Dissolved carbon biogeochemistry and export in mangrove-dominated rivers of the Florida Everglades

Manuscript; Page 2 Abstract. The karstic marshes of the Everglades and flow through the largest contiguous mangrove forest in North America. In November 2010 and 2011, dissolved carbon source-sink dynamics were examined in these rivers during SF 6 tracer release experiments. Approximately 80% of the total dissolved carbon flux out of the Shark and Harney Rivers during these experiments was in the form of inorganic carbon, either via air-water CO 2 exchange or longitudinal flux of dissolved inorganic carbon (DIC) to the coastal ocean. Between 42 and 48% of the total mangrove-derived DIC flux 20 into the rivers was emitted to the atmosphere, with the remaining being discharged to the coastal ocean. Dissolved organic carbon (DOC) represented ca. 10% of the total mangrove-derived dissolved carbon flux from the forests to the rivers. The sum of mangrove-derived DIC and DOC export from the forest to these rivers was estimated to be at least 18.9 to 24.5 mmol m -2 d -1 , a rate lower than other independent estimates from Shark River and from other mangrove forests. Results from these experiments also suggest that in Shark and Harney Rivers, mangrove 25 contribution to the estuarine flux of dissolved carbon to the ocean is less than 10%. 1 and 2 studies are the first to provide estimates of longitudinal DIC export, air-water CO 2 fluxes, and mangrove-derived DIC inputs for the Shark and Harney Rivers. The results show that air-water CO 2 exchange and longitudinal DIC fluxes account for ca. 90% of the mangrove-derived dissolved carbon export out of the Shark and Harney Rivers, with the remainder being exported as dissolved organic carbon.


Introduction
In many tropical and sub-tropical regions, mangrove forests are a typical feature surrounding estuaries (Twilley et al., 1992;Bouillon et al., 2008a). Mangroves are thought to play an important role in tropical and subtropical coastal biogeochemical cycling and the global coastal carbon budget, due to their high productivity and 30 rapid cycling of organic and inorganic carbon (Twilley et al., 1992;Jennerjahn and Ittekkot, 2002;Dittmar et al., 2006). However, there remain uncertainties regarding the fate of mangrove-fixed carbon and the amount of carbon exported to the coastal waters from these ecosystems (Bouillon et al., 2008a;Bouillon et al., 2008b;Kristensen et al., 2008). Bouillon et al. (2008a) showed that over 50% of the carbon fixed by mangroves through photosynthesis could 35 not be accounted for by growth in biomass, accumulation in soils, and export of organic carbon, and suggested that a large fraction of this missing organic carbon may be mineralized to dissolved inorganic carbon (DIC) and either lost to the atmosphere or exported to the surrounding waters. In fact, several studies have shown that the lateral advective transport of interstitial waters through tidal pumping represents a major carbon export pathway from mangroves into adjacent waters, both for DIC (Koné and Borges, 2008;Miyajima et al., 2009;Maher et al., 2013) and dissolved 40 organic carbon (DOC) (Dittmar and Lara, 2001;Bouillon et al., 2007c). However, to date, lateral mangrove derived aquatic carbon fluxes (as a proportion of overall forest carbon mass balance) have only been estimated for short time periods and over limited spatial (e.g., plot) scales (e.g., Troxler et al., 2015). These studies also typically do not determine the fate of mangrove-derived carbon once it is exported from the forest through tidal pumping and drainage.
Additional measurements of the magnitude and fate of mangrove carbon export at the basin scale are needed to help 45 quantify connections between inter-tidal, estuarine and coastal ocean carbon cycles.
Rates of lateral dissolved carbon export from tidal mangrove forest are heterogeneous over space and time due to variability in inundation patterns, forest structure, topography, and soil hydraulic properties. Direct, plot-scale measurements of dissolved carbon export therefore may not represent rates quantified at the basin-scale. However, mangrove-derived dissolved carbon fluxes may be estimated in some systems using information on the spatial 50 distribution of carbon-related measurements in adjacent waters. For example, the carbon balance of tidal riverine systems adjacent to mangrove forests should integrate the spatial and temporal variability of these lateral fluxes.
The objective in the study is to quantify dissolved carbon source-sink dynamics in a subtropical estuary dominated by two tidal rivers, the Shark and Harney Rivers in Everglades National Park, Florida, USA. These rivers Annotated Manuscript; Page 4 are centrally-located within the largest contiguous mangrove forest in North America and they discharge to the Gulf 55 of Mexico. The total dissolved carbon inventories and fluxes in these rivers are determined using a series of discrete and continuous measurements of carbon-related parameters along a salinity gradient, and the mangrove contribution separated using measurements of stable isotopic composition of dissolved organic and inorganic carbon. The results are then scaled by the area of mangrove forest that surrounds these rivers to express dissolved carbon fluxes on an aerial basis for comparison to independent measurements of dissolved carbon fluxes from this forest. 60 2 Methods

