Role of carbonate burial in Blue Carbon budgets

Calcium carbonates (CaCO3) often accumulate in mangrove and seagrass sediments. As CaCO3 production emits CO2, there is concern that this may partially offset the role of Blue Carbon ecosystems as CO2 sinks through the burial of organic carbon (Corg). A global collection of data on inorganic carbon burial rates (Cinorg, 12% of CaCO3 mass) revealed global rates of 0.8 TgCinorg yr−1 and 15–62 TgCinorg yr−1 in mangrove and seagrass ecosystems, respectively. In seagrass, CaCO3 burial may correspond to an offset of 30% of the net CO2 sequestration. However, a mass balance assessment highlights that the Cinorg burial is mainly supported by inputs from adjacent ecosystems rather than by local calcification, and that Blue Carbon ecosystems are sites of net CaCO3 dissolution. Hence, CaCO3 burial in Blue Carbon ecosystems contribute to seabed elevation and therefore buffers sea-level rise, without undermining their role as CO2 sinks.

M angrove forests and seagrass meadows have the capacity to elevate the seabed through the accretion of inorganic and organic particles 1 at global rates of~0.5 and~0.2 cm yr −1 , respectively 1 . Sediment accretion in mangrove forests and seagrass meadows leads to the sequestration of organic carbon (C org ) 2,3 originating from within and outside of the vegetated ecosystem 4 . Although mangroves and seagrass ecosystems occupy only a small fraction of the total coastal area (< 2%), they contribute 10% and 25% to the yearly C org sequestration in the coastal zone 1,5 , respectively. Recognition of mangrove and seagrass meadows, together with saltmarshes, as sites of intense C org burial led to the formulation of Blue Carbon strategies to mitigate and adapt to climate change, through conservation and restoration of these ecosystems 1,[6][7][8] . The focus on Blue Carbon has provided substantial impetus to assess sediment C org concentrations and burial rates in vegetated coastal ecosystems, which recently have been widely reviewed 9 .
C org generally represents a minor fraction (2-3%) of buried material within mangrove and seagrass sediments 10,11 (although this is highly variable 12 ), the rest being siliciclastic and carbonate particles. A global assessment of the concentration of inorganic carbon concluded that C inorg can exceed C org concentration in seagrass sediments 13 . Seagrass and mangrove plants do not calcify per se; however, they provide habitats for an abundant associated calcifying fauna and flora (e.g., crabs, sea stars, snails, bivalves, calcified algae, foraminifera), whose shells and skeletons may be deposited and buried in the sediment along with the plant litter, and the organic and inorganic particles imported from adjacent ecosystems.
Counterintuitively, CaCO 3 production represents a source of CO 2 to the atmosphere, as calcification produces CO 2 with a ratio of~0.6 mol of CO 2 emitted per mol of CaCO 3 precipitated 14 . This has led to the argument that high CaCO 3 burial may partially offset CO 2 sequestration associated with C org burial in some seagrass meadows and mangrove forests 15 . However, there are several caveats that affect these arguments and render inferences on the role of Blue Carbon ecosystems as net CO 2 sinks or sources inconclusive 13,16 , based on the comparison of C org and C inorg sediment burial rates. To date, very few articles report the burial rates of CaCO 3 in mangrove and seagrass ecosystems [15][16][17] , and the role of CaCO 3 burial in sediments and CO 2 emissions depends on the balance between dissolution and production. If CaCO 3 dissolution equals local calcification, then the burial of CaCO 3 is supported exclusively by allochthonous inputs and is neutral in terms of CO 2 emissions or sequestration. If dissolution exceeds local calcification, then CaCO 3 dynamics add to the CO 2 sink capacity of Blue Carbon ecosystems, even if CaCO 3 , which must be subsidized from allochthonous sources, is buried in the sediments. Only if CaCO 3 dissolution is lower than local calcification does CaCO 3 burial result in CO 2 emissions.
