Mangrove forests are highly productive but globally threatened coastal ecosystems, whose role in the carbon budget of the coastal zone has long been debated. Here we provide a comprehensive synthesis of the available data on carbon fluxes in mangrove ecosystems. A reassessment of global mangrove primary production from the literature results in a conservative estimate of ∼218 ± 72 Tg C a−1. When using the best available estimates of various carbon sinks (organic carbon export, sediment burial, and mineralization), it appears that >50% of the carbon fixed by mangrove vegetation is unaccounted for. This unaccounted carbon sink is conservatively estimated at ∼112 ± 85 Tg C a−1, equivalent in magnitude to ∼30–40% of the global riverine organic carbon input to the coastal zone. Our analysis suggests that mineralization is severely underestimated, and that the majority of carbon export from mangroves to adjacent waters occurs as dissolved inorganic carbon (DIC). CO2 efflux from sediments and creek waters and tidal export of DIC appear to be the major sinks. These processes are quantitatively comparable in magnitude to the unaccounted carbon sink in current budgets, but are not yet adequately constrained with the limited published data available so far.
 Mangrove ecosystems thrive along coastlines throughout most of the tropics and subtropics. These intertidal forests play important ecological and socioeconomical roles, e.g., by acting as a nutrient filter between land and sea [e.g., Robertson and Phillips, 1995; Rivera-Monroy et al., 1999], contributing to coastline protection [e.g., Field, 1995; Vermaat and Thampanya, 2006], providing commercial fisheries resources [e.g., Constanza et al., 1997; Barbier, 2000; Diele et al., 2005], and as nursery grounds for coastal fish and crustaceans [Baran and Hambrey, 1998; Rönnbäck, 1999; Mumby et al., 2004]. Tropical forests in general are a disproportionately important component in the global carbon cycle, and are thought to represent ∼30–40% of the terrestrial net primary production [see Malhi and Grace, 2000; Clark et al., 2001a]. Although the area covered by mangrove forests represents only a small fraction of tropical forests, their position at the terrestrial-ocean interface and potential exchange with coastal waters suggests these forests make a unique contribution to carbon biogeochemistry in coastal ocean [Twilley et al., 1992]. The coastal zone (<200 m depth, covering ∼7% of the ocean surface, Gattuso et al. ) has an important role in the oceanic carbon cycle, and various estimates indicate that the majority of mineralization and burial of organic carbon, as well as carbonate production and accumulation takes place in the coastal ocean [e.g., Gattuso et al., 1998; Mackenzie et al., 2004; Duarte et al., 2005]. The potential impact of mangroves on coastal zone carbon dynamics has been a topic of intense debate during the past decades. In particular, the “outwelling” hypothesis, first proposed for mangroves by Odum  and Odum and Heald  suggested that a large fraction of the organic matter produced by mangrove trees is exported to the coastal ocean, where it should form the basis of a detritus food chain and thereby support coastal fisheries. A number of recent studies have indicated that a direct trophic link between mangrove forest production and offshore secondary production is unlikely for many mangrove systems [e.g., Lee, 1995].
 Despite the large number of case studies dealing with various aspects of organic matter cycling in mangrove systems [Kristensen et al., 2008], there is still no consensus on the magnitude and partitioning of mangrove primary production and the fate of the organic matter produced. Several authors have suggested that mangrove-derived organic matter is of global significance in the coastal zone: estimates indicate that mangrove forests could be responsible for ∼10% of the global export of terrestrial particulate and dissolved organic carbon (POC and DOC) to the coastal ocean (Jennerjahn and Ittekkot  and Dittmar et al. , respectively), and for ∼10% of the global organic carbon burial in the coastal ocean [Duarte et al., 2005]. The exchange of carbon between tidal wetlands such as mangrove forests or salt marshes and the coastal ocean, and its ultimate fate in the ocean is therefore increasingly recognized as a potentially important component in the ocean carbon budget [Twilley et al., 1992].
