Reduction in Riverine Freshwater Supply Changes Inorganic and Organic Carbon Dynamics and Air‐Water CO2 Fluxes in a Tropical Mangrove Dominated Estuary

Reduction in riverine freshwater supply due to climate change as well as anthropogenic activities are documented throughout the globe. How river discontinuity in upstream reaches and the subsequent reduction in freshwater influx alter inorganic and organic carbon dynamics in downstream estuaries adjacent to mangroves has been rarely reported. We investigated the dynamics of the inorganic carbon system and organic matter (OM) in two Indian estuaries near mangroves; riverine freshwater supply to the Matla Estuary was reduced and that to the Dhamra Estuary was uninterrupted. Seasonal sampling was conducted over an annual cycle. We used elemental and stable isotope signatures to delineate the sources of OM and dissolved inorganic carbon (DIC). We found that compared to the Dhamra, the reduced riverine freshwater supply to the Matla increased the marine influence in the estuary on the OM degradation pathways and decreased CO2 emissions to the atmosphere. In the Dhamra, higher seasonal variability in biogeochemical pathways, facilitated high internal carbonate buffering capacity; in contrast in the Matla, the greater marine influence increased the carbonate buffering capacity, resulting in retention of higher DIC concentrations and low CO2 emissions. Dissolved and particulate organic carbon concentrations were higher in the Dhamra than the Matla, indicating higher riverine supply of these. The present study can contribute an overlooked effect of long‐term changes in riverine freshwater supply on the carbon dynamics of mangrove‐dominated estuaries, which might help to improve the understanding of coastal carbon budgets in a changing world.

