Introduction

Mangroves, saltmarshes, and seagrasses, the most “actionable” blue carbon ecosystems (BCEs) towards climate mitigation, occupy less than 0.2% of the ocean’s surface but store over 30 Pg of organic carbon (Corg) in the top meter of soil globally1,2. However, when these ecosystems are converted or degraded, an estimated 0.15–1.02 Pg of CO2 is released into the atmosphere annually3, accounting for 3–23% of annual emissions from global land use change over the past decade4. This release occurs primarily through the decomposition of soil Corg stock2,5,6. Although rates of BCE area decline are slowing down or even reversing in some regions7,8,9, the global annual loss rate remains high at 0.13% for mangroves10, 0.28% for saltmarshes11, and 1–2% for seagrasses12. As a result, conserving BCEs to avoid greenhouse gas (GHG) emissions has gained global recognition as a cost-effective climate action, supported by major international climate and biodiversity policies such as the Paris Agreement and the Kunming-Montreal Global Biodiversity Framework13,14.

Financial resources for the conservation of BCEs, aimed at preventing soil Corg stock loss and the resulting GHG emissions, can be effectively mobilized through compliance and voluntary carbon markets that recognize and assign carbon credits to blue carbon projects14,15,16. These markets require rigorous accounting and verification of GHG emission reductions to demonstrate “additionality”14,15,16. However, within the science and policy communities, calculations of GHG emission reduction from blue carbon projects often assume that the conversion and degradation of BCEs result in uniform Corg stock loss across the entire top meter of soil (e.g., refs. 3,11,17; Supplementary Table 1). This assumption, exemplified by the Intergovernmental Panel on Climate Change (IPCC) Tier 1 standard18, has been perpetuated across studies estimating of global, national, and regional GHG inventories without thorough examination or verification19, despite IPCC guidelines advocating for modifying the depth of top meter at higher tiers18.

Estimates of GHG emissions from soil Corg stock loss following BCEs disturbance have been assumed to range from 22.7% to 100% of the Corg stock present in the top meter of soil3,6,11,17,20,21 (Supplementary Table 1). However, growing empirical evidence suggests that the actual depth and amount of Corg loss can vary substantially across ecosystems and disturbance regimes22,23, indicating substantial uncertainties in GHG emission estimates that rely on untested assumptions. Such problematic GHG emission accounting methodologies risk inflating credit issuances, thereby undermining the true effectiveness of nature-based solutions within nationally determined contributions (NDCs). Indeed, the issuance of carbon credits for emission reductions through terrestrial forest conservation has faced scrutiny due to failure in demonstrating genuine GHG reductions, contributing to skepticism about the effectiveness of voluntary carbon markets24,25. Understanding how different ecosystems and disturbance regimes affect soil Corg stock loss is therefore critical, especially when estimating GHG emission reductions in carbon market transactions.

The various anthropogenic and climate change disturbances affecting BCEs can be broadly classified into two main categories: (1) those that physically remove or alter both the living biomass and the soil, such as conversion of BCEs for aquaculture or agriculture and dredging5,26,27; and (2) those that affect BCE plants without directly disturbing the soils, such as harvesting, degradation, or grazing5,28,29. For over a decade, it has been hypothesized that the trajectories of soil Corg stock loss in BCEs may vary substantially depending on the types of disturbance5,30 (Fig. 1). Disturbances that physically disrupt the soil are expected to cause greater Corg stock loss, penetrating deeper soil layers, while disturbances with minimal soil impact are likely to result in lower Corg stock loss, primarily from the shallower soil layers5,30. Soil Corg stock loss is also anticipated to last longer for disturbances that penetrate deep soil layers relative to those that minimally affect the soil post-disturbance5. A comprehensive examination of these hypotheses is essential to understand how soil Corg stock loss varies among BCEs subject to different disturbance regimes, which is crucial for accurately estimating GHG emission reductions in blue carbon projects.

Fig. 1: Hypothesized impacts of disturbance regimes on soil organic carbon (Corg) stocks in blue carbon ecosystems (BCEs).
Fig. 1: Hypothesized impacts of disturbance regimes on soil organic carbon (Corg) stocks in blue carbon ecosystems (BCEs).
Full size image

The major disturbance drivers investigated for mangroves include agricultural reclamation for rice fields36,68,69,70,71, pastures37,72, coconut40,73,74, rubber40, or crop plantations75,76; aquaculture development for shrimp34,40,41,72,76,77,78,79,80,81,82,83,84,85,86,87,88, or fish22,88,89, along with salt ponds construction40 and sand mining90; the harvesting22,28,91,92,93,94, clearance40,95,96,97,98,99,100,101,102, or grazing of trees103; and climate impacts such as typhoons104,105,106,107,108, saltwater intrusion109, drought110,111,112,113, or strong El Niño-Southern Oscillation114, as well as hydrological changes like settlement construction115,116,117,118,119,120,121,122,123,124,125,126,127,128,129, sediment runoff130,131, or nutrient effluents from nearby aquaculture ponds87,132,133,134,135,136,137. For saltmarshes, the disturbance drivers include agricultural reclamation for rice48,49,138,139,140,141 or crop fields49,139,142,143,144,145,146,147; aquaculture development for shrimp41,88,148,149 or fish ponds41,49,147; climate impacts such as typhoons150, as well as hydrological changes like embankment construction120,148,151,152,153,154,155,156,157,158, or excessive flooding159,160; and the grazing29,161,162,163,164,165,166,167, coppicing168, shading169, or damage of plants170. In this study, aquaculture was combined with agriculture due to the limited and geographically biased data availability for aquaculture, with no significant difference observed between them (P = 0.217; see section “Methods”). For seagrasses, the disturbance drivers include unintentional dredging due to mooring51, boat grounding26,52, or clam harvesting171,172,173, and intentional dredging for sand174; and vegetation cover damage due to shading175,176,177, removal178,179, overgrazing180, loss23,181,182,183,184,185,186,187, or nutrient enrichment116,188,189,190,191,192,193. Created in BioRender. Fu, C. (2025) https://BioRender.com/q55g628.

