Quantification of blue carbon storage in seagrass meadows across different management strategies in Lamu, Kenya

Abstract

Seagrass ecosystems play a crucial role in coastal carbon dynamics, yet their contribution to soil organic carbon (SOC) storage is underexplored in many areas and especially in African waters. This study assessed seagrass cover, species composition, and SOC stocks in the Lamu Archipelago, Kenya, comparing Marine Protected Areas (MPAs), Locally Managed Marine Areas (LMMAs), and unmanaged sites. Additionally, it investigated the relationships between management stratgeies and pertinent physical and chemical parameters. SOC was measured using the Loss on Ignition method and expressed as Mg C ha⁻¹. SOC stocks were higher in MPAs (89.16 ± 30.88 Mg C ha⁻¹) than in unmanaged areas, while SOC stocks in LMMAs were comparable to those in MPAs. While the MPA was associated with higher SOC stocks, these patterns may also reflect underlying environmental differences among sites, including sediment texture. SOC variability was strongly associated with sediment type, canopy cover, and disturbance intensity, with fine-grained sediments, particularly clay, linked to enhanced carbon retention. Higher SOC stocks co-occurred with the presence of large, persistent seagrass species such as Thalassodendron ciliatum and Enhalus acoroides in managed areas. This study provides the first field-based SOC estimates for Lamu’s seagrass meadows, highlighting the potential of LMMAs within broader blue carbon conservation strategies.

Introduction

Seagrass meadows are among the most extensive and ecologically important coastal ecosystems, covering an estimated area of 177,000 to 600,000 km² globally^1–3. They support marine biodiversity, stabilise shorelines, enhance water quality, and provide critical habitat for fisheries^4–6. As ecosystem engineers, seagrasses influence sediment dynamics, nutrient cycling, and carbonate chemistry^7–9, making them foundational to the function and resilience of shallow coastal systems. Seagrasses also play a vital role in carbon sequestration. They are recognised, together with mangroves and salt marshes, as Blue Carbon Ecosystems (BCEs) which are coastal habitats capable of capturing and storing significant quantities of atmospheric carbon dioxide (CO₂) in both plant biomass and sediments^10,11. Despite occupying less than 0.2% of the global ocean surface, seagrass meadows contribute approximately 10–18% of annual oceanic organic carbon burial and store ~ 7.4 Tg C annually^12–14. Their carbon storage capacity, particularly in anoxic sediments, can equal or exceed that of terrestrial forests on a per-area basis^2,10,15. Recent estimates suggest that global seagrass carbon stocks range from 15,372 to 30,744 Mt CO₂e¹⁰, some single-species assessments demonstrate substantial storage capacity; for example, Cymodocea nodosa alone accounts for ~ 29.5 Mt CO₂e in Spanish waters¹⁶.While this species is not present in the Western Indian Ocean, the genus Cymodocea occurs widely in the region, highlighting the relevance of species-level variability when assessing seagrass carbon stocks. Despite this potential, seagrasses remain underrepresented in blue carbon research and policy, with most global BCE restoration initiatives and carbon stock assessments focusing predominantly on mangroves¹⁷. This underrepresentation is partly attributable to the high spatial variability of seagrass extent, species composition, and sediment carbon stocks, which complicates standardised assessment compared to more structurally uniform coastal ecosystems^18,19. Additional constraints such as limited and inconsistent research funding, infrastructure, and technical capacity further restrict sustained data collection, long-term monitoring, and the development of robust carbon stock estimates, despite strong potential for locally led and collaborative research initiatives.

Seagrass meadows are currently declining at ~ 1.5% per year¹², with over 29% globally degraded and ~ 14% of species are at risk of extinction^18,20,21. This loss converts seagrass systems from long-term carbon sinks into carbon sources, releasing remineralized carbon into the atmosphere^2,12. The drivers of degradation are diverse and often act synergistically, including nutrient loading, sedimentation, coastal development, physical disturbance, overfishing, invasive species, and climate change^18,22. In East Africa, these pressures are amplified by growing coastal populations and direct human activities such as pollution, eutrophication and sedimentation from port construction, intensive fishing, and mollusc collection for tourism²³. These activities reduce water clarity and limit the depth at which seagrasses can grow. Additionally, high densities of sea urchins, which graze on seagrass, further constrain recovery²⁴. Efforts to conserve and restore seagrass meadows are increasing, as well as the recognition of their ecosystem services and role in supporting coastal livelihoods^6,25. However, large areas of the world’s seagrass habitats remain unmapped³, and most carbon stock estimates are limited to Mediterranean and Australian systems¹⁷. Africa, in particular, is underrepresented in seagrass blue carbon research, thus likely to lead to inaccurate regional carbon stocks and accumulation baselines^2,10.

Kenya formally included BCEs in its Nationally Determined Contributions (NDCs) during the 2020 revision of its 2016 framework²⁶. However, implementation has been hindered by a lack of empirical data on seagrass extent and carbon storage capacity. While studies in the Lamu Archipelago have assessed mangrove carbon stocks, only one has reported on seagrass status²⁵ and none, to our knowledge, have quantified soil carbon storage. By comparing seagrass diversity, distribution, and below-ground carbon storage, soil organic carbon stores (SOC) in sediments, across these contexts, we provide the first field-based estimates of SOC in Lamu’s seagrass meadows. In Lamu, seagrass meadows occur under contrasting governance systems, ranging from fully protected Marine Protected Areas (MPAs) to community-led Locally Managed Marine Areas (LMMAs) and open-access, unmanaged sites. Management strategies differ in their objectives, enforcement mechanisms, and patterns of human use, which may indirectly result in the variability of seagrass carbon storage through their effects on disturbance intensity, sediment stability, and meadow structure. MPAs are designated under national legislation and enforce no-take regulations with restrictions on extractive activities, anchoring, and coastal development. In contrast, LMMAs are governed by local communities under co-management arrangements, allowing regulated resource use while maintaining conservation objectives. Although LMMAs are not designed explicitly for carbon conservation, reduced destructive practices, social enforcement, and rapid local responses to disturbance may help maintain sediment integrity and seagrass structure. Unmanaged areas, by comparison, are more frequently exposed to chronic anthropogenic pressures such as trampling, anchoring, and unregulated harvesting, which can disrupt sediments and mobilise stored carbon. These contrasting management strategies provide a natural experiment to test how different levels of protection and human use are related to variation in seagrass structure and soil organic carbon (SOC) storage.

