2 influx to the sediment persisted even after sediment sterilization."Incubation: Wet-Drying" cores initially displayed positive CO 2 fluxes, but by the end of the incubation two out of three "Incubation: Drying" cores also displayed negative CO 2 fluxes (Fig. 2).Change in CO 2 flux over time significantly correlated with drying, as measured by decline in core mass (p = 0.004, r 2 = 0.27).However, net CO 2 flux over the incubation period did not correlate with core-specific sediment organic carbon or water content.
Consistent nonzero mean CH 4 flux values were only observed for "Wet" and "Drying" treatments.Positive CH 4 fluxes were observed on Days 0, 1, and 6 and Day 1 for "Wet" and "Drying" treatments respectively (Figure S3).
Similarly, the low methane fluxes observed in this study stress the relevance of local sediment properties in controlling carbon gas fluxes from sediment.Here, "Incubation: Wet-Drying" sediment cores only displayed nonzero CH 4 flux on Day 0 (Fig. S3).This positive CH 4 flux (28.1 ± 2.0 µmol m −2 d −1 ) was much smaller than both the CO 2 efflux observed here and the CH 4 effluxes observed in other reservoir drying studies (Jin et al., 2016;Kosten et al., 2018).This suggests that sitespecific sediment properties such as organic carbon content or grain size and porosity promoting oxic conditions during drying prevented significant methanogenesis from occurring.

Drying sediment carbon loss
Sediment organic carbon data suggests that the CO 2 efflux in "Incubation: Wet-Drying" sediment cores was driven by organic matter decomposition.Statistical analyses showed that "Incubation: Wet-Drying" cores displayed lower organic carbon content than "Wet" cores (Table 1, Fig. 3).If this small-scale experiment was representative of in-situ reservoir drying, the carbon loss via organic matter decomposition implied by this discrepancy has significant implications for the reservoir's carbon budget.
The difference in organic carbon content between "Incubation: Wet" and "Incubation: Wet-Drying" treatments corresponds to an average organic carbon loss rate of 0.57 ± 0.14 mol m −2 d −1 over the course of the incubation and a net organic carbon loss of 3.07 ± 0.76 Mg ha −1 .This loss over just 45 days is comparable in magnitude to changes in soil organic carbon stock during the transition from tropical secondary forest to perenial crops (Don et al., 2011).Moreover, it is equivalent to reversing approximately 1 year of carbon burial at the average burial rate of 250 g C m −2 yr −1 reported by Mendonça et al. (2017) for inland waters.This significant carbon loss may undermine the notion of organic carbon burial in reservoirs as a long-term carbon sink, particularly in regions such as the western Mediterranean in which reservoirs are fairly dynamic ecosystems.
Instead, decomposition during prolonged drying events may mineralize a sizeable fraction of the organic carbon buried in sediment during the reservoir's lifetime.Thus, sediment carbon burial should not necessarily be considered a carbon sink in a reservoir's long-term carbon budget, specially in regions where drying events are expected to become more frequent in the future.
However, the large variability between replicate cores suggests significant spatial heterogeneity in sediment composition and highlights the need for spatial replication within and across reservoirs in future studies.The previously discussed evidence for organic carbon loss during drying implies that the initial carbon content would be lower in "Initial: Dry" cores than in "Initial: Wet" cores, but this was not the case (Table 1, Fig. 3)."Initial: Dry" sediment did not significantly differ from "Initial: Wet" sediment in organic carbon content.Considering the large variability among replicate cores collected from the same location (which would therefore be more accurately described as pseudo-replicates), greater spatial replication within the reservoir would likely be necessary to resolve differences in sediment carbon content between wet and dry sites.Therefore, although our dataset supports the presence of an enhanced mineralization process during drying, the potential implications at the whole water body scale cannot be fully resolved.

Decoupling of carbon gas efflux from sediment carbon loss
While the observed organic carbon loss from "Incubation: Wet-Drying" cores was consistent with the observed CO 2 fluxes in direction, approximately three times more organic carbon was lost than CO 2 was emitted.The difference in sediment organic carbon content in the surface layer (0-5 cm) of "Initial: Wet" and "Incubation: Wet-Drying" cores would correspond to a net efflux of 72.3 ± 17.8 mmol CO 2 (mean ± SE).However, observed net efflux was only 26.3 ± 6.1 mmol CO 2 (mean ± SE) per core.Thus, even considering the aforementioned variability in sediment organic carbon data, it appears that a significant portion of the organic carbon consumed via decomposition was not emitted as CO 2 but rather consumed by one or more other processes.This is also supported by the observed influx of CO 2 from the atmosphere into the sediment for "Incubation: Dry" treatment cores.Consistent CO 2 influxes in "Incubation: Dry" cores similar in magnitude to the CO 2 effluxes in "Incubation: Wet-Drying" cores were observed after one week of incubation across all replicates (Table 1, Fig. 2).Furthermore, by the end of the incubation two out of three replicate "Incubation: Wet-Drying" cores also displayed CO 2 influxes.These findings show the relevance of the CO 2 consumption pathway(s) active in these sediments.They also imply that the CO 2 effluxes observed in "Incubation: Wet-Drying" cores must be considered the net result of CO 2 production and consumption processes.