Here we address the current gap in global estimates of C inorg burial in seagrass and mangrove ecosystems by providing first estimates of contemporary (last century) C inorg burial rates. We rely on a compilation and analysis of data on sediment chronologies (i.e., including radiometric dating of sediment cores with 210 Pb) and C inorg concentrations from around the world (Fig. 1). We compare burial, calcification and dissolution rates in three locations where most of the carbon mass balance terms were available. We then address the role of CaCO 3 burial in CO 2 emissions by resolving the source of the CaCO 3 buried in seagrass meadows as either allochthonous or autochthonous (i.e., from associated flora and fauna). We conclude that the high amounts of CaCO 3 found in Blue Carbon sediments can not be converted into CO 2 emissions.

Results
Global disparities in Blue Carbon sediments. CaCO 3 supports an important part of sediment accretion rates (SARs) in seagrass ecosystems, although with large geographical disparities and a non-normal distribution (Shapiro-Wilks test, p < 0.001). Indeed, in 40% of global locations, the CaCO 3 concentration was under 10% dry weight (DW), whereas in 28% of locations the CaCO 3 content exceeded 80 %DW (see Supplementary Figure 1a). Overall, the median (interquartile range: IQR) global concentration of CaCO 3 in seagrass meadow sediments was 61 (56) %DW (mean ± SE of 54 ± 7).
In mangrove forests, we observe a large difference between the mean (± SE) and the median (IQR) CaCO 3 concentration: 21 ± 11% and 3 (31)%, respectively. This is explained by the strong non-normal distribution between the eight study locations examined, including a group of five locations with < 5 %DW CaCO 3 in their sediments and three locations with CaCO 3 contents between 20 and 75 %DW (Shapiro-Wilks test, p < 0.001, see Supplementary Fig. 1b). Converted into C inorg concentrations (after normalization for the sediment bulk density), we obtain median (IQR) C inorg concentrations in seagrass and mangrove sediments of 59 (66) and 1 (21) mgC inorg cm −3 , respectively (means ± SE of 63 ± 11 and 35 ± 17 mgC inorg cm −3 ) (Fig. 2a).
Using the median SARs in seagrass and mangrove ecosystems compiled in this study (0.22 and 0.23 cm yr −1 , respectively; Fig. 2b), we estimate median (IQR) C inorg burial rates in seagrass and mangrove ecosystems of 87 (154) and 6 (207) gC inorg m −2 yr −1 , respectively (means ± SE of 182 ± 94 and 90 ± 43 gC inorg m −2 yr −1 ) (Fig. 2c, Fig. 3). These values correspond to vertical accretion rates of CaCO 3 of the order of 0.1 and 0.001 cm yr −1 in seagrass and mangrove ecosystems, respectively. Our mean SAR values agree with previously reported global values 1,3 . However, our new estimates of burial rates are lower than the previous, indirect median estimate of C inorg burial rate of 108 gC inorg m −2 yr −1 (mean ± SE of 126 ± 31 gC inorg m −2 yr −1 ) reported by Mazarrasa et al. 13 .
Global annual burial rates of C inorg . Scaling up to the global seagrass coverage (150,000-600,000 km 2 ) 9 , the annual burial rate of C inorg ranged from 13 to 52 TgC inorg yr −1 for the twentieth century (Table 1). Partitioning between tropical and non-tropical seagrass meadows as in Mazarrasa et al. 13 showed that 90% of the global C inorg burial occurs in the tropics (Table 1). In seagrass meadows, our estimates of global burial of C inorg are equivalent to 31-55% of the available estimates of contemporary C org burial rates (48-112 TgC org yr −1 ) 1,7 , depending on the estimated global seagrass coverage considered. If all buried CaCO 3 is locally produced (i.e., of autochthonous origin), the burial rates of C inorg in seagrass meadows would represent emissions of 8-37 TgC yr −1 to the atmosphere and thus would offset their role as CO 2 sinks through the sequestration of C org by~17-33%.
The extent of global mangrove coverage yields a median burial rate of 0.8 TgC inorg yr −1 (mean of 13 TgC inorg yr −1 ) ( Table 1). This value should be considered as a first-order estimate because of the scarcity of data available on C inorg burial rates in mangroves and because of the non-normal distribution between CaCO 3 -rich and CaCO 3 -poor mangrove sediments (Supplementary Figure 1b). When comparing with the global C org burial rates estimate of 31 TgC org yr −1 3 , the median C inorg burial rates would correspond to a negligible reduction of net atmospheric CO 2 uptake. However, assuming that all sedimentary CaCO 3 was produced in situ, the C inorg burial rates can largely outweigh C org burial in CaCO 3 -rich mangroves. For example, the C inorg burial rate corresponds to 10-20 times the C org burial rate in the Arabian Peninsula [17][18][19] .