physical forcing can also govern the air-water CO 2 fluxes from estuaries (Van Dam, Crosswell, Anderson, & Paerl, 2018). Recent estimates suggest that global estuaries emit CO 2 at a rate of about 0.25 PgC yr −1 (Ward et al., 2017;Tanner and Eyre, 2020, and references therein). Friedlingstein et al. (2019) reported a global oceanic CO 2 sink of about 2.6 ± 0.6 Pg C yr −1 ; hence, estuarine CO 2 emissions to the atmosphere may counterbalance on the order of 10% of oceanic CO 2 absorption. However, production of CO 2 in the estuarine water column along with the fate of several other biogeochemical processes depends upon various factors that are related to the geophysical characteristics of each specific estuary, including freshwater loads (Li et al., 2019;Yao et al., 2020); tidal amplitude (Crosswell et al., 2012;Joesoef et al., 2015); water residence time (Joesoef et al., 2017); vertical stratification (Koné et al., 2009); connectivity with salt marshes or tidal flats (Cai, 2011); and biogeochemical characteristics, such as trophic status (Cotovicz Jr. et al., 2015) and net community production (Jiang et al., 2019).
In the last century, both climate change and anthropogenic interventions have drastically transformed the geomorphology, structure, and function of estuaries throughout the world (Adel, 2002;Hooke, 2000;Jaffe et al., 2007;Poirier et al., 2011). Land-use changes like the construction of dams and the destruction of vegetation often lead to drastic increases in soil erosion (Adel, 2002;Hewawasam et al., 2003), which in turn enhance sediment transport toward rivers and seas (Walling, 2006) with the consequent rapid siltation and drying-up of estuarine and marine water bodies (Dearing & Jones, 2003). At present, many rivers and estuaries throughout the world are under environmental stress due to factors like climate change, impoundments, and water pollution. Several studies have strongly emphasized that human-as well as climate change-induced perturbations should be considered as key drivers that can modulate riverine and estuarine biogeochemistry (Jin et al., 2018;Kaushal & Belt, 2012;Park et al., 2010;Regnier et al., 2013). Several major rivers of the world are experiencing dramatic change in discharge rate due to both climate change and population pressure, and this water stress is diminishing these rivers' natural ability to adjust to and absorb disturbances (Palmer et al., 2008). These unprecedented changes in rivers and estuaries may create discontinuities in the river-estuary-sea continuum, which are capable of substantially altering the metabolic activities and CO 2 dynamics of these water bodies . However, globally, the observations of the biogeochemical aspects of estuaries are skewed in favor of the northern hemisphere, with fewer studies in the tropics (Vieillard et al., 2020).
Studies on the carbon dynamics of river systems that are impacted by climate change, anthropogenic perturbations, or both are still few in number and are especially lacking in Asian rivers and estuaries (Jin et al., 2018;Park et al., 2018;Ward et al., 2017), despite the fact that Asian rivers account for 40%-50% of the global organic and inorganic lateral fluxes to the open ocean (Dai et al., 2012). Most of the literature related to the changes in carbon dynamics in rivers are those affected by construction of dams (Maavara et al., 2017) or other impoundments (Hu & Cheng, 2013), wastewater draining through urbanized watersheds (Yoon et al., 2017), and eutrophication (Billen et al., 2007). However, changes in carbon dynamics in mangrove-dominated estuaries affected by reduced riverine freshwater supply have been overlooked.
The main objectives of this study were to investigate differences in the dynamics of inorganic carbon and organic matter (OM) in two mangrove-dominated estuaries, one for which the inflowing river is interrupted in the upstream reaches, which reduces riverine freshwater inflow, and the other which receives an uninterrupted perennial flow of freshwater. We focus on (i) the seasonal dynamics of dissolved inorganic carbon (DIC) and the diagenetic OM mineralization pathways that affect DIC and total alkalinity (TAlk), (ii) the differences of organic carbon pools and OM sources in the water column particulate organic matter (POM) and riverbed sedimentary OM, and (iii) the spatial and seasonal variability of the partial pressure of CO 2 in surface waters (pCO 2 [water]) and air-water CO 2 flux measured at high temporal resolution (1-min) to minimize the uncertainty that often arises due to omission of the flux data during the tidal maxima and minima (Rosentreter et al., 2018). The two Indian estuaries studied here are surrounded by mangroves: The Matla Estuary (freshwater interrupted, within the Sundarbans mangrove forest) and the Dhamra Estuary (freshwater uninterrupted, adjacent to the Bhitarkanika mangrove forest) (Figure 1). We hypothesized that dissolved and particulate organic matter, CO 2 dynamics, and air-water CO 2 fluxes differ substantially between the interrupted and the uninterrupted estuaries. We tested this hypothesis by sampling inner, middle, and outer stations of the two estuaries four times throughout an annual cycle covering all the seasons to minimize spatial and seasonal biases. AKHAND ET AL. freshwater flow into the Matla and Dhamra Estuaries to compare the carbon dynamics and other variables. The mangrove forest area is larger in the Sundarbans (4,263 km 2 ) than Bhitarkanika (672 km 2 ). We argue that since the present study has not estimated the lateral flux of materials, but rather described the different phenomena occurring due to reduced riverine freshwater supply, the effect of different mangrove area does not hamper the main purpose of the comparison.
Major estuaries of the Sundarbans region, like the Matla, Thakuran, Bidya, and Saptamukhi, have almost completely lost their connection with upstream rivers and at present act as tidal creeks (Roshith et al., 2018). The Hooghly River is the only river that still acts as a perennial source of riverine freshwater to the Indian Sundarbans, mainly through the Hatania Doyania Canal (Ray et al., 2018). After entering the Indian Sundarbans, the riverine freshwater from the Hooghly River spreads to the Indian part of the Sundarbans through other waterways. The Matla Estuary is located in the central part of the Indian Sundarbans, which experiences semi-diurnal tides with amplitudes ranging from 2.5 to 7 m and rapid tidal currents (>1 m s −1 ) (De et al., 2011). The upstream of the Matla has become clogged due to heavy siltation, continuing since the late fifteenth century (Chaudhuri & Choudhury, 1994), and the riverine freshwater supply decreased (Mitra et al., 2009;Raha et al., 2012). This estuary used to receive direct freshwater flow from the Hooghly; however, the natural eastward tilt of the Hooghly toward the Padma (currently flowing in Bangladesh) initiated decreases in water flow in the upper reaches of the Matla and other tributaries flowing through the Indian Sundarbans (Hist, 1915). To maintain the perennial flow of the Hooghly, the Farakka barrage was constructed 400 km upstream from the river mouth. However, the amount of water that flows through the barrage into the Hooghly Estuary was insufficient to rejuvenate the already decayed distributaries of the Indian Sundarbans (Bhadra et al., 2017).
At present, the Matla acts merely as an arm of the Bay of Bengal, and its estuarine nature (annual mean salinity about 20, Akhand et al., 2016) is mainly maintained by monsoon-driven runoff and tidal activity (Chatterjee et al., 2013). The Matla is at present experiencing heavy sedimentation, making the entire estuary shallow (2-10 m), except for some tracts where deeper water channels are found. Near the river mouth, the Matla is about 26 km wide, which allows substantial seawater to flow into the estuary during high tide. In the upper reaches, this estuary has steadily converged, and in the northern end at Port Canning, the width is only around 1 km (Chatterjee et al., 2013). This northern end of the Matla mostly remains exposed during low tide and practicing agriculture in the riverbed sediments of this region is also a common sight. Several other seaward bifurcations of the Matla have experienced human encroachment. Thus, it can be inferred that both natural and anthropogenic factors have led to the reduction of freshwater flow in the Matla.
The Dhamra, situated next to the Bhitarkanika Mangrove Forest, is actually a small part of the Mahanadi estuarine system formed by the union of the Brahmani and Baitarini Rivers (Sangita et al., 2014). This estuary is typically micro-mesotidal in nature, and it can be classified as a monsoonal wave to mixed energy wave-tide dominated type (Davis, 1987). Both the Brahmani and Baitarini Rivers act as perennial freshwater sources to the Dhamra, with peak discharge of 22,640 m 3 s −1 and 14,150 m 3 s −1 , respectively, during the June-September monsoon (Asa et al., 2013). The water depth in the Dhamra varies from 20 to 60 m (Mahapatro et al., 2011). Unlike the Matla, the immediate upper reaches of the Dhamra have never experienced drying up due to siltation (Sundaray et al., 2006). Compared to the Matla, the Dhamra is a much narrower estuary, with width varying from 1 to 2 km. Anthropogenic encroachment is not visible in the upper reaches of the Dhamra Estuary (i.e., Brahmani and Baitarini Rivers). Thus, the freshwater flow is well maintained in the Dhamra throughout the year (Swain et al., 2020). However, both of these estuaries connect to the Bay of Bengal, flow through dense mangrove forests, and share a similar and typical tropical climate (the mouths of these estuaries are only 120 km apart).
The waters adjacent to mangroves are recognized as substantial sources of CO 2 to the atmosphere throughout the world, although mangrove ecosystems on the whole are considered to be net autotrophic (Alongi, 2002) due to their high above-ground productivity (Bouillon et al., 2008;Donato et al., 2011). However, the pCO 2 (water) values in the Matla were lower than those observed in other estuaries; consequently, air-water CO 2 evasion rates from the Matla were lower than those observed in most other estuaries adjacent to mangroves (Akhand et al., 2016;Biswas et al., 2004;Dutta, Kumar, Mukherjee, Sanyal, & Mukhopadhyay, 2019). In contrast, the Dhamra, which is situated in the Mahanadi River delta (the third largest AKHAND ET AL. 10.1029/2020JG006144 4 of 22 river of the Indian peninsula), receives ample riverine freshwater from distributaries of the Mahanadi, the Brahmani and Baitarani Rivers. The climate of both the Dhamra and the Matla is tropical, and the seasons can be demarcated as pre-monsoon (February-May), monsoon (June-September), and post-monsoon (October-January) (Akhand et al., 2016;Asa et al., 2013). Species diversity in both the Sundarbans and Bhitarkanika mangroves is remarkably high. Avicennia sp., Excoecaria agallocha, Bruguiera sp., Ceriops sp. and Phoenix paludosa are the dominant plants in the Sundarbans (Gopal & Chauhan, 2006), whereas in Bhitarkanika, Rhizophora sp., Avicennia sp., and Heritiera sp. are dominant (Mishra et al., 2005).   (Figure 1). The Matla is surrounded mainly by mangrove vegetation from its upper to lower stretch (M1-M3 station); in contrast, the Dhamra is surrounded by mangroves only at its lower stretch (D3). Sampling was conducted for one complete diel cycle (24 h) at each station during each season.

