Estuarine, Coastal and Shelf Science Variability in the organic carbon stocks, sources, and accumulation rates of Indonesian mangrove ecosystems

± 124MgC ha − 1 ) than in the central lagoon (167 ± 36MgC ha − 1 ). The predominant accumulation of autochthonous mangrove organic matter likely contributed to the higher C org stocks in Berau and the eastern SAL. Interestingly, the CAR distribution pattern in SAL is opposite to that of its C org stocks. The central SAL that receives high sediment inputs from the hinterland has a much higher CAR than the eastern SAL (658 ± 311gC m − 2 yr − 1 and 194 ± 46gC m − 2 yr − 1 , respectively), while Berau has one of the highest CAR (1722 ± 183gC m − 2 yr − 1 ) ever measured. It appears that these large di ﬀ erences are driven by the environmental setting and conditions, mainly sediment dynamics and hydrodynamics, landform, and vegetation conditions. It is inferred that quantifying carbon accumulation in sediments is a useful tool in estimating the present-day carbon storage of mangrove ecosystems. This is a precondition for taking measures under REDD+ (Reducing Emissions from Deforestation and Forest Degradation and the role of conservation, sustainable management of forests and en- hancement of forest carbon stocks in developing countries) schemes.


Introduction
Situated in the transition zone between land and ocean, mangrove ecosystems are one of the most biogeochemically active areas in the biosphere due to permanent exchange of nutrients and organic matter with adjacent ecosystems. Although mangrove forests occupy only a small fraction of the global coastal area, they are highly productive, with a net primary production rate of 92-280 Tg C yr −1 (Twilley et al., 1992;Jennerjahn and Ittekkot, 2002;Bouillon et al., 2008), and they contribute up to 15% of the total carbon accumulation in marine sediments (Jennerjahn and Ittekkot, 2002). Mangrove ecosystems are also among the most carbon-rich ecosystems in the tropics . They sequester and store high amounts of organic carbon in plant biomass and sediment, which makes them an important natural https://doi.org/10.1016/j.ecss.2018.12.007 Received 18 May 2018; Received in revised form 10 December 2018; Accepted 12 December 2018 T carbon sink (Twilley et al., 1992;Jennerjahn and Ittekkot, 2002;Bouillon et al., 2008). The mean global carbon stock of mangrove ecosystems is estimated at 956 Mg C ha −1 , which is much higher than those of rainforests, peat swamps, salt marshes, and seagrass meadows (Alongi, 2014). Their advantage over terrestrial forests is their much higher carbon accumulation in sediments/soils which is mainly due to high autochthonous and allochthonous inputs and low decomposition rates of organic matter because of the mostly anoxic conditions in the sediment Kristensen, 2000). The rate of organic carbon accumulation in mangrove ecosystems is estimated to be around 20-24 Tg C yr −1 (Twilley et al., 1992;Jennerjahn and Ittekkot, 2002;Duarte et al., 2004).
However, climate change and anthropogenic disturbances can strongly impair the ecosystem service of sequestering and storing carbon (Grellier et al., 2017;Jennerjahn et al., 2017;Pérez et al., 2017). Over several decades, mangrove areas have been declining worldwide mainly due to deforestration and land use change (Valiela et al., 2001;Alongi, 2002). The impacts are not only the loss of biodiversity and coastal protection, but also the loss of the carbon sink function (Nellemann et al., 2009). Thus, preventing mangrove loss is considered an effective strategy for climate change mitigation (Pendleton et al., 2012;Murdiyarso et al., 2015).
While it is commonly accepted that mangrove ecosystems are a carbon sink, there are large uncertainties in the magnitude of this sink with a global carbon stock estimated at 4-20 Pg . The magnitude of carbon storage in mangrove ecosystems varies geographically, which is largely determined by the type of environmental setting (Twilley et al., 1992;Woodroffe, 1992;Saintilan et al., 2013;Rovai et al., 2018). Initially, Thom (1982) defined five common environmental settings which were later described in more detail by Woodroffe (1992); these are (i) river-dominated, (ii) tide-dominated, (iii) wave-dominated, (iv) composite river and wave dominated, (v) drowned bed rock valley, and (vi) carbonate settings. Each of these settings contains suites of landforms and physical processes which control the distribution of mangrove species, transport and deposition of sediment, and the biogeochemical conditions. These processes thus influence the efflux, accumulation, storage, and composition of sediment organic matter (Thom, 1982;Woodroffe, 1992;Saintilan et al., 2013;Rovai et al., 2018). Given the large differences between environmental settings and the high within-system spatial variability of environmental conditions of mangrove ecosystems, it is conceivable that these uncertainties are to a large extent related to the scarce data base on carbon storage and the lack of knowledge on the underlying mechanisms.
Moreover, identification of organic matter (OM) sources is important for evaluating the effectiveness of the 'blue carbon' sink in mangrove ecosystems, as various OM contribute differently to carbon sequestration, and their distribution and deposition vary geographically (Saintilan et al., 2013;Watanabe and Kuwae, 2015). Furthermore, with regard to the CO 2 sink function of mangrove ecosystems and the estimation of the global carbon budget, identification of the origin of allochthonous OM inputs will help to determine whether the OM is produced from newly fixed CO 2 or relocated "old" OM from terrestrial or marine reservoirs.
In order to better understand the spatial pattern of mangrove carbon storage and composition, we measured organic carbon stocks (aboveground and sediment) and accumulation rates, and we identified the sources of organic matter in three Indonesian mangrove ecosystems  Kusumaningtyas et al. Estuarine, Coastal and Shelf Science 218 (2019) 310-323 with different environmental settings and conditions: (i) a degraded estuarine mangrove forest in the Segara Anakan Lagoon, Central Java, (ii) an undegraded estuarine mangrove forest in the Berau region, East Kalimantan, and (iii) a pristine marine mangrove forest on Kongsi Island, Thousand Island Marine National Park, Jakarta. We hypothesize that the total organic carbon stock is higher in the undegraded than in degraded mangrove ecosystems.

