The amount of organic carbon locked away from connection with the atmosphere, or – in the case of sediments, the ocean water column – is the fundamental parameter determining if and where shelf sea sediments satisfy the blue carbon criterion of “significant carbon storage”. The distribution of carbon stocks must be known in order to consider their importance, vulnerability and manageability.

The Blue Carbon Initiative ¹ ’s manual designed to provide standardized guidance to managers for assessing carbon stocks and emissions factors (Howard et al., 2014), enumerates the three elements required to quantify stocks: (i) (total) organic soil thickness (depth to underlying hard substrate), (ii) dry bulk density, (iii) organic carbon content (%C_org). In the context of continental shelf sediments, dry bulk density and organic carbon content remain critical, while total soil (sediment) depth is unlikely to be determined and furthermore not immediately relevant in the context of climate mitigation. The depth of interest is that which is likely to be disturbed by human activities, and thus contains “vulnerable” carbon stocks. In traditional blue carbon ecosystems the organic soil thickness is expected to vary from between 10 cm to over 3 m (Howard et al., 2014), and IPCC guidelines recommend the use of 1 m as a standard depth for reporting stocks in the absence of information on total soil depth (IPCC, 2008; IPCC, 2013). In marine sediments, the relevant depth of vulnerability will be influenced by the natural setting (e.g., water depth, sediment type, fauna) and the influence of potential disturbances (bottom trawling, deep-sea mining, dredging and disposal, etc., see Anthropogenic pressures). Many, though certainly not all, existing datasets are limited to near-surface sediments (e.g., upper 10 cm; Diesing et al., 2017; Wilson et al., 2018; Smeaton et al., 2021a) due to standard sampling procedures in routine monitoring programmes. Significant (and unquantifiable) uncertainty in estimated stocks is introduced where data are extrapolated beyond their measurement depth (e.g., to 1 m, Atwood et al., 2020) as it is not expected that C_org content remains constant between near-surface and deeper sediments. Generally, C_org is expected to decrease with depth due to remineralization/diagenesis (e.g., Burdige, 2007; Arndt et al., 2013; LaRowe et al., 2020) such that extrapolating near-surface carbon stocks to e.g., 1 meter depth would yield an overestimation of stock, and that deeply buried organic matter may be less labile and less likely to be remineralized to release CO₂ if disturbed (see Characterising stock reactivity)

Determination of dry bulk density and organic carbon content in marine sediments are standard analyses which are routinely carried out by geologists, geochemists, and biogeochemists. Dry bulk density is the ratio of the mass of a dried sediment sample to its original volume. Porosity, the ratio of the volume of ‘voids’ (between sediment grains) to the total (wet) sediment mixture volume, is more frequently reported in marine sediments. Porosity is generally calculated by mass difference of wet and dried sediment, assuming a density for pore fluid (based on salinity) and sediment grains (typically 2650 kg cm^-3 for quartz/feldspar) (e.g., Jenkins, 2005; Diesing et al., 2017). Empirical relationships have been used to estimate bulk density from sediment grain size (% mud) and C_org content itself (e.g., Diesing et al., 2017; Silburn et al., 2017; Atwood et al., 2020). The organic carbon content of marine sediments is commonly determined by elemental analyzer, alongside total carbon, inorganic carbon, and organic matter carbon: nitrogen (C:N) ratio (Verardo et al., 1990; Nieuwenhuize et al., 1994; Harris et al., 2001). This well established method, and its advantages over the lower-cost loss on ignition (LOI) approach, which relies on determining a habitat and/or site-specific conversion between organic matter and organic carbon, is described in detail in Howard et al. (2014) in the context of its application in coastal blue carbon habitats. A methodologically simplified approach based on ramped thermal heating alongside CO2 detection by infrared spectrometry has recently been applied to marine sedimentary carbon (Smeaton, et al., 2021a).