Sediment carbon consumption via calcium carbonate chemical weathering
Sediment mineralogy results suggest that the observed sediment carbon consumption was likely caused by an increase in during sediment drying, which is consistent with the elevated pore water Ca 2+ ions expected to accompany an increase in calcium carbonate dissolution.
SEM imagery provides further evidence of sediment calcium carbonate chemical weathering on the time-scale of the incubation.Calcium carbonate crystals in the "Incubation: Wet" treatment were euhedric, but crystals in the "Incubation: Dry" and "Incubation: Drying" treatments were visibly corroded (Figure 5).Similarly, most carbonate in the "Incubation: Wet" treatment was present in the form of discrete crystals and visually biogenic in origin, while most carbonate in the "Dry" and "Drying" treatments appeared as a thin calcium carbonate coating covering all sediment surfaces.This suggests the existence of calcium carbonate precipitation and chemical weathering cycles, probably occurring as a response to sediment drying and flooding cycles.This supports the hypothesis that an increase in chemical weathering relative to precipitation may occur later in the drying process due to the common-ion effect.
Few soil or sediment science investigations link sediment CO 2 influx to an increase in calcium carbonate chemical weathering relative to precipitation.Those that do generally link chemical weathering to factors that are not applicable in the context of this investigation, i.e. climate (Lapenis et al., 2008)), high sediment alkalinity (Lapenis et al., 2008;Emmerich, 2003;Xie et al., 2009;Wang et al., 2016;Ma et al., 2014), and diurnal cycling (Roland et al., 2013;Hamerlynck et al., 2013;Chen and Wang, 2014;Fa et al., 2016).This raises the question of what conditions caused the calcium carbonate chemical weathering hypothesized to occur here.One possible explanation is a combination of 1) high carbonate sediments (29.6 ± 1.4 % CaCO 3 mean ± SE for wet cores) and 2) high air flow and thus CO 2 availability due to sediment dryness.The role of dryness in promoting air flow would explain the lack of both chemical weathering in "Incubation: Wet" sediments and CO 2 influx to dry sediments during the first week of the incubation.Core collection was performed within 48 hours of a rain event, so even exposed sediments were relatively humid at the beginning of the incubation.We posit that sediments may need to reach a certain dryness threshold to establish sufficient air flow and thus CO 2 availability for chemical weathering to occur.Under this scenario, sediment humidity is crucial to determining chemical weathering; sediments must be dry enough to establish sufficient air flow but humid enough for water to be available for the chemical weathering reaction to proceed.
Regardless of the precise mechanism(s) causing chemical weathering, high sediment calcium carbonate content and intermittently dry conditions are the most likely driving factors in this context.Thus, this process may regularly occur in this western Mediterranean reservoir as well as in similar systems around the world.Drying reservoir and lake sediments are understudied, so the calcium carbonate chemical weathering and precipitation observed here may be prevalent in a wide variety of contexts.Given the limited spatial replication and laboratory nature of this investigation, further work is needed to determine the relevance of this process under natural conditions.

Implications for drying reservoir carbon dynamics
The decoupling of organic carbon loss from CO 2 efflux and proposed consumption of inorganic carbon via calcium carbonate chemical weathering in reservoir sediments may have important implications for our understanding of sediment carbon dynamics.First, it indicates that the common strategy of equating carbon gas flux with organic carbon decomposition may be flawed.Thus, if calcium carbonate chemical weathering in sediments is geographically widespread, organic carbon mineraliza- tion rates in dry sediments may be significantly underestimated by studies that only measure CO 2 efflux as a proxy for carbon mineralization.Although we cannot make conclusions regarding the prevalence of this decoupling in other reservoirs with our experiment, our findings constitute a warning that further research is necessary to understand the significance of this process to overall freshwater carbon cycling.
Further research on the fate of the alkalinity produced by the calcium carbonate chemical weathering process is also needed to determine its impact on the carbon budget of lakes and reservoirs that seasonally or permanently dry.While CaCO 3 chemical weathering decreases CO 2 efflux, it is unlikely to constitute a long-term carbon sink if the bicarbonate ions produced by dissolution eventually transfom to CO 2 and re-enter the atmosphere through equilibration (Wang et al., 2016).However, if the bicarbonate ions produced by CaCO 3 dissolution are sequestered in either sediment or groundwater, it is also possible that the CO 2 influx observed in this study constitutes the basis of a previously unrecognized long-term carbon sink.The rate of carbon dioxide intake shown here is comparable to rates from a variety of Mediterranean and temperate forest soils (Baldocchi et al., 2018).Such a large sediment carbon sink would therefore carry considerable implications for our understanding of the freshwater carbon cycle, and this question also merits further research.