 Published global estimates on central components of the mangrove carbon budget (summarized in Table 1) have two main shortcomings: (1) information on mangroves is very limited and carbon budgets are therefore based on relatively small data sets; and (2) while there is a wealth of data on litter fall, estimates on belowground allocation and wood production are still scarce because of methodological difficulties and some published budgets are consequently biased because they ignore wood and belowground production. Moreover, mangrove systems occur in a wide range of environmental settings, and the degree of organic matter retention and export can therefore be expected to vary considerably according to factors such as geomorphology, tidal amplitude, local climate, vegetation type, and biotic influences, e.g., the activity of litter-retaining crabs [e.g., Twilley et al., 1986; Smith et al., 1991; McIvor and Smith, 1995; Twilley et al., 1997; Nordhaus et al., 2006]. This inherent variability among and within mangrove systems evidently complicates global extrapolations.
Table 1. Summary of Literature Estimates of Various Components in the Global Mangrove C Budgeta
Fluxes are expressed in Tg C a−1, except for the estimates by Duarte and Cebrián  which are in percent of the overall net primary production. Note that the areal extent of mangroves (in km2) differs between some of the data sources.
Net Primary Production
280 (litter + wood)
9.1 ± 2.4%
40.1 ± 6.5%
10.4 ± 3.6%
Organic carbon export
30–50 (POC + DOC)
46 (POC + DOC)
26.4 (as DOC)
29.5 ± 9.4%
 Nevertheless, when comparing some of the budget estimates (Table 1) with the global riverine organic carbon export (∼300–400 Tg C a−1 [see Schlünz and Schneider, 2000, and references therein]), it is clear that the potential role of mangrove forests in global coastal zone carbon budgets is significant, and that a more refined assessment of global mangrove carbon budgets is due. Moreover, mangrove forests are being cleared and converted worldwide at alarming rates (∼1% of the area a−1 [FAO, 2003]) and the past few decades have witnessed a significant decrease in the global mangrove forest cover (with estimates as high as ∼35% during the past 20 years according to Valiela et al. ).
 The purpose of this study is to provide an overview of the current knowledge on quantitative aspects of mangrove carbon dynamics. We performed an intensive literature search for data related to mangrove primary production and the various sinks of organic carbon (Figure 1). The results of our analysis stress the need to include belowground and wood production in primary production estimates. Furthermore, the currently available estimates on organic carbon export, burial and mineralization (see Table 1) account for <50% of mangrove net primary production (NPP), indicating that major unaccounted loss pathways exist and/or that some of the currently available budget components are severely biased. We propose a number of processes and pathways that might have been overlooked or underestimated in current budgets, and discuss how these important gaps in the global carbon budget for these systems may be resolved.
2. Data Sources
 The literature was screened for data relevant to mangrove primary production, assimilation/respiration by fauna, and organic carbon export, burial, and mineralization. A number of earlier reviews [Twilley et al., 1992; Saenger and Snedaker, 1993; Lee, 1995; Duarte et al., 2005; Jennerjahn and Ittekkot, 2002] provided a basic set of data, which we supplemented with a significant amount of (mostly recent) new data from case studies. When data were only presented as figures in the original publications, these were digitized and the data extracted using PlotDigitizer (http://plotdigitizer.sourceforge.net/). Estimates of carbon burial were taken from Twilley et al. , Jennerjahn and Ittekkot , and Duarte et al. . CO2 fluxes from sediments are typically used as a proxy for benthic mineralization, and we compiled data on benthic CO2 fluxes from a variety of literature sources (see Table S2). Similarly, estimates of CO2 exchange between mangrove creek surface waters and the atmosphere have been compiled (see Table S3). Moreover, we compiled a number of data sets on the distribution of dissolved organic and inorganic carbon (DOC and DIC) in mangrove creek waters, to provide a preliminary estimate of the relative importance of mangrove C export as dissolved inorganic carbon.
 Standard error propagation procedures were used to place uncertainties on our budget estimates; that is, standard deviations on budget components were propagated where possible. Since the area of mangrove cover (160,000 km2; based on FAO ) and carbon concentrations in different tissues were taken as constants (i.e., with no uncertainty ascribed), some of the uncertainty estimates should be considered conservative.