Here, we address these knowledge gaps by examining the loss of soil Corg stocks across BCEs subject to different disturbance regimes. Our dataset includes paired reports of soil Corg stock from both disturbed and intact BCEs from 118 mangrove sites, 82 saltmarsh sites, and 39 seagrass sites worldwide (Supplementary Fig. 1). We categorized disturbances into those causing deep soil disruption—such as the conversion of mangroves and saltmarshes to agriculture or aquaculture and dredging of seagrass meadows—and those causing minimal soil disruption, including mangrove harvesting, saltmarsh grazing, climate/hydrological change in mangroves and saltmarshes, and vegetation cover damage in seagrass meadows (Fig. 1). We first analyzed how soil Corg stock loss varied across BCEs and with depth following disturbance to resolve the distribution of Corg stock loss both across ecosystems and along the soil depth profile under various disturbance regimes. We then examined the relationship between soil Corg stock loss and initial Corg stock, as well as its relationship with the duration of disturbance across different disturbance regimes. Our results provide empirical evidence that both soil depth and the extent of Corg stock loss differ among BCEs and are controlled by the type of disturbance regimes. The estimates of Corg stock loss reported here address a knowledge gap and enable more accurate and robust estimates of GHG reduction in blue carbon projects as a nature-based climate solution.

Results

Top-meter soil Corg stock loss under different disturbance regimes

When extrapolating soil Corg stock to the top one meter globally, we found that disturbances in BCEs resulted in significantly higher losses of Corg stock in mangroves (−54.4 ± 8.7%, mean ± 95% confidence interval [CI]) and seagrasses (−36.4 ± 27.6%) relative to saltmarshes (−4.4 ± 36.3%, P < 0.001), though substantial variability was observed across all ecosystems. Soil Corg stock losses in mangroves were more pronounced and less variable in areas reported as mangrove loss hotspots, including Southeast Asia, and Central and South America10 (Fig. 2). In saltmarshes, soil Corg stock losses were similarly less variable and more prominent in China relative to Europe and USA, despite historical declines in saltmarsh area across these regions11,31. However, no clear geographic pattern of Corg stock loss was observed in seagrasses under disturbance. We could not detect an influence of plant species on soil Corg stock loss under disturbance across BCEs (P > 0.05; Supplementary Table 2), except in mangroves, where species interacted with disturbance regimes (disturbance × species, P = 0.003).

Fig. 2: Spatial distribution of reported and extrapolated top-meter soil organic carbon (Corg) stock changes in blue carbon ecosystems (BCEs) under different disturbance regimes.
Fig. 2: Spatial distribution of reported and extrapolated top-meter soil organic carbon (Corg) stock changes in blue carbon ecosystems (BCEs) under different disturbance regimes.
Full size image

a The longitudinal distribution of soil Corg stock change (%) in BCEs; b global distribution of soil Corg stock change (%) in BCEs, c the latitudinal distribution of soil Corg stock change (%) in BCEs. Shapefile of the world obtained from www.naturalearthdata.com. Source data are provided as a Source Data file.

Disturbance regimes exerted significant control on soil Corg stock loss in mangroves (P < 0.001), but their effects were less pronounced in saltmarshes and seagrasses (P > 0.05). Major soil Corg loss in mangroves were caused by extensive soil disruption, in contrast to the variability in soil Corg stock loss under limited soil disturbance. Specifically, the conversion of mangroves to agriculture and aquaculture led to significant reductions in top one-meter soil Corg stocks, averaging −68.4 ± 13.4% and −65.4 ± 10.6%, respectively (Fig. 3). It should be noted, however, that the estimated Corg stock loss from converting mangroves to aquaculture might be conservative if soil excavation was involved during the initial “construction” phase18. In contrast, climate/hydrological change (−32.1 ± 23.0%) and harvesting activities (+0.8 ± 46.2%) in mangroves, which cause limited soil disturbance, result in less or nonsignificant Corg stock changes. In saltmarshes, no significant changes in soil Corg stock were detected in the top meter, regardless of the disturbance regime (agriculture, −1.1 ± 36.1%; climate/hydrological change, −25.9 ± 30.7%; grazing, +48.6 ± 78.7%). Similarly, in seagrass ecosystems, soil Corg stock losses driven by dredging (−27.4 ± 33.6%) were comparable to those from damage to the vegetation cover of seagrass meadows (−34.2 ± 22.4%).

Fig. 3: Responses of top-meter soil organic carbon (Corg) stocks in blue carbon ecosystems (BCEs) to different disturbance regimes.
Fig. 3: Responses of top-meter soil organic carbon (Corg) stocks in blue carbon ecosystems (BCEs) to different disturbance regimes.
Full size image

Bars around the means denote 95% CIs. The first and second numbers in parentheses indicate, respectively, the number of observations and the number of studies included in each calculation. Filled circles indicate significant effects of soil Corg stock loss, while circles indicate nonsignificant effects (95% CIs overlapping zero). Mangrove soil Corg stock changes resulted from agricultural and aquaculture conversions, climate/hydrological change, and harvesting; saltmarsh soil Corg stock changes resulting from agricultural conversions, climate/hydrological change, and grazing; seagrass soil Corg stock changes resulting from dredging and vegetation cover damage. Source data are provided as a Source Data file.

Depth patterns of soil Corg stock loss under different disturbance regimes

The depth patterns of soil Corg stock loss varied among different disturbance regimes. Extensive soil disruptions, such as the conversion of mangroves to agriculture or aquaculture and the dredging of seagrass meadows, led to significant Corg loss reaching soil depths of 100–200 cm and at least 50 cm, respectively (Fig. 4). Specifically, for mangroves converted to aquaculture, soil was excavated to a mean depth of 66 ± 24 cm (± 95% CI, Supplementary Fig. 2), potentially extending soil Corg stock loss to approximately 166 cm. In contrast, saltmarsh conversion to agriculture showed no significant soil Corg stock loss across depth increments (Fig. 4b). For disturbance regimes with limited soil impact, soil Corg stock loss was either negligible across the depth profiles, as observed in the harvesting of mangroves, or largely confined to top 10–30 cm layers (climate/hydrological change in mangroves and saltmarshes, grazing of saltmarshes, and vegetation cover damage in seagrass meadows). However, deeper soil data (>30 cm) for saltmarshes and seagrasses impacted by grazing and vegetation cover damage, respectively, were underrepresented in the literature, underscoring the need for further research. These divergent depth patterns of soil Corg stock loss nevertheless suggest that the current assumption of uniform Corg loss across the top meter of soil requires reconsideration.