Sample Collection and In-Situ Measurements
pCO 2 (water), salinity, water temperature, dissolved oxygen (DO), and pH were measured with in situ sensors. pCO 2 (water) was measured continuously (1-min temporal resolution) with an automated CO 2 analyzer (CO 2 -09, Kimoto Electric Co., Ltd., Osaka, Japan), a non-dispersive infrared (NDIR) sensor after equilibration through a gas-permeable membrane Tokoro et al., 2014). pCO 2 (air) was also measured using the same instrument by keeping the sensor in the air for about 15 min at each site. The instrument was calibrated every day at the beginning of the measurements using certified reference gases (pure N 2 gas [0 ppm CO 2 ] and span gas [600 ppm CO 2 gas in a N 2 base], Chemtron Science Laboratories, Mumbai, India). The accuracy and precision of the pCO 2 (air) measurements were within ±5 μatm and ±2 μatm, respectively, and we assumed this same accuracy and precision holds for the pCO 2 (water) measurements, because the NDIR sensor used in this study, measured pCO 2 (air) after equilibration through a gas-permeable membrane. These accuracy and precision were similar to previous work using the same pCO 2 analyzer system (Kayanne et al., 2005;Saito et al., 1995). Other physicochemical parameters were measured onboard at 3-h intervals (pH, water temperature, and DO). Salinity and water temperature were measured using a Multikit (WTW Multi 340i Set; WTW Measurement Systems Inc., Florida, USA) fitted with a WTW Tetracon 325 probe. DO was measured using a FiveGo Portable F4 Dissolved Oxygen Meter (Mettler Toledo, Greifensee, Switzerland). pH was measured using an Orion PerpHecT ROSS Combination pH Micro Electrode (Thermo Scientific, Massachusetts, USA) with a resolution of 0.001 attached to a data logger (1112000 3-star benchtop pH meter, Thermo Scientific). The pH meter was calibrated daily using the pH buffers 4.01 (Part no: 1.09475.0500; Merck, New Jersey, USA), 7.00 (Part no: 1.09477.0500; Merck) and 9.00 (Part no: 1.09476.0500; Merck). Wind velocity and atmospheric pressure were measured using a portable weather station (WS-2350, La Crosse Technology, Wisconsin, USA). Gross primary production (GPP) and community respiration (CR) in the water column were measured in situ in triplicate using the lightand dark-bottle method via Winkler titration described by Strickland and Parsons (1972). Both the light and dark bottles were incubated for 3 h starting from 12:00 h. The initial DO and the DO of the incubated bottles after 3 h were measured. The incubated bottles were kept floating submerged at the surface. Special care was taken to wrap the dark bottle with black plastic followed by aluminum foil to prevent the penetration of light. GPP and CR were estimated from the difference in DO of the initial, light, and dark bottles according to the formula given in Strickland and Parsons (1972).
Four samples for TAlk and DIC were collected during two peak ebb and two flood tides covering a complete semidiurnal tidal cycle over one day. Surface water samples for POM and dissolved organic carbon (DOC) were collected twice in duplicate, at peak ebb tide and at the following peak flood tide during the daytime.
In addition, offshore surface-water samples were collected at one station about 60 km off the coast in the Bay of Bengal; these samples served as the marine end member (MEM) for both of the estuaries. Samples collected from the station Diamond Harbor (salinity close to 0) in the Hooghly River served as the riverine freshwater end member (RFWEM) for the Matla, and samples collected from station Tintara Ghata (almost 10 km upstream from D1) served as the RFWEM (annual mean salinity 0.8) for the Dhamra. The MEMs and RFWEMs were collected during Pre M1, Monsoon, and Post M. RFWEM was collected to determine the OM degradation pathway using salinity-normalized TAlk and DIC, whereas both RFWEM and MEM were used in three endmember-mixing models of OM (methods described in Section 2.6). In addition, a stainless steel Van Veen type Grab Sampler was deployed from the boat to collect riverbed sediment samples from each station in triplicate during each season.
Samples for TAlk and DIC were collected into 250-mL Duran bottles (Schott AG, Mainz, Germany), filtered through glass fiber filters (GF/F; Whatman, Maidstone, Kent, UK) and poisoned with mercuric chloride (200 µL saturated aqueous solution per bottle) to prevent changes in TAlk and DIC due to biological activity. The filtration of the water samples for TAlk analysis was done to avoid the interference of the high concentrations of suspended sediments in samples, specifically to avoid interference of particulate inorganic carbon and POM (cf. Chanson & Millero., 2007;Frankignoulle et al., 1996;Kim et al., 2006). Doctor et al. (2008), stated that filtration is the best method to avoid fractionation of DIC due to biological activity. However, we ascertained a 2-fold uncertainty on our TAlk and DIC data, than that of the accuracy of the measurement (described in section 2.4) for atmospheric CO 2 exchange during filtration and assume that the total uncertainty of the TAlk and DIC data were ±2 μmol kg −1 and ±4 μmol kg −1 , respectively. Samples for dissolved organic carbon (DOC) were filtered through 0.2 µm polytetrafluorethylene filters (DISMIC-25HP; Advantec, Durham, NC, USA) into pre-combusted (450°C for 2 h), 100-mL glass vials. DOC samples were acidified with H 3 PO 4 to pH < 2 and then frozen at −20°C until analysis. Samples for analysis of particulate organic carbon (POC), particulate nitrogen (PN), Chlorophyll-a (Chl-a), and stable isotope analysis (δ 13 C POC ) were obtained by filtration (approx. 1-2 L) onto pre-combusted (450°C for 2 h) glass fiber filters (GF/F; Whatman) and stored in the dark at −20°C until analysis.
Mangrove leaves were collected to determine the stable isotopic signature (δ 13 C TOC ) as an end member. Leaves of the dominant mangrove species from the Bhtarkanika (Mishra et al., 2005) and Sundarbans (Gopal & Chauhan, 2006) were collected and stored at −20°C until analysis. Leaves of A. marina, E. agallocha, B. gymnorrhiza, and P. paludosa were collected from the Indian Sundarbans, whereas leaves of A. officinalis, E. agallocha, and Heritera fomes were collected from Bhitarkanika (mangrove forest near the Dhamra).