Study site
The tidal-dominated Shark and Harney Rivers (river and estuary are used interchangeably in this contribution) are surrounded by mangrove forests and located on the southwest coast of Florida ( Fig. 1), within Everglades National Park. The subtropical climate in southern Florida is characterized by a May to October wet 65 season, when approximately 60% of the annual precipitation occurs (Southeast Regional Climate Center, http://www.sercc.com). The Shark and Harney Rivers together discharge approximately 50% of the flow from the Shark River Slough (SRS), the primary drainage feature of Everglades National Park, to the Gulf of Mexico (GOM) (Levesque, 2004). Seasonal variation of the water discharge from SRS mostly follows the precipitation patterns (Saha et al., 2012), and influences the transport of nutrients to the mangrove ecotone . The 70 Shark and Harney Rivers are each approximately 15 km long, and connected in Tarpon Bay (Fig. 1). The mean depths of Tarpon Bay, Shark River, and Harney River at mid tide are 1.4 ± 0.3, 2.8 ± 0.4, and 2.6 ± 0.4 m (Ho et al., 2014), respectively, and the surface areas are 1.48 x 10 6 , 2.54 x 10 6 , and 2.75 x 10 6 m 2 , respectively. The inter-tidal zones bordering the Shark and Harney Rivers are dominated by Rhizophora mangle (red mangrove), Avicennia germinans (black mangrove), Laguncularia racemose (white mangrove), and Conocarpus erectus (buttonwood). Semi-diurnal 75 tides in this region inundate the forest as often as twice a day. River discharge to the GOM is primarily influenced by tides, wind, and freshwater inflow from SRS (Levesque, 2004).
Discharges are determined by the US Geological Survey at stations near the midpoints of Shark River (USGS 252230081021300 Shark River) and Harney River (USGS 252551081050900 Harney River) (Fig. 1). Discharges are generally lower during March-May than the rest of the year. Hourly mean residual discharge values (i.e., filtered for 80 tides) from March to May of the 5-year period from 2007 to 2011 ranged from -21.9 to 24.1 m 3 s -1 , with a mean of 0 Annotated Manuscript; Page 5 m 3 s -1 for Shark River, and ranged from -28.9 to 38.5 m 3 s -1 , with a mean of 4.4 m 3 s -1 for Harney River. Positive values indicate flow towards the GOM. For the rest of the year (i.e., June to February), these values ranged from -46.2 to 89.2m 3 s -1 , with a mean of 8.8 m 3 s -1 for Shark River, and -41.6 to 75.0 m 3 s -1 , with a mean of 11.3 m 3 s -1 for Harney River. 85
During both campaigns, an inert tracer (sulfur hexafluoride; SF 6 ) was injected in the river near the point where the rivers diverge just downstream of Tarpon Bay (25.4092,) to determine the rates of longitudinal dispersion, and the water residence time. Each day, longitudinal surveys were made along the Shark and Harney Rivers from Tarpon Bay to the GOM, and included continuous underway measurements of temperature, salinity, SF 6 , 95 dissolved O 2 (DO; µmol kg -1 ), and partial pressure of CO 2 (pCO 2; µatm), and discrete measurements of total alkalinity (TAlk; µmol kg -1 ), dissolved inorganic carbon (DIC; µmol kg -1 ), dissolved organic carbon (DOC; µmol kg -1 ), stable carbon isotopic composition of DIC and DOC (d 13 C DIC and d 13 C DOC , respectively; ‰).

Discrete measurements
During SharkTREx 1, three to five surface water samples were collected daily in the Shark River with a 5-L 100 Niskin bottle at ~0.5 m below the surface for the analysis of TAlk, DOC, δ 13 C DIC , and δ 13 C DOC . At each sampling site, vertical profiles of temperature, salinity, and DO were recorded using a conductivity, temperature, and depth sonde (Sea-Bird SBE 19plus V2) equipped with a Clark type polarographic O 2 sensor (SBE 43). These profiles showed that the water column was vertically well mixed. No discrete samples were collected in the Harney River during SharkTREx 1. During SharkTREx 2, discrete samples for DIC, TAlk, DOC, δ 13 C DIC , and δ 13 C DOC were collected daily 105 at 20 stations distributed within the Shark and Harney Rivers (Fig. 1 from the volume of acid added at the inflection point closest to a pH of 4, and reported as µmol L -1 HCO 3 since the original pH of the water samples was near neutral. The precision of the measurements was ±2% from replicate analysis (n = 5) with an accuracy of ±2% as determined by analysis of certified reference material (Dickson, 2010). DIC and pH were computed from TAlk and pCO 2 using the dissociation constants of Cai and Wang (1998) for estuarine waters.
During SharkTREx 2, samples for TAlk and DIC were collected in 550 mL borosilicate glass bottles, 115 poisoned with HgCl 2 , and sealed with hydrocarbon grease (Apiezon M). The samples were stored at room temperature in the dark for travel to the laboratory at NOAA/AOML. Samples for TAlk were measured in an open thermostated cell (25 °C) with an automated titrator (Metrohm 765 Dosimat) connected to a pH glass-reference electrode system (Orion), using 0.2 M HCl as a titrant, and determined from the equivalence point of the titration curve using a nonlinear least-squares fit. For DIC analysis, water samples were first acidified to convert all the carbonate species to CO 2 120 in a DIC analyzer (Apollo SciTech), and then measured with a NDIR detector (LI-COR LI-7000). Calibrations for DIC and TAlk were performed using certified reference material (Dickson, 2010). The analytical uncertainty of the DIC and TAlk measurements based on replicate samples are 0.1 and 0.2%, respectively.
The measured TAlk and pCO 2 from SharkTREx 2 were used to calculate DIC using CO2SYS (Pierrot et al., 2006) and the dissociation constants of Cai and Wang (1998), and the results were 1.3 ± 1.1% (mean ± s.d.; n = 77; 125 range: -2.4 to +4.4%) higher than the measured DIC, possibly indicating a slight contribution (ca. 1%) to TAlk from organic or particulate material, as the samples were not filtered.

Dissolved organic carbon
The samples analyzed for DOC were filtered with pre-combusted 0.7 µm GF/F filters and collected in precleaned, acid-washed, brown high-density polyethylene bottles (HDPE; Nalgene). Containers were rinsed three times 130 before sample collection, transported on ice to the FIU SERC Nutrient Analysis Lab, and stored in a refrigerator until analyses within three weeks of collection. DOC was measured using the high-temperature catalytic combustion method on a total organic carbon analyzer (Shimadzu TOC-V), and standardized using 10, and 50 ppm of potassium Annotated Manuscript; Page 7 hydrogen phthalate (KHP), with reagent water as a blank. The analytical precision based on replicates of KHP is ca.

