Background & Summary

Coastal ecosystems are unique in global change science because they form soil mass as a dynamic response to sea-level change1,2,3,4. In tidal wetlands, soils accrete via organic and inorganic pathways5, forming organic soil mass by subsurface root turnover6 and high rates of organic matter preservation6,7, and promoting mineral sediment layers by trapping suspended sediment associated with incoming tides8 and storm surges9. Both of these have some, but limited, capacity to increase with accelerating sea-level rise, causing a stabilizing feedback10. The burial of organic matter also causes a continuous sink, with necromass potentially stored on millennial scales, effectively sequestering carbon1.

Coastal wetland restoration, preservation, and enhancement are of interest to many parties, including land managers and policymakers, because of the potential for wetland conservation and interventions to improve carbon sequestration11, minimize methane emissions12, and enhance a suite of other ecosystem services13. Multiple projects on the U.S. Pacific Coast have incorporated large tidal reintroduction projects, such as in the Nisqually Delta in Washington State14 and the San Francisco Bay and Sacramento Delta of California15.

Projecting whether tidal wetlands will fail to keep pace with sea-level rise and convert to open water requires information about their accretion history, in tandem with process-based models16. Therefore, accretion, soil carbon, and bulk density information are vital to planning for conservation and management action to prevent loss. Data is required to plan interventions to stop the net losses of carbon stocks by strategies such as sediment augmentation17 and facilitating the upland migration18,19 of wetlands.

Tidal wetland restoration is a vital part of national and sub-national policies to offset climate impacts. Fargione et al.20 highlighted methane emission reduction from reintroducing saline tides to artificially freshened impoundments as a “Natural Climate Solution”. Importantly, carbon sequestration and methane emission reduction synergize with other co-benefits such as the conservation of habitats for endemic and endangered species, protection of coastal infrastructure from storm surges, improvements in water quality, and support for economically important fishery species13.

The State of California has a particular interest in managing coastal systems for carbon benefits. The State’s greenhouse gas reduction plan calls for a suite of carbon removal strategies, including natural carbon sequestration from protecting and restoring tidal wetlands21. A critical requirement for both supporting coastal restoration and conservation policies from the top down and management decisions from the bottom up is the curation of transparent disaggregated datasets for both carbon stocks and carbon accumulation rates22. A previous iteration of the State of California Natural and Working Lands Scoping Plan did not include tidal wetlands outside of the Sacramento Delta because data were limited23.

Limited tidal carbon stock data exist for the outer coastline of California (Fig. 1). According to the Coastal Carbon Library24,25, there have been 15 previous studies published for California associated with 27 citations4,15,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50. For example, Ward et al.45 sampled carbon stocks from seagrasses and tidal marshes from multiple sites in Southern, Central, and Northern California. Two studies have sampled tidal marshes in Humboldt Bay, Northern California33,36. Nahlik and Fennesey sampled multiple tidal wetlands within the state as part of the National Wetland Conditions Assessment40.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Map of the California, USA coast shows the latitudinal transect, with sampling sites labeled. The number of soil cores per location is indicated by the size of the point (n). Triangles indicate the locations of previous studies and sites.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Example depth profiles from a single core from this dataset (SB15-11) including organic matter content, dry bulk density, a radiocesium peak indicating 1963 CE, and 210Pb, both total and supported (226Ra), as well as the age-depth model with propagated uncertainty.

As far as we know, current carbon accumulation rate data are almost entirely located in the San Francisco Bay and Sacramento Delta, and are absent from the Outer Coast in the literature. Currently available data for the San Francisco Bay and Sacramento Delta include ~50- to 100-year records of natural tidal marshes4, longer-term (centennial to millennial-scale) records of natural tidal marshes47, comparisons between restoration and reference wetlands15,31, and both nearer-28 and longer-term51 records of tidal freshwater forests, also known as swamps.

In this paper, we present a new dataset containing carbon stock estimates, radioisotope dating information, and accumulation rate estimates from 83 coastal marsh soil profiles from 15 wetland sites in California, mostly from the outer coastline. The tidal wetlands extend in latitude from near the Oregon border southward to the border with Mexico. We document the hierarchical, disaggregated data structure showing depth-series measurements nested within coring locations, nested within sites.

In addition to providing the dataset, we provide carbon stock calculations to a depth of 1 m and carbon accumulation rates. Finally, we provide R code demonstrating these calculations.