Therefore, this study assessed seagrass abundance (canopy cover and shoot density) and species diversity (richness) of seagrasses, and quantified the soil organic carbon stocks in the seagrasses of Lamu under different management strategies. We hypothesised that seagrass structural attributes, species diversity, and SOC stocks would be higher in MPAs and LMMAs than in unmanaged areas, recognising that outcomes are likely to depend on the specific management objectives, regulations, and enforcement of each MPA. This is based on the expectation that the non-extractive use and access to MPAs would reduce exposure to physical disturbances (e.g., anchoring, trampling, sediment resuspension), enhances sediment stability, and supports greater seagrass canopy cover, all of which promote organic matter accumulation and carbon retention in sediments. Our study addresses the following research questions:

1. Does soil organic carbon differ significantly in seagrass meadows found in marine protected areas, locally managed areas and unmanaged areas?

2. What environmental and/or ecological variables best explain the variation in SOC stocks across sites?

Results

Environmental conditions

Environmental characteristics varied across sites (Table 1). The LMMA exhibited moderate disturbance and mixed sediment types—52.4% sand and 47.6% loam. It supported the highest proportion of polyspecific meadows (4.4%), with an average water temperature of 32.0 ± 0.65 °C (mean ± SD) and a mean depth of 1.01 ± 0.48 m. The MPA experienced low disturbance and was dominated by fine sediments, including 42.1% clay and 36.8% loam. It had the lowest water temperature (31.3 ± 1.1 °C) and an average depth of 1.1 ± 0.70 m. In contrast, the Unmanaged Area was subject to high disturbance, was characterised by sand-dominated sediments, the shallowest average depth (0.65 ± 0.68 m), and the highest water temperature (32.5 ± 1.17 °C). Sediments classified as 100% sand reflect samples in which particle size analysis indicated negligible silt and clay fractions below analytical detection limits. Monospecific meadows were present in all sites but were most prevalent in the Unmanaged Area (66.7%).

Table 1.

Summary of biophysical and environmental variables across study sites (upper and lower estimates are mean ± SD and categorical variables presented as percentages).

	LMMA	MPA	Unmanaged Area
Water Temperature (°C)	32 ± 0.65	31.3 ± 1.10	32.5 ± 1.17
Water Depth (m)	1.01 ± 0.48	1.1 ± 0.70	0.65 ± 0.68
Disturbance Level	moderate	low	high
Soil Type
sand	52.40%	21.10%	100%
loam	47.60%	36.80%	-
clay	42.10%	-	-
Meadow Type
monospecific	65.20%	54.20%	66.70%
oligospecfiic	30.40%	41.70%	33.30%
polyspecific	4.40%	4.20%	0.00%

Seagrass meadow characteristics

On average, Lamu’s overall seagrass cover was 61.26 ± 23.01% (mean ± SD) of the sampled area. This indicates that seagrass occupies a substantial portion of the surveyed areas, with considerable variability across locations. Six out of the possible twelve seagrass species found in Kenya were recorded in the study area, namely Cymodocea serrulatta (C. serrulatta), Enhalus acoroides (E. acoroides), Halodule uninervis (H. uninervis), Syringodium isoetifolium (S. isoetifolium), Thalassodendron ciliatum (T. ciliatum), and Thalassia hemprichii (T. hemprichii). The most dominant species across the study area was T. ciliatum, which recorded the highest average percentage cover (50.42 ± 25.76%), and T. ciliatum recorded the highest shoot density (84.83 ± 116.04 shoots/m²) (Table 2). Overall, seagrass shoot density was an average of 810 ± 1098 shoots/m².

Table 2.

Mean shoot density and canopy cover for seagrass species in Lamu across all surveyed sites (± SD).

Species	Cover (%)	Density (shoot/m2)
Cymodocea serrulatta	32.55 ± 22.85	78.91 ± 50.50
Enhalus acoroides	40.53 ± 24.12	52.23 ± 153.20
Halodule uninervis	39.67 ± 28.80	41.64 ± 102.60
Syringodium isoetifolium	20 ± 21.15	18.39 ± 63.56
Thalassodendron ciliatum	50.42 ± 25.76	84.83 ± 116.04
Thalassia hemprichii	31.11 ± 24.81	56.39 ± 94.37

Seagrass cover and density varied between different meadows and sites, with T. ciliatum exhibiting the highest relative abundance in the MPA (56.0%). In comparison, T. hemprichii exhibited the highest relative abundance in the Unmanaged Area (46.36%) (Fig. 1). T. hemprichii, E. acroides and T. ciliatum were recorded in all three locations, while S. isoetifolium was recorded only in the MPA. All sites exhibited multi-species assemblages, but the differences between sites were not statistically significant (Kruskal-Wallis = 1.19, df = 2, p = 0.55).

Fig. 1.

Relative abundance of seagrass species across the three study sites. Proportions are based on total canopy cover per species within each management strategy.

Seagrass canopy cover exhibited distinct patterns across different depths and management zones (Fig. 2). The shallow zone (0–1 m) revealed marked variability across the sites: the LMMA exhibited significantly higher and more consistent seagrass cover compared to the MPA (p = 0.0001) and the Unmanaged Area (p = 0.0063). In contrast, seagrass cover in the mid zone (1–2.5 m) was relatively uniform across all sites, with no statistically significant differences detected (p > 0.7). In the deep zones (> 2.5 m), seagrass cover was higher in the LMMA compared to the MPA. No seagrasses were found in the deep zones of the Unmanaged Areas.