3. Primary Production
 A total of 178 litter fall measurements were compiled. Previous studies have indicated that litter fall exhibits a pronounced geographical trend, with highest litter fall rates near the equator and decreasing with increasing latitude [e.g., Twilley et al., 1992; Alongi, 2002]. Our extended data set confirms this trend (Figure 2; for full data, see Table S1), although it clearly shows more scatter than earlier compilations [Twilley et al., 1992]. This is not surprising, since primary production is influenced by a range of other factors, e.g., nutrients such as N and P [e.g., Chen and Twilley, 1999; Feller et al., 2002]. Using the same latitudinal zones as those used by the latter authors (i.e., 0–10°, 10–20°, 20–30°, and >30°), litter fall rates were found to be significantly higher in the 0–10° region (10.4 ± 4.6 t ha−1 a−1, n = 53) when compared to other latitudes, and significantly lower in the >30° latitudes (4.7 ± 2.1 t ha−1 a−1, n = 16). No significant differences were found between 10 and 20° (9.1 ± 3.4 t ha−1 a−1, n = 47) and 20–30° (8.8 ± 4.2 t ha−1 a−1, n = 62) latitudes (Figure 2). In order to scale these data to a global litter fall estimate, we assumed a global mangrove cover of 160,000 km2 [FAO, 2003], and a latitudinal distribution similar to that presented by Twilley et al. , i.e., we assumed that the loss of mangrove cover over the past decade was similar in each latitudinal region. This results in a global litter fall estimate of 156 ± 45 Tg a−1, equivalent to 68.4 ± 19.7 Tg C a−1 using a carbon content of 44% (the latter is based on a compilation of literature and our own unpublished data, 2002–2006). This estimate is ∼25 Tg C a−1 lower than the most recent previous estimate [Jennerjahn and Ittekkot, 2002], which is largely due to a different areal extent used (160,000 km2 in this study, versus 200,000 km2 used by Jennerjahn and Ittekkot ).
 Estimates of aboveground wood production are relatively scarce. Twilley et al.  mentioned 11 measures of wood production and estimated the total global wood production at 160 Tg C a−1 (for a global area of 200,000 km2). With 20 additional estimates of wood production (see Table 2 for data and sources), wood production estimates were found to range between 1.1 and 24.1 t ha−1 a−1, with no clear latitudinal trends in the still limited data set (Figure 3a). We compare two different approaches to estimate wood production by mangroves on a global scale: (1) On the basis of the average of all available wood production estimates (10.0 ± 5.6 t ha−1 a−1, n = 31), wood production can be estimated at 66.4 ± 37.3 Tg C a−1 (using a %C of 41.5%, based on a compilation of literature data) and, alternatively, (2) for a selection of these data where concurrent litter fall estimates are also available (Table 2 and Figure 3b), we estimate an average wood/litter production ratio of 1.03 ± 0.54 (n = 23). This ratio, which is within the range reported for other tropical forest ecosystems [Malhi et al., 2004] can subsequently be used to convert global litter fall rates (see above paragraph, this section) to an estimated global wood production of 161 ± 95 Tg a−1, or 66.7 ± 39.6 Tg C a−1. Both estimates are very similar, but significantly lower than the previously mentioned estimate (160 Tg C a−1 [see Twilley et al., 1992]), even when correcting for a 20% decrease in global area of mangroves (128 Tg C a−1) that occurred between the two estimates. It should also be noted that the data on which our extrapolations are based are from studies measuring biomass increments over time, and that this approach does not consider processes such as natural gap formation and regeneration, which can result in significant natural biomass turnover [Duke, 2001].
Table 2. Synthesis of Wood Production Data (and Litter Fall, if Available) in Mangrovesa
Expressed in t ha−1 a−1. Data are sorted per latitudinal zone.