Study area
The Segara Anakan Lagoon (SAL) is a mangrove-fringed shallow coastal lagoon located in southern Central Java (108°46′E − 109°03′E, 8°35′S -8°48′S; Fig. 1), which is separated from the Indian Ocean by the rocky mountainous Nusakambangan Island (Yuwono et al., 2007). The lagoon receives water input from the rivers in the west and by tidal exchange with the Indian Ocean through two channels in the western and eastern parts ( Table 1). The extensive mangrove forest surrounding the lagoon, with a total area of approximately 9000 ha (Ardli and Wolff, 2009), is the largest remaining single mangrove forest in the south coast of Java, which is inhabited by 21 mangrove species and 5 understorey genera (Hinrichs et al., 2009). Mangrove density and diversity are higher in the eastern than in the central part of the lagoon, mainly as a consequence of environmental degradation due to high sedimentation, overexploitation of natural resources and land use conversion (Hinrichs et al., 2009;Lukas, 2017). The lagoon became narrower and shallower due to heavy sedimentation from several rivers, mainly the Citanduy River (Yuwono et al., 2007). Because of large differences in hydrodynamics and sediment dynamics, flora and fauna composition, and redox conditions and porewater nutrient biogeochemistry in sediments, the Segara Anakan Lagoon is considered to be consisting of two parts in our study, the western and central part (collectively named central part in the following) vs. the eastern part (Table 1; Hinrichs et al., 2009;Holtermann et al., 2009;. The Berau Regency is located in East Kalimantan (118°05′E, 2°25′N; Fig. 1). This region has a unique coastal system that displays a good example where mangroves, seagrasses and corals interact. In 2005, the Berau Marine Protected Area (1321*10 6 ha) was established, which has the second highest coral reef biodiversity in Indonesia and the most extensive remaining mangrove forest in Kalimantan covering an area of 80,277 ha (Bengen et al., 2003;Wiryawan et al., 2005). Mangrove forests are found along the mainland coast (Tanjung Batu to Biduk-Biduk) and surrounding the Berau River estuary (Tomascik et al., 1997;Hoekstra et al., 2007). The forests are relatively undisturbed. Although some area near the estuary has been cleared for aquaculture, these activities are still in an early development stage. The hinterland is characterized by tropical rainforest, which has been partly converted into palm oil plantations. The hydrodynamics of the Berau coastal area are driven by rivers in the west, mainly the Berau River, and strong currents in the east influenced by the Indonesian Throughflow (ITF) that connects the Pacific and Indian Oceans through the Makassar Strait, with a tidal range of up to 3 m (Gordon, 2005;Wiryawan et al., 2005;Hoekstra et al., 2007). The rivers carry a considerable amount of sediments that are dispersed along the coast by the tidal and longshore currents, and form elongated mudflats seawards of the river mouth (Table 1; Tomascik et al., 1997;Hoekstra et al., 2007).
Kongsi Island is a pseudo atoll and forms part of the Pari Island cluster (106°36′E, 5°51′S; Fig. 1), situated in the Thousand Islands Marine National Park. The Thousand Islands are a group of small islands located in the Java Sea, approximately 40 km northwest of Jakarta Bay, in which 13 rivers debouch after passing Jakarta city (Damar, 2003). The hydrodynamics of the island is tidal-dominated. The mangrove forest stretches along the northern and eastern part of the coast and occupies up to 80% (807 m) of the coastline (Salim and Ahmad, 2013). It is a typical marine mangrove ecosystem with a Table 1 Summary of site characteristics. microtidal reef setting, and due to a lack of rivers, mangroves receive no or minimum allochthonous input from the hinterland. Mangroves grow densely and are still in a good condition. However, the island is affected by abrasion due to strong wave actions (Salim and Ahmad, 2013), which may wash off the autochthonous organic matter, and thus will affect sediment accumulation.

Field sampling and measurement
A sampling campaign was conducted from October to December 2016 in the Segara Anakan Lagoon and on Kongsi Island, while sampling in Berau was conducted in May 2013. In Segara Anakan, data and sample collection were undertaken at 12 stations, with seven stations located in the central area (C24, C49, C45, C41, C26, C42, C43) and five stations located in the eastern area (E40, E46, E44, E47, E16). These stations had been established during previous studies (Hinrichs et al., 2009;Jennerjahn and Yuwono, 2009). On Kongsi Island, data and samples were collected at two stations (K1 and K2), while in Berau, data and samples were collected at four stations located along the coast of Tanjung Batu (B1, B2, B3) and the Berau estuary (B4).

Aboveground tree biomass
At each station, a 100 m transect was laid from the water edge perpendicular towards the mangrove forest. Square plots sized 10 × 10 m were set along the transect in 10 m intervals (5 plots in total). Mangrove plants inside the plots were identified at species level (species identification based on Noor et al., 1999), counted, and diameter at breast height (DBH) of each tree was measured for the estimation of aboveground biomass.

Sediment
Sediments were collected with a 1 m long semi-cylindrical auger. The sediment corer was inserted vertically into the sediment, twisted several times to cut through any fine roots, and then gently pulled out. Sediments were sampled only down to 1 m depth. After the sediment was successfully extracted, it was, sampled in 5 cm intervals, collected into plastic bags, and then preserved in a coolbox before the samples were transferred to the laboratory.

Elemental C, N and stable carbon isotope analysis
All samples were dried at 60°C for 48-72 h, weighed, ground to fine powder and then flash-combusted in an elemental analyzer (Eurovector EA3000 Elemental Analyzer) to obtain the total C content, total N content and organic carbon (C org ) content. The C org content was determined after removing carbonate carbon by acidification with hydrochloric acid and subsequent drying at 40°C. The samples were then measured for the stable carbon isotope composition (δ 13 C org ) using a Thermo Finnigan Delta plus mass spectrometer coupled to a Flash EA1112 Elemental Analyzer. All data were expressed in a conventional delta (δ) notation, where the isotopic ratio of 13 C/ 12 C was reported relative to the international VPDB standard as defined below: The precision of instruments was determined by replicate analysis of the standards which resulted in standard deviations as follows: C = ± 0.04%, N = ± 0.006%, δ 13 C org = ± 0.08‰.