Laboratory Analysis
TAlk and DIC were determined on a batch-sample analyzer (ATT-05; Kimoto Electric Co., Ltd.) in a closed path mode implementing the Gran Plot method (Dickson et al., 2007). The pH probe used in the same analyses was a Radiometer Analytical PHC2401-8 Combination Red-Rod pH Electrode (glass body, BNC, Product No. E16M400, Hach, Colorado, USA). The accuracy of TAlk and DIC measurements was estimated to be ±1 μmol kg −1 water and ±2 μmol kg −1 water, respectively, by triplicate measurements of the certified reference material (CRM) for TAlk and DIC (Kanso Co., Ltd., Osaka, Japan) with known concentrations. DOC concentrations were determined by high-temperature catalytic oxidation with a total organic carbon (TOC) analyzer (TOC-L; Shimadzu, Kyoto, Japan). Potassium hydrogen phthalate (Wako Pure Industries, Osaka, Japan) was used as the measurement standard, and the coefficient of variation of the analyses was less than 2%. Chl-a was measured following standard spectrophotometric procedures using a Shimadzu UV-Visible 1,600 double-beam spectrophotometer after extraction in 90% acetone overnight in the dark (Parsons et al., 1992).
Filtrate samples for analysis of POC, PN, and δ 13 C of OM were dried in an oven at 60°C. To remove inorganic carbon, we acidified the samples with 1-N HCl and dried these again at 60°C. POC and PN concentrations and the stable isotope signatures were measured with an isotope-ratio mass spectrometer (Delta Plus Advantage; Thermo Electron, Bremen, Germany) coupled with an elemental analyzer (Flash EA 1112; Thermo Electron). The stable isotope ratio is expressed in δ notation as the deviation from a standard in parts per thousand (‰) according to the following equation: where R is the 13 C/ 12 C ratio. Vienna PeeDee Belemnite (VPDB) was used as the carbon isotope standard. Based on the standard deviation of internal reference replicates (l-histidine [δ 13 C VPDB = −10.19‰]; l-alanine [δ 13 C VPDB = −19.6‰]; Shoko Science, Yokohama, Japan), analytical precision was within ±2% for TOC and TN, and ±0.2‰ for δ 13 C. The δ 13 C of DIC (δ 13 C DIC ) was also measured with the same isotope-ratio mass spectrometer following the method of Miyajima et al. (1995).

