Rising CO2 concentration in the atmosphere, global climate change, and the sustainability of the Earth's biosphere are great societal concerns for the 21st century. Global climate change has, in part, resulted in a higher frequency of flooding events, which allow for greater exchange between soil/plant litter and aquatic carbon pools. Here we demonstrate that the summer 2011 flood in the Mississippi River basin, caused by extreme precipitation events, resulted in a “flushing” of terrestrially derived dissolved organic carbon (TDOC) to the northern Gulf of Mexico. Data from the lower Atchafalaya and Mississippi rivers showed that the DOC flux to the northern Gulf of Mexico during this flood was significantly higher than in previous years. We also show that consumption of radiocarbon-modern TDOC by bacteria in floodwaters in the lower Atchafalaya River and along the adjacent shelf contributed to northern Gulf shelf waters changing from a net sink to a net source of CO2 to the atmosphere in June and August 2011. This work shows that enhanced flooding, which may or may not be caused by climate change, can result in rapid losses of stored carbon in soils to the atmosphere via processes in aquatic ecosystems.
 Approximately two-thirds of the total suspended matter and total dissolved solids transported from the conterminous U.S. to ocean waters are carried by the Mississippi-Atchafalaya river system [Milliman and Meade, 1983]. The unique magnitude of the 2011 flooding of the Mississippi River basin required opening the Morganza Floodway near Baton Rouge, and other smaller floodgates in the delta plain, to divert water into the Atchafalaya River basin. Since its completion in 1954, the Morganza Spillway has been opened only one other time, in 1973. The 14 May 2011 opening of the spillway began with the lifting of a single floodgate with a flow of about 3500 m3 s-1 and was then ramped up to about of 4900 m3 s-1 followed by a ramping down until officially closed on 7 July 2011. Flooding events are important for the transfer of carbon from soils and plant litter to inland waters [Raymond and Saiers, 2010]. During this transport, carbon in the water can be exchanged with the atmosphere with approximately ~100 Tg of CO2 degassed from U.S. streams and rivers per year [Butman and Raymond, 2011]. Here we show that the 2011 Mississippi River basin flooding event enhanced export of radiocarbon-modern terrestrially derived dissolved organic carbon (TDOC) causing shelf waters in the northern Gulf of Mexico between the Atchafalaya and Mississippi river plumes to change from being a net sink to a net source of CO2 to the atmosphere.
2 Materials and Methods
 We collected water samples at three stations on the inner shelf adjacent to the Atchafalaya for analyses of DOC and dissolved inorganic carbon (DIC), total alkalinity (Talk), chlorophyll, primary production, RNA-based microbial community characterization, and continuous surface water dissolved pCO2. Samples were collected in April 2011 (pre-flood), June 2011 (flood), and August 2011 (post-flood; Figure 1). Continuous measurements of dissolved pCO2 in surface waters were made from the mouth of the Atchafalaya River to Morgan City and through Atchafalaya Bay estuary to the first shelf station in April, June, and August 2011. In June, during the flood period, samples were also collected for chromophoric dissolved organic matter (CDOM) and excitation emission fluorescence, radiocarbon (on select samples of DOC and DIC), dissolved lignin, and δ18O (as a water tracer; Supplementary Information, Table A1). The δ18O and salinity relationship between offshore waters and Atchafalaya River (R1–R5) and near-shore ATCH waters confirmed that the near-shore Gulf waters were solely derived from Mississippi waters and the Atchafalaya River and its estuary (Supplementary Information, Figure S1).