Stable carbon isotopic composition
Samples for δ 13 C DIC were collected in 40 ml glass bottles after passing the sample through a GF/F filter, and then poisoning with HgCl 2 . In the laboratory at RSMAS, vials with 0.5 ml 103% H 3 PO 4 were flushed for 60 s with He. Approximately 2 ml of sample were then injected into the vial, and after sonification the accumulated CO 2 was analyzed by a gas chromatograph (GC) coupled to an isotope ratio mass spectrometer (GC-IRMS; Thermo Delta V). 140 The δ 13 C was calibrated using two standards of NHCO 3 with differing δ 13 C values dissolved in H 2 O whose isotopic compositions had been previously calibrated relative to NBS-19 using conventional dual inlet mass spectrometry (Finnigan-MAT 251). The δ 13 C values are reported relative to the Vienna Pee Dee Belemnite (VPDB) standard, and has a reproducibility of ±0.2 ‰ as determined by repeated analysis of internal DIC standards.
Samples for δ 13 C DOC were collected in 60 ml brown HDPE bottles and stored on ice until returned to the lab at 145 FIU. δ 13 C DOC samples were filtered with GF/F (0.7 µm) filter, and then stored in pre-cleaned 40 ml bottles until analyses. Measurements for δ 13 C DOC were made using a total organic carbon (TOC) analyzer (Aurora 1030W, OI Analytical) coupled to a cavity ring-down spectroscopy system (CRDS; G1111-i, Picarro) following the approach of Ya et al. (2015). DIC was removed by adding H 3 PO 4 and sparging with N 2 . 1.5 ml of sample was chemically oxidized to CO 2 at a temperature of 98°C in the presence of sodium persulfate (Na 2 S 2 O 8 ). The CO 2 generated was detected by 150 non-dispersive infrared absorption (NDIR) for determination of DOC. The CO 2 was collected in a gas-tight bag and then pulsed into the CRDS for the δ 13 C measurement. In order to measure the different isotopic ranges within the collected samples, an isotopic calibration was based on two external standards of potassium hydrogen phthalate (KHP -29.8‰, OI-Analytical) and glutamine (-11.45‰, Fisher) with a concentration range of 0-25 ppm. These standards were prepared in synthetic seawater to match the salinity of the sample matrix. The isotope values of these two 155 standards were determined by using an elemental analyzer isotope ratio mass spectrometer (EA-IRMS). Analytical precision based on replicated standards ranged from ±0.15 to ±1.52 ‰ for this study.

Underway measurements
Surface water was continuously pumped from an intake located near the bow of the boat at a water depth of approximately 1 m during tracer recovery operations. Water temperature and salinity were continuously recorded 160 using a thermosalinograph (SBE 45 MicroTSG). During SharkTREx 1, DO was measured underway with a membrane covered galvanic sensor (WTW Cellox 325) calibrated with saturated air. During SharkTREx 2, DO was measured using an oxygen optode (Aanderaa 3835) calibrated against Winkler titration.
Underway measurements of atmospheric and waterside pCO 2 were made. Waterside pCO 2 were obtained with a showerhead type equilibrator coupled to a non-dispersive infrared (NDIR) analyzer (LI-COR 840A). 165 Measurements of underway SF 6 were made with an automated SF 6 analysis system (Ho et al., 2002), which is Annotated Manuscript; Page 8 comprised of gas extraction (membrane contactor), separation (molecular sieve 5A), and detection units (gas chromatograph equipped with an electron capture detector). Both the underway pCO 2 and SF 6 measurements are described in greater detail in Ho et al. (2014)

Inventories of DIC, DOC and DO 170
The inventories of DIC, DOC and DO were calculated in the same way that SF 6 inventories were determined in Ho et al. (2014). The river was divided into 100-m longitudinal sections, and the measured concentrations, corrected for tidal movement to slack before ebb for each day, were assigned to each section i and then summed over the entire length of the river. For example, to calculate the inventory of DIC, denoted DIC $%&'()'* (mol): where DICis the mean concentration (mol L -1 ) in section i, / -is the volume of the river (L) in section i at mid-tide, and n is the number of sections in each river (n = 273 for Shark River and Tarpon Bay; n = 152 for Harney River).
DOC and DO inventories were also calculated using Eq. (1), by substituting [DOC] The inventories of DIC and DOC were separated into contributions from estuarine and non-estuarine sources, first by determining inventories for DIC assuming conservative mixing between the freshwater and marine end members and 180 then subtracting these inventories from the total observed inventories while correcting for air-water gas exchange.
The estuarine DIC inventory, [DIC] '&567(8 , representing the DIC from all estuarine sources, was calculated as follows: where [DIC] :$0&'() is the inventory of DIC assuming conservative mixing between freshwater and marine end 185 members (i.e., from non-estuarine sources), and [DIC] <7&'= is the inventory of DIC lost to air-water gas exchange from the estuary, due to pCO 2 in the water being above solubility equilibrium with the atmosphere (see section 2.6).
The freshwater and marine end-members were assigned to the values measured at the lowest (Tarpon Bay) and highest salinities, respectively.
The total O 2 deficit in Shark River during the experiments was determined by examining the difference in O 2 190 inventories for conservative mixing and actual measurements, correcting for O 2 influx due to gas exchange using a formulation similar to Eq. To enable comparison between different gases and different aquatic environments, it is customary to normalize gas transfer velocities to a Schmidt number (Sc; kinematic viscosity of water divided by diffusion 195 coefficient of gas in water) of 600, k(600), corresponding to that of CO 2 in freshwater at 20 °C. k(600) for SharkTREx 1 and 2, determined from the wind speed and current velocity parameterization proposed in Ho et al. (2016), were 3.5 ± 1.0 and 4.2 ± 1.8 cm h -1 , respectively. To determine k for O 2 and CO 2 at the temperature and salinity measured in the rivers, the following equation was used, assuming a Sc -1/2 scaling (Jähne et al., 1987): where k and Sc of CO 2 could be substituted in Eq. (3) for O 2 , and Sc for O 2 and CO 2 were calculated as a function of temperature and salinity using data compiled by Wanninkhof (2014).
Air-water O 2 fluxes (M A B ; mmol m -2 d -1 ) were calculated as follows: where @ A B (cm h -1 ) is the gas transfer velocity for O 2 , O L NOPQR (mmol m -3 ) is the equilibrium concentration of O 2 in the 205 water at a given temperature and salinity (Garcia and Gordon, 1992), and O 2 is the measured oxygen concentration in the water.
Similarly, air-water CO 2 fluxes (M SA B ; mmol m -2 d -1 ), which were used to determine changes in DIC due to gas exchange, were calculated as follows: where @ SA B (cm h -1 ) is the gas transfer velocity for CO 2 , K 0 (mol atm -1 m -3 ) is the aqueous-phase solubility of CO 2 (Weiss, 1974), and DpCO 2 (µatm) is the difference between the measured pCO 2 in air equilibrated with water and atmospheric pCO 2 .
As with the inventories, M SA B were separated into estuarine and non-estuarine contributions. Because of the non-linearity in the relationship between pCO 2 and other carbonate system parameters, the pCO 2 in the river expected 215 from conservative mixing was calculated by assuming conservative mixing for DIC and TAlk, and then calculating pCO 2 using CO2SYS (Pierrot et al., 2006), with the dissociation constants of Cai and Wang (1998). Then, the nonestuarine M SA B was calculated as above with Eq. (5), and the M SA B attributed to estuarine sources was determined as the difference between total and non-estuarine M SA B . Annotated Manuscript; Page 10