Estimation of Air-Water CO 2 Flux
The air-water CO 2 flux (F CO2 , µmol CO 2 m −2 h −1 ) was determined with the following equation: where k is the gas transfer velocity (cm h −1 ), K 0 denotes the solubility coefficient of CO 2 (mol m −3 atm −1 ) and is computed based on the equation given by Weiss (1974), and ΔpCO 2 denotes the difference in the partial pressure of CO 2 between water and air (pCO 2(water) -pCO 2(air) ). The pCO 2 values were converted from the unit of µatm to atm, and values of k were converted from cm h −1 to m h −1 to compute the fluxes. A positive F CO2 value indicates CO 2 efflux from the water to the atmosphere and vice versa.
where U 10 is the wind speed at a height of 10 m, and a is a constant accounting for gas transfer from bottom-shear-driven turbulence. Ho et al. (2011) observed a negligible "a" value of 0.06 cm h −1 in the Hudson River Estuary; based on the depth and dimensions of the estuaries considered in this study, a was taken as 0 (cf. Chanda et al., 2020). Wind speed data were corrected for 10-m height using wind gradient formula (Kondo, 2000). Sc is the Schmidt number of CO 2 as given by Wanninkhof (2014). We calculated one mean magnitude of gas transfer velocity (one for each formula) for each estuary per season.

Data Analysis
To characterize the pathway of mineralization of organic matter in this study, TAlk and DIC were both normalized with respect to salinity. We analyzed the stoichiometric relationship (the slope of linear regres-sion) between salinity-normalized TAlk (nTAlk) and salinity-normalized DIC (nDIC). DIC was normalized according to the following equation (Friis et al., 2003): where DIC meas is the measured DIC, DIC s = 0 is the DIC of the RFWEM (i.e., where salinity is close to 0), S meas is the measured salinity, and S mean is the mean salinity of the samples collected in each estuary during each sampling event. TAlk was also normalized using the same equation, replacing DIC meas and DIC s = 0 with TAlk meas and TAlk s = 0 , respectively.
We used the Bayesian isotopic modeling package Stable Isotope Analysis in R (SIAR) (Parnell et al., 2010) to partition the proportional contributions of potential organic matter (OM) sources to the water column POM and riverbed sediment OM based on their δ 13 C and N/C signatures. We chose N/C rather than C/N ratios in the model because the former was statistically more robust; the higher number (TOC concentration) is the denominator and behaves linearly in end-member mixtures (Perdue & Koprivnjak, 2007). The contribution from the source with a low-N/C to sedimentary C org might be overestimated because the N/C ratio generally declines while OM is decomposing (Krüger et al., 2015). In contrast, there is little change of δ 13 C during decomposition (Fry, 2006). We used only δ 13 C and N/C in the mixing model because we assumed δ 15 N signatures might be affected by denitrification and ammonification, as these diagenetic pathways were found to be dominant in these estuaries (Dutta, Kumar, Mukherjee, Sharma, et al., 2019).
The SIAR model works by determining the probability distributions of the sources that contribute to the observed mixed signal while accounting for the uncertainty in the signatures of the sources. We assumed an isotopic fractionation of zero and ran the model through 1 × 10 6 iterations. For each potential source, we reported the median and 95% credible interval (CI) of the estimated proportional contribution to the calculated value. We defined three sources-RFWEM, mangrove-plant-derived OM, and MEM-as OM end members for the isotopic and elemental mixing model. The elemental and stable isotopic signatures reported for RFWEM for the Hooghly Estuary in previous studies (Ray et al., 2015 and the references therein) suggest that the RFWEM mainly originated from C3 land plants (N/C ratio about 0.085 and δ 13 C POC between −25.9‰ and −24.1‰). A negative shift of δ 13 C was also reported in the terrestrial organic matter along the Ganga-Bramhaputra delta from the last glacial maximum to the mid-Holocene mainly due to the transition from C4 to dominant C3 plants (ranging from −29.0‰ to −25.4‰; Galy et al., 2007). Data for the MEM are consistent with mainly phytoplankton origin (between −22.0‰ and −20.0‰, Rosentreter et al., 2018;molar C/N ratio about 7, Redfield et al., 1963). Since we did not collect RFWEM samples during the pre-monsoon season for Matla, we used the δ 13 C POC and the N/C ratio reported in Ray et al. (2015) as RFWEM for the pre-monsoon season; Ray et al.'s samples were collected during May 2014 from the same Diamond Harbor station we used.
Statistical analyses were done using SPSS software version 16.0 (SPSS Inc.). We checked for normality of the biogeochemical parameters by applying the Shapiro-Wilk test. In order to test the significance of differences in these parameters between the Matla and Dhamra Estuaries, we applied the student's t-test and non-parametric Mann-Whitney U test (also known as Wilcoxon rank-sum test) for the parameters with normal distribution and nonnormal distribution, respectively.
We used the previously reported salinity increase rate and the regression equation of salinity with TAlk and DIC observed in the present study to estimate (using CO2SYS software, Lewis & Wallace, 1998) a tentative future scenario of the pCO 2 (water) and associated air-water CO 2 flux for the Matla Estuary, where the riverine freshwater input is reportedly decreasing. The salinity of the Matla Estuary along with the central part of the Indian Sundarbans is increasing at present (Chowdhury et al., 2019). Trivedi et al. (2016) projected that by the year 2043, the salinity in the estuaries of the central part of Sundarbans might be similar to that in the open ocean (an increase of about 33% relative to that observed in the year 2013). We used the same gas transfer velocities (k, Equations 3-6) for estimating future air-water CO 2 flux values as was used to estimate the present air-water CO 2 flux. The significance of the differences in mean between the Matla and Dhamra for different parameters were tested using parametric student's t-test or non-parametric Mann-Whitney U test based on the normality of distribution.
The annual mean TAlk, DIC, and pH were significantly higher in the Matla than in the Dhamra (p < 0.001; Figures 2 and S3). δ 13 C DIC was heavier in the Matla (−3.80 ‰± 1.53‰) than the Dhamra (−5.96‰ ± 2.59‰) (p < 0.001; Figure S3). TAlk, DIC, and δ 13 C DIC were significantly, positively correlated with salinity in each estuary (p < 0.001; Table 2). TAlk/DIC was significantly correlated with salinity for the Matla, whereas the correlation was not significant for the Dhamra (Table 2). pCO 2 (water), pH, TAlk, DIC, δ 13 C DIC , and air-water CO 2 flux showed significant seasonal variations in each estuary (p < 0.001, ANOVA), whereas only in case of pCO 2 (water) was there significant spatial variability amongst the stations within the estuaries (Table S1).
In all the seasons, the slopes of the linear regressions of nTAlk on nDIC and nTAlk/nDIC ratios varied over a narrower range for the Matla than for the Dhamra (Figure 3 and Table S2). The linear regressions between nDIC and nTAlk in the Matla were significant during Pre M1, Post M, and Pre M2 (p < 0.01). In the Dhamra, the regressions were significant during all the seasons (p < 0.01). The annual mean TAlk and DIC of the RF-WEM used to calculate nTAlk and nDIC for Matla were 2,627 ± 543 µmol kg −1 and 2,597 ± 507 µmol kg −1 , respectively, whereas for Dhamra the same was 1,345 ± 456 µmol kg −1 and 1,300 ± 445 µmol kg −1 , respectively.