Earlier releases contained some of the information herein. Carbon stock data have been used in previous analyses52,53, but not published alongside the radioisotope dating information. Some radioisotope dating information was presented by Thorne et al.18 but is expanded upon here. Some radioisotope information was analyzed and presented in, but not published in disaggregated form, by Fard et al.54 (2021) and Brown55 (2019). This data release is the first attempt to synthesize carbon stock and carbon accumulation rate from a set of related studies taking place over the same time frame in California.

Methods

Field collection

This study presents data from a collection of 83 soil cores from 16 tidal wetland sites in California (Fig. 1). Cores were collected between 2012 and 2017. Five sites are located within Humboldt Bay, including White Slough (WTS), Mad River Low (MRL) and Mad River High (MRH), Jacoby Slough (JCB), Hookton Slough (HKS), and Eureka (ERK). Two additional sites were sampled on the outer coastlinenear San Francisco Bay, Bodega Bay (BOD) and Bolinas Lagoon (BOL). Two sites were within San Francisco Bay, Petaluma Marsh (PTL) and Triangle Marsh (TRM). The remaining sites were in central and southern California, including Morro Bay (MOB), Point Mugu (MGL), Seal Beach (SB), Upper Newport Bay (UNB), Mission Bay (MB), and Tijuana Estuary (TJE).

Cores were collected using a ‘Russian’- style peat corer, a meter-long half cylinder with a pivoting flap, that minimizes compaction. In the field, dominant plant species present at the sampling location were noted based on a visual inspection. General flooding zones were classified qualitatively based on known associations with vegetation species56. Latitude and longitude were recorded for each sampling point using a handheld GPS.

Laboratory processing

Cores were transported intact as collected, usually in 1 m drives, to the University of California, Los Angeles. They were stored in plastic wrap to reduce moisture loss and refrigerated until processing. We subsectioned cores into 1 cm increments down to a depth of 1 m, and stored processed material in sealed plastic bags, which were refrigerated.

For each 1 cm depth increment, we measured organic content and dry bulk density. Dry bulk density was measured by allocating a standardized volume of 1 cm−3 of wet sample and drying samples in an oven set at 105 °C until constant mass was achieved. Organic content was measured using loss-on-ignition, the mass lost after 4 hours of ignition in a 550 °C muffle furnace57.

For select depth increments, we measured various radioisotope activities useful in radiometric dating techniques: radiocesium dating (137Cs), lead-210 (210Pb) dating, and radiocarbon (14C) dating. 137Cs and 210Pb dating were done at three different lab facilities: Core Scientific International, Queen’s University, and the University of Southern California. Activities of 137Cs, total 210Pb, and proxies for supported 210Pb (226Ra) were all measured using gamma detectors.

Material for 14C dating was recovered from select 1 cm subsamples and included plant and shell fragments. 14C dates were run at the Keck Lab at the University of California, Irvine. Because they use different calibration curves to convert radiocarbon ages to calendar ages, we note whether dates originate from carbonate or organic material alongside radiocarbon data.

Data analysis

In analyzing the data, we estimated 1 m depth equivalent carbon stocks, created age-depth models from 137Cs and 210Pb profiles, and calculated the resulting carbon accumulation rates.

Carbon stocks

Depth increment-level carbon density was estimated as the product of modeled organic carbon content and dry bulk density. We estimated organic carbon from loss-on-ignition by using a conversion function published by Craft et al.58 (Eq. 1).

$${OC}=0.40\,{LOI}+0.0025\,{{LOI}}^{2}$$
(1)

For each core, we summarized carbon stocks as the weighted average of the individual depth increments scaled to one meter depth. One meter depth is the default depth for existing global tidal carbon stock maps59,60 and the International Panel on Climate Change’s assumptions for the depth of carbon lost when wetlands are converted to open water or other land cover types61. Sensitivity analysis shows that the U.S. coastal wetland greenhouse gas inventory is highly sensitive to uncertainties in this assumption62,63.

We assessed 53 cores from 12 tidal wetland sites for carbon stocks that had a depth of at least 90 cm for carbon stocks (Fig. 3). In our dataset, mean carbon stock extrapolated to a depth of 1 m was 27.8 kgC m−2 (Fig. 3). There was little deviation between site level means and an IPCC default emission factor --the amount of carbon assumed lost per square meter when the top 1 meter of soil in a wetland is converted to open water --of 25.5 kgC m−2 for mineral and organic soils combined61. Average stocks were slightly lower than those estimated using the same method for a previous cores collected in California for tidal marshes (27.1 kgC m−2, n cores = 37)27,32,36,40,43,47.