Total carbon stocks are usually derived by extrapolating spatially limited observations over a relevant spatial area to generate an inventory. The validity of, or numerical uncertainty in, the estimated total stock then depends on how representative the sampled sites are of the larger area of interest (Howard et al., 2014). Where this is partially addressed by estimating a different stock for different spatial sub-units (e.g., Burrows et al., 2014; Watson et al., 2020; Gregg et al., 2021), additional uncertainty is introduced by the estimation of the extent of the area for each. It is important to recognize the large uncertainty in values obtained by multiplying estimated stock over a large estimated area (discussed for soil C_org in Goidts et al., 2009). Since measurements in marine sediments require sampling from a research vessel, at high cost, achieving adequate spatial resolution for high confidence inventories is particularly challenging. Furthermore, there is a tendency for observations to be biased to near-coast and, because sediment coring is more challenging in coarse sediments, to higher mud-content sediments which are expected to contain high C_org (Figure 5 in Smeaton et al., 2021a). Linking carbon content to explanatory parameters (e.g., particle size, hydro-acoustic reflectance; Hunt et al., 2020) and applying statistical models or machine learning approaches to generate spatially resolved maps (Serpetti et al., 2012; Diesing et al., 2017; Wilson et al., 2018; Atwood et al., 2020; Smeaton et al., 2021a), can help to overcome issues of spatial bias in observations but assumes carbon processing, which controls stock, occurs in the same way across the region of interest. Co-measurement of the key physical and carbon content properties is a fundamental requirement for producing accurate stock assessments therefore determining additional sediment and environment parameters alongside carbon stocks adds significant value to carbon stock measurements.

Marine carbon stocks can also be determined by modelling. Process-based biogeochemical modelling studies focused on shelf sea carbon cycling (e.g., Kühn et al., 2010; Wakelin et al., 2012) generally assume marine carbon ultimately derives from marine primary production and, crucially, that most organic detrital carbon in the benthic system is remineralized over an annual cycle. With this assumption, models typically predict stocks of benthic C_org one to two orders of magnitude lower than observed (Aldridge et al., 2017). This discrepancy is resolved if it is assumed the models follow only the relatively labile carbon most important for correctly modelling the annual recycling of carbon and nutrients, whilst most of the observed carbon stock is highly refractory: does not degrade significantly over annual timescales. However, this means that model results do not, at present, provide appropriate stock values for sediment carbon inventories. Long term storage (sequestration) of highly refractory (non-mineralizable) carbon has been implemented within models, for example by explicitly assuming a background flux (Soetaert et al., 1996) or assuming a fraction of more labile carbon becomes deeply buried and no longer undergoes degradation (Butenschön et al., 2016).

To apply the blue carbon accounting framework, organic carbon in marine sediments must be linked back to the source of photosynthesis to avoid “double-counting”. For example: macroalgal carbon being accounted for in a carbon valuation of the coastal macroalgae habitat, and then again in offshore marine sediments where it is eventually buried. To protect or enhance the carbon storage potential of shelf sediments, management measures need to be applied not only locally, but also to the key carbon source ecosystems. Furthermore, the vulnerability of organic carbon in marine sediment to disturbance, in terms of release back towards the water column, is determined in part by its provenance as this influences how readily it is broken down and remineralized to CO₂.

The primary production of much of the organic carbon subsequently stored in marine sediments occurred elsewhere: on land or in coastal habitats. In shelf seas marine phytoplankton are an important organic carbon source, though much of this carbon is expected to be recycled back to CO₂ within ocean waters (e.g., Humphreys et al., 2019; Dang, 2020). In shallow waters, even away from the coast, microphytobenthos also fix carbon at the sediment surface (e.g., Reiss et al., 2007). Fundamentally, linking stored carbon to its source(s) relies on constraining the properties of the source itself: potential sources should be identified and analysed wherever possible, which is more challenging when carbon storage occurs at a distance from the carbon source and molecular properties are modified between source and sink.

In being transported to marine sediments, and after having been deposited there, organic matter is typically transformed by physical (Ausín et al., 2021) and biogeochemical processes (Middelburg, 2018; Kharbush et al., 2020). This chemical alteration makes it more difficult to identify the provenance of different components of sediment organic matter. Moreover, it often renders the organic material molecularly uncharacterizable. Within the sediments, interaction with the mineral surface (Kleber et al., 2021), and with iron minerals in particular (Lalonde et al., 2012) have been suggested as a key mechanisms for this. This fraction of the input organic carbon that is sequestered long-term in marine sediments, largely significantly altered from its original molecular form, is that which is largely resistant to remineralization to inorganic carbon. Organic carbon which is less altered, and thus more reactive/vulnerable to disturbance, can also be buried in marine sediments in environments of low biological activity (e.g., under low oxygen or low temperature conditions; Canfield, 1994).