Fig. 2.

Boxplots showing total seagrass cover (%) across management strategies (LMMA, MPA, and Unmanaged Area) stratified by depth zone (Deep, Mid, and Shallow). Significant pairwise differences (Wilcoxon test) are annotated with p-values.

Although variability was also exhibited by seagrass shoot density across the sites, the difference was not statistically significant (Kruskal-Wallis = 0.15, df = 2, p = 0.93). Relative abundance patterns showed clear differences in species composition among management regimes (Fig. 3). Thalassodendron ciliatum dominated shoot density in unmanaged and LMMA sites, accounting for 46.42% in LMMAs. In contrast, MPA sites showed a more even distribution among species, with two dominant species, T.ciliatum and T. Hemprichii. Although T. Hemprichii remained the major contributor in the MPA (27.85), overall species contributions were more evenly distributed compared to other sites (Fig. 3). The Unmanaged Area supported fewer seagrass species and showed strong dominance of T.ciliatum.

Fig. 3.

Relative abundance of seagrass species across the three study sites. Proportions are based on total shoot density per species within each management strategy.

The Shannon-Wiener Species Diversity Index (H) and Evenness Index (E) were used to gain further insights into the biodiversity and distribution of seagrasses. Seagrass diversity was the highest in the LMMA (H = 0.36 ± 0.34; E = 0.51 ± 0.43), which indicates an evenly distributed and species-rich seagrass community. In contrast, the LMMA and Unmanaged Areas showed reduced species richness and increased species dominance under no management (H = 0.19 ± 0.30, E = 0.28 ± 0.41).

Soil organic carbon characteristics

The soils in the seagrass meadows mainly consisted of clay (17.4%), loam (34.8%), and sand (47.8%). Soil types varied by site: LMMA consisted of nearly equal parts loam (47.62%) and sand (52.38%), the MPA was composed of a mix of clay (42.11%), loam (36.84%), and sand (21.05%), and the Unmanaged Areas consisted exclusively of sandy soils. Dry Bulk Density (DBD) was highest in the Unmanaged Area (1.52 ± 0.20 g cm⁻³), lower in the LMMA (1.27 ± 0.21 g cm⁻³), and comparable between LMMA and MPA (p = 0.94). SOC stocks were highest in the MPA (89.16 ± 30.88 Mg C ha⁻¹), intermediate in the LMMA (68.58 ± 19.96 Mg C ha⁻¹), and lowest in the Unmanaged Area (39.56 ± 22.53 Mg C ha⁻¹). Carbon density followed a similar trend: highest in the MPA (0.018 ± 0.006 g cm⁻³), intermediate in the LMMA (0.014 ± 0.003 g cm⁻³), and lowest in the Unmanaged Area (0.008 ± 0.005 g cm⁻³). Organic matter was lowest in the Unmanaged Area (0.57 ± 0.43%OM), higher in the LMMA (1.18 ± 0.35%OM), and highest in the MPA (1.45 ± 0.58%OM), though this increase in the MPA was not statistically significant (Table 3). When averaging across all surveyed sites, mean SOC stocks in the top 30 cm of seagrass soils were 74.3 ± 27.9 Mg C ha⁻¹, ranging from 20.16 to 135.73 Mg C ha⁻¹.

Table 3.

Results of the ANOVA used to test for significant effects of site (LMMA, MPA, Unmanaged Area) on bulk density, SOC, carbon density, and organic matter in Lamu.

Variable	Factor	F	df	p-value
Dry bulk density (g cm− 3)	Site	3.752	2	0.032
SOC stocks (Mg C ha− 1)	Site	10.623	2	0.00018
Carbon Density (g cm− 3)	Site	10.623	2	0.00018
Organic Matter (%OM)	Site	8.129	2	0.001

Sediment properties varied significantly across site types (Table 3). ANOVA results showed significant effects of different sites on bulk density (p = 0.032), soil organic carbon (SOC; p = 0.00018), carbon density (p = 0.00018), and organic matter (p = 0.001) (Table 3). Post hoc tests revealed that unmanaged areas had significantly lower SOC than LMMAs (p = 0.0191) and MPAs (p = 0.00069), and similarly lower carbon density (p = 0.0124 and p = 0.00013) and SOC stock (p = 0.0124 and p = 0.00013). Bulk density was significantly higher in unmanaged areas compared to LMMAs (p = 0.0333) and MPAs (p = 0.0403). No significant differences were found between MPA and LMMA for any variable (p > 0.05). SOC stocks differed across sites, with a Kruskal-Wallis test approaching significance (p = 0.05394). To aid interpretation of this near-significant result, exploratory pairwise Wilcoxon rank-sum tests were conducted. These analyses indicated that SOC stocks were lower in unmanaged areas than in LMMAs and MPAs, while no significant difference was detected between LMMAs and MPAs (Fig. 4). P-values were adjusted for multiple comparisons.

Fig. 4.

Comparison of soil carbon stocks (mg C ha-1) among seagrass sites (LMMA, MPA, and Unmanaged Area). Boxplots show medians, interquartile ranges, and individual data points. Wilcoxon pairwise comparisons reveal significant differences.

We used a general linear model (multiple linear regression) to assess the influence of several environmental and biological variables (soil type, disturbance, meadow type, total shoot density, percentage cover, water depth, and water temperature) on SOC stocks in Lamu (Table 4). An ANOVA was applied to the fitted model to evaluate the significance of each predictor. The results show that soil type was the strongest predictor of SOC (p = 3.60 × 10⁻¹⁵). Disturbance (p = 0.011), water temperature (p = 0.00382), and percentage cover (p = 0.00289) were also associated with statistically significant differences in SOC stocks, although they explained a smaller proportion of the observed variation compared to sediment texture. Meadow type, shoot density, and water depth were not significant predictors.