Table 1 Mean ± Standard Deviation With Range (Minimum to Maximum) of ΔpCO 2 (µatm) and Air-Water CO 2 Flux (µmol m −2 h −1 ) (F) along with the gas transfer velocities (k) observed during the seasonal surveys carried out in the Matla and Dhamra Estuaries.
No significant seasonal variability (p > 0.05) of TAlk/DIC ratio in RFWEM was observed in each of these two estuaries.
Mean chlorophyll-a was significantly higher in the Dhamra than the Matla during all seasons (p < 0.001, Table S3). However, there were no significant difference in the GPP and CR between the estuaries (p > 0.05, Table S3).

Organic Matter Characteristics
The POC concentrations differed significantly between the two estuaries except during Pre M1, but differences in the values of POC/PN and δ 13 C POC were not significant during Pre M1 and Post M (Figure 4). The annual mean POC was lower in the Matla (72.9 ± 49.5 µM) than the Dhamra (210.3 ± 158.1 µM) (p < 0.001). However, the annual mean POC/PN and δ 13 C POC in the Matla did not significantly differ from those in the Dhamra (p > 0.05). The seasonal variability of POC, POC/PN, and δ 13 C POC in both of the estuaries was significant (p < 0.001; Table S1). Significant positive correlations were found between POC and salinity in both the Matla and the Dhamra (p < 0.05; Table 2). The correlation between POC/PN and salinity was also significant in both the Matla (p < 0.05) and the Dhamra (p < 0.001). δ 13 C POC was significantly positively correlated with salinity in both the Matla (p < 0.05) and the Dhamra (p < 0.001).
The annual mean DOC was lower in the Matla (119.1 ± 25.8 µM) than the Dhamra (150.1 ± 51.6 µM) (Figure 4, p < 0.05). However, during Pre M1, the DOC concentration in the Dhamra was lower than in the Matla (Figure 4). A significant negative correlation was found between DOC and salinity in both estuaries ( Table 2).
The mean sedimentary OC (SOC) concentrations did not differ significantly between the Matla and the Dhamra (p > 0.05) ( Figure 5). Between estuaries, δ 13 C SOC differed significantly (p < 0.05); however, the sediment C/N ratio did not (p > 0.05) ( Figure 5). The seasonal variability of SOC, C/N, and δ 13 C SOC of sedimentary OM in the riverbed of the Matla and the Dhamra was significant (p < 0.001; Table S1).

Mixing Models for Three End Members of Water Column POM and Riverbed SOM
The isotopic and elemental signatures showed that the water column POM and riverbed SOM were composed of three OM sources: Mangrove tissue, RFWEM, and MEM ( Figure 6 and Table S4). There was no significant (95% CI) difference in POM composition between the two estuaries ( Figure 6). In SOM, the annual mean percentage of mangrove-derived OM was higher in the Matla than in the Dhamra.

Future Projections
The linear regressions of salinity with TAlk and DIC calculated for the Matla Estuary are as follows: TAlk = 55. 408 × salinity + 909.47 and DIC = 41. 668 × salinity + 1,063.6, respectively (Table S5). In the Matla, the annual mean salinity and pCO 2 (water) were 19.8 and 735 µatm, respectively (Figure 2). When we assumed that salinity will increase to ≈30 in 2040 (Trivedi et al., 2016), we estimated a tentative pCO 2 (water) decrease by the year 2040 to ≈534 µatm from the present annual mean value in the Matla Estuary and associated air-water CO 2 fluxes between 26 and 205 µmol m −2 h −1 .