Methods for determining the provenance of organic carbon in coastal blue carbon ecosystems (tidal marshes, mangrove and seagrass meadows) were recently reviewed by Geraldi et al., (2019). These authors concluded that the more different properties of organic matter that can be determined - in other words: applying a multi-tool approach - the more likely it is a single source can be unambiguously identified. When characterizing organic matter, the methods that provide greater source specificity are those that are performed on the smallest fractions of the bulk material (e.g., compound-specific analyses; see Figure 2.1 in Bianchi and Canuel, 2011). Here we place relevant methodology in the context of continental shelf sediments and consider issues and sources of uncertainty.

Not all organic carbon present in marine sediments at a given time will be stored over timescales greater than weeks to months to years (LaRowe et al., 2020). The total stock of organic carbon can be classified by its biogeochemical reactivity (Alderson et al., 2016), a key biogeochemical model parameter (Arndt et al., 2013). Organic matter reactivity or degradability (which can be defined as susceptibility to removal of a fraction by either biotic or abiotic processes, measured by lifetime e.g., Hansell, 2013) is understood to be a continuum (e.g., LaRowe et al., 2020) from labile (reactive; easily degraded; short lived) to recalcitrant (resistant to rapid microbial degradation; Hansell, 2013) and refractory. The term refractory is defined variability in different studies and used as “a one-fits-all word” for various properties related to C_org reactivity, typically as the opposite of labile in a certain context, see discussion in Hansell, 2013 and Baltar et al., 2021. Understanding sedimentary organic matter’s re-mineralization potential and ability to resist degradation (Doetterl et al., 2016) is vital in estimating its vulnerability to potential natural or anthropogenic disturbance and remineralization of the associated organic carbon.

The balance of organic carbon fractions with various reactivities determines the biodegradability of sedimentary organic matter and in turn the stability of the associated carbon. The rate of microbially-mediated degradation of organic matter can be impacted by the addition of more or of less labile material to a mixture (the priming effect; Bianchi and Canuel, 2011). If disturbed, the organic carbon in sediments dominated by labile organic matter is highly likely to be remineralized. Conversely if sediments dominated by recalcitrant organic matter are disturbed this material would likely be re-deposited as particulate organic carbon on or in the seabed. Only organic carbon of some intermediate degradability is thus relevant to the climate mitigation role of marine sediments.

Organic carbon reactivity is fundamentally linked to the organic matter’s source, discussed in Characterising stock provenance. Black carbon, a refractory carbon fraction formed from incomplete combustion of organic carbon, accounts for 15-30% of the organic matter buried in marine sediments (Middelburg et al., 1999). The most refractory fraction of the black carbon is highly condensed soot, formed from combustion of fossil fuels at high temperatures (Masiello, 2004), which may remain in the atmosphere for months before eventually being buried in sediments. More degradable lightly charred black carbon formed during low temperature biomass burning may reach the marine environment via rivers through soil leaching or erosion (Coppola et al., 2014). Once deposited in sediments, factors including the depth of burial, microbial and redox conditions as well as subsequent disturbance influence the degree of black carbon’s long-term storage. In general, small soot particles withstand degradation and result in long-term storage whereas lower formation temperature/larger size fragments are more susceptible to biotic and chemical degradation (Masiello, 2004).

Chew and Gallagher (2018) show that black carbon can be a major component of organic carbon in seagrass, mangrove and saltmarsh sediment cores and because it is resistant to being re-released back into the atmosphere, as CO₂ for example, assert that it should be discounted from any blue carbon storage mitigations being proposed. Taking a different perspective, Ren et al. (2019) highlight that the relatively slow cycling of black carbon in Arctic Ocean sediments makes it an important fraction of inert carbon representing a significant sink of atmospheric CO₂, which is likely to increase with global warming and increased melting of permafrost. There is thus a clear need to understand both the overall reactivity of the marine organic matter as well as to identify specific types and sources of organic carbon which are resistant to degradation.