Fig. 4: Responses of soil organic carbon (Corg) stocks to different disturbance regimes across various depth intervals in blue carbon ecosystems (BCEs).
Fig. 4: Responses of soil organic carbon (Corg) stocks to different disturbance regimes across various depth intervals in blue carbon ecosystems (BCEs).
Full size image

a Mangrove soil Corg stock changes due to agricultural and aquaculture conversions, climate/hydrological change, and harvesting, b Saltmarsh soil Corg stock changes due to agricultural conversions, climate/hydrological change, and grazing, c Seagrass soil Corg stock changes due to dredging and vegetation cover damage. Bars around the means denote 95% CIs. The bars with arrows denote 95% CIs that extend beyond the X-axis range. The numbers alongside the arrows indicate the maximum values of the 95% CIs. The first and second numbers in parentheses indicate, respectively, the number of observations and the number of studies included in each calculation. Filled circles indicate significant effects of soil Corg stock loss, while circles indicate nonsignificant effects (95% CIs overlapping zero). Source data are provided as a Source Data file.

Relationship between soil Corg stock loss and initial Corg stock

Soil Corg stock losses resulting from the conversion of mangroves to agriculture and aquaculture were strongly dependent on the initial Corg stock (P < 0.001; Fig. 5 and Supplementary Fig. 3), indicating that mangroves with higher initial Corg stocks are more susceptible to greater Corg losses following disturbances. This dependence also held for the conversion of saltmarshes to agriculture. While converting saltmarsh to agriculture did not show significant soil Corg stock loss in the top meter overall, it did result in significant loss—reaching up to −60%—in saltmarshes with higher initial soil Corg stock levels (above ~75 Mg C ha−1, P < 0.001). In contrast, the conversion of saltmarshes to agriculture with lower initial Corg stocks could enhance, rather than reduce, Corg stock levels. Grazing of saltmarshes showed a notable gain of soil Corg stock, which positively increased with higher initial Corg stocks (P = 0.028). In seagrass meadows, both gains and losses of soil Corg stock were observed following dredging (P = 0.143) or vegetation cover damage (P = 0.109), independent of initial Corg stock. When compared to the GHG emission factor commonly used in previous studies (taking 25–75% of soil Corg stock loss as a conservative range; Supplementary Table 1), we found that as initial Corg stocks increased, losses driven by the conversion of mangroves and saltmarshes to agriculture and aquaculture were more aligned with or exceeded this range. However, as initial Corg stock increases, soil Corg stock losses resulting from seagrass disturbances were not well aligned with the range of the GHG emission factor.

Fig. 5: Top-meter soil organic carbon (Corg) stock change in response to initial Corg stock in blue carbon ecosystems (BCEs) and disturbance duration.
Fig. 5: Top-meter soil organic carbon (Corg) stock change in response to initial Corg stock in blue carbon ecosystems (BCEs) and disturbance duration.
Full size image

a−c Top-meter soil Corg stock change (Mg C ha−1) in response to initial Corg stock, d–f Top-meter soil Corg stock change (%) in response to disturbance duration. Solid lines represent linear or logarithmic regression, with shading representing the 95% CI of regression fits. The carbon emission factor of 25–75% was taken as a conservative range commonly used in previous studies (Supplementary Table 1). Top-meter soil Corg stock change (%) in response to initial Corg stock are shown in (Supplementary Fig. 3). Source data are provided as a Source Data file.

Relationship between soil Corg stock loss and disturbance duration

Soil Corg stock losses generally increase with disturbance duration for activities that cause extensive soil disruptions, including the conversion of mangroves and saltmarshes to agriculture and aquaculture (P < 0.05), and the dredging of seagrass meadows (P = 0.035, Fig. 5). Specifically, losses in soil Corg stock after converting mangroves to agriculture and aquaculture asymptotically reached −68.2 ± 13.4% and -86.3 ± 14.0%, respectively, 30 years post-conversion. Reclamation of saltmarshes to agriculture resulted in an average soil Corg stock reduction of −49.9 ± 9.2% approximately 60 years post-conversion, despite showing both positive (net gain) and negative (net loss) changes at shorter intervals. Available reports indicate that dredging of seagrass meadows leads to a soil Corg stock loss of −77.1 ± 8.4% after 80 years since the disturbance. Overall, these patterns are consistent with estimates based on previous modeling, which calculated that cumulative GHG emissions from disturbed BCEs account for 70–80% of the initial top-meter soil Corg stocks over 40 years32. In contrast, for disturbances with minimal soil impact, the response of soil Corg stock loss to disturbance duration lacks a clear pattern.

Discussion

Within the latest NDCs under the Paris Agreement, 41 countries have prioritized ocean or BCEs as key sector for reducing GHG emissions33. However, the lack of empirical support for the long-held assumption that disturbance of BCEs leads to uniform loss of Corg stock across the top meter of the soil—ranging from 22.7% to 100%—poses a challenge in effectively informing and operationalizing NDCs34,35. Our synthesis and quantitative analysis of available evidence demonstrates that the widely used assumption that the top meter of soils in BCEs is uniformly susceptible to loss, regardless of disturbance regimes (e.g., ref. 3), should be rejected. The magnitude and dynamics of soil Corg loss in disturbed BCEs are highly dependent on ecosystem type, disturbance regimes, baseline soil Corg stock, and the time since disturbance. Furthermore, our estimations provide GHG emission factors for anthropogenic and natural disturbance drivers that were not previously considered by the IPCC (Table 1), such as climate/hydrological change in mangroves and saltmarshes, which increasingly lead to BCE losses10,11,12,13,14. This work marks a critical advancement in refining global and national inventories of GHG emissions from BCEs using the IPCC Tier 1 standards, providing robust support for GHG emissions accounting and abatement policy settings for future blue carbon projects and NDCs commitments.