 While belowground roots, pneumatophores and prop roots can form a substantial fraction of the total mangrove biomass [e.g., Komiyama et al., 1987; Mall et al., 1991], estimates of belowground production are even scarcer, which is undoubtedly due to the methodological difficulties involved. Of the four studies that have reported belowground root production estimates [McKee and Faulkner, 2000; Gleason and Ewel, 2002; Sánchez, 2005; E. Castañeda-Moya et al., unpublished data, 2002–2006], three were used for further analysis, since the rates presented by Gleason and Ewel  referred only to the top 15 cm. Given the limited number of data (n = 16), we first compared these root production estimates (all for mixed species assemblages) with concurrently reported litter fall rates, which resulted in a root/litter production ratio (in C equivalents) of 1.20 ± 0.76 (n = 16). Applying this ratio to our global litter fall rate, global root production can be estimated at 82.8 ± 57.7 Tg C a−1. Although this may appear to be high, our estimated root/litter production ratio for mangroves is considerably lower than ratios reported for productive terrestrial forest ecosystems (typically 2.2–2.5, see Raich and Nadelhoffer ). This indicates that our root production estimate is likely to be conservative. The partitioning of biomass between belowground and aboveground biomass may also be dependent on nutrient availability, yet the precise impact of nutrient limitation or nutrient additions on belowground allocation and root turnover are not well understood [Nadelhoffer, 2000; Hendricks et al., 2006] and to our knowledge, no studies have assessed such relationships in mangroves. Moreover, it must be stressed that because of the measurement approach (ingrowth of roots), the published root production estimates refer only to fine root production. Although fine roots may not always be the dominant root fraction in terms of biomass (e.g., Komiyama et al.  and Fiala and Hernandez ; depending on how fine roots are defined), they are considered the most active component and may therefore contribute more to NPP than coarse roots [Clark et al., 2001b]. Coarse root production in mangroves has to our knowledge not been measured or estimated separately, and will require careful consideration of methodological issues, given the variety of different root structures formed by various mangrove types (e.g., stilt roots, cable roots, pneumatophores). Production of stilt roots by Rhizophora spp., however, is in most cases included in the existing estimates of wood production, since these are included in allometric relationships to estimate total aboveground biomass [e.g., see Sherman et al., 2003].
 When comparing our estimated rates of litter, wood, and (fine) root production, it becomes evident that litter production only amounts to ∼32% of the total mangrove NPP. A few reports have made a direct comparison of litter fall rates and total NPP, and indeed suggested that litter fall represents ∼30% of the overall NPP [see Alongi et al., 2005, and references therein]. These estimates match well with our global extrapolations, and hence, indicate that our numbers are realistic and likely to be valid within reasonable limits. The overall global NPP of mangroves (using an areal extent of 160,000 km2) can therefore be estimated at ∼218 ± 72 Tg C a−1, using the indirect approach described above for the wood production component.
 This estimate of NPP should be considered conservative, since (1) coarse root production data are not available and their contribution is therefore not included, (2) we do not account for direct herbivory in this context, although it represents a formal component of NPP (i.e., litter fall rates were not corrected for leaf damage or direct consumption, see further), and (3) a number of other components of NPP such as the exudation of dissolved organic carbon by roots [Clark et al., 2001b] are as yet unknown for any mangrove system.
4. Sinks of Mangrove Production
4.1. Organic Carbon Export
 Despite the longstanding debate on the role of mangrove forests in exporting organic carbon to the coastal zone [e.g., Lee, 1995; Jennjerjahn and Ittekkot, 2002; Dittmar et al., 2006], the number of quantitative estimates remains very limited: we only found six and seven estimates for DOC and POC export, respectively, and 11 estimates for total organic carbon export (see Table 3). If the average rates of DOC export (150 ± 134 g C m−2 a−1) and POC export (137 ± 172 g C m−2 a−1) are combined, they are within the range of independent estimates for total organic carbon (TOC) export (252 ± 277 g C m−2 a−1). Scaling these to a global level results in estimated exports of 44.6 ± 30.5 Tg C a−1 and 40.3 ± 44.3 Tg C a−1 based on the sum of DOC (24 ± 21 Tg C a−1) and POC (22 ± 27 Tg C a−1) export estimates or when scaling the TOC export rates, respectively. Although based on a small data set with significant shortcoming, our estimates are within the same range as previous global estimates, which are either partially based on the same data [Twilley et al., 1992] or were derived independently [Jennerjahn and Ittekkot, 2002; Dittmar et al., 2006]. Note that these export estimates only refer to direct export of organic carbon, and do not include movement of animal biomass dependent on mangrove-derived carbon, which has also been proposed as a potential mechanism for organic matter export [e.g., see Lee, 1995].
Table 3. Summary of Literature Data on Particulate, Dissolved, and Total Organic Carbon (POC, DOC, and TOC) Export From Mangrovesa
Adapted from Lee  and updated with more recent data. Fluxes are expressed in g C m−2 a−1.