2.1 DOC and Dissolved Lignin
 Acidified filtered water samples (100 μL of 2 N HCl added to remove inorganic carbon) were analyzed for DOC on a Shimadzu TOC-VCSH/CSN using high-temperature catalytic oxidation [Guo et al., 1994]. Solid-phase extraction (SPE) was used to collect DOM and lignin according to the method of Louchouaran et al. . Freeze-dried SPE dissolved lignin samples were analyzed for lignin-phenols using the cupric oxide method of Hedges and Ertel , as modified by Goni and Hedges . Sigma-6 (Σ6) is defined as the sum of vanillyl (vanillin, acetovanillone, and vanillic acid) and syringyl (syringaldehyde, acetosyringone, and syringic acid) phenols, and Sigma-8 (Σ8) includes the cinnamyl (p-coumaric and ferulic acid) phenols. Mississippi River (Belle Chase, LA) and Atchafalaya River (Morgan City, LA) samples were collected as part of the U.S. Geological Survey (USGS) National Stream Quality Accounting Network (http:/water.usgs.gov/nasqan) and analyzed for DOC concentration, specific UV absorbance at 254 nm (SUVA254), and percent HPOA according to methods described in Spencer et al. .
2.2 Molecular Characterization of Microbial Populations
 Bacteria were collected on 0.2 µm pore diameter polycarbonate filters. Total RNA was extracted using the Mills extraction method [Mills et al., 2012]. Extracts were treated with Turbo DNA Free (Ambion, Austin, TX) according to the manufacturer's instructions to remove residual DNA coextracted with the RNA. Reverse transcription of RNA to cDNA and subsequent PCR amplification was performed according to Mills et al.  and Reese et al. . The resulting cDNA was sequenced at the Research and Testing Laboratory (Lubbock, TX) following standard laboratory procedures, quality control, and analysis as described in Reese et al. .
2.3 DOM Absorption and Fluorescence
 Filtered water samples (0.2 µm) were measured for absorbance on a Varian 300 UV spectrophotometer in a 10 cm quartz cell and converted to Napierian absorption coefficients. DOM fluorescence was measured on a Varian Eclipse spectrofluorometer, corrected for inner-filtering effects, lamp intensity (excitation mode), and detector response (emission mode), and reported in quinine sulfate equivalents (ppb QSE).
 Continuous underway measurements were made of surface water pCO2 and atmospheric pCO2 using an instrument similar to that described by Pierrot et al. . Seawater was drawn from an intake 3 m below the water surface and flowed continuously through an equilibrator while the headspace was recirculated. Air was drawn at ~6 L min-1 through a 3/8-inch tube (Synflex) from an elevated location forward on the ship. The automated instrument allowed dried air to pass through the NDIR detector (LiCor) for 15 min and then switched to dried equilibrator headspace for 105 min. Periodic calibrations were performed with blended gas standards (Air Liquide).
2.5 DIC and Talk
 DIC was determined coulometrically following the methods described in SOP 2 [Dickson et al., 2007]. Talk was determined by Gran titration [Gran, 1992] following the methods described in SOP 3b [Dickson et al., 2007].
2.6 Radiocarbon of DIC and DOC
 The method used for carbon isotopic analyses (14C, 13C) of DOC and DIC is described in detail in Bauer et al. .
2.7 Statistical Comparison of Sequence and Geochemical Data
 Differences between sites, operational taxonomic units, functional groups, and geochemical composition were evaluated through one-way analysis of variance. A post-hoc least squares difference test was performed when significant differences were identified. Correlation analysis was conducted to identify strength of relatedness in geochemical and physical variables. Statistical analyses were performed using Microsoft Excel 2008 and SPSS version 16.0. Principal component analysis (PCA) was conducted in Matlab version 7.4.3.