Stable isotope mixing model
The stable carbon isotope composition (δ 13 C org ) was used to identify the sources of organic matter in the mangrove sediment. The natural variation in the isotope composition of organic matter occurs as a consequence of distinct photosynthetic pathways by different groups of plants; the C3, C4 and CAM plants (Smith and Epstein, 1971;Fry, 2006). However, due to a strong overlap of δ 13 C between C4 vegetation and marine-derived organic matter, δ 13 C was used in combination with the carbon to nitrogen (C/N) ratio to distinguish organic matter (Bouillon et al., 2003;Fry, 2006;Khan et al., 2015). The C/N ratio can be used to distinguish aquatic from terrestrial organic matter due to a higher relative N content of aquatic organic matter (Tyson, 1995).
The Bayesian isotopic modeling package, Stable Isotope Mixing Model in R (SIMMR) was used to estimate the proportional contribution of potential sources to sedimentary organic carbon based on their isotopic and elemental signatures. The SIMMR model is an updated version of the SIAR package (Parnell et al., 2010), which has been used elsewhere to analyze source contribution of organic matter in coastal ecosystems (Sarma et al., 2014;Watanabe and Kuwae, 2015). This model uses the Markov Chain Monte Carlo (MCMC) algorithm to determine the probability of source proportions to the observed mixture while incorporating the uncertainty for correction (e.g. isotopic fractionation, concentration-dependency, residual error, etc.). Two variables (δ 13 C org and C/N atomic) were used and five endmembers (mangrove root and leaf litter, general C3 plants, soils, riverine surface sediment, and coastal Particulate Organic Matter = POM) were considered in this study. The number of endmembers was selected carefully in order to maximize the estimation of source contribution by checking a diagnostic matrix plot of proportional source contributions. The isotopic fractionation and concentration-dependency were set to zero, assuming that the early degradation of organic matter in mangrove sediment does not significantly alter the stable isotope composition (Meyers, 1994;Khan et al., 2015). A model was run through 1 × 10 3 iterations, and the results are presented using a boxplot which depicts the 50% credibility interval.
The endmember's signatures were taken from region-relevant literature, e.g. mostly collected from Segara Anakan such as mangrove leaf litter (Herbon and Nordhaus, 2013;supplementary information in Nordhaus et al., 2017), surface sediment from the Citanduy River and rice field soils (Yuwono et al., 2007). In addition, data from other areas were also used, for surface sediment from the Berau River (Weiss et al., 2016), for coastal POM from Madura Strait, Indonesia (Jennerjahn et al., 2004), for land plants from Trang, Thailand (Kuramoto and Minagawa, 2001), and for mangrove roots (fresh and decomposed) from Segara Anakan and a Kenyan mangrove forest (Huxham et al., 2010;Weiss et al., 2016). We found that the values of δ 13 C org and C/N ratio between mangrove leaf litter and dead roots were overlapping, therefore, we did not separate these endmembers.

Biomass and carbon stocks
Tree biomass was determined by using published allometric equations ( Table 2). The aboveground C org stock was then calculated by multiplying the biomass of the individual tree by their specific C org content. For sediment, carbon density of the sediment was determined by multiplying sediment dry bulk density with C org content (wt%) at a specific depth. Dry bulk density was obtained by dividing the mass of the dried sample by the initial volume of the sample. The sediment C org stock per sampled depth interval was then calculated as follows: The total sediment C org stock from one core was determined by summing up C org stocks at all depth intervals from the entire core (to 1 m depth).

Radionuclide analyses
Natural and artificial gamma emitting radionuclides were analyzed in three sediment cores, C24 (central SAL), E40 (eastern SAL) and B2 (Berau) in order to estimate sediment accumulation rates. Their activity concentrations were determined by reverse electrode coaxial HPGe detector (Canberra Industries, 50% rel. efficiency) with a carbon-composite window housed in a 10 cm Pb shielding with Cu, Cd and plastic lining. Dry samples were pressed into pellets with a diameter of 35 mm, placed on a polystyrene dish without a lid and sealed using radon-tight aluminum barrier foil. Before measurement, they were stored for a minimum of 3 weeks to establish equilibrium between 226 Ra and 222 Rn and its daughters to prevent underestimation of "supported" 210 Pb (Pittauerová et al., 2011). The counting times ranged from a minimum of 24 h to almost 7 days for samples with low radionuclide concentrations. Spectra were analyzed in Genie 2000 software and the efficiency calibration was performed with the inbuilt LabSOCS (Laboratory SOurceless Calibration System) for a characterized detector.
The values for "supported" 210 Pb ( 210 Pb sup ) were derived from the concentrations of 226 Ra, determined via its daughter product 214 Pb. Only the peak area of the strongest and cascade correction free gamma line at an energy of 352 keV was used. "Excess" 210 Pb ( 210 Pb xs ), e.g. the part of 210 Pb originating from the atmospheric deposition, was calculated for each sample by subtracting the 210 Pb sup from the total 210 Pb measured via its 46 keV gamma line. 137 Cs was determined using the peak area of the 137m Ba line at 662 keV. In most of the samples, no 137 Cs could be detected and the decision threshold varied depending on measurement times and sample masses. Spectra were also analyzed for 241 Am (59 keV), which can sometimes be used as a marker for bomb test fallout. However, no traces of 241 Am could be detected.
The carbon accumulation rate (CAR) was calculated by multiplying the sediment accumulation rates with sediment dry bulk density (BD) and organic carbon content (C org content) as follows:

Statistical analysis
The differences of a single variable (C org content, N content, bulk density and δ 13 C org ) between areas (SAL East, SAL Central, Berau, Kongsi), and their differences with depth were analyzed by using a oneway ANOVA. The normality of data distribution was tested using Shapiro-Wilk prior to statistical analysis. If the data were not normally distributed, the respective variables were log-transformed to fit a normal distribution when testing linear models. The analyses were performed in R programming (R Core Team, 2017).