 Three different studies have estimated global organic carbon burial rates in mangrove systems, each using a different approach. Twilley et al.  compiled the available data on organic carbon burial rates, while Jennerjahn and Ittekkot  estimated global carbon burial rates on the basis of a number of assumptions, i.e., (1) that litter fall provides the dominant organic carbon input in the sediments, (2) a global litter fall of 92 Tg C a−1, and (3) that 50% of the litter is exported and 25% mineralized. Duarte et al. , on the basis of the data compiled by Chmura et al. , presented both a bottom-up estimate (i.e., upscaling of carbon burial rates) and an estimate derived from global mangrove community mass balance. Despite the different approaches used, they all result in a very similar estimate at ∼23 Tg C a−1. Since no significant new data are available, we have scaled this number to the surface area used here (160,000 km2), which results in a global organic carbon burial rate of 18.4 Tg C a−1. Note that none of the literature sources provided an estimate of uncertainty on these global burial rates, hence no error was propagated in further calculations.
4.3. CO2 Efflux as a Proxy for Mineralization
 Total mineralization in mangrove sediments has typically been estimated through measurements of CO2 fluxes from sediments. Such data are available for a wide range of mangrove systems, and have been measured both under inundated conditions (i.e., as the increase in DIC through time in overlying water during sediment incubations, e.g., Alongi et al. ) and exposed conditions (typically as gaseous CO2 increase through time in a closed or flowthrough benthic chamber, e.g., Kristensen and Alongi ). Since the flux rates in our data set were not significantly different between exposed and inundated sediments (p > 0.05 and >0.2 for light and dark conditions), data were pooled for further analysis (Figure 4). Dark fluxes from sediments range between 6 and 241 mmol CO2 m−2 d−1, with an average of 61 ± 46 mmol m−2 d−1 (n = 82). Under light conditions about half of the available flux data show a net CO2 uptake (Figure 4), with an average influx of −15 ± 54 mmol m−2 d−1 (n = 14). These lower CO2 fluxes result from CO2 uptake during photosynthesis by benthic primary producers and should therefore not be taken into account since we are interested in assessing mineralization of mangrove-derived organic matter. Moreover, this light flux is based on a very limited data set and should be used with caution given the extreme variability in these data.
 Mangrove creek waters have consistently been found to show high CO2 oversaturation, and hence, are a net source of CO2 to the atmosphere, with average CO2 flux estimates for a given system generally >20 mmol m−2 d−1 [Borges et al., 2003]. We compiled 21 estimates on water-air CO2 fluxes (see Table S2) with an overall average of 59 ± 52 mmol m−2 d−1. For most of these data, CO2 fluxes were estimated on the basis of water column pCO2 and gas transfer velocities estimated from field wind speed measurements (see Borges et al  for discussion).
 It must be stressed that these sediment and water column estimates relate only to net CO2 fluxes, and not to overall mineralization rates (see discussion below, section 5.1). Upscaling CO2 fluxes for sediments and the water column separately is somewhat problematic, since the surface areas to be used are not static in these intertidal systems. However, given the similar magnitude in CO2 efflux from both sediments in the dark and water column, we estimated the overall CO2 efflux from mangrove systems to be in the order of ∼60 ± 45 mmol m−2 d−1. This results in a global estimate of 42 ± 31 Tg C a−1.
5. Gaps in the Carbon Budget
 All the abovementioned estimates of organic carbon sinks (export, burial, and CO2 efflux) together only accounted for ∼45% of the estimated NPP (Figure 5), thus leaving a surprisingly large part of the mangrove production unaccounted for (112 ± 85 Tg C a−1, which is equivalent in magnitude to ∼30–40% of the global riverine flux of organic carbon). Assuming that mangrove forests are in equilibrium and are currently not showing net biomass accumulation, this implies that either some sink components of the global mangrove C budget are severely underestimated, and/or that some quantitatively important pathways for mangrove-derived organic matter removal have been overlooked in previous budget studies. Although we must stress that the existing literature does not provide sufficient data to fill this gap, a closer look at several mechanisms and processes may explain and solve part of this discrepancy. In particular, the potential importance of export of mangrove carbon as dissolved inorganic carbon (DIC), the emission of CO2 from sediments via burrows and along pneumatophores, and the consumption and respiration of organic matter by faunal communities will be discussed below, sections 5.1 and 5.2.