Table 1 Changes in soil organic carbon (Corg) stock of blue carbon ecosystems (BCEs) due to various disturbances and their comparison with Intergovernmental Panel on Climate Change (IPCC) guidelines
Full size table

Among the various disturbance drivers examined here, none strictly conformed to the assumption of significant soil Corg stock loss to the depth of one meter. For example, the conversion of mangroves to agriculture resulted in soil Corg loss up to 200 cm. This disturbance regime involves substantial hydrological manipulation and soil remobilization through drainage and dredging, which reverse hypoxic soil conditions and increase the exposure of Corg to aerobic decomposition, thereby extending Corg stock losses to deeper layers5,36,37,38. For instance, the drainage of mangrove forests in East Trinity Inlet, Australia, for agricultural production between 1976 and 2000 resulted in a soil elevation loss of 1.3 m—from 0.9 m above sea level to 0.4 meters below—driven by substantial soil Corg remineralization5. Thus, considering soil Corg stock loss only from the top meter may underestimate GHG emissions due to such disturbance. Furthermore, our results indicate that soil Corg stock loss from agricultural conversion persists over decades, albeit at a decreasing rate. Therefore, removing seawalls or dikes to reintroduce tidal flow—thereby inhibiting aerobic Corg decomposition—is key to prevent further Corg stock loss and enable cost-effective mangrove recovery14,39.

In contrast to agricultural conversions, converting mangroves to aquaculture typically involves more intensive soil excavation during pond construction, reaching down to depths of 50–250 cm, which substantially enhances Corg destabilization and loss under aerobic conditions5,38,40,41. We estimated that aquaculture conversions roughly excavated soil to depths, averaging ( ± 95% CI) 66 ± 24 cm, with additional soil Corg stock loss extending to 100 cm in post-excavated soil, culminating in soil Corg stock loss down to approximately 166 cm. This indicates that considering only Corg stock loss in the top meter of soils during the initial “construction”, without accounting for losses during the “use” and “discontinued” phases18, may substantially underestimate GHG emissions. Indeed, our results suggest that soil Corg stock during these subsequent phases may be completely depleted over 30 years. Overlooking soil Corg stock loss in the “use” and “discontinued” phases could result in the allocation of insufficient carbon credits to the project, thereby limiting the scale and ambition of blue carbon projects, as well as diminishing interest in funding these initiatives. Furthermore, our findings indicate that three out of 16 converted aquaculture ponds in our dataset of elevation changes exhibited gains in soil mass (Supplementary Fig. 2). This counterintuitive result may be attributed to the collapse of soil structures due to belowground Corg decomposition, combined with bottom leveling during pond construction and soil erosion from berms, or regional variance in soil depth across mangroves and aquaculture ponds42. Therefore, clarifying soil depth dynamics is essential to robustly estimate GHG emissions following mangrove conversion to aquaculture43.

Our findings further demonstrate that soil Corg stock losses due to agriculture and aquaculture conversions in mangroves are positively correlated with initial Corg stock. This likely reflects the role of hydrogeomorphic processes in shaping Corg levels in soils of BCEs, along with the mineral-dominant nature of soils with low Corg stock, which attenuates Corg remineralization after disturbance through mineral protection44. In contrast, soils with higher Corg stocks contain greater proportions of particulate organic matter with weaker mineral associations, making them more susceptible to Corg remineralization and increased GHG emissions following disturbance44,45. As such, conserving mangroves with high soil Corg stock against conversion to agriculture and aquaculture yields the greatest reduction in Corg loss and, therefore, potential GHG emissions. The variance in initial soil Corg stock is influenced by environmental factors, particularly hydrogeomorphic settings46. Estuarine interior mangroves generally have higher soil Corg stock than open-coast or fringe mangroves22 but are highly vulnerable to conversion for agriculture and aquaculture due to their fertile soils and flat landscapes, resulting in deep soil Corg stock loss47. Therefore, incorporating geomorphic context, either alongside or as a proxy for initial soil Corg stock can contribute to the development of robust conservation projects, enabling better prioritization of the most vulnerable areas and maximizing GHG emission reductions.

The context-dependent effect of initial Corg stock on soil Corg stock loss is particularly pronounced in conversion of saltmarshes to agriculture. The assumption of significant soil Corg stock loss in the top one meter was only met in saltmarshes with initially higher soil Corg stocks. In contrast, reclamation of saltmarshes with lower initial soil Corg stock to agriculture leads to the net gain in soil Corg stock, likely due to the application of organic or chemical fertilizers that promote the accumulation of plant-derived organic material in soils48,49. This tendency is particularly conspicuous in China, where the newly formed saltmarshes, created after long-term sequential reclamation, have low soil Corg stock31,50. Consequently, the reclamation of these saltmarshes has enhanced their soil Corg stock after converting them to paddy or crop lands31,49. Similarly, aquaculture production in converted saltmarshes with low initial soil Corg stock may introduce substantial organic inputs, such as excessive fodder and feces from cultured fish or shrimp, contributing to soil Corg accumulation. Therefore, conservation projects for saltmarshes that do not account for background soil Corg stocks may not achieve the anticipated avoidance of GHG emissions. Our results further show that soil Corg stock losses following conversion of saltmarshes to agriculture are sustained over centuries. Hence, rapid deployment of restoration efforts where conversion of saltmarshes with high soil Corg stock to agriculture already occurred is critical for maximizing GHG emission reduction, while restoring Corg sequestration.

