Human population density and blue carbon stocks in mangroves soils

Mangrove soils provide many important ecosystem services such as carbon sequestration, yet they are vulnerable to the negative impacts brought on by anthropogenic activities. Research in recent decades has shown a progressive loss of blue carbon in mangrove forests as they are converted to aquaculture, agriculture, and urban development. We seek to study the relationship between human population density and soil carbon stocks in urban mangrove forests to quantify their role in the global carbon budget. To this end, we conducted a global analysis, collecting mangrove soil carbon data from previous studies and calculating population density for each study location utilizing a recent database from the European Commission. Results indicate population density has a negative association with mangrove soil carbon stocks. When human population density reaches 300 people km−2, which is defined as ‘urban domains’ in the European Commission database, mangrove soil carbon is estimated to be lower than isolated mangrove forests by 37%. Nonetheless, after accounting for climatic factors in the model, we see the negative relationship between population density and soil carbon is reduced and is even non-significant in mixed effects models. This suggests population density is not a good measure for the direct effects of humans on mangrove ecosystems and further implies mangrove ecosystems in close proximity to very high population density can still possess valuable carbon stocks. Our work provides a better understanding of how soil carbon stocks in existing mangrove forests correlate with different levels of population density, underscores the importance of protecting existing mangroves and especially those in areas with high human population density, and calls for further studies on the association between human activities and mangrove forest carbon stocks.


Introduction
Mangrove forests cover approximately 0.1% of the Earth's land surface (Hamilton and Casey 2016) but provide a disproportionate amount of ecosystem services (Sanderman et al 2018), including acting as important habitats for wildlife (Mozumder et al 2018) and regulating global climate stability (Cummings and Shah 2018).More recently, mangroves have been recognized for their capacity to store a large amount of carbon (blue carbon stocks) and regulate carbon cycling (Inoue 2019, Zhu and Yan 2022), containing three to four times the mass of carbon typically found in some boreal, temperate, or tropical forests (Donato et al 2011).
Although mangrove trees store a significant amount of carbon in their biomass (Hutchison et al 2014), most of that carbon is typically found in their soils (Donato et al 2011, Sanders et al 2016, Sanderman et al 2018).The high soil carbon stocks of mangrove ecosystems are mainly attributed to their high carbon inputs and relatively low turnover rates (see table 1).It is estimated that mangrove soils store 49%-98% of total mangrove ecosystem carbon (Donato et al 2011, Liu et al 2014, Xiong et al 2018), which is around 5.0-10.4Pg  The numbers in the estimate column are shown in mean values, and their units are based on the references directly.Abbreviations: CO2 means carbon dioxide; CH4 indicates methane, DOC denotes dissolved organic carbon; DIC is dissolved inorganic carbon, mostly in carbonate minerals.
globally (Duarte et al 2013, Jardine and Siikamäki 2014, Atwood et al 2017).However, since so many mangrove forests have been replaced by agriculture, aquaculture, and urban land management, carbon emissions from those soils have substantially increased (Hamilton and Casey 2016, Richards and Friess 2016, Richards et al 2020) and their potential as carbons sinks has decreased (Sasmito et al 2019).
Over the past 20 years, global mangrove carbon stocks have declined by 158.4 Mt due to land cover changes and anthropogenic activities (Richards et al 2020).As a result, the annual rate of soil carbon emissions from mangrove loss is estimated to be 7.0 Tg carbon dioxide (CO 2 ) (Atwood et al 2017).Moreover, by the end of the century, global emissions from mangrove loss are projected to reach approximately 3,392 Tg CO 2 when considering foregone soil carbon sequestration (Adame et al 2021).These projections speak to the important role of existing mangrove forests in maintaining global carbon sequestration, particularly as they are significant carbon sinks subject to a variety of human disturbances.
Existing mangrove forests are occasionally used for fishing (Hamdan et al 2019), clam and oyster collecting (Asare et al 2019), and recreation and ecotourism (Spalding andParrett 2019, Intyas et al 2021), leaving them with low to intermediate levels of disturbance.However, mangrove forests in urban areas are associated with coastal development (e.g.Singapore and southeastern China) and are more vulnerable to environmental pollution of trace metals (Hong et al 2021), microplastics (Kannankai et al 2022), and organic compounds (Chai et al 2019) in both water and soils.These anthropogenic activities may impact mangrove soil carbon differently.For instance, collection of wood and leaves from mangrove forests reduces in situ carbon inputs (Bhomia et al 2016, Rasquinha andMishra 2021), while mangrove forests surrounded by urban settlements may have reduced or heavily polluted allochthonous carbon sources (Wigand et al 2021).Additionally, accumulation of metals and plastics can impair soil health and functioning, further influencing carbon storage capabilities (Wang et al 2022).Understanding the soil carbon stocks in those mangrove forests susceptible to degrees of human impacts can inform our understanding of mangrove ecosystem health and services, thus developing a clearer picture of the global carbon budget (Arnaud et al 2023).Therefore, a comprehensive quantification of carbon stocks in mangrove soils is needed, especially with respect to mangrove forests made vulnerable by urbanization and human land use interactions.
Here, we conducted a global meta-analysis by collecting mangrove soil carbon data from existing literature, showing how different levels of population density correlate with mangrove soil carbon stocks.Our analysis indicates that population density is negatively associated with mangrove soil carbon.Moreover, for mangrove forests that are in urban areas where population density is higher than 300 people km −2 , our analysis shows that the soil carbon stocks are likely to be over 37% lower than isolated mangrove forests ('isolated' is defined as mangroves with nearby population densities equal to 0 people km −2 ).However, when accounting for climatic factors such as temperature, our model shows the negative relationship between population density and soil carbon is reduced.This may suggest population density is not a good proxy for direct effects of human on mangrove ecosystems.This work informs our understanding of carbon stocks in existing mangroves experiencing intense anthropogenic impacts, provides reference for sustainable management of remaining mangroves, and speaks to the importance of protecting mangroves for their ecosystem services and socioeconomic values.