The process of assigning ages to organic carbon present in specific layers (depth intervals) of marine sediment can be achieved by measuring the activity concentration of co-occurring ²¹⁰Pb (Tolosa et al., 1996; Pappa et al., 2019). In relatively undisturbed depositional aquatic ecosystems ²¹⁰Pb abundance decreases exponentially with depth from the seawater-sediment interface due to the radioactive decay of the tracer, allowing the ²¹⁰Pb depth profile to estimate a sedimentation rate for any substances of interest associated with the sediment. Application of ²¹⁰Pb dating remains a challenge because of unresolved issues associated with the variety of radioanalytical methodologies available as well as different modelling approaches that play an essential role in deriving sediment accumulation rates from radiotracer measurements (MacKenzie et al., 2011; Garcia-Tenorio et al., 2020).

Disturbances can complicate the straightforward determination of sediment accumulation rates from ²¹⁰Pb depth profiles. The surface 10-15 cm of the seabed are likely to be disturbed by, e.g., bioturbation, fishing activities and natural resuspension by bottom water currents (see Anthropogenic pressures). Furthermore, dating must account for the additional sources of ²¹⁰Pb within the sediment, from the in-situ decay of radium-226 (²²⁶Ra) and, more problematically, from industrial activities including offshore oil and gas and other mineral processing industries (Sahu et al., 2014; Ahmad et al., 2021). The proportion of ²¹⁰Pb present in the sediment and from human activity (termed supported ²¹⁰Pb) does not decay at the same rate as the unsupported ²¹⁰Pb fraction from atmospheric deposition (Gulliver et al., 2001; Cook et al., 2004). Measuring ²²⁶Ra in addition to ²¹⁰Pb enables the subtraction of supported ²¹⁰Pb from the total ²¹⁰Pb measured prior to estimating sedimentation rates and dating sediment layers.

With a half-life of 22.3 years, ²¹⁰Pb dating is limited to relatively recent sediment ages (up to 150 years). It can therefore be challenging to use this technique in low-sedimentation rate marine environments (<0.15 cm y^-1) (De Haas et al., 1997). There are several ²¹⁰Pb dating models available, that can be applied and help interpret different ²¹⁰Pb core profiles (Arias-Ortiz et al., 2018; Blaauw et al., 2018). These models have been comprehensively described over the last forty years. The Constant Initial Concentration (CIC) (Robbins, 1978), Constant Rate of Supply (CRS) (Appleby and Oldfield, 1978) and Constant Flux: Constant Sedimentation (CF : CS) (Appleby and Oldfield, 1978) models are the most widely used to estimate sedimentation accumulation rates, but they are associated with numerous assumptions that can sometimes limit their appropriateness in more complex and disturbed aquatic ecosystems (Arias-Ortiz et al., 2018). Very often, the mass accumulation rate, expressed in mg cm^-2 yr^-1, is also estimated to account for compaction effect (Schmidt et al., 2014). More recently the advantages of a Bayesian approach, as applied in radiocarbon dating (Blaauw et al., 2018), is gaining attention (Aquino-López et al., 2020; Blaauw et al., 2021).

²¹⁰Pb in environmental matrices, including marine sediments, is commonly determined directly by gamma spectrometry. However, gamma spectrometers equipped with Well-type germanium (Ge) detectors that allow the direct determination of low levels of ²¹⁰Pb with little sample mass and preparation required are not commonly available. More common and cheaper flat end Ge detectors with much higher limits of detection can provide insufficient levels of accuracy and precision for measuring the very low levels of ²¹⁰Pb naturally present in the environment (Dal Molin et al., 2018). Alternatively, ²¹⁰Pb can also be determined indirectly by measuring polonium-210 (²¹⁰Po), its granddaughter decay product, via alpha spectrometry. Although this alternative traditional radiometric technique requires samples to be digested in strong acid mixtures and treated prior to analysis, the limits of detection and level of accuracy and precision that can be achieved are more appropriate (Hassen et al., 2020). The underlying assumption, that ²¹⁰Pb is in secular equilibrium with ²¹⁰Po such that both radionuclides are present at the same concentration, can be verified by performing a second analysis of ²¹⁰Po at a later stage, i.e., generally, by determining ²¹⁰Po ingrowth from the decay of ²¹⁰Pb after three to four months.