Soil Corg stock loss due to dredging of seagrass meadows can penetrate to depths of at least 50 cm, aligning with observations that dredging—whether unintentional or intentional—disturbs soil to depths ranging from 10 to 160 cm26,51,52. However, significant Corg stock loss was not observed when extrapolating to the top meter of soil, owing to the variable response of soil Corg stock following dredging. This variability may be attributed to constant seawater inundation of disturbed seagrass soil, which limits oxygen penetration and Corg decomposition1,5,32. Nutrient release from soil disturbance can also stimulate phytoplankton and bacterial blooms53,54, which may either lead to the accumulation of labile Corg or enhanced Corg decomposition through the priming effect55,56. Further research is needed to delineate the impact of soil Corg stock loss due to dredging, with a particular focus on sampling deeper soil layers in seagrass meadows.

In contrast to disturbance regimes that cause extensive soil disruption, our results show that climate/hydrological changes in mangroves and saltmarshes, harvesting of mangroves, grazing of saltmarshes, and vegetation cover damage in seagrass meadows result in Corg stock loss primarily within the top 10–30 cm, or lead to minimal Corg stock losses or even net gains due to limited soil disruption. Climate/hydrological changes in mangroves and saltmarshes and vegetation cover damage in seagrass meadows lead to the cessation of new organic material accumulation, while the root decay may trigger a priming effect, enhancing the decomposition of surface soil Corg and/or soil erosion55,57. Harvesting of mangroves and grazing on saltmarshes substantially reduce aboveground biomass and disturb the surface soil via trampling28,29. However, surface soil Corg stock losses may be partially offset by increased belowground biomass production29 and ecosystem recovery post-disturbance, which re-promotes Corg accumulation22,28. Trampling may also compact surface soil, enhancing the development of anaerobic conditions that weaken Corg mineralization58. Consequently, conservation efforts targeting these disturbances may not achieve the anticipated avoidance of GHG emissions. These findings underscore that overestimated emission factors, based on unverified assumptions perpetuated across studies3,11,17,20,21, may have led to the issuance of excessive carbon credits to blue carbon projects achieving additionality through avoidance of these impacts. Failure to achieve the committed GHG emission avoidance undermines the credibility of nature-based solutions, including blue carbon projects, despite their important role in contributing to climate change mitigation efforts.

While the avoided GHG emissions from conservation projects focusing on BCEs affected by disturbances causing limited soil disruption are modest, these efforts can still be motivated by the broader benefits of the ecosystem services they provide, such as biodiversity enhancement and coastal protection14,15,16. Management regimes that include sustainably managed harvesting practices, such as logging of mangroves and grazing of saltmarshes, can minimize GHG emissions while supporting provision services that benefit communities14,29. Incentivizing such sustainable management practices can achieve modest GHG avoidance while supporting co-benefits, representing a new dimension of blue carbon projects beyond those focused on conservation and/or restoration that deserves consideration.

While our global synthesis provides a compelling analysis of soil Corg dynamics across different disturbance regimes, it is restricted by data scarcity for some regions, such as South America and Africa, particularly for saltmarshes and seagrasses. To further reduce uncertainty, there is a need to expand empirical evidence on Corg dynamics over long time scales and across a broader range of disturbance regimes. Nevertheless, the disturbance regimes explored in this study can serve as approximations for unstudied disturbance drivers, based on their potential effects on ecosystems and their convergence with studied drivers. For example, disturbances leading to plant die-off and loss due to insects or diseases59,60,61 could be approximated using climate/hydrological change category for mangroves and saltmarshes, as well as the vegetation cover damage category for seagrasses, as described in this study. Notably, due to the lack of paired data reporting Corg dynamics, we could not explore coastal erosion as a disturbance driver increasingly causing BCE loss, such as reported in Bangladesh, Brazil, USA and Russia10,11. Moreover, losses of soil Corg are not necessarily equivalent to GHG emissions but provide a useful upper ceiling to resulting GHG emissions. Available evidence suggests that some of the eroded Corg from saltmarshes and terrestrial sources delivered to estuaries may be reburied elsewhere in the coastal ocean62,63. Therefore, further assessment of the fate of eroded Corg will be key to constrain the GHG emission driven by direct or indirect erosion.

The results presented provide robust, empirically validated estimates across BCEs, allowing for the development of improved emission factors and the assessment of potential GHG emission reductions from actions that protect BCEs at risk from various disturbances. Major soil Corg stock losses due to the conversion of mangroves and saltmarshes to agriculture or aquaculture, and the dredging of seagrass meadows, highlight the urgent need for preventing such disturbances, particularly in regions with higher Corg stocks, which can be achieved through including habitats highly vulnerable to such disturbances into marine protected areas14. Given the enduring nature of Corg loss after disturbances, restoration efforts underpinned by major global instruments should prioritize projects in converted areas, improving land tenure and providing economic alternatives for affected communities to prevent further GHG emissions and restore Corg sequestration14,16,64. Climate/hydrological change results in relatively lower soil Corg stock loss due to lesser soil disturbance, imparting optimism that restoring degraded BCEs can recover Corg stock while preventing further losses. Furthermore, sustainable, non-destructive use of BCEs, such as selective harvesting of mangroves, can sustain human benefits without impairing soil Corg storage. These findings contribute to strengthening blue carbon projects as a robust and effective nature-based solutions, supporting the inclusion of BCEs as an important asset in the NDCs’ more ambitious commitments to address the growing magnitude and pace of climate change.

Methods

Literature search and screening

We systematically searched all peer-reviewed literature using the Web of Science (http://apps.webofknowledge.com) with the keywords listed in Supplementary Table 3 published before 20 November 2023, following the PRISMA protocol65, to systemically investigate the effects of disturbance on soil Corg stock loss in BCEs. A follow-up search was conducted on 25 December 2024 using the same keywords to include recent studies published after 20 November 2023. To prevent bias in selection from the 4886 potentially relevant articles obtained from the two systematical searches, we first removed a significant number of articles through title screening, leaving 1943 articles for further inspection (Supplementary Fig. 4). The title, abstract, and full text were screened based on the following criteria: first, they must report the anthropogenic or natural driver of soil Corg stock loss in BCEs, such as, agricultural reclamation, aquaculture development, dredging, clearance, and climate/hydrological change; second, they must contain reference sites for the disturbed sites, such as an undisturbed, natural, or controlled plots without apparent human and/or climate change impacts, sharing the same parent material and environmental conditions, including at least one undisturbed site if multiple disturbed sites were present; third, they must report at least the soil Corg (or soil organic matter) content for the paired sites; and fourth, they must report the sampled soil depth to determine soil Corg content. If an article included multiple sites or different habitat types, each was treated as a separate entry in the database. We excluded studies that did not specify a clear disturbance driver, reported multiple disturbance drivers affecting the same site, or investigated soil Corg stock losses based on area changes of BCEs, modeling with scaled‐up data (i.e., GIS and remote sensing modelling), or laboratory manipulation experiments. We also excluded studies that investigated soil Corg stock changes due to ecological invasion (e.g., the Spartina alterniflora invasion in Chinese BCEs66), encroachment between BCEs (e.g., mangroves encroached into saltmarshes near the poleward limits67), given that the loss of the original habitats is compensated by the gain of another BCE.