Elemental and isotopic composition of sediment organic matter
On average, the highest bulk density was measured in marine mangroves on Kongsi Island (1.35 ± 0.18 g cm −3 ). It was lower in the estuarine mangroves in Berau (1.20 ± 0.36 g cm −3 ) and in the Segara Anakan Lagoon (0.62 ± 0.11 g cm −3 in the east, 0.74 ± 0.11 g cm −3 in the central area). While C org and N contents were higher in the estuarine than in the marine mangroves, δ 13 C org was higher in marine mangroves than in estuarine mangroves (Table 3).

Radionuclides and accumulation rates
210 Pb xs in core C24 decreased with depth (Fig. 2). The mass accumulation rate was calculated using the Constant Flux -Constant Sedimentation (CF-CS) model (e.g., Appleby and Oldfield, 1983;Sanchez-Cabeza and Ruiz-Fernández, 2012), using a single exponential fit, for which all samples were considered. No mixing was assumed in this model, as there was neither a direct observation of it, nor an indirect proof in a form of a deep penetration of a shorter lived radioisotope excess 228 Th (Pittauerová et al., 2014), which can be used as a mixing Table 2 Allometric equations used in this study to determine aboveground biomass (D is tree DBH in cm; ρ is wood density in g cm −3 ).  (2017). Table 3 Mean values ( ± SD) of sediment dry bulk density, elemental C org and N content, atomic C/N ratio and δ 13 C org from all locations.
Location Bulk density indicator. The average mass accumulation rate was estimated to be 2.4 ± 1.4 g cm −2 yr −1 (Table 4). High fluctuation of values around the model is responsible for high uncertainty of the mean accumulation rate. No 137 Cs could be detected with a decision threshold between 0.50 and 1.25 Bq·kg −1 . Detailed results of radionuclide analyses are presented in Table S1. The 210 Pb xs depth profile of core E40 is less regular and does not allow the use of the CF-CS model. A single positive value of artificial 137 Cs was detected at the depth of 22.5 cm (equivalent to the cumulative mass depth of 14.8 g cm −2 , Fig. 2). In all other layers 137 Cs was below the decision threshold (0.3-1.2 Bq·kg −1 ). The radioisotope 137 Cs was introduced in the environment globally mainly through the atmospheric nuclear testing. Therefore, we can assume this layer younger than 1950 and the estimated average mass accumulation rate is > 0.2 g cm −2 yr −1 . This is also consistent with the fact that in the deeper layers (27 g cm −2 ) 210 Pb xs was still present, marking the layers likely younger than 100 years. In the rest of the core, below 30 g cm −2 , 210 Pb activities were very low, not detectable by the gamma spectrometry setup.
In core B2, 210 Pb xs is present in the entire core and does not show any trend with depth. At the cumulative mass depth of 80 g cm −2 the 210 Pb xs concentration did not decrease in comparison to the top of the core, so the sediment is very likely younger than 50 years, assuming the top is recent. The mass accumulation rate is therefore > 1.6 g·cm −2 yr −1 , possibly much larger than that and likely accompanied with significant reworking. Similar with core C24, 137 Cs was not detected (decision threshold 0.3-1.3 Bq·kg −1 ), which might be a result of dilution due to large mass accumulation.

Carbon accumulation rates (CAR)
The highest carbon accumulation rate (CAR) was measured in the mature and undisturbed forest in Berau (B2: 1722 ± 183 g C m −2 yr −1 ), followed by a disturbed forest that receives high sediment input from river discharge in the central SAL (C24: 658 ± 311 g C m −2 yr −1 ). The lowest was measured in the eastern SAL (E40: 194 ± 46 g C m −2 yr −1 ) which does not receive significant input from the hinterland. No CAR could be calculated for Kongsi Island as radionuclides were not analyzed. For carbonate-rich sediments like these, the atmosphereborne radionuclide activities are usually extremely low and age dating is impossible. However, as Kongsi Island has little allochthonous input and the C org concentration is low, it is likely that the CAR is also very low.

Biomass and carbon stocks
The aboveground tree biomass varied between 7.5 Mg ha −1 (E46) to 329 Mg ha −1 (B1), with the highest mean biomass measured in Berau (262.2 ± 65.3 Mg ha −1 ), followed by Kongsi Island (161.4 ± 43.6 Mg ha −1 ) and Segara Anakan Lagoon (33.7 ± 18.5 Mg ha −1 in the eastern and 21.7 ± 6.3 Mg ha −1 in the central lagoon, Table 5). Biomass values in the eastern lagoon are similar with the values reported by Nordhaus et al. (2018, this volume). However, the aboveground biomass in the central lagoon is higher in this study due to a higher DBH, which is likely as a result of a slight decrease of logging. A higher aboveground tree biomass resulted in a higher aboveground carbon stock in Berau (130.1 ± 32.1 Mg C ha −1 ), followed by Kongsi Island (74.3 ± 20.2 Mg C ha −1 ), and Segara Anakan Lagoon (15.8 ± 8.6 Mg C ha −1 in the eastern and 10.3 ± 3.0 Mg C ha −1 in the central lagoon). The mean sediment C org stock was also highest in Berau (485 ± 197 Mg C ha −1 ), but it was almost similar with the eastern area of Segara Anakan Lagoon (467 ± 118 Mg C ha −1 ), while the sediment C org stock in the central area was lower (161 ± 34 Mg C ha −1 ). Kongsi Island had the lowest sediment C org stock (37 ± 3 Mg C ha −1 ) due to a low sediment depth of 40 cm and a low C org content.
The total C org stocks (aboveground and sediment) varied among stations, from the lowest of 95 Mg C ha −1 on Kongsi Island (K1) to the   highest of 869 Mg C ha −1 in Berau (B2). The mean total C org stock in the undegraded estuarine mangroves in Berau (615 ± 181 Mg C ha −1 ) was twice as high as that measured in the degraded estuarine mangroves in the Segara Anakan Lagoon (298 ± 181 Mg C ha −1 ). However, the mean total C org stock in the eastern SAL was higher (483 ± 124 Mg C ha −1 ) than that in the central SAL (167 ± 36 Mg C ha −1 ). While marine mangroves on Kongsi Island had the lowest C org stocks (112 ± 24 Mg C ha −1 ).