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

Carbon stock data for each core aggregated at 1 m depth. Sites are sorted by latitude; left to right indicates north to south. Each point represents an individual core, and bar and whisker plots represent site-level distributions. Central bars represent median values, box edges represent 25 and 75% quantiles, and whisker edges represent the most extreme values within 1.5 times the interquartile range. The total dataset’s mean value is displayed relative to the IPCC global default (Tier I) emissions factor for tidal marshes converted to open water, and the mean of 1 m integrated carbon stock estimates from previously measured locations in California.

Mean organic matter content in our new dataset was 11% organic matter by dry weight, slightly lower than the threshold above which soil bulk density becomes insensitive to further decreases in mineral content52.

For each core, we fit a linear model of organic matter content as a function of depth, using R64. Typical organic matter percentage in a core decreased with depth (Fig. 4). Average organic matter content was 19.5% for the top 5 cm of soil, which decreased to 8.1% at the 95 to 100 cm increment. Of the cores with at least 1 m depth, 82% had organic matter content that significantly decreased with depth, 9% had organic matter content that significantly increased with depth, and 9% displayed no significant depth-wise trends. Average slope for significant decreases indicated an organic matter loss of 0.13% per centimeter of depth.

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Relationships between depth and organic matter content for each core and each site. White points and lines indicate individual depth series. Red lines indicate smoothed site-wide trends.

Carbon accumulation rates

In addition to carbon stocks, we provide decadal- to centennial-scale carbon accumulation rate calculations, comparing these new values to averages from existing data. To create decadal to centennial-length age-depth models, we used a combination of 137Cs and 210Pb dating.

In North America, peaks in 137Cs activity correspond to peaks in above-ground nuclear weapons testing ahead of the 1963 nuclear weapons test ban treaty65. However, a previous review by Drexler et al.66 showed that the decline of clarity of 137Cs peaks with time and geography results in decreasing confidence in the resulting accretion rate estimates. Therefore, we quantified the quality of these dates using three independent metrics and propagated uncertainty in the resulting accretion rates and age-depth models.

137Cs peaks were used only if they met three levels of automated statistical quality control. First, we determined whether peaks were significantly different from a background level, given the total distribution of 137Cs per core. We determined this using the pnorm function in R64 with the peak radioactivity as the value, the mean and standard deviation of the activity counts for all cores at the site, and a significance threshold of 97.5%. Second and third, we determined whether peaks were significantly greater than both their adjacent upper and lower samples, given radioactivity counting uncertainty. We calculated these as the proportional overlapping area shared by two normal distributions defined by each sample’s mean estimated radioactivity and counting uncertainty. We classified peaks as being significantly elevated compared to the shallower and deeper adjacent measurements if the difference in activity was more than 2 standard deviations. Pairs that shared less than 5% proportional area were considered significantly different.

Uncertainty in the 137Cs-based age was propagated by treating the age as a uniform distribution and substituting age for depth. We estimated the mean and standard deviation of the 137Cs age using a commonly available moment matching formula (Eqs. 2, 3)67.

$${\mu }_{A}=({A}_{{\max }}+{A}_{{\min }})/2$$
(2)
$${\sigma }_{A}=\sqrt{{({A}_{{\max }}-{A}_{{\min }})}^{2}/12}$$
(3)

\({\mu }_{A}\) and \({\sigma }_{A}\) refer to the mean and standard deviation of the age. \({A}_{{\max }}\) and \({A}_{{\min }}\) refer to the minimum and maximum potential age given a uniform distribution.

We expanded the age-depth modeling and uncertainty propagation from the decadal scale to the centennial scale by combining 137Cs dates with 210Pb dating using PLUM68. PLUM is a Bayesian age-depth modeling algorithm based on the constant rate of supply (CRS) model69 that requires dry bulk density, a depth series of total 210Pb values, and associated radioactivity counting errors as input variables.

To prepare cores for PLUM dating, we interpolated any missing dry bulk density or loss-on-ignition values from neighboring values using the approx function in R64. We specified the year of core collection as the surface age.

The algorithm is flexible in that it can either estimate supported 210Pb from the depth series or use proxy measurements of 226Ra to calculate unsupported 210Pb. It can also integrate other types of dates as long as they have associated uncertainties. In the majority of cases, 226Ra was measured directly, and we specified that the algorithm assume radon profiles were variable. In one case, no radon was measured, so we assumed the deepest two increments established background unsupported 210Pb.