All data for the analysis were derived from paired intact and disturbed BCEs, from 140 research articles encompassing 118 mangrove sites, 82 saltmarsh sites, and 39 seagrass sites worldwide (Supplementary Fig. 1 and Supplementary Fig. 4). The major disturbance drivers investigated in these studies were categorized based on two key principles: (1) the purpose of the disturbance to BCEs and the extent of soil disruption, and (2) the availability of sufficient reports to perform statistical analysis. When the number of studies was insufficient, disturbance drivers were combined according to the first principle. For mangroves, the disturbance drivers include: (1) agricultural reclamation for rice fields36,68,69,70,71, pastures37,72, coconut40,73,74, rubber40, or crop plantations75,76; (2) aquaculture development for shrimp34,40,41,72,76,77,78,79,80,81,82,83,84,85,86,87,88, or fish22,88,89, along with salt pond construction40 and sand mining90; (3) harvesting22,28,91,92,93,94, clearance40,95,96,97,98,99,100,101,102, or grazing of trees103; and (4) climate impacts such as typhoons104,105,106,107,108, saltwater intrusion109, drought110,111,112,113, or strong El Niño-Southern Oscillation114, as well as anthropogenic activities like settlement construction115,116,117,118,119,120,121,122,123,124,125,126,127,128,129, sediment runoff130,131, or nutrient effluents from aquaculture ponds87,132,133,134,135,136,137. For saltmarshes, the disturbance drivers included: (1) agricultural reclamation for rice48,49,138,139,140,141 or crop fields49,139,142,143,144,145,146,147; (2) aquaculture development for shrimp41,88,148,149 or fish ponds41,49,147; (3) climate impacts such as typhoons150, as well as hydrological changes like embankment construction120,148,151,152,153,154,155,156,157,158 or excessive flooding159,160; and (4) the grazing29,161,162,163,164,165,166,167, coppicing168, shading169, or damage of plants170. We combined (1) and (2) due to only one study reporting soil Corg dynamics to one meter soil depth due to fish aquaculture conversion148. Furthermore, all reports on converting saltmarshes to aquaculture come exclusively from China41,49,88,147,148,149, and no significant difference was observed between (1) and (2) (P = 0.217). We have categorized the combined disturbance driver as agriculture as the data is dominating by agricultural conversions. For seagrasses, the disturbance drivers included: (1) unintentional dredging due to mooring51, boat grounding26,52, or clam harvesting171,172,173, and intentional dredging for sand174; and (2) vegetation cover damage due to shading175,176,177, removal178,179, overgrazing180, loss23,181,182,183,184,185,186,187, or nutrient enrichment116,188,189,190,191,192,193.

These disturbance drivers cover all common disturbances occurring in BCEs10,12,14,194 and were categorized based on their potential impact on soils, following the framework proposed by Lovelock et al.5. Consequently, the drivers for mangrove disturbance were divided into four categories: agriculture, aquaculture (extensive soil disturbance), climate/hydrological change, and harvesting (limited soil disturbance). For saltmarsh disturbance, the drivers were divided into three categories: agriculture (extensive soil disturbance), climate/hydrological change, and grazing (limited soil disturbance). For seagrass disturbance, the drivers were divided into two categories: dredging (extensive soil disturbance), and vegetation cover damage (limited soil disturbance). All the studies were conducted in paired sites using the “space for time” approach27. Given that soil Corg may take several years or decades to reach a new equilibrium32,195, there were limited studies with time series data extending back to pre-disturbance conditions. Consequently, for each paired site, it was assumed that soil conditions were similar before the disturbance. For chronosequences, data from all time points were utilized in this study.

Data extraction and standardization

Data on latitude and longitude, mean annual temperature (MAT), mean annual precipitation (MAP), sampling depth (and intervals), sampling number, soil Corg (or soil organic matter) content, dry bulk density (DBD), and duration of disturbance were extracted from the eligible literature. For those data not provided numerically but graphed, we determined values from figures with Web Plot Digitizer (https://automeris.io/). For studies reported soil organic matter, we used the conversion factor reported for mangroves (soil Corg = 0.21 × soil organic matter ∨ 1.12, ref. 196), saltmarshes (soil Corg = 0.52 × soil organic matter −0.21, ref. 196), and seagrasses (soil Corg = 0.40 × soil organic matter −1.17, ref. 197) to recalculate soil Corg content to facilitate the calculation of soil Corg stock. For studies providing only age intervals (e.g., 10–25 years or >66 years), we averaged the range (e.g., 17.5 years) or use the lower range if the upper range was not reported (e.g., 66 years). If a publication presented data for multiple control sites, we combined those control sites and recalculated their mean, standard deviation and sample size.