3 Results and Discussion
3.1 DOC and CO2 Fluxes
 Typically, rivers are sources of CO2 to the atmosphere [Butman and Raymond, 2011; Raymond et al., 1997] and continental shelves act as CO2 sinks removing up to 0.5 to 1 Pg C year-1 from the atmosphere [Cai, 2003; Borges et al., 2005; Lohrenz and Cai, 2006; Cai et al., 2006]. Just before the flood in April 2011, large dissolved pCO2 concentrations (maximum = 1875 ppm) and fluxes (max = 1948 mmol m-2 d-1) to the atmosphere were observed in the Atchafalaya and a short distance from the river mouth (Figure 1; Figure A2). In surface waters on the shelf, dissolved pCO2 was very low (minimum = 66 ppm) with substantial fluxes of CO2 (-368 mmol m-2 d-1) into the water from the atmosphere due to increased net autotrophy (Figure 1; Figure A2 and Table A1). Although we do not have primary production measurements for June 2011, primary production in April and August (pre- and post-flood) was greater from east to west (Table A1), except at the stations closest to Atchafalaya Bay in April. We did measure similar rates of primary production at station AB5 (~2.1–2.4 mg C m3 h-1) in both months, with 70% higher rates at station 10B (0.43 and 0.64 mg C m3 h-1) and 50% higher rates at station 8C (0.71 and 1.4 mg C m3 h-1) in August. By contrast at station ATCH, August primary production (0.86 mg C m3 h-1) was significantly lower than that measured in April (2.0 mg C m3 h-1). These differences in seasonal primary production rates before and after the flood reflect spatial changes in potential biological uptake of CO2, which has been shown to be the dominant process in controlling this region as a net sink for CO2, under nonflooding conditions [Cai, 2003; Lohrenz and Cai, 2006; Quigg et al., 2011]. Thus, our primary production measurements support the CO2 flux data with higher primary production on the shelf stations where lower CO2 fluxes to the atmosphere were observed (Figure A2 and Table A1).
 During the flood in June 2011, pCO2 concentrations were higher in the river reaching 4382 ppm (max flux = 669 mmol m-2 d-1) with supersaturation extending from the river onto the shelf for about 60 km. It should be noted that, when compared with April, June CO2 fluxes were lower for higher dissolved pCO2 concentrations due to lower overall wind speeds. In August after the flooding had subsided, whereas the river pCO2 concentrations declined to near April levels (max = 2132 ppm), pCO2 concentrations in the surface waters on the shelf did not decline as much and remained slightly supersaturated with respect to the atmosphere (Figure 1; Figure A2 and Tables A1 and A2). By dividing the region into areas characterized by relatively constant pCO2 concentrations observed during the June cruise, we are able to determine the regional net flux of CO2 to the atmosphere during the flood event and compare it with the regional net fluxes for April and August (Table A2 and Figure A4). The regional net fluxes not including the river are -2.66, 4.36, and -0.22 Gg-C d-1 for April, June, and August, respectively. The net fluxes including the river are -2.16, 4.71, and -0.009 Gg-C d-1 for April, June, and August, respectively. A molecular biological link to the reported June pCO2 concentrations and primary production measurements was determined by a lower detection frequency of phototrophic, CO2-consuming lineages and a higher detection frequency of heterotrophic, CO2-producing lineages in the river compared with the shelf.
 In May and June 2011, the combined summed fluxes of DOC from the Atchafalaya and Mississippi rivers to the coast were 1041.1 Gg (1.0 Tg; Figure 2), which is 33% and 58% of the overall DOC flux estimates to the Gulf of Mexico per year, reported to be 3.1 and 1.8 Tg, by Bianchi et al.  and Cai and Lohrenz , respectively. DOC concentrations in the Mississippi River typically range from 223 to 380 μM [Bianchi et al., 2004; Benner and Opsahl, 2001]. Before the flood, in April 2011, DOC on the shelf was 122 ± 37 μM, significantly lower than in June 2011 (227 ± 39 μM; during flood) and August 2011 (223 ± 45 μM; post flood; Table A1). In June 2011, the average DOC concentration in the lower Atchafalaya River was 419 ± 35 μM (Table A1). In August 2011, 62 days after sampling in June 2011, the average DOC in the lower Atchafalaya River had decreased to 294 ± 41 μM. Similarly, the Atchafalaya Bay estuarine region was higher in June 2011 during the flood (374 ± 36 μM) than August (251 ± 12 μM). These were also the regions with highest CO2 concentrations and sea-to-air fluxes (Tables A1 and A2), suggesting that DOC consumption in inland waters was driving CO2 supersaturation. Molecular analysis of metabolically active microbial community supported this conclusion by indicating a statistically significant increase in the bacterial populations capable of consuming DOC in regions with higher CO2 concentrations. Radiocarbon measurements made on select samples from the lower Atchafalaya demonstrate that this DOC was predominantly modern (FM = 1.08 and 1.09 for April and June, respectively). DOC concentrations on the shelf between the Mississippi and Atchafalaya river plume regions, which generally range from 80 to 300 μM (values > ~150 μM are found in plume regions [Benner and Opsahl, 2001; Wysocki et al., 2006]) were also higher after the flood.