210 Pb inventories
Expected inventories of 210 Pb xs in sediments of the study region can be calculated from the atmospheric deposition of 210 Pb. While the authors are not aware of any estimates for Indonesia, literature values for north Australia, Torres Strait, and Malaysia can provide an indication. Brunskill et al. (2004) estimated deposition of 210 Pb in north Queensland based on rain collections and 15 soil profiles as 50 Bq·m −2 yr −1 and at Sassie Island in Torres Strait of 18-22 Bq·m −2 yr −1 based on three soil profiles. Bonnyman and Molina-Ramos (1971, cited in Brunskill et al., 2004 reported an average of 90 Bq·m −2 yr −1 deposition for Darwin. Gharibreza et al. (2013) derived an atmospheric 210 Pb flux of 90 Bq·m −2 yr −1 to Bera lake, Malaysia, based on sediment cores. From these published values we expect the maximum 210 Pb flux in our study area to be 100 Bq·m −2 yr −1 , therefore the upper limit of inventories in sediment cores is 3.2 kBq·m −2 , which can be compared to the experimentally determined radionuclide inventories. The observed values are reported in Table 4. They are significantly higher than expected in cores C24 and B2, which suggests an important additional contribution of 210 Pb xs other than from the direct deposition, e.g. from the catchment area or through lateral transport and focusing.

Sources and composition of sedimentary organic matter
In the Segara Anakan Lagoon, differences between the central and the eastern areas occurred not only in mangrove communities (Hinrichs et al., 2009), but also in the composition of sedimentary organic matter. Variation of C org and N contents, δ 13 C org , and C/N ratio were observed between these two areas (p < 0.05). Based on the δ 13 C org (−29.5 to −23.9) and the C/N ratio (9-26.6), the sedimentary OM that accumulated in the central SAL is mainly allochthonous (Fig. 3), with 34-60% (median = 49%) derived from riverine sediment (Fig. 4). Conversely, sedimentary δ 13 C org (−29.4 to −25.8) and the C/N (18-79) values in the eastern SAL are fairly similar to those of mangrove root and leaf litter, which contributed about 40-55% (median = 48%) to sedimentary OM (Fig. 4), a similar portion as reported from many other mangrove ecosystems where 58% of sediment OM was mangrove-derived . Mangrove sediments in Berau displayed δ 13 C org (−28.4 to −26.9) and C/N (20.1-62.8) values similar to those of the eastern SAL (p > 0.05), and the 30-50% (median = 41%) mangrove root and leaf litter contributions indicate an important accumulation of autochthonous organic matter. However, due to overlapping values of δ 13 C org and C/N ratio between mangrove leaf litter and roots, the proportion of contribution from each of these endmembers was uncertain. Mangrove leaf litter and roots are important sources of organic matter. In some mangrove ecosystems, organic carbon content in sediment is determined more by the capacity of root production than by litter fall (Chen and Twilley, 1999). There, slow degradation of dead roots in combination with high rates of root production are the critical processes controlling organic carbon accumulation in the sediment (Middleton and McKee, 2001;Robertson and Alongi, 2016).
The hydro-geomorphics likely influenced the variations of sedimentary OM in the studied sites, as also reported from other studies (Kuramoto and Minagawa, 2001;Saintilan et al., 2013). Mangrove sediments growing in the river-dominated estuary in the central SAL, accumulated larger portions of allochthonous OM (mainly riverine sediment) than those in the tide-dominated estuaries in the eastern SAL and in Berau. Due to less dilution from river discharge, sediment mangroves in the eastern SAL and in Berau contained higher organic carbon content (7.7 ± 1.8% and 5.7 ± 3.7%, respectively) than the central SAL (2.4 ± 0.8%) which received high sediment inputs from the hinterland, with a large portion of mineral sediment diluting the C org content.
In contrast to estuarine mangroves, sediment on Kongsi Island had a higher δ 13 C org (−25.9 to −22.1‰), which is closer to that of marinederived organic matter, and the highest contribution comes from coastal POM (13-38%; Fig. 4). This carbonate setting is distinctly different from the estuarine systems, mainly because of the absence of river influence, and often being dependant upon in situ production of either mangrove peat or calcareous sediment (Woodroffe, 1992). In many cases, island mangroves accumulate a thick peat-muck layer overlying hard coral sand, and thus contain high sedimentary organic carbon, such as reported for the Yap and Palau islands, Micronesia, with around 14.2% and 18.4% of C org content, respectively . Despite having similar characteristics with Yap and Palau islands where mangroves grow over former reef flats and coral sands fringing the island , the mangrove ecosystem on Kongsi Island has a lower sedimentary C org content (0.8 ± 0.2%). This Fig. 3. Bi-plot of δ 13 C org and the C/N ratio of sediments from the study area and potential endmembers. M.A. Kusumaningtyas et al. Estuarine, Coastal and Shelf Science 218 (2019) 310-323 is probably because mangroves on Kongsi Island grow in the low intertidal zone that is frequently flushed by tides, which prevents the accumulation of autochthonous organic matter. Tides might have also delivered organic matter from sea water, and a large amount of eroded coral rubble that diluted the autochthonous organic matter, as indicated by a very high calcium carbonate content of the sediment (94.6 ± 1.0%).