Despite its importance for determination of soil Corg stock, DBD was reported in only 55% of the included studies (34/77 for mangroves, 19/39 for saltmarshes, and 10/28 for seagrasses). We therefore used a random forest algorithm to estimate the missing DBD for each soil depth which reported soil Corg data, using all available predictor variables, including soil Corg content, BCE species (only for intact sites), middle depth of each sampling depth interval, MAT, MAP, disturbance driver (only for disturbed sites) and disturbance duration (only for disturbed sites), in a manner analogous to a pedotransfer function198 (Supplementary Fig. 5). For observations where MAT and MAP were not reported, we filled in the missing data using the WorldClim 2.0 dataset (spatial resolution: 30 s, https://www.worldclim.org/data/worldclim21.html) by averaging values within a 1 km buffer of each site’s longitude and latitude using ArcGIS 10.8 (ESRI, Redlands, CA). Random forest model combines a large number of regression trees, trained using bootstrap aggregation, to build a robust predictive model resistant to noise in the data199. The DBDs predicted using the random forest algorithm show significant linear agreement with the observed DBDs for both intact and disturbed BCEs, with an overall R2 of 0.89 (range: 0.73–0.99; P < 0.001) and a root mean square error of 0.11 ± 0.02 g cm−3 (mean ± standard error; range: 0.05–0.16 g cm−3; Supplementary Fig. 6), supporting the validity of this imputation method.

Standardization of soil depth was necessary due to the depth pattern of soil Corg stock loss being a core focus of this study, but there was considerable variation in soil depth and intervals reported, with depth classes ranging from 5 to 300 cm across eligible studies. To ensure comparability, we identified the most common sample depths for each ecosystem: 0–15, 15–30, 30–50, 50–100, 100–200, and 200–300 cm for mangroves; 0–10, 10–20, 20–30, 30–40, 40–60, 60–80, and 80–100 cm for saltmarshes; and 0–5, 5–10, 10–15, 15–20, 20–25, 25–30, 30–40, 40–50, 50–60, 60–70, 70–80, 80–90, 90–100, 100–120, 120–140, 140–160, and 160–180 cm for seagrasses. Data from the included research articles that reported sample depths not aligning with these distinct ranges were standardized using two methods198,200, depending on the maximum depth intervals reported. First, if at least five depth intervals were reported in the study, we fitted the soil Corg content and DBD to soil depth for each profile using various regression models, including linear regression, second- and third-order polynomial regressions, exponential function, loess regression, and spline regression200. The midpoint of each soil layer (e.g., 7.5 cm for the 0–15 cm soil layer of mangroves) was used as the representative depth. The best-fit model, determined by the smallest residual standard error, was then used to predict soil Corg content and DBD within the sampled profile200. The average residual standard error of the best-fit model for DBD in intact and disturbed BCEs are 0.05 ± 0.004 (range: 0.002–0.16) g cm-3 and 0.05 ± 0.004 (range: 0.002–0.22) g cm−3, respectively. Meanwhile, the average residual standard error of the best-fit model for soil Corg in intact and disturbed BCEs are 0.58 ± 0.08 (range: 0.01–3.02) % and 0.48 ± 0.07 (range: 0.002–4.41) %, respectively (Supplementary Fig. 7).

It is worth noting that for mangrove soils, studies often report soil Corg content and DBD for the entire 100–300 cm depth range based on a published protocol201. However, the actual sampling depth is typically just 5 cm in the middle of this range, specifically from 197.5 cm to 202.5 cm. To avoid introducing uncertainty in assessing Corg stock changes for the 200–300 cm depth range, we regarded this data as representing the 100–200 cm depth. Therefore, only data specifically sampled from the 200–300 cm range were included for this depth category.

Second, if fewer than five depth intervals were reported, we assigned soil Corg content and DBD to specific depth categories by first finding the median of the reported depth increment and then determining the appropriate depth category as best as possible198. For studies reporting multiple soil Corg content and DBD measurements within one depth category, a single value was calculated using a weighted average198. The weighted average for soil Corg (Corg,weighted, g C kg-1) was calculated by using the normalized depth intervals as the weighting factor for Corg of each layer (Corg,i and Corg,i+1). The depth intervals were normalized by dividing each depth interval (Ti and Ti+1) by the sum of the depth intervals, ensuring that the sum of the weighting factors equaled one (formula 1). The same approach was applied to DBD (g cm−3).

$${C}_{{{org}},{{weighted}}}={C}_{{{org}},i}\times \frac{{T}_{i}}{{T}_{i}+{T}_{i+1}}+{C}_{{{org}},i+1}\times \frac{{T}_{i+1}}{{T}_{i}+{T}_{i+1}}$$
(1)

Soil Corg stock calculation and extrapolation

Converting mangroves to aquaculture ponds typically involves soil excavation, but the actual depth of excavation is seldomly reported in literature43. To estimate the potential soil excavation depth, we filtered soil depth data from our database for paired intact mangroves and aquaculture ponds reported to have been sampled to the bedrock or parent material. Studies that did not sample to these depths were excluded from this analysis due to inability to accurately capture soil depth changes. We first calculated the soil mass to the bedrock or parent material of intact mangroves (Mintact, Mg ha−1) and aquaculture ponds (Maquaculture, Mg ha−1) by multiplying DBD, the thickness of the corresponding soil layer (Ti, m), and a unit conversion factor (104 m2 ha-1) (formula 2). Then, we calculated changes in soil mass and estimated the potential excavation depth (Texcavation, m) by dividing the soil mass loss by the average DBD of the intact mangroves (formula 3).

$${M}_{intact}\,({{\rm{or}}}\,{M}_{aquaculture})={\sum }_{i=1}^{n}{DBD}_{i}\times {T}_{i}\times {10}^{4}$$
(2)
$${T}_{Excavation}=\frac{{M}_{intact}-{M}_{aquaculture}}{\overline{{{DBD}}_{intact}}}$$
(3)

We caution against directly comparing the soil Corg stock at the original depth for intact mangroves with that at excavated soil depths in aquaculture ponds, as this may introduce uncertainty in estimating soil Corg stock loss. However, our investigation into the depth pattern of soil Corg density in intact mangroves using a linear mixed model revealed that soil Corg density did not vary significantly across different soil depths (Supplementary Fig. 8; P = 0.104). This supports the direct comparison of soil Corg stock between intact mangroves and aquaculture across soil profiles without correction.