Discussion
CaCO 3 burial in Blue Carbon ecosystems is the balance between inputs (autochthonous and allochthonous) and losses (dissolution and export). Assessments of the mass balance of CaCO 3 in seagrass meadows are few and none have been reported, to our knowledge, in mangrove forests. For seagrass ecosystems, we assessed the balance between calcification, dissolution and burial of CaCO 3 in three locations: the Balearic Islands, Spain 20,21 , Shark Bay in Western Australia 22 and Florida Bay, USA 23,24 ( Table 2).
The most comprehensive assessment of seagrass carbon budgets is that reported for a Mediterranean Posidonia oceanica meadow at Magalluf (Mallorca Island, Spain) 20,21,25,26 . In this meadow, Barrón et al. 21 estimated a net CO 2 uptake by primary production of 8.4 gC m −2 yr −1 . This estimate was corroborated by the C org burial rate, estimated independently, at 9 ± 2 gC org m −2 yr −1 27 . Barrón et al. 21 also estimated net calcification rates of 51 gCaCO 3 m −2 yr −1 , corresponding to 6 gC inorg m −2 yr −1 . This amount of calcification would result in a CO 2 emission of 3.6 gC m −2 yr −1 (0.6-fold the net calcification 14 ). The CO 2 emission by calcification therefore represents an offset of 40% of the CO 2 uptake from net primary production (thereby yielding a total CO 2 sequestration of 4.8 gC m −2 yr −1 21 ). However, the C inorg burial rate in this meadow is estimated here at 226 gC inorg m −2 yr −1 . This is two orders of magnitude greater than the net calcification rate of 6 gC inorg m −2 yr −1 21 (Table 2). This implies that about 90% of the CaCO 3 burial in this seagrass meadow must be supported by allochthonous inputs. Therefore, calculation of the CO 2 sequestration by comparing C org and C inorg burial rates or stocks would have concluded that this meadow is a strong source of CO 2 , whereas estimates of calcification rates and net primary production concludes that it is a sink (as confirmed independently through air-sea flux measurements 26 ).
Similarly, in Shark Bay, the burial of C inorg is four times higher than the independently reported net calcification rate 22 (Table 2). This again could require large allochthonous carbonate inputs.
In Florida Bay, the low ratio between C org and C inorg concentration in the sediment (g cm −3 ) implied that seagrass meadows may be a net source of CO 2 to the atmosphere 15 . However, such assessment requires consideration of the full carbon mass balance, including accounting for allochthonous inorganic carbon inputs and the balance between calcification and dissolution in the meadows. The contemporary C inorg burial rates in Florida Bay are approximately ninefold higher than the global median, whereas median SAR is about fourfold higher than estimated globally, in an area where 80% of the sediment dry mass is composed of CaCO 3 . However, attempts to assess the gross or net calcification rates in the area yielded values one and two orders of magnitude lower than the estimated CaCO 3 burial rates ( Table 2) 23,24 . In contrast, past geological work in the Bay has suggested that it is a net producer of CaCO 3 28 . It is likely to be that some areas within this large Bay act as sources of CaCO 3 and some others as sinks, helping explain the discrepancy between reported production and burial estimates. Hence, internal redistribution of CaCO 3 production within Florida Bay needs to be considered when drawing inferences on the role of seagrass meadows from sediment composition.