Downcore variation of organic matter composition
Variation of organic carbon in the sediment is the result of changes in deposition from multiple sources and the decomposition of organic matter by microbes Kuramoto and Minagawa, 2001). The downcore enrichment of 13 C org (p < 0.001) coupled with a constant C/N ratio (p = 0.13) in the central SAL indicates a shift in the organic matter source, because decomposition of organic matter usually does not cause a significant enrichment of 13 C org (Saintilan et al., 2013). It seems there was a shift in OM accumulation from a predominance of riverine sediment (δ 13 C org −26.2 ± 1.6‰, C/N 10.5 ± 1.9) and hinterland soils (δ 13 C org −25.9 ± 1.6‰, C/N 15.4 ± 3.9; Kuramoto and Minagawa, 2001;Yuwono et al., 2007;Weiss et al., 2016) in the bottom layer, to more C3 vegetation (δ 13 C org −28.2 ± 2.2‰, C/N 22.8 ± 3.8; Kuramoto and Minagawa, 2001) in the upper layer (Fig. 5). Moreover, the gradual decrease of C org content and C org density downcore (p < 0.001) reflects the dilution with mineral matter which is generally high in riverine suspended matter and soils (Yuwono et al., 2007). In the eastern SAL, the low δ 13 C org and high C/N ratio fall in the range of mangrove leaf litter (δ 13 C org −31 to −28‰, C/N 28-84; Herbon and Nordhaus, 2013;Nordhaus et al., 2017), and mangrove roots (δ 13 C org −29 to −28‰, C/N 28-60 of decomposed roots; Huxham et al., 2010;Weiss et al., 2016). Their increase with depth (p < 0.001), in combination with a downcore increase of C org content and density indicates the predominance of autochthonous mangrove OM in the sediments. The somewhat lower concentrations at the surface are probably caused by an admixture of allochthonous sediment input by the tides. In Berau, downcore profiles of C org content, C org density, and the C/N ratio did not show variations (p = 0.67, 0.93, 0.62, respectively) and δ 13 C org increased only slightly with depth (p < 0.001). Values were in the same range as in the eastern SAL, and indicate a predominance of autochthonous mangrove OM throughout the core. Downcore variations at Kongsi Island are negligible.

Variation of organic carbon stocks
A variation of C org stocks in different mangrove ecosystems was observed in this study, where the majority of C org is stored in the sediment (from 79% in Berau up to 96% in the eastern SAL), which is in accordance with other studies Kauffman et al., 2011;Adame et al., 2013;Murdiyarso et al., 2015). An exception was observed on Kongsi Island, where only 33% of the total C org stock is allocated in the belowground pool because of dilution with eroded carbonate rubble which makes up 95% of the sediment.

The impact of ecosystem degradation on carbon storage
The aboveground C org stock was higher in the undisturbed mangrove ecosystems in Berau (130.1 ± 32.1 Mg C ha −1 ) and Kongsi Island (74.3 ± 20.2 Mg C ha −1 ) than in the degraded mangrove ecosystem in the Segara Anakan Lagoon (15.8 ± 8.6 Mg C ha −1 in the eastern and 10.3 ± 3.0 Mg C ha −1 in the central lagoon, (Fig. 6)). The mangrove forest in SAL is dominated by a small-stature vegetation, where the highest mean DBH was only 6.30 ± 3.57 cm (from species S. caseolaris), while DBH was up to 50.2 ± 22.5 m (from species S. alba) in Berau (Kusumaningtyas, 2017). According to a previous study (Hinrichs et al., 2009), most of the large mangrove trees in SAL had M.A. Kusumaningtyas et al. Estuarine, Coastal and Shelf Science 218 (2019) [310][311][312][313][314][315][316][317][318][319][320][321][322][323] already been cut. Even during our data collection, at some stations in the central area, many cut trees were found. As a consequence, the aboveground C org stocks are very low in SAL. The mangrove habitat in SAL has undergone transformation due to temporal dynamics of riverine sediment input and deforestation, which caused the decline of mangrove tree density and the overall species number and diversity, particularly in the central area (Yuwono et al., 2007;Hinrichs et al., 2009;Lukas, 2017). In the past, some mangrove species such as A. alba, A. corniculatum, S. caseolaris, Rhizophora and Bruguiera were found in higher numbers in the central SAL, but the changing environmental conditions caused a shift from mangrovedominated to understorey-dominated communities such as Derris and Acanthus. The abundance of Acanthus, which characterizes degraded mangrove areas in Java (Whitten et al., 2000;Hinrichs et al., 2009), was attributed to uncontrolled deforestation, high freshwater discharge, and high sedimentation as a result of extensive erosion in the hinterland due to land conversion for agriculture and settlements, deforestration for timber production, and volcanic eruptions (Yuwono et al., 2007;Hinrichs et al., 2009;Lukas, 2017;Nordhaus et al. this volume). Excess sediment input can reduce seedling numbers and bury aerial roots, thus inhibiting mangrove growth (Ellison, 1999;Sidik et al., 2016), and being aggravated with deforestation practices, these disturbances can prevent mangrove forests to reach a mature state.
Previous studies (Sun, 2011;Wang et al., 2013) found organic carbon content and density in the upper layers (down to 100 cm) increasing along with biomass growth, as primary production increased and input of dead roots and leaf litter also increased. In our study, sediment C org stock in the upper meter layer in the eastern SAL (467 ± 118 Mg C Ha −1 ) was similar to that in Berau (485 ± 197 Mg C Ha −1 ), and much higher than that on Kongsi Island (37 ± 3 Mg C Ha −1 ), despite a lower aboveground biomass. This is probably due to a combination of two factors. The relatively high tree density in the eastern SAL (Hinrichs et al., 2009), despite an absence of large trees, probably reflects a moderately high productivity that provided a large supply of mangrove litter and dead roots. High vegetation density also can inhibit resuspension by water motion and trap particles in the forest floor (Alongi, 2012). Moreover, the very low allochthonous input does not dilute the autochthonously produced OM. By contrast, the low sediment C org stock in the central SAL is probably the result (i) of a lower tree density and low primary production due to an uncontrolled deforestation, and (ii) of a dilution with allochthonous mineral sediment input from the Citanduy River.