2.7
Estuarine and mangrove contributions to DIC 220 DIC in the Shark and Harney Rivers may originate from several sources in addition to input from the freshwater marsh upstream and the coastal ocean, including: 1) mangrove root respiration; 2) organic matter mineralization in sediments or in river water; 3) dissolution of CaCO 3 in sediments or in river water; and 4) groundwater discharge. Groundwater in this region is likely to contain DIC from CaCO 3 dissolution that occurs when saltwater intrudes into the karst aquifer that underlies this region (Price et al., 2006), as well as DIC from sediment 225 organic matter mineralization. In this setting, the combination of #1 and #2 represents the mangrove source of DIC Measurements of d 13 C DIC and estuarine DIC/TAlk ratios were used to determine the mangrove sources to estuarine DIC. Fixation of CO 2 through photosynthesis is neglected in both models as these rivers are characterized 235 by low chlorophyll-a concentration and low phytoplankton biomass (Boyer et al., 1997). During SharkTREx 1 and 2, there was a negligible difference between pCO 2 measured during the day and night (ca. 3%).

2.7.1
Determining mangrove contribution from d 13 C DIC Processes 1 through 4 listed above influence d 13 C DIC in the estuary differently due to the differences in the d 13 C values originating from respiration of mangrove-derived organic matter, and CaCO 3 dissolution. The isotopic 240 fractionation during respiration of organic matter is small, and the d 13 C DIC values produced via this pathway should be approximately equivalent to the d 13 C of the organic matter respired (DeNiro and Epstein, 1978). The isotopic fractionation during dissolution/re-precipitation of CaCO 3 is also considered to be negligible (Salomons and Mook, 1986).
The expected d 13 C values of DIC in the rivers as a result of conservative mixing (d 13 C conserv ) of the marine 245 and freshwater end-members of the Shark and Harney Rivers were calculated as follows (Mook and Tan, 1991): where [DIC] is the observed DIC concentration, S is the measured salinity, and M and F subscripts refer to the marine and freshwater end-members, respectively.
where the d 13 C conserv value is the DIC isotopic composition expected for conservative mixing (Mook and Tan, 1991), is the equilibrium isotope fractionation between DIC and CO 2 gas (~8‰; Zhang et al., 1995).

Determining mangrove contribution from TAlk/DIC
An independent approach to separate the mangrove contribution from CaCO 3 dissolution is to use the co-  (Alongi, 1998;Alongi et al., 2005)

Determining mangrove contribution to DOC
In the Shark and Harney Rivers, dissolved organic matter may be derived from upstream freshwater wetland species such as periphyton and sawgrass, from seagrass communities and marine phytoplankton, or from mangrove vegetation inside the estuary (Jaffe et al., 2001). The estuarine contributions to DOC ([DOC] estuary ) in the rivers was determined in the same way as for DIC above using Eq. (6), by substituting DOC for DIC accordingly, without the 270 correction for gas exchange: Then, measurements of d 13 C DOC were made to ascertain the mangrove source of DOC in the river, in order to determine the proportion of [DOC] estuary that is of mangrove origin. The expected d 13 C values of DOC as a result of conservative mixing ( δ 2Z C :$0&'() ) were calculated using Eq. (7) where d 13 C mangrove is the isotopic composition for mangrove-derived material (-30‰).

Longitudinal dispersion
The longitudinal SF 6 distribution was corrected for tidal movement to slack water before ebb for each day using a method described in Ho et al. (2002). The absolute magnitudes of the average daily corrections were 2.0 and 285 2.7 km for SharkTREx 1 and 2, respectively, with a range for individual measurements of 0 to 5.8 km and 0 to 7.3 km for SharkTREx 1 and 2, respectively. Longitudinal dispersion coefficient K x (m 2 s -1 ) was calculated from the change of moment of the longitudinal SF 6 distribution over time as follows (Fischer et al., 1979;Rutherford, 1994): where σ x L is the second moment of the longitudinal SF 6 distribution for each day. 290

Longitudinal fluxes to the Gulf of Mexico
The longitudinal fluxes of DIC and DOC from Shark and Harney Rivers to the Gulf of Mexico were calculated using the averaged DIC or DOC inventories, and the residence time of water (τ; d), which was determined from the decrease in the inventory of SF 6 after correcting for air-water gas exchange (Ho et al., 2016). For example, the longitudinal DIC flux (M [\S ; mol d -1 ) can be calculated as follows (e.g., Dettmann, 2001): 295 Equation (12) can be used to calculate the fluxes of any other dissolved or suspended substance in the river by substituting its inventory in place of DIC. In addition, using the estuarine and non-estuarine fractions of the inventories in equation (12) allowed the estuarine and non-estuarine proportions of the longitudinal carbon fluxes to be quantified.
The advantage of this method to calculate longitudinal flux in a tidal river over a method that uses net 300 discharge and constituent concentration is that the effect of tidal flushing is implicitly accounted for by the residence Annotated Manuscript; Page 13 time, and therefore there is not a need to explicitly define the fraction of river water in the return flow during each flood tide.
Both pCO 2 and DO showed large spatial variability within the Shark and Harney Rivers during SharkTREx 1 and 2 (Fig. 2). Measured pCO 2 values were well above atmospheric equilibrium along the entire salinity range, with 310 values ranging from ca. 1000 to 6200 µatm. Maximum pCO 2 values were observed at intermediate salinities, decreasing towards both end-members, while DO showed the opposite pattern, with saturations ranging from 36 to 113%.
The patterns of TAlk and DIC along the salinity gradient followed the same trend as pCO 2 and were clearly non-conservative ( Fig. 3a-f). TAlk varied between ca. 3400 and 5000 µmol kg -1 during SharkTREx 1 and between ca. 315 3000 and 3900 µmol kg -1 during SharkTREx 2. DIC ranged from ca. 3400 to 5100 µmol kg -1 during SharkTREx 1, and ca. 2800 to 4000 µmol kg -1 during SharkTREx 2. d 13 C DIC values ranged from -10.3 to -6.6 ‰ and from -11.4 to -5.8 ‰ during SharkTREx 1 and 2, respectively. Higher DIC, TAlk and pCO 2 coincided with lower O 2 saturation, more depleted d 13 C DIC , and lower pH values (Fig. 3g-i), indicative of mineralization of mangrove-derived organic matter within the estuary. 320 During SharkTREx 1, the DOC concentrations in the freshwater end member were higher than SharkTREx 2 (Fig. 4). For both experiments, DOC concentrations followed a non-conservative pattern (see also Cawley et al., 2013), but this trend was less apparent during SharkTREx 1 compared to SharkTREx 2 (Fig. 4).
The inventories of DIC, DOC, DO, TAlk, and pCO 2 were relatively constant in the Shark and Harney Rivers, indicating quasi steady state conditions during SharkTREx 1 and 2. Under these conditions, carbon inputs and exports 325 are balanced, and fluxes and concentrations may be examined interchangeably. K x during the experiments (16.4 ± 4.7 and 77.3 ± 6.5 m 2 s -1 for Shark River during SharkTREx 1 and 2, respectively, and 136.1 ± 16.5 m 2 s -1 for Harney Annotated Manuscript; Page 14 River during SharkTREx 2) were relatively large, and suggest that any perturbations (such as export of DIC from mangroves) would be quickly mixed thoroughly in the estuary.
In the following, for brevity, fluxes and inventories are summarized as ranges, which cover the two rivers 330 and two experiments so they reflect both temporal and spatial variability. The individual values are given in Tables 1   and 2. DIC was the dominant form of dissolved carbon in both rivers and accounted for 79 to 82% of the total dissolved carbon in the rivers. The contribution of DOC to the total carbon pool varied between 18 and 21% (Table   1). 335