We set priors for PLUM to be informative enough to impose reasonable constraints on potential accretion and atmospheric 210Pb flux while allowing the profiles to determine the modeled accretion rates. Mean accretion was set to 2.86 years per cm and was informed from a review of 210Pb-based accretion rates70 that are independently held out from an existing disaggregated data synthesis25. We set the accretion prior to be lightly informative by setting the shape to 1. We set the memory prior using a 0.66 mean and strength of 10, indicating that any potential depth-wise autocorrelation is possible, but we assume it is more likely that we observe high autocorrelation given that tidal wetlands accrete over long periods due to changes in relative sea-level. We modeled atmospheric deposition of 210Pb using a global synthesis assembled by Zhang et al.71, using ordinary kriging72,73 to estimate the mean and standard error of log-transformed average annual 210Pb fallout at core locations. We then used the mean and a calculation of shape (mean/variance) as priors for 210Pb fallout. For the rest of the priors, we used PLUM generic inputs.

We calculated carbon accumulation rates from the carbon density depth profiles and PLUM age-depth models, using the raw Markov Chain Monte Carlo (MCMC) output of PLUM to propagate uncertainty in accretion. We discarded the initial 2,000 iterations of the MCMC and stored an additional 10,000 iterations. For each iteration of PLUM MCMC, we omitted all ages older than 100 years relative to the surface, assuming that the age-depth model becomes unreliable when older than that. For each MCMC iteration, we calculated carbon stock down to the bottom of the age-depth profile and divided carbon stock by the age of the bottom of the profile relative to the surface age.

We estimated the mean accretion rate of this dataset to be 3.40 mm yr−1. Accretion rates from individual cores ranged from 2.41 mm yr−1 to 3.97 mm yr−1 per year. In and south of San Francisco Bay, mean estimated accretion was higher than long-term relative sea-level rise over the instrumental period (2.22 mm yr−1 in San Diego and 1.98 mm yr−1 in San Francisco, CA)74. For three cores from Humboldt Bay, accretion was lower than relative sea-level rise over the instrumental period (5.04 mm yr−1 at North Spit, CA)74.

Mean uncertainty in accretion, expressed as a critical value (standard deviation/mean * 100%), was 17.4%. Uncertainty in the modeled accretion rate could arise from having too few dated samples, non-contiguous measurements, or an incompletely observed stock of supported 210Pb.

We estimate carbon accumulation rates to range from 39.2 to 130.0 gC m−2 yr−1 (Fig. 5). The mean estimate from our dataset (102.1 gC m−2 yr−1) is not markedly different from the default IPCC value for estimating carbon removals from rewetted tidal wetlands (91 gC m−2 yr−1)61, a figure which originates from a separate review of dated salt marsh sediment cores. It was slightly higher than carbon accumulation rates calculated from other 210Pb dated cores in the San Francisco Bay region of California (78.4 gC m−2 year−1) originating from 28 cores presented by Callaway et al.4 and three cores presented for Eden Landing by Arias-Ortiz et al.15 (2021). 95% credible intervals overlapped default IPCC removals factors for 9 out of 14 cores, and the California mean value for 6 out of 14 cores (Fig. 5).

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Carbon accumulation rates were calculated for a subset of 14 cores for which 210Pb and 137Cs profiles were present and passed all data quality control steps. We present mean and 95% credible intervals from probabilistic calculations, the dataset mean, the IPCC global default (Tier I) removals factor for tidal marshes remaining marshes, and the mean of reported 210Pb-based carbon accumulation rates from two previous studies from the San Francisco Bay region.

Data Records

The dataset presented is titled “Tidal Wetland Soil Carbon Accumulation Rates for Coastal California”75. It is hosted on Smithsonian Libraries’ FigShare and is formatted by modifying community-generated data entry templates and controlled vocabulary put forward by the Coastal Carbon Network24.

The data release contains a hierarchy of data related to wetland soil carbon profiles. The data itself is housed in four separate .csv files, which can be joined by the core_id and/or site_id attributes. Derivative calculations are kept separate from core and depth series tables and can be recreated with an R-code workflow.

  • brown_et_al_2026_metadata - Definitions and units of attributes and variables within each table.

  • brown_et_al_2026_materials_and_methods.csv - Contains information on materials and methods for this study, formatted so that it can be readily compared to others formatted using the same template.

  • brown_et_al_2026_cores.csv - Positional and descriptive information on coring locations.

  • brown_et_al_2026_depthseries.csv - Original measurements for the sampling intervals of each core.

  • brown_et_al_2026_species.csv - Information on the dominant plant species at coring locations.

  • brown_et_al_2026_site.csv - Site descriptive information.

  • brown_et_al_2026_derivative_values.csv - Carbon stock and carbon accumulation rate information calculated from other tables.