To determine ²²⁶Ra and correct for supported ²¹⁰Pb, gamma spectrometry is the most common technique. The sample must be placed in a radon proof container for a minimum of three weeks prior to analysis to ensure secular equilibrium is established before indirect measurement via its two decay products (bismuth-214 and lead-214). ²²⁶Ra can be determined indirectly by either measuring emission of radon-222 gas, or directly by alpha-spectrometry, and more recently by ICP-MS, following appropriate sample preparation steps (IAEA, 2010).

Caesium-137 (¹³⁷Cs) is commonly used as a complementary tool to validate ²¹⁰Pb dating models. Characteristic abundance peaks of ¹³⁷Cs observed in marine sediments in the Northern hemisphere can be directly related to nuclear weapon testing events of the late 1950s and early 1960s, and more locally in the UK seas to peak discharges from Sellafield nuclear reprocessing plant in the late 1970s (Environment Agency et al., 2021). Caesium-137 can easily be measured by gamma spectrometry, however, because its concentration is constantly decreasing over time [half-life of 30 years and no major additional characteristic peak since Chernobyl; see Figure 7.9 in Environment Agency et al. (2021)], it is important to optimize the counting time to ensure reliable and accurate results by gamma spectrometry (Garcia-Tenorio et al., 2020).

The radioactive ¹⁴C isotope has a half-life of 5,700 ± 30 years (Kutschera, 2013) which allows for the determination of the age of a carbon bearing material formed over the past 55,000 years (Hajdas et al., 2021). Since the detection of ¹⁴C in 1977 using particle accelerators (Bennett et al., 1977), Accelerator Mass Spectrometry (AMS) has become the most used technique in ¹⁴C analysis. In AMS, the ¹⁴C/¹³C or ¹⁴C/¹²C ratio in the sample is directly measured resulting in shorter measurement times (days to minutes) and reduced sample requirements (g to µg of C) than previous scintillation methods (Litherland et al., 2011). Recent developments in AMS technologies have resulted in ¹⁴C detections at significantly lower energies (Synal et al., 2000), resulting in the reduction in size of AMS’s and the establishment of new sample introduction techniques for direct analysis of CO₂ resulting from combustion (EA-AMS) (Haghipour et al., 2019), wet oxidation (Leonard et al., 2013), acid hydrolysis (Molnár et al., 2013) and laser ablation (LA-AMS; Welte et al., 2016).

The data produced by the AMS (Fraction ¹⁴C or F¹⁴C) allows for conventual ¹⁴C ages to be calculated following the approach of Libby (1955) where conventional ¹⁴C age = -8,033ln(F¹⁴C). To determine the true age of the sample, the conventual ¹⁴C age must be calibrated to a curve based on measurements of ¹⁴C from samples of known age and location (e.g., northern hemisphere, southern hemisphere, or the marine environment). The IntCal radiocarbon calibration curves (Heaton et al., 2020; Hogg et al., 2020; Reimer, 2020) are a resource constructed from measurements of ¹⁴C made in different laboratories on a range of dated samples. The last 13,900 years are constrained by dated tree rings (Reimer, 2020), while beyond this (13,900-55,000 years) lake and marine sediments, speleothems or corals are statistically integrated to create the most accurate representation of past atmospheric ¹⁴C (Reimer, 2020).

Today, conventional ¹⁴C ages are calibrated using advanced Bayesian statistical tools that include dedicated routines for the different calibration curves, the most commonly used modelling tools are OxCal (Ramsey, 1995; Ramsey et al., 2020) and MatCal (Lougheed and Obrochta, 2016). Calibrated ¹⁴C ages are referred to as a calendar age counted backwards from 0 cal BP (before present) which corresponds to 1950 Common Era (CE, previously referred to as AD) or to calendar years Before Common Era (BCE, previously referred to as BC). Post 1950 CE correspond to the age of nuclear testing where atmospheric ¹⁴C concentrations measured in CO₂ of air and dated tree rings, display significantly elevated ¹⁴C concentration (bomb spike) hindering the use of conventional ¹⁴C dating approaches (Hua et al., 2013; Santos et al., 2020). To calculate sedimentation and accumulation rates an age-depth model is constructed using the calibrated dates. Classical age-depth models can be constructed in software such as CLAM (Blaauw, 2010), but more commonly Bayesian statistical packages such as OxCal (Ramsey, 1995; Ramsey, 2009), and BACON (Blaauw and Christen, 2011), are applied to create models that fully account for uncertainties allowing for the calculation of sedimentation and accumulation rates.