5.1. Mineralization and Export of Inorganic Carbon
 Mineralization is generally considered to be a major fate of plant production in coastal ecosystems [Duarte and Cebrián, 1996]. However, when based on sediment CO2 fluxes, benthic mineralization only represents ∼15% of the total mangrove NPP. For a number of reasons, however, these CO2 emission rates are unlikely to serve as appropriate proxies for mineralization:
 1. Core incubations typically use sediments which are free of pneumatophores and crab burrows. Recent data, however, indicate that these structures are important as vectors for enhancing CO2 exchange, and add considerably to the total CO2 flux from the sediment [Kristensen et al., 2008]. Crab burrows can be very complex, abundant, and deep structures, and thus greatly enhance the surface area of the sediment-air or sediment water interface where exchange of CO2 or DIC can take place [Thongtham and Kristensen, 2003]. Similarly, CO2 emission rates were found to be greatly enhanced when Sonneratia or Avicennia pneumatophores were included in the measurements, with CO2 emission rates equivalent to 0.2–0.7 mmol CO2 d−1 per individual pneumatophore (E. Kristensen, unpublished data, 2005).
 2. An unknown part of CO2 released from pneumatophores and rooted sediments might represent plant respiration rather than heterotrophic respiration (i.e., mineralization). Since our mangrove carbon budget is based on net primary production, the CO2 release due to autotrophic respiration should not be included, but there are currently insufficient data to estimate this contribution [Lovelock et al., 2006]. This stresses the need for novel methodological approaches that allow quantitative estimates of the partitioning of measured CO2 fluxes between autotrophic and heterotrophic respiration.
 3. Lateral transport of DIC resulting from mineralization via pore water drainage is not taken into account using the standard approach used to estimate sediment mineralization: the latter is based on the assumption that efflux of carbon dioxide equates with depth-integrated mineralization. Mangrove pore waters are typically rich in total alkalinity and DIC [e.g., Ovalle et al., 1990; Bouillon et al., 2007a], indicating that buildup of inorganic carbon resulting from mineralization occurs. Lateral transport of DIC-enriched mangrove pore water to creeks thus results in an underestimation of mineralization rates if based on sediment-water carbon dioxide effluxes only. The impact of mineralization in intertidal sediments and subsequent lateral transfer on the DIC dynamics in mangrove creeks has been demonstrated in a number of studies [e.g., Ovalle et al., 1990; Borges et al., 2003; Bouillon et al., 2007c].
 In order to assess the potential magnitude of lateral CO2 transport, we evaluated the water column concentrations of DIC and DOC along the salinity gradient in a number of mangrove systems. Because of lateral inputs into creek waters, DIC and DOC typically show a nonconservative pattern in the mangrove creek water column (see Figure 6). The relative amounts of “excess” DIC and DOC (ΔDIC and ΔDOC) can thus be considered proportional to their relative export rates, on the condition that the tidal variations in both parameters are similar, so that their overall relative fluxes are not influenced by the variations in flow rates during the tidal cycle [e.g., Dittmar and Lara, 2001]. We compiled data from several mangrove creeks in different regions (see Table 4), and calculated the amount of excess DOC and DIC for each of these data sets. This was based either on deviations from conservative mixing scenarios, or by comparing the amounts of DIC and DOC relative to reference concentrations at the outer boundary, i.e., where lowest DIC and DOC concentrations were found (see Figure 6). Although such data sets are only available from a limited number of sites (n = 5), the resulting calculations consistently show that ΔDIC exceeds ΔDOC, by a factor of ∼3–10 (Table 4). Under the assumption that both originate mainly from the tidal exchange and therefore follow the same tidal variations [e.g., Bouillon et al., 2007c], this implies that DIC export should exceed DOC export to the same degree. Assuming a global DOC export rate in the order of 24 ± 21 Tg C a−1 (see above, section 4.1), and using the available data on ΔDIC/ΔDOC ratios, tidal DIC export would represent 178 ± 165 Tg C a−1. Other processes may obviously affect DIC and DOC pools after their export from the system boundaries (e.g., CO2 efflux, mineralization or photodegradation of DOC), but data to assess this further are currently not available. It must be noted that the estimated tidal DIC export in principle includes the cumulated DIC released from mangrove sediments during tidal inundation, DIC released from permanently inundated sediments, and DIC from organic carbon degradation in the mangrove creek waters (which has seldom been measured). However, it does not include the sediment CO2 emission during exposure or CO2 exchange between the water column and atmosphere. Thus, although there are insufficient data to fully constrain CO2 emission from sediments and tidal DIC export, our analysis clearly illustrates that sediment mineralization is likely to be much higher than suggested by standard CO2 flux measurements alone. We thus suggest that a much larger fraction of mangrove production than previously anticipated may be mineralized and either emitted directly to the atmosphere, or exported as DIC to adjacent waters.