3.2 Physical and Biological Controls on pCO2 Sea-to-Air Fluxes
 At this point, we have demonstrated that the flood resulted in higher DOC fluxes than normal and that CO2 fluxes reversed from a net influx to an efflux to the atmosphere in lower and inner shelf waters. We also know that dissolved pCO2 in the lower Mississippi River did increase during the flooding event, albeit not to values as high as other times before the last year (Figure A5). Although there have been higher concentrations in past decades, this suggests that higher pCO2 values likely also occurred further upriver from our sampling stations. Although efflux of CO2 can be controlled by both physical and biological processes, we now attempt to decipher the possible role of TDOC sources in controlling in part, this change in net sink to source of CO2, by examining (1) the role of respiration, (2) changes in microbial community, and (3) linkages between CO2 efflux and TDOC abundance.
 Using the observed DIC concentrations along with dissolved pCO2 concentrations, the air-water fluxes, and some primary production rates (Table A1 and Figure A6), we estimated the amount of respiration that could have occurred in the bays in April and June (Figure A7). In April, the calculated respiration rate suggested that there was an additional carbon sink of 1.21 × 108 mol-C d-1 (e.g., mixing or primary production was underestimated) present at that time. In June, if we assume a primary production rate equal to that observed in April, the estimated respiration rate is 3.70 × 108 mol-C d-1, suggesting that up to 83% of the CO2 flux to the atmosphere could be sustained by in situ respiration. If we assume that the additional carbon sink estimated for April was also present in June, the estimated respiration rate needed to sustain the carbon concentrations in the bays was 3.70 × 108 mol-C d-1, suggesting that approximately 56% of the flux of carbon to the atmosphere could be sustained by respiration. Finally, DIC isotopes made in the river (14C = 1.03, δ13C = 6.6) and near coast (14C = 1.06, δ13C = 4.6) during the flood were also enriched in 14C and depleted in 13C, which is consistent with a DOC source.
3.3 Characterization of DOC Sources
 SUVA values (defined as absorption coefficient at 254 nm normalized to DOC) and the proportion of hydrophobic organic acids in the DOC pool (HPOA%), which are suggestive of contributions of more complex DOM from surface soils and plant litter, showed no significant changes (Figure 2A). To better constrain the contribution of vascular plant inputs from surface soils and litter to the enhanced DOC loading during flood conditions, we analyzed for lignin chemical biomarkers (Σ8 and Σ6) in our river, bay, and shelf samples (only in June). The highest Σ8 and Σ6 values were found at the uppermost station in the river, 25.9 and 12.0 µg L−1, respectively (Table A1). In addition to SUVA, which was strongly correlated with Sigma 6 in river samples and humification index (HIX; from excitation and emission [EEM] data) in near-shore Gulf samples (Figure 2B), we examined several other spectral parameters including maximum fluorescence (Fmax), biological index (BIX and EEM data), and the spectral slope ratio (SR) [Shank and Evans, 2011] to gain insight into the contribution of terrestrial DOM sources (Table A1). Spectral data for Gulf samples from June 2011 including high HIX, low SR, and high Fmax values clearly showed a strong terrestrial signature (Table A3). Incorporating the spectral data above with DOC and pCO2 data into a PCA model revealed that ~60% of variability in the data set could be explained by SUVA (as PC1). Also, positive scores along PC1 coincided with positive loadings for terrestrial DOM and for pCO2, whereas negative loadings along PC1 were found for SR and for BIX (proxy for autochthonous DOM [Huguet et al., 2009]). Along PC2, about 30% of the variability in the data was driven by date of sampling and by HIX, which is consistent with the CO2 flux being lower in August than in June. The PCA results also indicate that, in August, more humified DOM on the shelf produced less CO2 than did fresher DOM in the estuarine Atchafalaya River. The combination of PCA results and spectral data from June (Table A3) suggests that CO2 efflux was higher when there was more TDOC material in the surface waters; the role of TDOC was not investigated.