Air-water CO 2 fluxes
As shown by Ho et al. (2014), pCO 2 observed during SharkTREx 1 and 2 fall in the upper range of those reported in other estuarine (Borges, 2005) and mangrove-dominated systems (Bouillon et al., 2003;Bouillon et al., 2007a;Bouillon et al., 2007b;Koné and Borges, 2008;Call et al., 2015). The mean air-water CO 2 fluxes in Shark River for SharkTREx 1 and 2 were 105 ± 9 and 99 ± 6 mmol m -2 d -1 (Ho et al., 2016). The analysis is taken further 340 here by including data from Harney River. The mean air-water CO 2 fluxes in Harney River were 150 ± 8 and 114 ± 21 mmol m -2 d -1 for SharkTREx 1 and 2, respectively. Borges et al. (2003) summarized all available pCO 2 data from mangrove surrounding waters, and calculated CO 2 fluxes to the atmosphere that averaged 50 mmol m -2 d -1 (with a range of 4.6 to 113.5 mmol m -2 d -1 ), and Bouillon et al. (2008a) estimated a global CO 2 flux from mangroves of ca. 60 ± 45 mmol m -2 d -1 . One reason that the fluxes 345 from SharkTREx 1 and 2 are on the upper end of those estimates may be that the Shark and Harney Rivers receive a large input of DIC from the freshwater marsh upstream (Table 1), causing higher pCO 2 in the estuary compared to the global average.

Scaling the air-water CO 2 fluxes by the area of open water in the Shark and Harney Rivers, where Tarpon
Bay is included with Shark River, suggests that the total carbon emissions to the atmosphere through air-water gas 350 exchange in Shark River was 4.2 ± 0.4 x 10 5 and 4.0 ± 0.2 x 10 5 mol d -1 during SharkTREx 1 and 2, respectively, and were 4.1 ± 0.2 x 10 5 and 3.1 ± 0.6 x 10 5 mol d -1 from the Harney River during SharkTREx 1 and 2, respectively ( Fig.   5), which is remarkably consistent, both spatially and temporally.
These fluxes were incorporated into the DIC mass balance of the Shark and Harney Rivers (Eq. 2) by calculating the total CO 2 degassed over the residence time of water in the rivers. Given the mean air-water CO 2 fluxes 355 Annotated Manuscript; Page 15 (Table 2), the total CO 2 degassed in the Shark River represents approximately 13 and 21% of [DIC] $%&'()'* during SharkTREx 1 and 2, respectively, and the CO 2 degassed from the Harney River during SharkTREx 2 represents 20% of [DIC] $%&'()'* , indicating that air-water CO 2 exchange removes a non-negligible fraction of the inorganic carbon in these rivers. Exclusion of [DIC] <7&'= from the mass balance in Eq. (2) would lead to an underestimation of [DIC] '&567(8 of between 33 and 44%. 360

Mangrove contribution to DIC inventory
The highest DIC concentrations were correlated with low DO (Fig. 2) and characterized by 13 C-depletion ( Fig. 3j, k, l). Observations of elevated DIC and pCO 2 in the middle of the estuary, coupled with d 13 C DIC and O 2 depletion may indicate the importance, noted by other authors, of lateral transport of pore water from the peat-based mangrove forest into the river via tidal pumping (Bouillon et al., 2008a;Maher et al., 2013). However, as demonstrated 365 below, the observed DIC and d 13 C DIC distributions in these rivers cannot be explained solely by mineralization of mangrove-derived organic carbon.

Evidence from d 13 C DIC
The distributions of DIC and d 13 C DIC cannot be explained solely by the addition of mangrove-derived DIC and air-water gas exchange. Solving Eq. (8) for d 13 C DIC , assuming that [DIC] dissolution is negligible and that the only 370 source of DIC in the rivers is of mangrove origin, would result in d 13 C values significantly lower than those observed.
The low pH in interstitial waters of mangrove sediments due to organic matter mineralization processes may be favorable to CaCO 3 dissolution in mangrove sediments, and this process could have an effect on estuarine d 13 C DIC .
Groundwater discharge could also influence DIC and d 13 C DIC . Inputs of DIC derived from CaCO 3 dissolution from either of these sources may explain the differences in observed d 13 C DIC and those expected if [DIC] estuary was entirely 375 of mangrove origin.
Solving Equations 6 and 8, the mineralization of mangrove-derived organic matter is estimated to account for ca. 60 ± 6 % of [DIC] '&567(8 (Table 3), with the remainder originating from the dissolution of CaCO 3 . This estimate is sensitive to the end member value chosen for d 13 C mangroves and d 13 C dissolution . For instance, if d 13 C mangroves were -29‰ instead of -30‰, the mangrove contribution would increase to 62%. 380 Annotated Manuscript; Page 16

Evidence from DIC and TAlk
In the Shark and Harney Rivers, the high correlation (r 2 = 0.99; Fig. 6 (Table 3), with the remainder due to the dissolution of CaCO 3 . These estimates are in reasonable agreement with those based on 385 the carbon isotopic mass balance.
The [TAlk] estuary vs.
[DIC] estuary ratios were 0.84 and 0.92 for Shark River during SharkTREx 1 and 2, and 0.90 for the Harney River during SharkTREx 2 (Fig. 6). The TAlk to DIC ratios for CaCO 3 dissolution, sulfate reduction, and aerobic respiration are -0.2, 0.99, and 2, respectively. Hence, in order to achieve the observed ratios, and given the estimated contribution of CaCO 3 dissolution to [DIC] '&567(8 of ca. 30%, sulfate reduction and aerobic 390 respiration were estimated to contribute 32 to 39% and 31 to 38%, respectively.