Table 4.

Summary of ANOVA results from a multiple linear regression model assessing the effect of biophysical and environmental variables on soil organic carbon (SOC) stocks in seagrass meadows of Lamu. Significant predictors (p < 0.05) are shown in bold. Soil type, disturbance level, percentage cover, and water temperature were significant drivers of SOC variability.

Parameters	Df	Sum Sq	Mean Sq	F value	p
Soil Type	2	26,127	13,064	93.166	3.60E-15
Disturbance	1	1002.44	1002.44	7.149	0.011
Meadow Type	1	14	14	0.097	0.75677
Total Density	1	27	27	0.196	0.66078
Percentage Cover	1	1428	1428	10.181	0.00289
Water Depth	1	17	17	0.118	0.73267
Water Temperature	1	1336	1336	9.527	0.00382
Residuals	37	5188	140

Discussion

This study provides the first field-based assessment of soil organic carbon (SOC) stocks in seagrass meadows across contrasting management contexts in the Lamu Archipelago, Kenya. SOC stocks varied substantially among sites, with higher values observed in Marine Protected Areas (MPAs) and Locally Managed Marine Areas (LMMAs) compared to unmanaged areas. However, these differences should be interpreted as associations rather than direct causal effects of management, given the absence of baseline data on management duration and pre-protection conditions. SOC variability was strongly associated with sediment type, canopy cover, and disturbance intensity, while shoot density and water depth had limited explanatory power. Notably, the MPA site was characterised by a higher proportion of clay-rich sediments, which are known to enhance organic carbon retention through greater surface area and particle binding capacity. Such pre-existing sedimentary characteristics may have predisposed this site to higher SOC storage irrespective of protection status.

The mean SOC stock recorded in Lamu (74.3 ± 27.9 Mg C ha⁻¹ in the top 30 cm) is lower than global and regional estimates of approximately 140 Mg C ha⁻¹^10,27, many of which are derived from sediment cores extending to 1 m depth. Similarly, higher values reported from the Caribbean (241 Mg C ha⁻¹)^27,28 and West Africa (140.2 Mg C ha⁻¹)²⁹ reflect the inclusion of deeper sediment layers in those estimates. In contrast, our estimates are restricted to the upper 30 cm and were not extrapolated beyond this depth, limiting direct quantitative comparison with 1 m assessments. Notably, even where comparable shallow-depth (0–30 cm) measurements were reported in Gazi Bay, Kenya, SOC stocks remain higher than those observed in Lamu². These findings emphasise the necessity of region-specific baselines and rigorous methodological standardisation, as differences in sampling depth and environmental context can substantially influence SOC estimates and lead to both over- and underestimation of seagrass carbon stocks.

Soil composition varied substantially across the study sites, and this variation was reflected in the measured SOC and bulk density values. The MPA, which exhibited a higher proportion of clay (42.11%), exhibited the highest SOC stock, the highest carbon density, and relatively low bulk density. The LMMA, dominated by loam and sand, showed intermediate SOC and carbon density values, consistent with site conditions that favour retention of finer sediments (e.g., lower hydrodynamic energy)². In contrast, the Unmanaged Area, composed exclusively of sandy soils, exhibited the lowest SOC stock (39.56 ± 22.53 Mg C ha⁻¹), the lowest carbon density (0.008 ± 0.005 g cm⁻³), and the highest bulk density (1.52 ± 0.20 g cm⁻³). This pattern is consistent with findings from Marbà et al. (2015)³⁰, who reported low carbon storage in sandy seagrass meadows due to rapid drainage and reduced capacity to trap organic particles. Notably, organic matter content was also significantly lower in unmanaged areas, contradicting earlier assumptions of uniform organic input quality across seagrass systems³¹. This suggests that both the quantity of organic material in sediments vary with site-level disturbance and environmental context, rather than protection status alone.

Biophysical variables such as water depth, sediment texture, and temperature are intrinsically linked to disturbance regimes through their influence on hydrodynamic exposure, sediment stability, and human accessibility, all of which affect sediment resuspension and carbon retention^10,31,32. The qualitative disturbance classification therefore serves as an integrated proxy for chronic physical and anthropogenic pressure, rather than a standalone driver, reflecting the combined influence of environmental context and human use^33,34. Environmental conditions varied across sites, with the LMMA supporting more diverse soil types and meadow structures under moderate disturbance. In contrast, the MPA and Unmanaged Areas showed reduced soil and meadow complexity, particularly where disturbance levels were high. These differences are consistent with site-level environmental conditions and lower chronic disturbance often observed in areas under community-led stewardship, where local rules and social enforcement can reduce destructive use^35,36. Local stewardship typically reduces destructive harvesting and improves compliance through social enforcement, while enabling more rapid responses to disturbances³⁷. These processes collectively minimise sediment disruption and canopy damage, key drivers of seagrass degradation, allowing heterogeneity in soil composition and meadow structure to persist^34,38.

Among abiotic drivers, soil texture emerged as the most significant determinant of SOC variability, consistent with findings from other coastal systems^39–41. These results align with earlier studies, indicating that sediment characteristics strongly predict SOC stocks^33,39,41. Recent work emphasises that finer-grained sediments, which have higher porosity and particle surface area, are associated with elevated organic carbon due to enhanced organic matter binding and reduced decomposition^42,43. Similarly, hydrodynamic exposure gradients have been shown to influence carbon accumulation, with more sheltered settings supporting higher sedimentary C_org²⁸. The MPA was composed predominantly of clay-rich sediments. Soils with fine textures, such as those rich in clay and silt, typically contain more organic matter than sandy or coarse-textured soils due to their higher porosity, surface area, and organic matter-binding capacity^39,41. Organic matter in sands is often below 1%, loam soils usually range from 2% to 3%, and clays can contain between 4% and over 5%⁴⁴. Higher SOC in the MPA co-occurred with a greater proportion of fine sediments and lower bulk density, consistent with established relationships between sediment texture and carbon retention^39,41,45,46. Overall, these patterns align with previous studies showing sediment properties as strong predictors of SOC stocks in seagrass meadows^33,42,43.