These three example locations are in areas close to coral reefs and/or terrestrial lithogenic sources of CaCO 3 . We could not find estimates of calcification rates (net or gross) in areas without external sources of CaCO 3 . The discrepancies between calcification rates and burial rates in the three example locations could indicate that an important fraction of CaCO 3 burial (> 90%) is supported by CaCO 3 produced elsewhere and trapped in the seagrass sediments. This conclusion is consistent with comparable C inorg concentrations within and outside seagrass meadows, whereas, in contrast, C org concentrations are higher in seagrass sediments 13 . A large role of C inorg import is also consistent with the large CaCO 3 export from coral reefs to adjacent waters, equivalent to 25-50% of the CaCO 3 production, predominantly to reef lagoons 27 , where seagrass meadows and mangroves often grow. Mangroves, seagrass and saltmarsh ecosystems are likely to be sites of net carbonate dissolution. Roots of marine plants release organic compounds and oxygenate the sediments during the day, promoting microbial aerobic remineralization of organic matter, thereby increasing sedimentary respiratory CO 2 29,30 and/or stimulating the re-oxidation of reduced metabolites. These processes result in the release of strong acids (e.g., H 2 SO 4 , HNO 3 ) [31][32][33] , which leads to CaCO 3 dissolution in the sediment 34,35 (although re-precipitation can occur 34 ).  Dissolution of CaCO 3 might also be influenced by the CO 2 system in the water column of Blue Carbon ecosystems. Respiration and photosynthesis of the flora and fauna, together with sediment redox processes in seagrass and mangrove ecosystems, strongly influence the chemistry of the water column, generating large diel amplitudes of the saturation state for CaCO 3 (Ω) with a tendency towards dissolution or the reduction of calcification at nighttime, amplified at low tide [36][37][38][39][40] . The dissolution of allochthonous CaCO 3 in carbonate platform areas, caused directly or indirectly by metabolism of the marine vegetation and associated biota, leads to a reduction in pCO 2 through the release of fossilized total alkalinity. This sink of atmospheric CO 2 should be incorporated into the Blue Carbon framework. A recent assessment considers alkalinity addition through the dissolution of allochthonous carbonate as a very effective geoengineering approach to remove atmospheric CO 2 and mitigate climate change 41,42 .
Similarly, saltmarshes are not known to host high levels of calcifying organisms but can accumulate CaCO 3 from allochthonous sources. In arid tropical saltmarshes of the Western Arabian Gulf, dominated by succulent shrubs, a concentration of CaCO 3 of 57 ± 8% in sediments and a contemporary burial rate of C inorg of 100 ± 15 gC inorg m −2 yr −1 (mean ± SE) were found 17 . In a temperate saltmarsh of the Western Scheldt estuary in the Netherlands, a concentration of CaCO 3 of 14 ± 1% and a high contemporary burial rate of 467 ± 99 gC inorg m −2 yr −1 , mostly due to high SAR (1.1 ± 0.3 cm yr −1 ), were measured 43 . Yet, this does not imply that CaCO 3 dynamics have a negligible role in saltmarsh carbon budgets, as they may still act as sites of net dissolution of CaCO 3 , adding to the removal of CO 2 associated with C org burial. A dissolution rate of 24-96 gC inorg m −2 yr −1 was estimated in the sediment of the saltmarshes of the Eastern Scheldt estuary, corresponding to~85% of the C inorg burial rate 44 .
To further examine the conclusion that Blue Carbon ecosystems are sites of substantial allochthonous CaCO 3 deposition, based on existing mass balances for Blue Carbon sediments, we examined (qualitatively) the relationship between the CaCO 3 % DW in sediments and the presence/absence of sources of CaCO 3 adjacent to the coring locations, including coral reefs and terrestrial lithogenic sources of CaCO 3 (Fig. 4, see dataset in Supplementary Information). Seagrass and mangrove ecosystems without potentially large adjacent allochthonous CaCO 3 sources have a remarkably lower median (IQR) sediment CaCO 3 content of 4 (15) and 1 (1) %DW (means ± SE of 11 ± 4 and 1.7 ± 0.8 % DW), respectively, compared with59 (51) and 61 (27) %DW (means ± SE of 56 ± 5 and 53 ± 16 %DW) when at least one allochthonous CaCO 3 source was present (Fig. 3). For sediments in seagrass meadows, the presence of coral reefs (t-value = 4.68, df = 48.5, p < 0.0001) and lithogenic sources (t-value = 4.76, df = 57.3, p < 0.0001) increased the amount of CaCO 3 in the sediment. However, there was a significant interaction between these factors (t-value = − 3.29, df = 53.2, p = 0.0018), because the CaCO 3 % DW in the presence of both allochthonous sources was less than would be expected if these variables were additive. The presence/ absence of coral reefs and lithogenic sources accounted for 36% of the variation in CaCO 3 , whereas the random variables (study, lithology grouping and marine province) accounted for 54% of the variation in CaCO 3 (see Methods for model description). Mangrove sediment samples showed a similar pattern to the seagrass meadows and the presence of allochthonous sources had a marginally significant positive effect on the amount of CaCO 3 in the sediment (t-value = 4.29, df = 1.81, p = 0.0596). The presence/absence of a CaCO 3 source accounted for 71% of the variation of in CaCO 3 within mangrove sediments, whereas the random variables accounted for 20% of the variation in CaCO 3 . In testing for biases of outlying cores and studies, we found that one study from Western Florida had an outlying data point that disproportionality skewed the results. The study from Western Florida had relatively low CaCO 3 but did have an allochthonous source of CaCO 3 . When this study was removed from the analysis, the presence of an allochthonous source became significant (t-value = 7.92, df = 4.16, p = 0.0012). This highlights the need for more studies in mangrove sediments to determine the global influence of allochthonous sources on CaCO 3 content.