Evidence from DO
The deficit of O 2 in Shark River was found to be 2.7 ± 0.7 x 10 6 and 3.7 ± 0.3 x 10 6 mol during SharkTREx 1 and 2, respectively. Assuming a stoichiometric ratio of ca. 1.1 for O 2 to CO 2 during degradation/remineralization of terrestrial organic matter (Severinghaus, 1995;Keeling and Manning, 2014), the maximum contribution of aerobic 395 respiration to the DIC added to the estuary was estimated to be 57 to 69%. However, O 2 may also be consumed during oxidation of reduced products from anaerobic metabolism, such as H 2 S, Mn 2+ or Fe 2+ , with similar O 2 to CO 2 stoichiometry as aerobic respiration. Hence, the numbers derived above represent an upper limit for aerobic respiration, and if there were complete re-oxidation of metabolites from anaerobic respiration, the O 2 deficit would represent total mineralization of terrestrial organic matter instead of just aerobic respiration. The mangrove 400 contributions estimated from d 13 C DIC (section 3.3.1) and TAlk/DIC (section 3.3.2) are consistent with this analysis of the O 2 deficit, which indicates that a minimum of 57-69% of [DIC] '&567(8 derived from the mineralization of organic matter.

Mangrove contributions to DOC inventory
During both experiments, the d 13 C DOC was highly depleted, indicative of contribution from higher plants, 405 including mangroves. During SharkTREx 1, the lowest observed d 13 C DOC value (-31.6‰) was in the mid-estuary (i.e., from salinity of ca. 10 to 20) (Fig. 4d). Previous studies of DOC from mangrove-dominated systems have reported Annotated Manuscript; Page 17 values as low as -30.4‰ (Dittmar et al., 2006), and some of the more depleted samples from SharkTREx 1 might have DOC sourced from algae associated with mangrove roots, which can have relatively depleted values (Kieckbusch et al., 2004). The overall d 13 C DOC depletion was less during SharkTREx 2, and the overall distribution was indicative of 410 a stronger marine influence and/or mixing (Fig. 4e, f). The marine end member had a more enriched d 13 C DOC , indicating a greater contribution of seagrass and/or marine phytoplankton derived organic matter to the marine DOC pool (Anderson and Fourqurean, 2003). These observations are consistent with the greater longitudinal dispersion observed during SharkTREx 2 compared to SharkTREx 1.
The calculations of mangrove contribution using d 13 C DOC mass balances (Eq. 10) also suggest that the 415 majority of [DOC] estuary , but only a small percentage of the total DOC inventory, was derived from mangroves (7 and 5% in the Shark River during SharkTREx 1 and 2, and 7% in the Harney River during SharkTREx 2).

Longitudinal fluxes to the Gulf of Mexico and comparison with previous studies
Residence times of Shark River (including Tarpon Bay) for SharkTREx 1 and 2 were, 5.8 ± 0.4 and 8.1 ± 1.1 days, respectively (Ho et al., 2016), and that of Harney River was 4.7 ± 0.7 days for SharkTREx 2. The resulting 420 longitudinal DIC fluxes to the Gulf of Mexico (15.8 to 33.6 x 10 5 mol d -1 ) were significantly larger than the longitudinal DOC fluxes (3.3 to 7.5 x 10 5 mol d -1 ) at salinity of ca. 27 ( Fig. 5; Table 2).
There are no previously published DIC inventories or fluxes for the Shark and Harney Rivers, so comparison with previous studies is focused on the DOC results. The DOC flux from the Shark River to the coastal ocean in SharkTREx 1 (7.5 ± 0.2 x 10 5 mol d -1 ) is in very good agreement to that estimated by Bergamaschi et al. (2011) in an 425 experiment conducted in the Shark River from 20-30 September 2010 (7.6 ± 0.5 x 10 5 mol d -1 ). However, the net discharge during the Bergamaschi et al. (2011) study was higher than SharkTREx 1 (mean ± s.d.: 9.1 ± 7.1 vs. 6.9 ± 5.3 m 3 s -1 ), which would lead to a shorter residence time of 4.6 days using a relationship presented in Ho et al. (2016).
Using the DOC concentration data presented in Bergamaschi et al. (2011) yields an inventory that is ca. 3% higher than the DOC inventory in Shark River during SharkTREx 1. Calculations using the shorter residence time and higher 430 DOC inventory yields a DOC flux of 9.7 ± 0.2 x 10 5 mol d -1 , which is ca. 30% higher than the estimates of Bergamaschi et al. (2011).

Annotated Manuscript; Page 18
Mangroves contributed 4% to 6% of the total longitudinal DOC flux in the Shark River and 7% in the Harney River during SharkTREx 2 (Tables 1 and 4). Cawley et al. (2013), estimated a mangrove contribution to DOC flux of 3 ± 10% for Shark River and 21 ± 8% for the Harney River during November 2010, the same time period as SharkTREx 1. DOC measurements were not made in Harney River as part of SharkTREx 1. However, using the November 2010 DOC data from Harney River collected by Cawley et al. (2013) for inventory calculations, along with 440 residence time derived from the tracers, a mangrove contribution of 19% to the total DOC longitudinal flux to the Gulf of Mexico was obtained.