There is a long history of using ¹⁴C to chronologically constrain sedimentary records allowing the calculation of long-term (100s-1000s of years) sedimentation and accumulation rates on the seabed (Kershaw, 1986; Darby et al., 1997; Colman et al., 2002; Santschi and Rowe, 2008; Smeaton et al., 2021c). In marine sediments both organic matter and carbonate (inorganic) material (e.g., shells or foraminifera) can be dated, but organic matter cannot be used to constrain the age of the sediment as the carbon it contains originates from multiple (many unknown) sources with varying compositions and ages. Carbonate material is therefore the preferred sample medium for ¹⁴C analysis and age constraint. Living benthic foraminifera are found in the upper most layers of most sediments and therefore the shells of foraminifera buried after death provide material ideally suited for dating as they are representative of past surface sediments from their time of deposition. The number of foraminifera required to measure ¹⁴C using a conventional AMS can vary but it is not uncommon to need between 500-1000 individuals, whereas the recent developments in AMS technology (Synal et al., 2000) means that far smaller numbers are now required (<100 individuals) with state-of-the-art systems allowing single foraminifera dating (Wacker et al., 2013; Lougheed et al., 2018). Shells can also be dated but greater care must be taken to ensure they are representative of past surficial sediments and not shells associated with mobile burrowing species (Cage and Austin, 2010; Baltzer et al., 2015).

When calibrating ¹⁴C ages derived from material from the marine environment the marine reservoir effect must be considered. The marine reservoir effect occurs when the C in tissues of organisms or inorganic deposits reflects a mix of C sources and ages (Hajdas et al., 2021). The marine reservoir effect was previously accounted for by adding 400 years to an age calibrated using the northern hemisphere calibration curve (Reimer, 2020) which led to large uncertainties. With the production of the Marine20 calibration curve (Heaton et al., 2020) the marine reservoir age can now be robustly accounted for and is directly integrated into the calibration stage. The Marine20 calibration curve has significantly improved age calibration of marine samples, but it remains a global calibration; for greater accuracy local reservoir age corrections can be applied (i.e., Cage et al., 2006) if available.

For 500 to 100,000-year timescales, thorium-230 (²³⁰Th) dating, also referred to as U/Th dating or ²³⁸U- uranium-234 (²³⁴U)-²³⁰Th dating, can be applied. This method relies on the disequilibrium due to solubility differences of uranium and thorium nuclides and involves calculating ages by studying the radioactive relationships between ²³⁸U, ²³⁴U, and ²³⁰Th (Henderson and Anderson, 2003). It is appropriate in closed carbonate systems with no initial ²³⁰Th, for example, fossil carbonates present within the terrestrial environment (e.g., rocks), but also marine carbonate formations such as corals (Cheng et al., 2000) and authigenic carbonates. It is very often used as a complementary method to ¹⁴C dating of fossil carbonates and for validating results from ¹⁴C dating.

Extending thorium-230 (²³⁰Th) dating to marine sediments is challenging mainly due to the presence of additional ²³⁰Th in materials scavenged from the seawater column. In addition, the typical levels of carbonates found in marine sediments as well as the potential post-depositional weathering, transport and mixing processes can further limit its application (Chen et al., 2020). Nevertheless, modelling and state-of-the-art analytical tools can potentially overcome some of these challenges and thus provide valuable insights into sediment accumulation rates over longer timescales. Geibert et al. (2019) have recently presented an alternative approach using a CRS model, similarly to ²¹⁰Pb dating and assuming constant rate of supply of ²³⁰Th from the seawater. Although alpha spectrometry can provide accurate measurement of all relevant U and Th isotopes, these techniques require labor intensive preparation steps and lengthy counting times in comparison to mass-spectrometry coupled with a laser ablation system. These state-of-the-art techniques have become the preferred analytical option (Robinson et al., 2004; Mas et al., 2020).