In seagrass meadows, the median (IQR) C inorg burial rate found in areas where no allochthonous sources were identified was 1 (13) gC inorg m −2 yr −1 (mean ± SE of 8 ± 4), only 1.1% of the global median. This contrast is consistent with our hypothesis that much of the C inorg buried in seagrass and mangrove sediments is allochthonous. It explains the non-normal distribution of CaCO 3 concentrations observed in sediments of seagrass and mangroves (Supplementary Figure 1), and indicates that the import of CaCO 3 from carbonate-forming ecosystems or adjacent karstic areas is the norm 27 . The global burial rate of C inorg in seagrass meadows is between a third to a half of their C org burial rate 1,7 , whereas for mangroves our first estimate of global C inorg burial rates is only 3% of the C org burial rate 3 . If the buried CaCO 3 and C org in seagrass sediments were produced entirely in situ, C inorg burial would offset up to a third of CO 2 sequestration through C org burial, particularly in tropical seagrass ecosystems where~90% of the global C inorg burial occurs. However, imbalances between production, dissolution and burial, and the observation of much higher CaCO 3 concentrations in sediments near lithogenic formations and coral reefs, suggest that, where present, allochthonous CaCO 3 inputs are substantial and support most of the net CaCO 3 burial.
Locally, despite supporting significant CaCO 3 burial, Blue Carbon ecosystems may be sites where imported CaCO 3 dissolves, strengthening rather than weakening the capacity of these ecosystems to sequester CO 2 . Whereas there is emphasis on apportioning the sources of autochthonous and allochthonous C org in Blue Carbon sediments (up to 50% of the buried C org ) 4,9 , determining the sources of CaCO 3 in Blue Carbon sediments is just as important, to resolve the role of vegetated coastal ecosystems as CO 2 sinks and, hence, their potential to support climate change mitigation. The current focus on C org budgets in vegetated coastal ecosystems needs to be augmented with integrative assessments of organic and inorganic carbon fluxes and budgets, including both allochthonous and autochthonous inputs. Moreover, these assessments must consider the sources and fate of carbon exchanged between Blue Carbon and adjacent ecosystems, as Blue Carbon ecosystems export important amounts of their organic production [45][46][47] but also import significant amounts of CaCO 3 and organic matter from adjacent sources. A comparison of paired vegetated and unvegetated sediment CaCO 3 % DW showed that vegetated and adjacent unvegetated sediments have similar carbonate concentrations, both using standard parametric statistics (general linear model (GLM), t-value = 1.32, df = 83.1, p = 0.191) and meta-analysis (z-value = 0.88, p = 0.379; Supplementary Fig. 2A,B), which also showed no evidence for reporting bias (all points within the 95% confidence lines of the funnel plot, Supplementary Fig. 2C. This provides further support to the hypothesis that much of the carbonate buried in vegetated coastal sediments derives from allochthonous sources rather than being produced within the habitat. Inorganic carbon burial in Blue Carbon ecosystems has been overlooked, with the rates compiled here representing the first direct estimates reported in the literature. These estimates confirm that seagrass ecosystems, and to lesser extent mangrove ecosystems, are intense sites of CaCO 3 burial, supporting sediment accretion. CaCO 3 burial is a fundamental process supporting the role of Blue Carbon ecosystems in climate change adaptation, which is underpinned by their capacity to rapidly accrete sediments, reducing relative SLR by raising the seafloor 1,17 .