We recalculated soil Corg stocks in disturbed BCEs using a soil mass equivalent approach rather than fixed depth and general soil Corg stock differences22,27,39. This soil mass equivalent approach compares soil Corg stock between intact and disturbed sites on an equivalent soil mass basis for each soil layer, reducing uncertainties associated with DBD variations. This method can address Corg density changes caused by processes such as soil compaction, subsidence, drainage, or flooding, which are often significant during or following disturbances22,27,37,202. Briefly, we first calculated the dry soil mass (Mi, Mg ha−1) of each layer by multiplying DBD, the thickness of the corresponding soil layer (Ti, m), and a unit conversion factor (104 m2 ha−1) (formula 4). Then, we calculated the areal Corg stock (Ci,fix, kg C ha−1) of each layer to a fixed depth by multiplying Corg content (Corg,i, g C kg−1) and Mi (formula 5). Next, we calculated the soil mass difference between the intact and disturbance sites for each soil layer by subtracting Mi from the selected minimum soil mass (Mi,equiv, Mg ha−1) (formula 6). Finally, the equivalent Corg mass (kg C ha−1) in a soil layer was calculated by subtracting the added Corg mass from top soil layer and adding Corg mass from the bottom soil layer (formula 7).

$${M}_{i}={{DBD}}_{i}\times {T}_{i}\times {10}^{4}$$
(4)
$${C}_{i,{fixed}}={C}_{{org},i}\times {M}_{i}$$
(5)
$${M}_{i,{add}}={M}_{i,{equiv}}-{M}_{i}$$
(6)
$${C}_{i,{equiv}}={C}_{i,{{fixed}}}-{C}_{{org},{{top}}} \times {M}_{i-1,{add}}+{C}_{{{org}},{{bottom}}}\times \left({M}_{i,{add}}-{M}_{i-1,{{add}}}\right)$$
(7)

We extrapolated soil Corg stock to the top one meter to allow the comparison with the current paradigm and literature values. For studies that reported soil Corg stock to depth less than 100 cm depth, we assumed that the Corg density of the unmeasured layer was the same as that of the deepest measured layer39. For example, if the deepest measured soil layer was 30–50 cm, the Corg density in the 50–100 cm layer was assumed to be the same as that in 30–50 cm. This approach, however, might overestimate soil Corg density in the deeper layers given that top soil section contains typically more Corg than the older bottom section31,197. Nonetheless, our compiled soil Corg content and DBD data showed limited depth trend from 30 cm to 100 cm (Supplementary Fig. 9), supporting the plausibility of this approach. Moreover, to further reduce uncertainty, we did not include studies that only measured surface soil Corg stock (<30 cm) when calculating top-meter soil Corg stock. In addition, if studies reported total soil Corg stock to a depth less than 100 cm, we projected Corg stock in the top meter by multiplying the Corg stock by the reciprocal of the ratio to 100, to extrapolate the Corg stocks to 100 cm depth54. To estimate emission factors for soil layers associated with significant Corg stock losses, we used only the measured or depth-standardized values without extrapolation to minimize uncertainty.

Data synthesis of disturbance on soil Corg stock

The natural log-transformed response ratio (lnRR) was employed to quantify the effect sizes of disturbance on soil Corg stock across different soil depths (formula 8; ref. 203):

$${ln}{RR}={Ln}\left(\frac{{C}_{disturbed}}{{C}_{intact}}\right)$$
(8)

where Cdisturbed and Cintact are the soil Corg stock in the disturbed and intact BCEs, respectively.

However, in our dataset, standard deviation or the standard error of soil Corg stock was not reported in 96 of the 140 articles, preventing us from estimating the variance of the effect size based on the standard deviations, sample numbers and mean values of the disturbed and intact groups. Therefore, we employed the number of replications for weighting (formula 9; refs. 204,205,206):

$${W}_{r}=\frac{{N}_{{disturbed}}\times {N}_{{intact}}}{{N}_{{disturbed}}+{N}_{{intact}}}$$
(9)

where Wr is the weight for each observation, Ndisturbed and Nintact are the numbers of replicates in the disturbed and corresponding intact BCEs, respectively.

To examine the influence of ecosystem or disturbance regimes on top-meter soil Corg stock, a linear mixed model was employed with ecosystem or disturbance denoted as fixed factors, while study as a random factor to account for the autocorrelation among observations within each study (formula 10). These analyses were conducted using restricted maximum likelihood estimation (REML) with the lme4 (version 1.1.29) and emmeans (version 1.7.5) packages with Wr as the weight for each corresponding observation in R (version 4.0.4, http://www.r-project.org/).

$${ln}{RR}={\beta }_{1}\times ({Disturbance}-1)+{\pi }_{{study}}+\varepsilon$$
(10)

where β1 is the coefficient to be estimated; πstudy is the random effect factor of study; ɛ is sampling error.

Similarly, to examine the influence of disturbance regimes on soil Corg stock across the depth profile, we employed depth interval denoted as fixed factors for each disturbance driver, while study was treated as a random factor as before (formula 11).

$${ln}{RR}={\beta }_{1}\times ({Depth}-1)+{\pi }_{{study}}+\varepsilon$$
(11)

For ease of interpretation, LnRR and its corresponding 95% CIs were further transformed into percent soil Corg stock change (ΔCorg,%; formula 12). If the 95% CI of ΔCorg does not cover zero, the effect of individual disturbances was significant (positive or negative, at P = 0.05) in affecting soil Corg stock.

$${\varDelta C}_{{org}}=({e}^{{ln}{RR}}-1)\times 100$$
(12)

To explore the response of soil Corg stock loss to initial Corg stock and disturbance duration, we first statistically compared the linear and logarithmic functions with the factor of interest as the fixed effect, and ‘site’ as the random effect, using the Akaike information criterion (AIC). We found that the logarithmic functions resulted in lower or similar AIC values for the response of soil Corg stock change (%) to initial soil Corg stock and disturbance duration (Supplementary Table 4), while linear functions resulted in lower or similar AIC values for the response of soil Corg stock change (Mg C ha−1) to initial soil Corg stock (Supplementary Table 5). Therefore, we used logarithmic or linear functions to construct the response of soil Corg stock change to initial soil Corg stock and disturbance duration and visualized with ggplot2 (version 3.3.6) packages in R.