Canopy cover was positively associated with SOC, consistent with prior work linking denser canopies to sediment retention and organic matter accumulation⁴⁷. The lowest SOC stocks coincided with the lowest canopy cover in Kipungani, consistent with the broader role of vegetated coastal habitats in coastal carbon storage. Duarte et al. (2005)¹⁴ found that in areas where seagrasses have been lost due to anthropogenic destruction, more than 25% of the CO₂ sink capacity of these ecosystems has been lost. These findings are consistent with those of Mazarrasa et al. (2015)⁴⁸, who found strong associations between seagrass canopy complexity and sediment carbon storage. However, our results contradict those reported in Belize, where no relationship was found between percentage cover and SOC stocks²⁸. It is worth noting that this study used remote sensing techniques to map the extent of seagrasses and reported that over half of the Central Lagoon’s seagrass cover was partially obscured by high turbidity (55%), which made it difficult to determine the percentage canopy cover remotely²⁸. While Lamu’s seagrasses cover a substantial area of 317 km², the results highlight regional variability, where factors such as species composition, environmental conditions, and anthropogenic impacts may influence SOC values.

The co-occurrence of higher SOC stocks with morphologically large, persistent species such as Enhalus acoroides and Thalassodendron ciliatum, characterised by large leaf blades, extensive below-ground biomass, and long-lived rhizome systems, is consistent with previous studies linking these species to greater sediment retention and organic matter accumulation^49,50. Additionally, Githaiga et al. (2017)² reported species-specific variations in SOC stocks in Gazi Bay, Kenya, within the same seagrass species assemblages observed in our study, with E. acoroides exhibiting the highest levels of SOC stocks. According to Rifai et al. (2021)⁵¹, E. acoroides is morphologically large, dominant, and possesses habitat characteristics suitable for soil accumulation because it is a large and persistent species⁴. Recent studies that explicitly incorporate above-ground biomass measurements demonstrate that large-bodied seagrass species, characterised by high canopy height and substantial leaf biomass, play a disproportionate role in sediment organic carbon storage by enhancing hydrodynamic attenuation, sediment trapping, and organic matter retention^42,52. While polyspecific meadows generally show higher SOC levels, mono-specific meadows can also contribute significantly to carbon sequestration^10,53. For example, studies of Mediterranean seagrass systems show that Posidonia oceanica monospecific meadows sustain long-term carbon burial rates ranging from 17 to 191 g C m⁻² per year, forming organic matter layers that can persist for millennia¹⁰. However, not all studies are consistent with our findings. For instance, Stankovic et al. (2018)¹² reported higher SOC stocks in uniform seagrass meadows than in mixed-species meadows, within a system dominated by functionally similar seagrass species. Similarly, Felgate et al. (2024)²⁸ found that species composition weakly influenced SOC stocks in Belize, where the dominant species was Thalassia testudinum. These contradictions suggest that factors beyond seagrass diversity, such as environmental conditions and meadow structure, significantly determine SOC stocks^39,40,54.

Disturbance significantly influenced SOC stock in seagrasses^52,55. Disturbances, such as boat anchor and propeller damage and herbivore grazing, can destroy seagrasses, resulting in the loss of carbon stored within plant structures^33,56. Disturbances can also decrease the filtering efficiency of seagrasses, thereby reducing the suspended allochthonous carbon that contributes to the below-ground sediment carbon stock⁵⁷. Furthermore, anthropogenic activities affect the carbon sink capacity of seagrasses through eutrophication and sedimentation, which both result in the reduction of light⁴⁰. Unmanaged Areas showed markedly lower SOC and reduced seagrass cover. These results support previous work showing that disturbances reduce sediment accretion, increase erosion, and lead to the export of stored carbon^33,57. In contrast, LMMAs showed SOC levels comparable to MPAs, an association that may reflect both governance context and site-level environmental differences^58,59.

While total SOC stocks, carbon density, and bulk density differed significantly across sites, organic matter concentration (%OM) did not. This finding indicates that the proportion of organic material per unit soil mass was relatively consistent, regardless of protection status. However, MPAs and LMMAs still stored significantly more carbon overall, reflecting their greater capacity for long-term carbon accumulation. These higher stocks may reflect differences in organic matter inputs, sediment stabilisation, and disturbance regimes rather than differences in organic matter concentration alone^33,48. Although dry bulk density was significantly different across sites, higher carbon stocks in MPAs and LMMAs were not necessarily associated with lower soil compaction, suggesting that multiple, interacting factors shape carbon storage. Higher carbon stocks in managed contexts highlight the importance of ecosystem-level processes—such as productivity, sediment retention, and below-ground biomass dynamics— that may differ between managed and unmanaged areas^2,41. This emphasizes the value of both fully protected and community-managed seagrass areas in supporting blue carbon outcomes^50,51. Moreover, these results align with those from earlier studies, which indicated that sediment characteristics strongly predict SOC stocks^33,39,41.