Distribution of carbon fluxes
During SharkTREx 1 and 2, [DIC] '&567(8 made up 20-28% of the total DIC in the rivers, and [DOC] '&567(8 made up only 4 to 7% of the total DOC in the rivers. Mangroves are estimated to contribute 13 to 19% 445 to the total DIC inventory. In all cases, the mangrove contribution to the DIC inventory is a factor of 3 greater than the mangrove contribution to the DOC inventory (Table 1).
During SharkTREx 1 and 2, the inventory of mangrove-derived DIC exceeded that of DOC by a factor of 15 to 17, which supports the idea that a large fraction of the carbon exported by mangroves to surrounding water is as DIC (Bouillon et al., 2008a), but is considerably larger than the estimates of ca. 3 to 10 compiled by Bouillon et al. 450 (2008a) for mangroves at 5 sites in Asia and Africa.
The total dissolved carbon fluxes from all sources (i.e., freshwater wetland, mangrove, carbonate dissolution, and marine input) out of the Shark and Harney Rivers during SharkTREx 1 and 2 are dominated by inorganic carbon (82-83%; see Tables 2 and 4), either via air-water CO 2 exchange or longitudinal flux of DIC to the coastal ocean (Fig.   5). The remaining 17-18% of the export is as DOC. This proportioning is remarkably similar between SharkTREx 1 455 and 2, and between the Shark and Harney Rivers (Table 1). The estuarine contribution to these fluxes is relatively small (generally <15%), with the exception of air-water CO 2 flux, where the estuary contribution was 49 to 63% (Table 4).
In this study, the particulate organic carbon (POC) flux was not examined. However, He et al. (2014) estimated the mangrove-derived POC flux in Shark River by taking the total volume discharge from the five major 460 rivers along the southwest coast of Everglades National Park from 2004 to 2008, and assuming that Shark River contributed 14% to the mean annual discharge. They then multiplied this discharge by the average POM concentration Annotated Manuscript; Page 19 (5.20 ± 0.614 mg L -1 ) in the middle of the estuary to yield an annual POM flux from Shark River. Based on analysis of organic matter biomarkers, He et al. (2014) estimated that mangrove-derived POM was 70-90% of the total POM pool in the Shark River. Using this contribution and further assuming that 58% of POM weight is POC (Howard, 465 1965), they estimated a POC flux of 1.0 to 2.2 x 10 4 mol d -1 . Because this estimate was based on biomarker and POM data from the mid-estuary, where the POM concentration and the mangrove contribution to POM are both likely to be much higher than either toward the freshwater end member or the marine end member, it is likely an overestimate of the mangrove derived POC flux. Nevertheless, the mangrove-derived POC flux determined by He et al. (2014) is still only a small fraction (3 to 7%) of the mangrove-derived dissolved carbon fluxes in Shark River during SharkTREx 1 470 and 2.

Mangrove contributing area and estuary carbon balance
One of the challenges of relating the results reported here to other studies is to scale the results to a mangrove contributing area, and thereby relate the findings to mangrove forest carbon balance, typically expressed on an aerial basis. Estimates of forest carbon export derived here are compared with other investigations in this estuary. The entire 475 area of mangroves surrounding the Shark and Harney Rivers region is ca. 111 km 2 , and the water area is ca. 17.5 km 2 (Ho et al., 2014). Scaling the forest area by the water area of Shark River (2.5 km 2 ) yields an associated forest area of 15.9 km 2 . The forest area associated with Harney River (2.8 km 2 ) is 17.4 km 2 .
Using the total forest area associated with Shark River to scale estimates of total export of mangrove-derived carbon (the combination of longitudinal fluxes and air-water gas exchange) suggests an average dissolved carbon 480 lateral export rate from the forest of 18.9 to 24.5 mmol m -2 d -1 , including both DIC and DOC. However, since it is unknown what fraction of the total forest area associated with these rivers exported dissolved carbon through tidal pumping (a function of tidal height and duration), this is considered to be a minimum estimate. Average water levels at high tide during SharkTREx 1 and 2 at the USGS Shark River station were 88% and 95% of maximum wet season water levels reported at this site over the period from November 2007 to December 2012 (U.S. Geological Survey, 485 2016), and 12 inundation events occurred during both SharkTREx 1 and 2. Water levels in the main river channel at the USGS Shark River station were above an estimate of the average minimum ground surface elevation derived from nearby groundwater monitoring wells in the estuary (sites SH3 and SH4; http://sofia.usgs.gov/eden/stationlist.php) for 21% and 28% of the time during the SharkTREx 1 and 2 experimental periods, respectively. These values indicate the export of dissolved carbon from flooded portions of the forest during the discontinuous inundation periods should be 490 Annotated Manuscript; Page 20 significantly greater than the dissolved carbon lateral export rate derived above in order to produce the observed inventories of mangrove-derived dissolved carbon in the main channel. Bergamaschi et al. (2011) proposed an annual total DOC export from the forest surrounding Shark River of 15.1 ± 1.1 mol m -2 y -1 and describe their method of calculating contributing area using a model based on the relationship between discharge volume and changes in water levels during tidal cycles. They do not provide a 495 contributing area, but this can be calculated from their results. They determined longitudinal DOC fluxes of 7.6 ± 0.5 x 10 5 and 1.3 ± 0.02 x 10 5 mol d -1 for the wet and dry seasons, respectively, and assumed that they are entirely of mangrove origin. Given the lengths of the wet and dry seasons, this would yield a mean annual DOC flux of 3.9 ± 0.2 x 10 5 mol d -1 , and 9.4 ± 0.7 km 2 of mangrove forest contributing to carbon fluxes thru tidal flushing in this segment of Shark River. However, data from SharkTREx 1 and 2 indicate that ca. 5% of the total longitudinal DOC fluxes 500 were of mangrove origin, with an average mangrove-derived DIC to DOC flux ratio of 10.5. Using this information, the Bergamaschi et al. (2011) results were recalculated to yield a wet season dissolved carbon lateral export rate of 46.5 ± 4.4 mmol m -2 d -1 (as DIC and DOC) from the forest.
The rate of mangrove-derived carbon exported to estuarine waters is likely to vary over space and time, as a result of factors that include tidal cycles, phenology, and forest and soil structural characteristics. For example, Bergamaschi et al. (2011) found that DOC fluxes were 6 times higher during the wet season (September) than the dry season (April), whereas Cawley et al. (2013) found that the DOC fluxes were 4 and 10 times higher during the wet vs. 515 dry season (November vs. March) in the Shark and Harney Rivers, respectively. Barr et al. (2013) showed that forest respiration rates derived from NEE data are greater during the wet than dry seasons. Higher respiration rates combined with increased inundation during the wet compared to dry seasons suggest that wet season DIC export will also be Annotated Manuscript; Page 21 greater than dry season values. For these reasons, the annual carbon export rates derived from the difference between NECB and NEE are expect to underestimate wet season values. If annual lateral carbon export rates are considered as 520 equivalent to a time-weighted sum of dry season (7 months) and wet season (5 months) values (after Bergamaschi et al. 2011), and wet season export is assumed to be, for example, 5 times greater than dry season values, the seasonal export rates (15 and 75 mmol m -2 d -1 for dry and wet seasons, respectively) that correspond with the difference between annual NECB and NEE can be calculated.
The discrepancies between the estimates of carbon export rates derived here, and those derived from 525 Bergamaschi et al. (2011) and the difference between NEE and NECB point out the need for additional studies to reduce the uncertainty in the relationships between riverine carbon fluxes, forest carbon export, and estimates of contributing areas. For example, Bergamaschi et al. (2011) conducted an Eulerian study at a single location in the middle of the estuary, where the mangrove influence might be higher than the Lagrangian study conducted during SharkTREx 1 and 2, which covered the entire estuary. Also, the estimate of forest carbon export based on the 530 difference between NEE and NECB is from a single location along Shark River (at FCE LTER site SRS6), and may not be representative of the entire forest. Furthermore, forest lateral carbon export rates and contributing areas should be considered dynamic, varying over semi-diurnal time scales with the extent and duration of inundation during individual tidal cycles. The correct interpretation of a single, static value for contributing area such as derived above is therefore uncertain, since the tracer-based results represent an integration of carbon sources and sinks calculated 535 over the water residence time and expressed on daily time scales. To improve understanding of how mangrove forest carbon balance and export influence riverine carbon inventories and fluxes to the Gulf of Mexico in this system, wet and dry season measurements over multiple years, information on the relationships between forest structure, productivity and lateral carbon export rates, and independent estimates of forest inundation area in relation to tidal height are needed. 540