Methods
Calculation of the C inorg accretion rate. We searched the peer-reviewed literature for data on sediment cores dated with 210 Pb, including CaCO 3 or C inorg concentration in seagrass and mangrove sediments. Search terms on Google Scholar were seagrass OR mangrove AND 210Pb OR SAR OR sediment accretion rate. We then searched returned articles that contained data on SAR and CaCO 3 or C inorg data. We found only one study presenting CaCO 3 content in a dated sediment core. However, we found 15 and 22 studies with SAR for seagrass and mangrove sediments, respectively. To obtain the CaCO 3 or C inorg concentrations needed to calculate C inorg burial rates, we used the database of Mazarrasa et al. 13 , which was the most recent exhaustive compilation of sediment cores from Blue Carbon habitats, for data on CaCO 3 in seagrass sediments. We also contacted experts in Blue Carbon studies (published studies using cores from Blue Carbon habitats) for unpublished CaCO 3 sediment concentration data (see data and references in Supplementary Data 1). In total, we compiled 42 and 53 210 Pb dated cores with CaCO 3 content in mangrove and seagrass ecosystems, respectively (see PRISMA checklist and flow diagram 48 in Supplementary Note 1).
The SARs (cm yr −1 ) from the literature were re-calculated according to the constant flux-constant sedimentation model 49 , to have a coherent and comparable dating system between all cores. The CaCO 3 concentration (% sediment DW) was calculated as the mean between all slices younger than 1900, for cores with the contemporary 210 Pb chronologies. The C inorg concentration in sediment (gC inorg m −3 ) was calculated from the dry bulk density (g m −3 ) and the percentage of CaCO 3 content (using sediment DW), considering a mass ratio of 12% carbon in CaCO 3 . The C inorg burial rate (gC inorg m −2 yr −1 ) was then calculated as the product of the SAR and the C inorg concentration for each sediment core. Cores with negligible content of CaCO 3 were also included in the calculation (see Supplementary Figure 1).
All cores from the same site or area and with similar presence or absence of allochthonous sources of CaCO 3 (see below) were treated as replicates for a global location and averaged for the analysis (geologic grouping). For seagrass, the 51 cores dated with 210 Pb were grouped into 17 locations (Figs. 2, 3). For mangroves, we compiled a total of 42 cores dated with 210 Pb in 8 locations (Figs. 2, 3). Seagrass locations ranged from tropical to sub-arctic locations, with 50% of estimates derived from tropical and subtropical locations and 50% from higher latitudes. Mangrove sediment derived mostly from subtropical locations (seven out of eight locations), particularly in Australia and the Arabian Peninsula (Supplementary Figure 2).
Determination of the influence of allochthonous sources of CaCO 3 . We analysed the influence of the presence/absence of proximity of coral reefs and continental surface lithology (qualitative data), as potential allochthonous sources of CaCO 3 in seagrass and mangrove sediments (in %DW) (see dataset in Supplementary Data 1). For seagrass, we expanded our dataset by including CaCO 3 concentrations from 264 cores compiled by Mazarrasa et al. 13 , reaching a total of 341 cores with measured CaCO 3 %DW.