  • brown_et_al_2026_associated_publications.bib - Bibliographic information for publications associated with this data release.

  • CA Pb 210 Eval.csv - Notes from visual inspection of 210Pb profiles.

  • analysis - A directory containing a scripted workflow for analyzing carbon stock data, preparing derivative statistics, and creating data visualizations.

The dataset contains 83 cores and soil depth profiles, from 15 tidal wetland sites (Fig. 1). All sampling locations were naturally occurring, intact wetlands. None were receiving active management of water levels, nor had a significant history of extraction, diking, dredging, or replanting and restoration. Two tidal wetland sites occurred within the San Francisco Bay Region, and 14 were from coastal California. Collection years ranged from 2012 to 2017.

The final dataset contains 5,436 individual measurements of both dry bulk density and loss on ignition. 362 samples from 41 cores were measured for 137Cs activity, 327 samples from 38 cores contain total 210Pb measurements, most with the addition of supported 210Pb (226Ra), and 154 samples from 63 cores have been 14C dated.

We observed a range of modern-day ecohydrological conditions. 48% of cores do not specify a general flooding zone. However, 15.7% are specified as having originated from high-elevation marsh, 1% from high to mid elevation, 12% from mid elevation marsh, and 22.9% from low-elevation marsh. Six major plant species were dominant at coring locations, including Salicornia pacifica, Distichlis spicata, Spartina foliosa, Spartina alterniflora, Frankenia grandifolia, and Jaumea carnosa.

Technical Validation

To quality control the dataset, we used a mixture of data visualization and automated statistical testing and filtering. First, we performed automated outlier filtering for loss-on-ignition by cores, omitting points that fall 1.5 times the interquartile range higher or lower than the 75th and 25th percentiles for a core (Fig. 2). Data visualizations that aided in detecting and correcting data entry errors included histograms of loss-on-ignition and dry bulk density, depth profiles of bulk density, loss on ignition, and radioisotopes (Fig. 2). We also visualized the relationship between loss-on-ignition and dry bulk density as it is well-constrained for tidal wetland soils5.

We performed both statistical and visual assessments of the data before calculating accretion rates. For 137Cs dates, we omitted any profiles that did not meet all three criteria under the scoring discussed in section 2.4. Of the 41 cores dated for 137Cs, 10 had statistically significant peaks representing 1963. For 210Pb data, we evaluated depth profile data using a methodology review by Arias-Ortiz et al.76 as a guide. We omitted any profiles that were likely to be indicative of past disturbance, that were ‘step-shaped’ or had ‘low activity’ for both 210Pb and 226Ra. These profiles are listed in a file in the workflow titled ‘CA Pb 210 Eval.csv’. Fourteen cores were classified as having profiles indicative of undisturbed conditions matching the assumptions of the CRS model.

Usage Notes

It is important to note that the carbon accumulation rates presented here are not directly equivalent to the net ecosystem carbon balances. Our calculation of carbon accumulation rates ignores down-core trends in decomposition. It is probably more appropriate to refer to this measure as a long-term apparent rates of carbon accumulation (LARCA) as opposed to true rates of carbon accumulation, a distinction that is made more often in the peatland literature77.

The observed apparent decrease in organic matter content with depth could be representative of two possible scenarios, each of which could affect overall projections of the strength of the tidal wetland carbon sink in opposite ways. First, the downcore decrease in organic content could be due to decomposition of organic matter during burial. If this is the case, then this would suggest that LARCA overestimates net ecosystem carbon balance because the observed carbon accumulation incorporates both slowly decaying organic matter, which is effectively removed from the system over centennial scale time scales, as well as organic matter pools such as live roots and faster decaying organic matter, which will be returned back to the atmosphere relatively quickly. Second, alternatively, the apparent downcore decrease in organic matter content could reflect a recent increase in organic matter burial. Some studies indicate that accelerating sea-level rise may also be increasing carbon stocks44 and net ecosystem carbon balance78. If the downcore decrease in organic matter content reflects a recent increase in organic matter burial, then LARCA could be underestimating the strength of the tidal wetland carbon sink.

Caveats aside, these results offer a first-order assessment of carbon burial rate, and future studies could integrate core data with process models to bridge the theoretical gap between these observations and carbon cycle processes. The dataset presented here has utility for estimating baseline carbon stocks and accumulation rates. Uses include developing scenarios for tidal wetland restoration projects15, inventorying carbon fluxes on managed lands79,80, and fitting models of soil carbon accretion to forecast future wetland elevation under rising sea levels16,81.