In contrast, meadow type, total density and water depth did not significantly affect the surveyed sites. This was unexpected, as several studies have reported shoot density and shallow water depth as important controls on below-ground biomass and sedimentation processes^2,60. For example, Githaiga et al., (2017)² found that Syringodium isoetifolium, which exhibited substantially higher shoot density and canopy cover than co-occurring species, supported higher below-ground biomass than even the larger species in Gazi Bay, Kenya. Recent studies have similarly linked seagrass density and meadow structure to sediment carbon storage and stability^42,52. However, higher below-ground biomass does not necessarily translate directly to higher sediment SOC stocks, particularly where sediment properties and disturbance regimes exert stronger control. In this respect our results are consistent with those of Stankovic et al. (2018)¹², who similarly reported no significant influence of seagrass density on carbon storage within seagrass sediments. Although water depth has been reported as a key driver of SOC in other studies— often with higher SOC observed at intermediate or shallow depths due to reduced hydrodynamic energy and enhanced sediment retention^17,42,56—it was not significant in our analysis. This is likely due to the narrow depth range across sampling sites, limiting variability in this parameter as all seagrass meadows surveyed were found in shallow waters (≤ 2.5 m), which may reflect a natural ecological pattern or a sampling bias. The lack of detailed bathymetric and seagrass distribution maps for Lamu further complicates the interpretation by limiting our ability to explicitly assess depth-related patterns in seagrass extent, meadow continuity, and sediment accumulation across the seascape^3,25. Future studies should incorporate a broader depth gradient and remote sensing tools to more fully explore depth-related SOC dynamics.

This study highlights the importance of localised carbon estimates when integrating seagrass ecosystems into climate mitigation strategies, as reliance on global averages risks misvaluing carbon storage and overlooking regional variability relevant to blue carbon finance. In the Lamu Archipelago, seagrass meadows within both Locally Managed Marine Areas (LMMAs) and state-designated Marine Protected Areas (MPAs) were associated with higher soil organic carbon (SOC) stocks than unmanaged sites. These patterns, however, should be interpreted in the context of site-specific environmental conditions and management characteristics, including sediment texture, disturbance regimes, governance arrangements, and enforcement capacity, rather than as evidence of universal management effectiveness. LMMAs, which maintain seagrass diversity and soil organic carbon (SOC) stocks under managed use, offer a promising nature-based solution that aligns with community-driven conservation and sustainable development. This work provides the first site-specific estimates of SOC stocks in the Lamu Archipelago, with a mean value of 74.3 ± 27.9 Mg C ha⁻¹ in the top 30 cm of sediment. While SOC stocks were higher in MPAs and LMMAs compared to unmanaged areas, variability was driven by soil texture, canopy cover, and disturbance level, with minimal influence from species richness, shoot density, and water depth.

These findings indicate that both community-led and government-protected areas are associated with seagrass meadows that retain higher SOC stocks relative to unmanaged sites, while acknowledging that underlying environmental heterogeneity contributes to these patterns. Although SOC values in Lamu are lower than regional and global averages, these meadows represent a potentially important carbon store. These findings emphasize the need for field-based, localized assessments to support blue carbon accounting and regional coastal planning, alongside improved spatial mapping and long-term monitoring. Future research should extend these estimates to deeper sediment layers and biomass pools, integrating high-resolution spatial mapping and management timelines to better disentangle management effects from environmental controls.

Methods

Study area

This study was carried out along the Kenyan coast within the Lamu Archipelago (2.2696 ° S, 40.9006 ° E). Lamu consists of a mainland and over 65 islands that form the Lamu Archipelago, covering a total land surface area of 6 607 km² and has an approximately 130 km long coastline (Fig. 5). It is characterised by coastal lagoons, expansive mangrove forests and seagrass meadows^53,61. The area lies along the equatorial coastal climatic zone characterised by two monsoon winds. The climate is generally warm, with variable precipitation that averages 1000 mm a year. Lamu is characterised by bimodal rainfall: the long rain period between March and May and the short rain period from mid-October to mid-December. The temperatures are usually high, ranging between 23 and 33°C⁶².

Fig. 5.

Map of the Lamu Archipelago showing the location of the study sites. Green polygons represent seagrass meadows mapped in 2016, while coloured dots indicate sampled points within each study site. The map was created using ArcGIS Pro 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Of the 12 seagrass species found in Kenya, nine of them have been recorded in Lamu coastal waters, namely, Cymodocea rotundata, Cymodocea serrula, Halodule uninervis, Halodule ovalis, Syringodium isoetifolium, Zostera capensis, Enhalus acoroides, Thalassodendron ciliatum and Thalassia hemprichii^2,25,49,63. The two commonly found species in Lamu are Thalassodendron ciliatum, followed by Thalassia hemphrichii. The presence and abundance of seagrasses are regarded as indicators of the overall environmental quality of the coastal zone^18,64,65. Lamu is characterised by numerous locally managed marine areas and a marine park in the northern parts of the County, Kiunga Marine National Reserve. These areas aim to improve communities’ livelihoods through conservation and sustainable use of natural resources⁵⁸. Despite this, seagrass cover in Kenya has declined by 0.85% yr^− 1 since 1986, and those in Lamu have suffered the most significant total decline of 1.5% between 1989 and 2016⁶³⁵⁷,.

Site selection

Satellite imagery of potential seagrass habitats around Lamu was accessed using WorldView-2. The habitats identified from the satellite image and the shapefile informed the preliminary site selection of possible seagrass presence. Local knowledge was also integrated to refine the selection, targeting areas known locally for seagrass presence alongside predicted sites. Four sites were selected for sample collection, namely Kipungani, Manda Toto, Pate Island and Mkokoni. These sites were identified based on their accessibility and lower security risk given that certain areas in Lamu are known to have high security concerns, such as north of Kiunga Marine National Reserve. Areas such as Dodori and Mongoni creeks, even though reported to have seagrasses, were excluded from the survey due to navigation challenges. In ArcGIS Pro 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview) each site’s seagrass meadow polygon—derived from the seagrass meadow shapefiles—was first imported. Using the Generate Tessellation tool, a fishnet of 5 km² square cells was overlaid across each polygon. Then, with the Create Random Points tool, one point per cell was generated, constrained by the meadow polygons so that every sampling location fell within the seagrass habitat. This resulted in a total of 47 sampling points across all sites distributed as follows: Kipungani (3), Manda Toto (3), Pate Island (21) and Mkokoni (20). The size and extent of the meadows determined the number of sampling points per site.