Table 4. Overview of Ratios of Excess DIC to Excess DOC (ΔDIC/ΔDOC) in Different Mangrove Creeksa
The n represents the number of data between end-members for each system for which ratios of excess DIC to excess DOC (ΔDIC/ΔDOC) could be determined. See text for details.
Ca Mau (Vietnam)
S. Bouillon and A. V. Borges, unpublished data, 2005
5.2. Secondary Production and Respiration by Fauna
 Although NPP formally includes biomass lost through direct feeding [Clark et al., 2001b], we did not include this component in our budget estimates. The generally low levels of direct herbivory (typically <5%, e.g., Robertson and Duke , Lee , and Saur et al. ) are unlikely to bias the overall budget considerations (see also section 3). It should be noted, however, that high levels of herbivory have been reported in a few cases (e.g., ∼30% in the work of Duke ). Particularly, the few available long-term measurements of leaf herbivory which take into account the consumption of entire leaves and abscission due to herbivory damage result in higher biomass loss than typically reported [Burrows, 2003].
 However, mangrove forests harbor a highly diverse and abundant invertebrate fauna (primarily crabs), which may rely directly or indirectly on carbon from mangroves. From a community perspective, the contribution of mangroves to invertebrates has been found to be less dominant than previously thought [e.g., Bouillon et al., 2008]. Nevertheless, considering the high abundance of fauna, their overall impact on mangrove carbon turnover is considerable [see Kristensen et al., 2008]. Individual species have been found to remove or consume a significant proportion of litter fall (most estimates range between 10 and 80%) in both Old World and New World mangrove systems [e.g., McIvor and Smith, 1995; Lee, 1998; Proffitt and Devlin, 2005]. Leaf litter retention by fauna can thus have a significant impact on organic matter dynamics and litter turnover, although the relative importance of fauna and geophysical processes (e.g., tidal inundation frequency or river flow) on litter dynamics vary substantially among different mangrove settings [Twilley et al., 1997].
 Quantitative estimates of carbon processing by the entire faunal community are scarce: the only detailed estimate available is that by Koch and Wolff , who studied the energy balance of key species of epifauna in a Brazilian mangrove forest. For the high intertidal site in their study (where Ucides cordatus, a crab species known to feed substantially on mangrove material, was dominant), secondary production and respiration of key faunal species indicate a total assimilation rate equivalent to 15–20 mmol C m−2 d−1. Even for individual species, the few data available confirm the substantial role of mangrove fauna. Thus, Thongtham and Kristensen  found that the assimilation (i.e., production + respiration) of the sesarmid crab Neoepisesarma versicolor at a density of 5 ind. m−2 is equivalent to ∼40–60 mmol C m−2 d−1 in a Thai mangrove forest. When taking into account that faunal communities rely only partially on mangrove-derived carbon, these numbers stress the potential role of faunal production and respiration for the mangrove carbon budget. Reliable quantitative extrapolations on a global scale, however, require substantial research efforts to provide a broader database on the faunal impact.
6. Uncertainties in Budget Estimates and Source Characterization
 The large fraction of mangrove primary production that is unaccounted for in current budgets also warrants a critical inspection of the available literature estimates. All estimates have inherent uncertainties due to methodological issues, the limited amount of data on many of the potential carbon sinks, and the intrinsic variability within and between different mangrove systems regarding nutrient availability and relevant biogeochemical processes [e.g., Twilley et al., 1997; Rivera-Monroy et al., 2004; Poret et al., 2007; Kristensen et al., 2008]. An additional caveat in budgeting efforts is that with few exceptions [e.g., Dittmar et al., 2001, 2006], most of the flux estimates refer to the total carbon pool, i.e., not taking into account that other potential carbon sources (e.g., riverine or marine inputs, photosynthetic and chemoautotrophic microbial production) may also contribute to the organic carbon flux measured. Such allocthonous sources have, however, been shown to contribute significantly to particulate and dissolved organic carbon pools, as well as microbial and faunal food webs [see Bouillon et al., 2008].