We estimated the presence/absence of coral reefs using the map of the global distribution of warm-water coral reefs compiled by the UNEP-WCMC 50  burial rate or CaCO 3 %DW were normally distributed (all p < 0.05). We therefore chose to use the median (IQR) as the most appropriate description of central tendency. Traditional meta-analysis tools, which calculate effect sizes to standardize the difference between control and experimental treatments, thereby allowing comparison among disparate response variables and weighting to account for unequal variance among studies, could not be used for this analysis for multiple reasons. These reasons include that the question posed and the studies available did not include experimental designs with paired control and experimental plots required for effect size calculations, that there was a single response variable facilitating direct comparison and data integration, and, most importantly, that we used the raw data for each core. Instead, we ran a statistical test using a mixed effect GLM to determine the effect of coral reefs and lithogenic sources on the CaCO 3 % DW of the sediment. For sediments within seagrass meadows, the GLM included two fixed factors (presence/absence of coral reefs and of lithogenic sources), as well as the interaction between the two factors. For sediment within mangrove forests, the GLM included one fixed factor (presence/absence of allochthonous sources), because replication did not exist for all combinations of the two factors. The data had unequal samples among studies and studies were not evenly distributed around the globe (Fig. 1), which could result in pseudo-replication and biased results. To account for the data structure and minimize non-independence, we included three separate random variables, which included study, lithology grouping and marine province. The marine province was determined for each sample location using the marine provinces of the world as defined by Spalding et al. 53 . Separate models where run for seagrass and mangrove sites. The statistical model was produced using the lmer function within the lme4 package 54 and p-values were calculated with the lmerTest package 55 . The R 2 was calculated for the fixed and random effects using the r.squared GLMM function in the MuMIn package 56 . The response variable was log transformed, which improved the model fit compared with raw data. The model fit was assessed by plotting the Q-Q plot (linear relationship) and the fitted values compared with the residuals (random distribution). To test whether individual cores or studies were biased and having a disproportionate influence on findings, we systematically removed any studies that contained outlying samples as determined from being outside the 95% confidence interval for the fitted values vs. residuals comparison using the plot model function from sjPlot package 57 . This analysis was conducted in R version 3.4.2.
Reporting bias and its effect on findings is an important consideration for metaanalyses 58 and when the result from a meta-analysis is not the same as it would have been if data from all correctly conducted studies were included in the analysis 59 . A main cause of reporting bias is not publishing research because of a lack of merit as determined by the researcher, reviewer or editor 60 . As indicated by the data inclusion flow diagram ( Supplementary Fig. 3), researchers often measured but did not publish data on soil CaCO 3 content and authors needed to be directly contacted for these results. In addition, the researchers not only provided information from published studies but also unpublished data on CaCO 3 content (10 of 51 seagrass studies included in the analysis were not published). For these reasons, it is unlikely to be that our findings were affected by reporting bias. A subset of data collected for this study included the appropriate information to run both a GLM and a traditional meta-analysis (effect size could be calculated between paired data). The data included information from nine studies that measured the CaCO 3 content of sediment from both vegetated and unvegetated habitats. There were 92 core samples with 32 from unvegetated and 60 from vegetated habitats (Supplementary Fig. 2A). The GLM followed the same procedures as detailed in the main text, except it had only two random factors, study and marine province, because study and lithology grouping differed in only one instance. For the metaanalysis, the data were paired for each study and the mean CaCO 3 %DW, number of samples and SD were calculated for vegetated and unvegetated cores for each study. Two studies included in the GLM were removed for the meta-analysis, because they only had one core for an unvegetated habitat and SD could not be calculated, leaving seven comparisons for this analysis. Hedges' g was calculated for the effect size following Borenstein et al. 60 (Eqs. 1-3) and a variance for each effect size was also calculated (V g ) 59 (Eq. 4), as indicated by equations: X E and X C are the mean (n is sample size) of vegetated and unvegetated sediments, along with SD pooled and J, which accounts for biases associated with different sample sizes. The meta-analysis included the same two random variables as the GLM and was conducted using the rma.mv function from the metafor package 61 .
Calculation of global yearly burial rates of C inorg . The global annual burial of inorganic carbon (TgC inorg yr −1 ) in seagrass meadows was calculated as the product of the global median C inorg burial rates and the estimated global seagrass area, which ranges from 150,000 to 600,000 km 29 . We also calculated the global annual burial of C inorg as the sum of separate calculations for tropical and arid climates and meadows at higher latitude climates. Median C inorg burial rates were calculated for tropical (core locations with tropical and hot desert climates) and non-tropical areas (temperate, continental and polar climates) and multiplied by the global seagrass cover range under the assumption that 2/3 of the seagrass area is in the tropical and subtropical zone 13 . The global annual burial of inorganic carbon (TgC inorg yr −1 ) in mangroves was calculated as the product of the global median C inorg burial rates and the estimated global mangrove cover of 137,760 km 262 .

Data availability
The dataset is available as Supplementary Data 1.