Study sites were categorised into three management strategies based on their governance structures and access restrictions: Marine Protected Areas (MPAs), Locally Managed Marine Areas (LMMAs), and Unmanaged Areas. MPAs correspond to gazetted marine parks under Kenyan national law, characterised by full protection with no extractive use permitted⁶¹. LMMAs are community-governed areas that are managed by Beach Management Units (BMU) which are legally recognised under the Fisheries Management and Development Act (2016)^52,62. Unmanaged Areas lack state-designated governance structures and are subject to open-access use. These categories were used as fixed factors in all comparative analyses Sites were further categorised based on their respective exposure to anthropogenic activities referred to as disturbance. Disturbance level was quantified using a semi-quantitative index based on direct field observations of anthropogenic activities at each sampling location. During sampling, the presence or absence of key disturbance indicators was recorded, including (i) boat anchoring or propeller scars, (ii) evidence of trampling or foot traffic, (iii) fishing activity (e.g., nets, traps, gleaning), (iv) proximity to settlements or infrastructure, and (v) sediment resuspension or physical damage to seagrass canopies. Each indicator was scored as present (1) or absent (0), and scores were summed to produce a disturbance index ranging from 0 to 5 for each sampling point. Sampling points were subsequently classified into three disturbance categories: low (index = 0–1), moderate (index = 2–3), and high (index = 4–5). Mkokoni was classified as having low disturbance because it is within Kiunga Marine National Reserve, an example of a conventional Marine Protected Area (MPAs). Marine parks in Kenya are designated as fully protected areas under the Wildlife Conservation and Management Act (2013), where extractive activities are prohibited and access is generally restricted to scientific research, education, and non-extractive tourism.⁶⁶. Pate Island was classified as moderately disturbed because it is located within a locally managed marine area (LMMAs). LMMAs refer to sections of nearshore waters and their associated coastal and marine resources that are primarily overseen by local coastal communities, landowners, partner organisations, or collaborating government officials who live or operate in the immediate vicinity^59,67. Unlike conventional MPAs, LMMAs emphasise local governance, ownership, and usage rights, often incorporating customary tenure systems and traditional management practices specific to the region⁵⁸. Manda Toto was classified as highly disturbed due to its proximity to the Lamu Port-South Sudan-Ethiopia-Transport (LAPSSET) port and the lack of any documented management structure. Kipungani was also classified as highly disturbed due to its exposure to tourist and residential areas. Both sites are herein referred to as Unmanaged Areas.

Seagrass canopy cover and density

Field sampling was conducted in line with the methods outlined by the IUCN Coastal Blue Carbon Manual^32,62 during the southeast monsoon period (Kuzi) associated with heavy rains and rough seas between March and April. Sampling was conducted during low tide to improve visibility and working conditions. In each site, a 50 × 50 cm quadrat was placed randomly on seagrass beds and water depth, temperature, and tidal zone were recorded using a dive computer (Suunto Zoop Dive Computer Scuba Diving Wristwatch). Canopy cover (%) was estimated using visual estimates by counting the number of squares containing seagrass. The species-specific cover was similarly assessed by identifying and counting squares containing specific seagrass species. Seagrass meadows within each site were classified based on their meadow type into either multi-specific (Fig. 6a), or monospecific, containing single species (Fig. 6b). Seagrass species identification was conducted in situ to the lowest taxonomic level using a field manual. Underwater photographs were taken of the quadrat from above and used to refine and confirm canopy cover estimates, as well as to confirm species identification. To estimate shoot density, a smaller 25 × 25 cm quadrat was placed in the same spot as the larger quadrat to count individual seagrass shoots, thus estimating total and species-specific density. Meadow type was classified at the sampling-point (quadrat) level. Each quadrat was classified as monospecific if only one seagrass species was present, oligospecfiic if two species were present or polyspecific more than two species occurred within the quadrat Photographs of each quadrat and its surroundings were taken to verify species identification and to identify other organisms present in the quadrat. Surrounding anthropogenic activities were also recorded.

Fig. 6.

Classification of meadow types. a). Mixed meadow consists of two seagrass species, Cymodocea serrulata and Syringodium isoetifolium. b) Uniform meadow consisting of a single species, Thalassodendron ciliatum.

For analysis, sampling points were categorised into three depth zones based on measured water depth: shallow (0–1 m), mid (1–2.5 m), and deep (> 2.5 m). These depth classes were used to examine depth-related patterns in seagrass cover and SOC stocks.

To calculate species diversity and richness, the Shannon-Weiner diversity index (H) and index for evenness (E) were used, as given by Eq. (1) and Eq. (2):

1

Where,

H; species diversity index, s; the number of species; and P_i; the proportion of individuals of each species belonging to the i^th species of the total number of individuals and,

2

Where,

E; Species evenness, H; Shannon-Wiener index; and S; Species richness.

While these indices are widely applied in seagrass ecology, their sensitivity is constrained in systems with low species richness, such as the Lamu Archipelago where only six seagrass species were recorded. In low-diversity assemblages, Shannon-based metrics are more strongly influenced by relative abundance and evenness than by species richness, which can reduce their ability to detect subtle differences among sites or management strategies. Consequently, observed similarities in diversity indices across management strategies should be interpreted with caution and do not necessarily imply ecological equivalence. However, these indices remain useful for characterising community structure and dominance patterns, particularly when interpreted alongside structural attributes such as canopy cover and shoot density, which were more responsive to management effects in this study.