 For the organic carbon export component (see section 4.1.), one of the main shortcomings is the relatively small number of well-documented studies and their high variability. Direct measurements of net carbon export require high-frequency sampling efforts coupled to knowledge on hydrodynamics, since the net result of import and export fluxes is often small compared to the gross fluxes [see also Lee, 2006]. Furthermore, the contribution of terrestrial and marine organic carbon to suspended and dissolved organic carbon can in some cases be substantial and would result in an overestimate of export of mangrove carbon if these are not considered.
 The three different burial estimates (see section 4.2) match remarkably well. Nevertheless, these may be partially skewed when the underlying data are based on direct estimates of sediment or carbon accumulation, since the potential contribution of nonmangrove sources may result in an overestimate of burial in the overall mangrove carbon budget. However, part of the underlying data is based on community mass balance budget [Duarte et al., 2005], which may ignore the potentially important carbon storage fuelled by belowground production [e.g., Chen and Twilley, 1999; Middleton and McKee, 2001; Marchand et al., 2003]. Recent evidence indeed suggests that belowground root accumulation contributes significantly to overall surface elevation changes in mangrove sediments [McKee et al., 2007].
 The CO2 emission data similarly reflect total CO2 emission rather than CO2 production linked to heterotrophic respiration fuelled by mangrove detritus. Some of the CO2 may be attributed to root respiration and there is recent evidence that other carbon inputs may contribute substantially to mineralization, both in the upper sediment layers [Bouillon and Boschker, 2006] and on an ecosystem level [Bouillon et al., 2007c]. With the data at hand, the bias caused by contributions of other carbon sources to the various sinks discussed here is difficult to constrain, yet this only adds to the conclusion that much of the mangrove production is unaccounted for in existing budget estimates.
7. Summary and Conclusions
 Despite the often cited role of mangroves in exporting organic carbon to adjacent waters, a reevaluation of the available data stresses the gaps in our knowledge on carbon cycling in these tropical coastal ecosystems. Net primary production by mangrove forests, assuming a global coverage of 160,000 km2, was estimated at 218 ± 72 Tg C a−1, with litter fall, wood and root production accounting for ∼31, 31, and 38% of the overall production, respectively (Figure 5). The available literature estimates on carbon burial, organic carbon export and CO2 emission from sediments and the water column are equivalent to ∼45% of the mangrove production, leaving ∼112 ± 85 Tg C a−1 unaccounted for in current budgets. Our analysis suggests that tidal export of carbon as DIC is a quantitatively important pathway (178 ± 165 Tg C a−1), and that CO2 fluxes from intertidal sediments are likely significantly higher than currently assumed. These processes are of the same order of magnitude as the unaccounted carbon sink, and in view of the limited data available, there is a need for more detailed measurements of these processes in a range of different mangrove systems. Besides the need for more quantitative process studies, complementary approaches to determine the contribution of mangroves and other carbon sources to various fluxes and process rates are required to better constrain the major sinks of mangrove carbon.
 S.B. was supported by a postdoctoral mandate from the Research Foundation Flanders (FWO-Vlaanderen). A.V.B. is a research associate with the FNRS, Belgium. This research was also supported by the Netherlands Organisation for Scientific Research (PIONIER), and part of the original data were obtained in the framework of the EC-STREP project PUMSPEA (contract 510863). E.K. was supported by the Danish Science Research Council (grant 21020463). S.Y.L.'s research on mangrove ecosystems has been funded by the Australian Research Council (grants A0000284 and DP0344546). Funding by the Jet Propulsion Laboratory–NASA (Land Cover Land Use Change program) and the NSF-FCE-LTER program (DEB-9901514) to V.H.R.-M. and R.R.T. contributed to the preparation of this manuscript. We thank F. Dehairs and D.P. Gilllikin for useful suggestions to an earlier draft of this manuscript, V. Koch for providing raw data on epifaunal production and respiration, and R. Jaffé for his constructive review of this manuscript. This is publication 4201 of the Netherlands Institute of Ecology (NIOO-KNAW).