Conclusions
The SharkTREx 1 and 2 studies are the first to provide estimates of longitudinal DIC export, air-water CO 2 fluxes, and mangrove-derived DIC inputs for the Shark and Harney Rivers. The results show that air-water CO 2 exchange and longitudinal DIC fluxes account for ca. 90% of the mangrove-derived dissolved carbon export out of the Shark and Harney Rivers, with the remainder being exported as dissolved organic carbon. 545 Annotated Manuscript; Page 22 The mangrove contribution to the total longitudinal flux was 6.5 to 8.9% for DIC and 4 to 18% for DOC. A lower bound estimate of the dissolved carbon export (DIC and DOC) from the forest surrounding Shark River during the wet season was 18.9 to 24.5 mmol m -2 d -1 with 15.9 km 2 of mangrove contributing area. This basin-scale estimate is somewhat lower by comparison than other independent estimates of lateral carbon export from this mangrove forest.
However, mangrove forest carbon export rates on an aerial basis are expected to vary with the spatial and temporal 550 scales over which they are calculated, and depend on factors such as tidal inundation frequency, distance from the riverbank and the coast, and forest and soil characteristics.
Future experiments should investigate the contribution of DIC from groundwater to the rivers, by making measurements of d 13 C DIC of groundwater, Sr and Ca concentrations in the river to quantify CaCO 3 dissolution and to separate carbonate alkalinity from TAlk, radon to quantify groundwater discharge, 14 C DIC to separate input of DIC 555 from remineralization of organic matter from dissolution of CaCO 3 . Experiments should also examine the seasonal variability in the carbon dynamics and export, by conducting process-based studies like SharkTREx during both wet and dry seasons. Also, time series measurement of current velocities, wind speeds, pCO 2 and pH (to calculate DIC), DO, chromophoric dissolved organic matter (CDOM, as a proxy for DOC), and radon will also allow the temporal variability of the sources and sinks of DIC in these rivers to be examined. 560

Author contribution
D. Ho, S. Ferron, and V. Engel conceived and executed the experiment, interpreted the data, and prepared the manuscript with input from the other authors. W. Anderson measured the samples for δ 13 C of dissolved organic carbon, P. Swart measured the samples for δ 13 C of dissolved inorganic carbon, R. Price measured the total alkalinity samples for SharkTREx 1, and L. Barbero measured the total alkalinity and dissolved inorganic carbon samples for SharkTREx 565 2.

Data availability
The pCO 2 data collected during SharkTREx 1 and 2 are available from the SOCAT database <www.socat.info>. The other data may be obtained by contacting the corresponding author. The uncertainty in the observed and non-estuarine inventories are the standard deviations of the inventories for all the days of the experiment. The estuarine 750 contribution is calculated from the observed and non-estuarine contribution, and the uncertainty is from propagating the errors of the two. The uncertainty in contribution from gas exchange is from propagating the uncertainty in CO 2 flux and the residence time. The uncertainty in mangrove contribution is calculated from propagating the error from the estuarine contribution. c Proportion of each form of carbon (i.e., DIC, DOC) relative to the total mangrove-derived carbon pool. 755 d Proportion of each form of carbon (i.e., DIC, DOC) relative to the total carbon pool. e Estuarine DOC is assumed to be entirely of mangrove origin.  SharkTREx 2 61 ± 6% 70 ± 3% Harney River SharkTREx 1 --SharkTREx 2 61 ± 6% 70 ± 2%   Harney Rivers during the 2010 (SharkTREx 1) and 2011(SharkTREx 2) campaigns. During SharkTREx 1, TAlk and pH were measured at FIU, and DIC was calculated using CO2SYS (Pierrot et al., 2006). During SharkTREx 2, DIC and TAlk were measured at NOAA/AOML, and pH was calculated using CO2SYS. The dashed lines indicate the distribution expected for conservative mixing.   Figure 6. (a) Covariation of DIC estuary and TAlk estuary . Black squares are samples from the Shark River during SharkTREx 1, and black and gray circles are from the Shark and Harney Rivers, respectively, during SharkTREx 2. Dotted lines represent the theoretical covariation of DIC and TAlk for different biogeochemical processes: 1) aerobic respiration; 2) CO 2 emission, 3) sulfate reduction, 4) CaCO 3 dissolution, 5) manganese reduction, and 6) iron reduction. 810