Carbon stocks in contrasting environmental settings
While carbon stocks also vary geographically, there are major differences between estuarine and marine settings, because of the different relative importance of allochthonous input by rivers vs. autochthonous production. The mean total C org stocks found in the estuarine mangroves in Berau and Segara Anakan are higher than those in marine mangroves on Kongsi Island. This is in accordance with the pattern reported by Murdiyarso et al. (2009) who found a higher mean total C org stock in the river-delta mangroves in Tanjung Puting, Central Kalimantan (1220 Mg C ha −1 ) than in the oceanic mangroves in Bunaken, North Sulawesi (939 Mg C ha −1 ). Donato et al. (2011) also found a higher C org stock in estuarine mangroves (alluvial delta) than in oceanic mangroves (1074 Mg C ha −1 and 990 Mg C ha −1 , respectively) in the SE Asia region. Conversely, Weiss et al. (2016) reported a higher sediment C org stock within the topmost meter of oceanic mangroves on the Togian Islands, Sulawesi (570 Mg C ha −1 ) than in the estuarine mangroves in the Berau Estuary (310 Mg C ha −1 ).
A major factor in explaining the large geographical variation of C org stocks is the simple fact that the length of obtained sediment cores can vary largely. Depending on the coring device used and the withinsystem natural variability, it is possible that not the full sediment record is covered by the core, hence C org stock calculations may have large uncertainties or even be underestimates. For example, in an Indonesiawide study, sediment cores from 8 mangrove ecosystems differed between 46 cm and 300 cm in length even in one system, the respective carbon stocks varied between 408 and 2208 Mg C ha −1 (supplementary information in Murdiyarso et al., 2015). Estuarine mangrove forests may have a higher C org stock due to a longer sediment record compared to most oceanic mangroves . This is mainly because the estuarine mangroves receive large allochthonous inputs from river discharge, which is to a large extent mineral sediment diluting the carbon content. However, depending on the proportion of mineral sediment, this dilution may also result in lower carbon stocks. In order to highlight geographical variations, we compare carbon stocks standardized to 1 m sediment depth. In cases where sediment cores have a maximum depth < 1 m, C org stocks were extrapolated to 1 m depth. We grouped carbon stocks into estuarine and marine mangroves (Fig. 7). Marine/oceanic mangroves have no river influence, and include island mangroves, intertidal fringe mangroves, as well as mangrove forests growing in a karstic or carbonate landscape. Estuarine mangroves were categorized based on the existence of rivers, and include riverine and deltaic mangrove forests.
The overall total carbon stock (aboveground biomass and sediment including belowground biomass within 1 m depth) is higher in marine mangroves (median: 535 Mg C ha −1 , mean: 465 ± 193 Mg C ha −1 ) than that in estuarine mangroves (median: 299 Mg C ha −1 , mean: 402 ± 242 Mg C ha −1 , Fig. 7). Even among the estuarine mangroves alone, variation of C org stocks is high. The mean total C org stock in Berau (615 ± 181 Mg C ha −1 ) is lower than in an undisturbed, mixed-stature mangrove forest in Sembilang, than in the undisturbed, tall-stature mangrove forests in Timika (Murdiyarso et al., 2015), and than in the Encrucijada Biosphere Reserve, Mexico (Adame et al., 2015). However, C org stocks in Berau and the eastern SAL (483 ± 124 Mg C ha −1 ) are higher than those of the rest. It indicates that not all mangrove ecosystems have the same carbon storage capacity and that local variations of C org stocks likely mainly depend on the differences in environmental settings and conditions (Woodroffe, 1992). For instance, among all estuarine mangrove ecosystems described above, lower C org stocks are mostly found in river-dominated settings, such as in the Mui Ca Mau National Park in the Mekong Delta, in the Ganges-Brahmaputra Delta, in the Indian Sundarbans, and in the central SAL, which are likely due to high river inputs of mineral sediment that diluted the organic matter. The large difference in C org stocks we found between central and eastern SAL demonstrates the large variability which can occur within one ecosystem. It is mainly a result of the differences in the dominant hydrodynamic processes, landforms and vegetation conditions (Hinrichs et al., 2009;Holtermann et al., 2009;. In a global-scale approach and using an ecogeomorphology framework, a recent study relates carbon storage in mangrove soils to distinct coastal environmental settings. Similar to our results, it finds lower carbon densities in river delta, estuarine and lagoon settings with a substantial allochthonous inorganic sediment input than in other settings (Rovai et al., 2018).
Carbon stocks are also highly variable among marine mangrove forests, in which, Kongsi Island has the lowest C org stock (Fig. 7). It implies that not all marine mangroves accumulate high sediment organic carbon, despite little or no dilution from river discharge. Local factors such as tidal amplitude, wave action, landform and elevation are the important drivers controlling organic carbon distribution and deposition in the intertidal mangroves. The restricted accumulation of organic carbon can be partly related to rapid water circulation that washes off autochthonous organic matter, high input from the ocean such as coral rubble that dilutes the organic matter, and a low residence time of water that increases exposure time to oxygen and promotes decomposition Ranjan et al., 2011). These typical overwash mangrove ecosystems are usually found in carbonate reef environments (Woodroffe et al., 2016), such as on Kongsi Island. With its very low sedimentary C org content and stock it appears to be an extreme example, but it highlights that the variability of carbon stocks for oceanic mangroves is larger than reported so far.
The standardization to 1 m sediment depth allows for a better comparison of C org stocks among sites and hence a better assessment of the factors responsible for any observed variability. However, it does not provide the total C org stock of a mangrove ecosystem. Moreover, sediment cores obtained for carbon stock assessments are usually not longer than 3 m although the total sediment record may be longer which adds more uncertainty to carbon stock estimates. In mangrove ecosystems with high allochthonous (mineral sediment) input, the sediment record may be much longer and hence the carbon stock much higher. In turn, if in such a case only a 3 m core is assessed, the dilution with mineral sediment will result in an underestimate of the total carbon stock of a mangrove ecosystem. In the case of the Segara Anakan Lagoon, for example, carbon stocks vary by a factor of three between two sites within one ecosystem (Fig. 7). While there is little allochthonous sediment input in SAL East and the carbon stock high, there is high allochthonous input from the Citanduy River in SAL Central which dilutes sedimentary organic matter there and results in a low carbon stock.
These variations are of particular importance for the calculation of the climate change mitigation potential of mangrove ecosystems (e.g. Donato et al., 2011;Kauffman et al., 2014;Murdiyarso et al., 2015). The climate change mitigation potential builds on keeping the high stocks of carbon. In contrast, degradation and deforestation of mangrove forests results in the decomposition of organic matter and the release of CO 2 , hence a sink for atmospheric CO 2 is ultimately turned into a source. However, while these calculations allow to get an idea on the potential release of CO 2 , they do not allow to assess how much CO 2 is actually accumulating in the form of organic carbon in mangrove ecosystems.