Effect of Reduced Riverine Freshwater on Inorganic Carbon Dynamics
The present study using our high temporal resolution data set showed that reduced riverine freshwater supply in the Matla Estuary lowered pCO 2 (water) and consequently CO 2 efflux compared to the Dhamra Estuary ( Figure 2 and Table 1) diurnally, seasonally, and spatially. More riverine freshwater input to estuaries leads to high pCO 2 (water) and subsequently high air-water CO 2 efflux (Figure 7; Jiang et al., 2008;Maher & Eyre, 2012). In Hooghly Estuary, near the Matla, monthly mean pCO 2 (water) was significantly correlated with monthly mean river discharge (Akhand et al., 2016). The pCO 2 (water) and corresponding air-water CO 2 efflux from the Matla are similar to those from previous studies of the same estuary AKHAND ET AL.
The carbonate buffering capacity was consistently higher in the Matla than the Dhamra (Figure 2), which we attribute to reduced riverine freshwater supply and lower pCO 2 (water) and air-water CO 2 efflux at the Matla, as suggested by our results obtained from the TAlk/DIC ratio ( Figure 2). The TAlk/DIC ratio is a little studied but important property that provides insights into the carbonate buffering capacity of the water column (Egleston et al., 2010;Joesoef et al., 2017;Wang et al., 2015). The higher TAlk/DIC at the Matla ( Figure 2) suggests that CO 2 evasion was suppressed by the concomitant supply of TAlk at a higher rate than that of DIC. Furthermore, higher values of δ 13 C DIC at the Matla than the Dhamra ( Figure S3) suggest greater marine influence on DIC source; greater influx of Bay of Bengal seawater would supply more carbonate buffering capacity than would riverine input. Although phytoplankton standing stock (Chl-a) differed significantly between these estuaries, the lack of a significant difference of gross primary production and community respiration suggests that biological metabolism in the water columns did not regulate the differences in carbonate chemistry (Table S3).
Our results show that reduced riverine freshwater supply reduced the variability in biogeochemical processes controlling carbonate chemistry (Figure 3). The processes affecting both DIC and TAlk varied over a narrower range in the Matla (Figure 3 and Table S2) due to reduced riverine freshwater supply, which is also evident from the narrower ranges of salinity as well as TAlk and DIC (Figure 2) than those in the Dhamra. Changes in the TAlk/DIC ratio follow different stoichiometric relationships depending on the biogeochemical processes (Figure 3; Borges et al., 2003;Krumins et al., 2013;Sippo et al., 2016). The slopes of the nTAlk/ nDIC regressions for the Matla (Figure 3) indicate that the plausible biogeochemical processes there could be ammonification and denitrification. However, the slopes in the Matla might also be an outcome of a combination of pathways, like aerobic respiration and sulfate reduction. In Dhamra, various biogeochemical processes, such as aerobic respiration, ammonification, denitrification, sulfate reduction, and calcium AKHAND ET AL.

10.1029/2020JG006144
15 of 22 carbonate dissolution, are consistent with the slopes. However, we could not determine the exact diagenetic pathway of organic matter decomposition of the two estuaries from this study due to the absence of direct evidence on these processes. Our main purpose was to understand the change in carbonate buffering capacity within the estuaries excluding the effect from conservative mixing and using the slopes of nTAlk versus nDIC. The slopes of the nTAlk versus nDIC regressions were steeper for the Dhamra than the Matla (Figure 3), which indicates that the biogeochemical processes in the Matla decrease the TAlk/DIC ratio, which offsets the effects of higher carbonate buffering capacity and reduced CO 2 evasion there.
Natural and/or anthropogenic processes like siltation and dam construction cause partial or full river abandonment (cut-off of a river from the main channel), which in turn alters inorganic and organic carbon AKHAND ET AL.
10.1029/2020JG006144 16 of 22 Figure 7. Schematic diagram of the effects of reduced riverine freshwater supply on inorganic and organic carbon dynamics, and air-water CO 2 flux in the two mangrove-dominated estuaries. POC and DOC concentrations were lower in the Matla than the Dhamra. The internal diagenetic processes within the estuaries produced higher carbonate buffering capacity (TAlk/DIC ratio) in the Dhamra than the Matla. However, the carbonate buffering capacity due to predominance of seawater in the Matla is much greater than the internal buffering capacity of the Dhamra; this greater carbonate buffering capacity in the Matla facilitated retention of higher DIC than in the Dhamra. As a result, air-water CO 2 efflux was lower in the Matla than the Dhamra. The Matla riverbed sediment contained more mangrove-derived blue carbon than did the Dhamra. DIC, dissolved inorganic carbon; DOC, dissolved organic carbon; POC, particulate organic carbon; TAlk, total alkalinity. The magnitudes displayed for different variables are annual average and range of air-water CO 2 flux are annual average according to Equation 3 to Equation 6. dynamics of estuaries differently in different places. Jin et al. (2018) observed that dam construction obstructed river flow and lower pCO 2 (water) near the dams resulted, as stagnant waters facilitate a higher degree of autotrophy, whereas the lower reaches received all of the resulting anthropogenic DOC and POC loads, which led to net heterotrophy. In contrast, pCO 2 (water) in the lower reaches of the Matla Estuary was low due to higher marine dominance.
We estimated a tentative decrease of mean pCO 2 (water) (from 735 µatm at present to ≈534 µatm) and associated air-water CO 2 fluxes (between 722 and 1,635 µmol m −2 h −1 at present to between 26 and 205 µmol m −2 h −1 ) by the year 2040 in the Matla Estuary. Given the present business as usual scenario, by the year 2040, the ambient concentration of CO 2 is expected to reach 450 µatm (NASA, 2021), which would decrease the ΔpCO 2 to only ≈84 µatm. Hence, we infer that if the present rate of salinity increase in the central part of the Sundarbans continues as in the recent past, the air-water CO 2 flux would significantly decline in the near future.