Conclusions
This investigation used a laboratory sediment core incubation to explore the effects of reservoir drying on sediment carbon dynamics.We directly linked organic carbon loss to carbon dioxide emissions in drying reservoir sediment for the first time, undermining the idea that organic carbon burial in active reservoir sediments represents a long-term carbon sink.However, we also found a decoupling between carbon loss and carbon gas fluxes and observed carbon dioxide influxes to most sediment cores analyzed.Mineralogical sediment composition suggests that these discrepancies were due to an increase in calcium carbonate chemical weathering.Together, these findings show that while reservoir sediment drying can cause organic carbon decomposition and thus carbon gas efflux to the atmosphere, other sediment processes can potentially offset or even reverse these fluxes.
. To determine mineralogical trends among treatments, we conducted a principal component analysis (PCA) on the correlation matrix of the arcsine √ (x) transformed percent abundance data for each mineral using the prcomp function from R core.Pearson correlation tests were conducted between CO 2 flux and core organic carbon content, water content, and change in core mass, and between the first Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-128Manuscript under review for journal Biogeosciences Discussion started: 14 May 2019 c Author(s) 2019.CC BY 4.0 License.two principal axes of the mineralogy PCA and organic carbon and water content using R core.Pearson correlation tests were also conducted between averages of the first two principal axes per replicate core and CO 2 flux and change in core mass.

Figure 2 .Figure 3 .
Figure 2. Sediment core CO2 + CH4 fluxes over the course of the 45-day-long incubation as determined by core headspace CO2 and CH4 measurements for "Incubation: Dry" and "Incubation: Drying" treatments, and overlaying water DIC and dissolved CH4 measurements for "Incubation: Wet" cores.All replicates are shown in the graph.CH4 fluxes composed less than 1% of carbon gas fluxes for all data points (see Fig. S3 for just CH4 fluxes).Color lines are splines fitted to the data, and are included only for visual reference.
, https://doi.org/10.5194/bg-2019-128Manuscript under review for journal Biogeosciences Discussion started: 14 May 2019 c Author(s) 2019.CC BY 4.0 License.3.3Mineralogical sediment transformationsX-ray diffraction results suggested mineralogical transformations occurred during drying(Figs.4, S4).A PCA run using percent abundances of the nine identified minerals revealed divergence between treatments (Figure4).The first two axes of the PCA explained 74.51% of variance.The first axis of the PCA explained 45.6% of variance, with a positive loading of calcite and kaolinite and a negative loading of clinochlorite and quartz.The second axis accounted for 28.8% of variance, with a positive loading of quartz and negative loadings of dolomite, muscovite, and kaolinite."Incubation: Wet-Drying" core samples were grouped by high quartz and clinochlorite, "Incubation: Dry" cores were grouped by high muscovite and dolomite content, and "Incubation: Wet" cores were grouped by high calcite.The first PCA axis scores correlated with organic carbon content (p < 0.001, r 2 = 0.14) and water content (p < 0.001, r 2 = 0.29), and the second PCA axis scores correlated with organic carbon content (p = 0.002, r 2 = 0.13).Average first and second PCA axis scores did not correlate with either CO 2 flux or change in 10 core mass.11 Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-128Manuscript under review for journal Biogeosciences Discussion started: 14 May 2019 c Author(s) 2019.CC BY 4.0 License.

Figure 4 .Figure 5 .
Figure 4. Principal component analysis (PCA) of mineralogy data for sediment samples from varying depths of "Incubation: Wet", "Incubation: Dry", and "Incubation: Wet-Drying" cores.Color identifies the treatment, while the size of the dots represent core sample depth.

Table 1 .
Summary of sediment core properties and incubation gas fluxes averaged across treatments types.All values are mean Treatment Initial: Control Wet Initial: Control Dry Incubation: Wet Incubation: Dry Incubation: Wet-Drying 8 Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-128Manuscript under review for journal Biogeosciences Discussion started: 14 May 2019 c Author(s) 2019.CC BY 4.0 License.