3.4 Microbial Changes During the Flood
 Although photooxidation of DOM in the coastal ocean has been shown to be an important process for the removal of the DOM [Stubbins et al., 2010, and references therein], the highest concentrations and fluxes of CO2 were in the most turbid waters, further suggesting that biological production of CO2 was likely more important than the photochemical processes during this flooding event. Although we only have microbial data for June, there were distinct differences in the microbial community within the river and plume compared with the inner shelf Gulf sites during the flood (Figure A8). Metabolically active microbial community structure and function [Mills et al., 2012; Reese et al., 2013]; singular value decomposition (described in Supplemental Data) was used to statistically compare biogeochemical and microbial observations. Interestingly, a diverse heterotrophic population including terrestrial, lignin-degrading lineages Polyangiaceae [Velicer and Hillesland, 2008], Gemmatimonadaceae [DeBruyn et al., 2011], and Anaerolineaceae [Narihiro et al., 2012] were detected more frequently in the river. Fewer sequences were detected offshore, outside of the river plume (Figure A8). More importantly, the decrease in abundance of lignin-degrading lineages was significantly correlated (p < 0.05) to a decrease in lignin abundance. The combination of these microbial changes, along with the possible role of photochemical breakdown in surface waters, and enhanced oxygen availability, via turbulence in floodwaters, all likely enhanced the breakdown of TDOC. Finally, the role of priming from algal exudates [Guenet et al., 2010; Bianchi, 2011] in enhancing the breakdown of TDOC may have also contributed to the rapid turnover of TDOC to CO2 in these floodwaters.
 In conclusion, characterization of DOC and the metabolically active bacteria populations provided a biological link to the geochemically described shift between CO2 source river water and the CO2 sink shelf water. Given the nature of the organic material fueling the respiration in the bays and the microbial population, the CO2 emitted to the atmosphere was likely from the oxidation of TDOC. This work shows how greater flooding associated with scenarios of climate change allows for the rapid transfer of soil carbon to the atmosphere through aquatic pathways. The faster breakdown of TDOC derived from leached litter in flooded waters compared with surface soils may suggest that greater oxidation, photochemical breakdown, and even priming mechanisms may be at work to enhance the decomposition of TDOC. From a global perspective, if certain regions continue to experience more flooding with global change as well as increased thawing of ice at the poles, this may allow for greater exchange between soil/litter-derived organic carbon and aquatic environments. Moreover, if this terrestrially derived organic carbon is then more rapidly converted to CO2 in aquatic than terrestrial systems [Cole et al., 2007], there should be a net increase in the transfer of stored organic carbon material in terrestrial systems to the atmosphere. This flooding event resulted in an enhanced flux of DOC to the Gulf of Mexico that was in only 2 months as much as approximately 50% of total annual DOC flux of this river system and changed the adjacent shelf region that is normally net sink to a net source of CO2 to the atmosphere for this brief period of time. The effects of enhanced removal of modern litter and surface soil carbon from terrestrial ecosystems that are experiencing more flooding (via climate change) on regional CO2 exchange, utilization of older soil carbon, and terrestrial carbon budgets remain largely unknown. This positive feedback resulting from global warming can be expected to occur where we see a greater turnover and loss of organic carbon from terrestrial environments that are experiencing more flooding and thawing.
 We thank Rik Wanninkhof for the analysis of DIC in the water samples collected in April. This work was funded, in part, by NASA (NNX08AW98G). We thank the USGS field personnel for collecting water samples from the Mississippi and Atchafalaya rivers. This research was supported by the USGS National Stream Quality Accounting Network. The use of brand name in this manuscript is for identification purposes only and does not imply endorsement by the USGS.