Effect of Reduced Riverine Freshwater Supply on OM Dynamics
Our findings indicate that riverbed sediment might be a previously overlooked but important carbon pool for mangrove-derived organic carbon. In the riverbed sediment, the contribution of mangrove derived OC was 47% (median of estimates) in the Matla (Figures 6 and 7). Moreover, we attribute increased storage of mangrove derived OC to reduced riverine input in the riverbed sediments of the Matla than those of the Dhamra. We attribute this to (i) reduced "flushing effect" on OM due to reduced freshwater supply from upstream, (ii) higher sedimentation rate, and (iii) larger mangrove coverage and catchment area in the Matla than in the Dhamra. However, delineating the exact mechanism of this phenomenon is not possible from the present study. We could not compare the riverbed OC percentage with other studies in mangrove-dominated estuaries because of the lack of such data. However, the SOC observed in the present study (0.55 ± 0.16% and 0.55 ± 0.29% in Matla and Dhamra, respectively) was within the range observed at other tropical and subtropical rivers like the Amazon (0.17 ± 0.27% to 1.17 ± 0.34%, Bouchez et al., 2014) and Yangtze River (0.08 ± 0.01% to 0.81 ± 0.08%, Li et al., 2015), respectively.
Our results suggest that reduced freshwater flow leads to significantly lower POC in the Matla. In contrast, the Dhamra having substantial freshwater from upstream, also receives substantial POC. However, mixing models for three end members suggest that the difference between the two estuaries in riverine freshwater contribution to POM is not significant ( Figure 6). Furthermore, the significant positive correlation of POC (and PN) with salinity (Table 2) in both estuaries indicates that POM increases toward the sea. This phenomenon can be explained by the input of mangrove-derived OM to the POM in the seaward reaches of the estuaries where salinity is high, as shown from the significant positive correlation between mangrove-derived OM and salinity in both the Matla and the Dhamra (Table 2). A significant negative correlation between contribution of RFWEM to POM and salinity was found in the Dhamra (Table 2), which shows that the contribution to POM from the upstream river to the Dhamra was more than that to the Matla.
We also attribute the lower DOC in the Matla to reduced riverine freshwater supply compared to the Dhamra ( Figure 7); in fact, DOC was negatively correlated with salinity in both estuaries (Table 2). In contrast to the significant positive correlation between POC and salinity, the negative correlation between DOC and salinity suggests that the riverine freshwater acted as the source of DOC in both estuaries.

Seasonal and Spatial Variations
In both estuaries, pCO 2 (water) and CO 2 efflux were significantly higher in the monsoon season compared to the dry seasons ( Figure 2, Table S1). The seasonality of the pCO 2 (water) and efflux data in the present study is in good agreement with the annual study of Akhand et al. (2016), conducted in the Hooghly and Matla Estuaries. However, a few studies have reported higher pCO 2 (water) and CO 2 efflux during the pre-monsoon season compared to the post-monsoon and monsoon seasons (Dey et al., 2013;Ganguly et al., 2011). We argue that the riverine freshwater and runoff from the catchment area increase the pCO 2 (water) due to increase in allochthonous CO 2 supply and CO 2 efflux more during the monsoon than other seasons.
Results of ANOVA show that unlike seasonal variability, between different stations in both of the estuaries there were mostly no significant differences among the OM and inorganic carbon parameters (Table S1). However, we observed significantly higher pCO 2 (water) in the upper estuary, which might be attributed to higher inorganic and organic carbon loading from the riverine freshwater and/or intense respiration along with reduced photosynthesis as revealed in previous studies (Abril et al., 2003;Jeffrey et al., 2018;Raymond et al., 2000).

Conclusion
Our comparative approach shows that reduced supply of riverine freshwater affects biogeochemical processes in mangrove-dominated estuaries, changing the balance of carbonate chemistry and CO 2 fluxes. Alteration of riverine freshwater discharge impacts coastal and estuarine ecosystems, and might have changed their functions in global carbon cycles. In recent decades, due to both climate-change-related alterations in river flow, and several anthropogenic activities, including construction of dams and barrages, river bank encroachment, and extraction of freshwater for domestic and industrial activities, more than half of the world's rivers have experienced significant reductions in both freshwater discharge and sediment load (Kumar & Jayakumar, 2020;Syvitski et al., 2005;Walling & Fang, 2003;Wang et al., 2016). The magnitude of anthropogenic stress in terms of hydro-geophysical change, pollution, and sediment mining, has irreversibly jeopardized the flow characteristics in most of the rivers of the world (Best, 2019). Partial river abandonment in any form and for whatever reason, can significantly alter the carbon dynamics in both the upstream and downstream reaches of a river. These alterations in river carbon dynamics can either increase or decrease greenhouse gas emission to the atmosphere. River abandonment can reduce in lateral carbon fluxes to the oceans in the future (Li et al., 2017), which would change the role of coastal and estuarine ecosystems in global climate. Our findings can help to improve understanding of the carbon dynamics of partially abandoned estuaries and their roles in global carbon budgets.