Carbon accumulation rates in the global context
In order to assess the present day sink function of mangrove forests for atmospheric CO 2 it is likely more useful to measure the current carbon accumulation rate (CAR) in mangrove sediments. In our study, the CAR of all dated cores (C24, E40, B2) is higher than the global average of 174 g C m −2 yr −1 (Alongi, 2012), in Berau (core B2) it is even an order of magnitude higher and among the highest rates ever measured (Fig. 8). While OM proportions in our endmember mixing model (Fig. 3) indicate a predominantly autochthonous origin of sedimentary organic matter, the high sedimentation and carbon accumulation rates and a C org concentration of around 5% (Fig. 5) suggest high allochthonous input mainly of mineral sediment. The high CAR in Berau is therefore probably a product of high primary production and high allochthonous input. Primary production usually increases with stand age, and so does carbon burial in sediment (Alongi, 2014;Marchand, 2017). The high sedimentation rate measured in Berau (18 mm yr −1 ) is likely related to high river floods, during which the mean river discharge can be up to 2000 m 3 s −1 (Hoekstra et al., 2007). Occasionally, strong currents and waves influenced by the Indonesian Throughflow (ITF) may cause resuspension of sediment and prevent its settlement. However, wave effects in the northern part of Berau are generally limited due to the breakwater effects of the delta front barrier reef and its shallow depth (Hoekstra et al., 2007), which may promote the settling of particulate organic matter in Berau.
In some cases the rate of sediment delivery may be a more important driver of CAR than primary production (Alongi et al., 2001), in particular in impacted areas, such as in the Jiulongjang estuary, China, which receives high amounts of sewage and has a massive accretion rate of 46.3 mm yr −1 (Alongi, 2005). In a highly impacted mangrove ecosystem in Cubatão, Brazil, the high CAR of 1023 g C m −2 yr −1 was to a large extent attributed to the eutrophication of coastal waters (Sanders et al., 2014, Fig. 8). Interestingly, the CAR distribution pattern in the Segara Anakan Lagoon is opposite to that of the C org stocks. While the latter is higher in SAL East, the CAR is much higher in the central than in the eastern lagoon. The high sedimentation rate (36 ± 22 mm yr −1 ) and CAR in the central SAL is most likely a result of high river input due to erosion in the agriculture-dominated hinterland (Yuwono et al., 2007;Lukas, 2017) as also illustrated by the high 210 Pb xs inventory. Sedimentary OM is diluted by high mineral sediment input leading to the low C org stock, as indicated by a C org concentration of 2-4%, and it consists of a mixture of autochthonous and allochthonous OM, in which mangrove OM appears to be a minor contribution only (Figs. 3 and 4). In contrast, in SAL East little river influence, hence little allochthonous input, is responsible for a lower, but still significant accumulation rate of OM that is mainly of mangrove origin (Fig. 4).
On a global scale, the CAR in Berau and in the central SAL is at the upper end of the range and even in the eastern SAL it is higher than the global average. High rates of 581 ± 130 g C m −2 yr −1 were also measured in Irian Jaya (now Papua), Indonesia, in the Ajkwa River estuary. There, the mangrove forest receives high inputs of allochthonous sediment from the mountainous hinterland and it has a macrotidal exchange with the Arafura Sea . Similarly, a CAR of up to 949 g C m −2 yr −1 was measured at the margin of a mangrove forest in Tamandaré, Brazil, which receives allochthonous sediment input from a small river and has a mesotidal exchange with the Atlantic Ocean (Sanders et al., 2010a). In contrast, very low carbon accumulation rates of 67 ± 31 g C m −2 yr −1 were obtained in Rookery Bay, USA. It has an annual average tidal range of 0.6 m and a very low supply of allochthonous sediment by Henderson Creek which has a discharge of 0.7 m 3 s −1 only (Lynch et al., 1989). The low CAR of 55-70 g C m −2 yr −1 in mangrove forests in Celestun Lagoon, Mexico, is also related to low allochthonous inputs (Lynch et al., 1989). The lagoon is located in a karstic landscape and therefore has almost no surface runoff and a maximum tidal range of 1.5 m (Herrera-Silveira and Morales-Ojeda, 2010). It appears that besides productivity, the environmental setting is a major driver of carbon accumulation in mangrove sediments (Woodroffe et al., 2016). Our study sites in Berau and the central SAL are river-and tide-dominated settings with high allochthonous inputs and hence high CARs, while SAL East and Kongsi have low allochthonous inputs, hence low CARs.
Our study, in line with other studies, demonstrates that the riverand tide-dominated mangrove settings are quantitatively most relevant in terms of carbon storage even despite the dilution with high amounts of mineral sediment. An average of global budgets estimates mangrove carbon storage to be on the order of 22 ± 6 Tg yr −1 , based on an area of 138,000 km 2 (Giri et al., 2011). Assuming that Indonesia's contribution to global mangrove carbon storage is equivalent to its portion of mangrove area of 22.6% (Giri et al., 2011), its total carbon accumulation would be at minimum 5 Tg yr −1 . However, as Indonesia is located in the zone of maximum natural weathering and erosion and hence high river fluxes of suspended sediments (Milliman and Farnsworth, 2011), it could be even higher.

Conclusion
Blue carbon storage is presently considered one of the most important natural sinks for annual anthropogenic greenhouse gas emissions. With respect to quantifying the recent sink function of mangrove ecosystems for carbon, it is likely more important to quantify current carbon accumulation rates rather than estimating carbon stocks. Stock assessments allow to calculate the potential of CO 2 being released upon degradation and deforestation, but they do not allow to estimate present-day carbon storage. While stock assessments are useful and numerous respective studies have been conducted in the past years, they may be misleading in cases where dilution with mineral sediment masks the real accumulation potential as illustrated by our example from the Segara Anakan Lagoon. There, SAL East has a much higher carbon stock than SAL Central, but the latter has a threefold higher CAR, hence presently stores carbon in a much higher rate than SAL East does. Quantifying carbon accumulation in mangrove sediments is therefore a useful tool in valuing ecosystem services and it is of particular importance for PES (payments for environmental services) and REDD+ (Reducing Emissions from Deforestation and Forest Degradation and the role of conservation, sustainable management of forests and enhancement of forest carbon stocks in developing countries) schemes.  Chmura et al. (2003).