Estimating soil organic carbon (SOC)

Soil organic carbon (SOC) stocks were quantified for the upper 30 cm of seagrass sediments following the IUCN Blue Carbon Manual and the framework of Fourqurean et al. (2012). No extrapolation to 1 m depth was applied, as vertical SOC distributions can be highly variable and site specific. Consequently, comparisons with global estimates are made cautiously and limited to studies reporting SOC stocks over comparable depth intervals. SOC stocks were calculated by combining measurements of sediment dry bulk density, organic carbon concentration, and sediment depth, and extrapolated from point-based sediment samples to area-based estimates (Mg C ha⁻¹) assuming spatial homogeneity within each sampled meadow polygon. Sediment samples for SOC analysis were collected using a 5-cm diameter bucket auger to a maximum depth of 30 cm. Preliminary field tests indicated that PVC push-corers experienced core loss, excessive compression, and poor penetration due to dense seagrass rhizome mats and heterogeneous sediments containing shell hash and compacted layers. Bucket augers are widely used in seagrass and coastal blue-carbon studies where push-corers cannot recover intact sediment column^68–70. The auger allowed for consistent recovery of minimally disturbed sediment sections. The soil samples were stored in Ziplock bags and stored in a cooler until laboratory analysis to assess carbon storage. A total of 47 soil samples across the study sites were collected and classified into soil types based on particle size. Sediment grain size distribution was determined using laser diffraction particle size analysis (Mastersizer, Malvern Instruments). Oven-dried sediment subsamples were homogenised and analysed to quantify the relative proportions of sand, silt, and clay based on particle size distributions. Grain size data were subsequently processed using GRADISTAT software, which classifies sediment texture according to standard particle size thresholds (sand: >63 μm; silt: 2–63 μm; clay: <2 μm). The percentage contribution of each fraction was calculated relative to the total sediment sample, and sediment texture classes were assigned accordingly. The aforementioned technique was executed thrice at each sampling point (three randomly selected points within the quadrant) to collect three replicates. All three soil samples obtained from each sampling point were amalgamated and meticulously blended to form a uniform composite sample, which was subsequently analysed as a singular representative soil sample. This compositing approach was used to minimise small-scale sediment heterogeneity and produce a representative SOC estimate for each sampling location. Surrounding anthropogenic activities and landmarks were also recorded at each site.

In the laboratory, samples were air-dried at room temperature to remove excess moisture. The dried soil samples were thoroughly homogenised using a pestle and mortar to ensure representative subsampling. Dry bulk density (DBD; g cm⁻³), which is the level of sediment compaction, was calculated as the dry mass of sediment divided by the volume of the sampled sediment core corresponding to the 0–30 cm depth interval.

3

An accurately measured portion of the dried sediment sample (1 to 5 g) was weighed using an analytical balance, and the initial weight (Wi) was recorded. The weighted sediment sample was placed in a porcelain crucible. The crucible with the sample was heated to a high temperature (450–550 °C) in a muffle furnace for 6 h. Proper ventilation was ensured to allow the combustion of organic matter and the release of volatile components. After heating, the crucible and residue were allowed to cool to room temperature in a desiccator to prevent moisture absorption. The crucible and residue were then weighed accurately, and the final weight (Wf) was recorded. The Loss on Ignition (LOI) was calculated as the difference between the initial weight of the sample and the weight of the residue after heating using Eq. 3:

4

Loss on ignition (LOI) was used to estimate organic matter content (%OM), which was converted to organic carbon concentration (%C) using established conversion factors for marine sediments (Eq. 5a–b)^10,32

5a

5b

Carbon density (g cm⁻³), which is the amount of organic carbon in a soil sample, was calculated as the product of dry bulk density and organic carbon concentration, providing a volumetric measure of sediment carbon storage. SOC stocks were calculated for the 0–30 cm sediment depth by multiplying carbon density by sediment depth and applying unit conversions to express carbon stocks per unit area (Mg C ha⁻¹), following the IUCN Blue Carbon Manual⁶² and recommended protocols and equations (Eqs. 6–8) for estimating carbon for marine ecosystems (Howard et al., 2014; Fourqurean et al. (2012).

6

7

8

Statistical analysis

To test our hypotheses, we used a combination of univariate and multivariate statistical methods. A Kruskal-Wallis test was employed to compare differences in SOC stocks, canopy cover, and shoot density across the three management strategies  (MPA, LMMA, Unmanaged), followed by pairwise Wilcoxon tests when significant differences were observed.

To determine the influence of biophysical and environmental variables on soil organic carbon (SOC) stocks, we fitted a multiple linear regression model in R using SOC as the response variable and soil type, disturbance level, meadow type, shoot density, percentage cover, water depth, and water temperature as predictors. Soil type, disturbance, and meadow type were treated as categorical variables, while the remaining predictors were continuous.

The model was fitted using the lm() function in R, and an ANOVA (Type I sums of squares) was performed using the anova() function to evaluate the contribution of each predictor sequentially. Model assumptions, including normality, homoscedasticity, and independence of residuals, were verified using diagnostic plots and were met. Predictor significance was assessed using F-tests, and assumptions of normality and homogeneity were checked using Shapiro-Wilk and Levene’s tests, respectively. Non-parametric methods were used where assumptions were violated.

All analyses were conducted in R (version 4.4.1), and statistical significance was assessed at p < 0.05.

Author contributions

S.S. developed the study design, methodology, and project administration; conducted fieldwork; performed data analysis and wrote the manuscript. C.O. supported study design, provided methodological guidance, assisted with fieldwork, and contributed to manuscript review and editing. M.O. contributed to the research conceptualization, supervision of the author, led project administration, secured funding, supervised the project. Contributed to the manuscript review and editing All authors reviewed and approved the final version of the manuscript.

Data availability

The raw data supporting the findings of this study are not publicly available at this time as they are still in use for ongoing research. Data are available from the corresponding author upon reasonable request at [Simangele.Sithole@students.unibe.ch](mailto: Simangele.Sithole@students.unibe.ch).

Declarations

Competing interests

The authors declare no competing interests.