Literature search and data collection
In this study, we refer to soil carbon as soil organic carbon (SOC), and ensure the soil carbon data we collected are all in the form of SOC.We conducted a literature search in Google Scholar to collect mangrove soil carbon data.We used the following search queries to locate studies: [mangrove OR mangrove forest] and [soil] and [carbon OR organic carbon OR carbon stock OR carbon storage OR carbon density] and [nature OR natural OR human OR human population OR human population density OR urban OR city OR metropolitan].In total, we conducted 82 searches.In each search, we collected the papers that satisfied the following criteria: (1) the paper included SOC data and (2) the paper was published between 2000-2021.This range was selected in order to avoid obsolete datasets.Within each query, we went through publications from the top 10 pages of results.After a filtering and screening process (supplemental figure 1), we included 69 journal papers, but because some papers conducted their research in more than one location (e.g.different location names, sets of coordinates, or geographic information) and contributed to more than two datasets, we ultimately accumulated 184 datasets.In each cited paper, the following data were recorded for each study location: latitude and longitude, mean annual temperature, mean annual precipitation, dominant mangrove tree species, SOC, and soil bulk density (if present).

Data processing for SOC
We standardized the unit of SOC data to stocks (Mg ha −1 ) at 100 cm of soil depth to normalize the data.When a paper presented its SOC data in concentration rather than stock, we used the following equation to transform the data (He et al 2008, Barros et al 2015): For the papers whose SOC data did not reach to 100 cm depth, we used the following two asymptotic equations to extrapolate those data (Jobbágy andJackson 2000, 2001): (3) where Y is the cumulative proportion of soil carbon stocks from the soil surface to its depth (cm); β indicates the relative decrease of soil carbon stocks as soil depth increases; d is soil depth; X 100 denotes the soil carbon stocks in the upper 100 cm; d 0 is the original soil depth (cm) in each soil study, and X d0 denotes the soil carbon stocks from soil surface to d 0 soil depth.We followed Shi et al (2018) and set the value of β as 0.9786.It was found that the soil depth distribution of carbon did not significantly change across a variety of ecotones (Jobbágy and Jackson 2000, Yang et al 2011).As a result, the β value remained constant throughout data processing.It is important to note that not making appropriate adjustment for SOC over soil depth may lead to misestimations or biased results (Post and Kwon 2000, Guo and Gifford 2002, Chien and Krumins 2022).Overall, we did not perform many data transformations or extrapolations because 158 of the total 184 datasets (around 86%) had soil data to 100 cm.Moreover, only 12 of the total 184 datasets (around 7%) reported SOC in concentration rather than stocks, and 9 of those 12 datasets (only around 5% of the total) did not reveal soil BD.

Criteria for defining human population density in each study location
To obtain the minimum population density value for each study location, we utilized the Global Human Settlement Layer (GHSL) dataset from the European Commission (see Pesaresi et al 2022).This new dataset provides estimates of built-up surface areas and human presence on the Earth at 10 m-resolution (Schiavina et al 2022).The data were analyzed in ArcGIS Pro (Esri, v3.1.0),and a final map product was visualized in QGIS (OSGeo,v3.34.0).
We then overlayed the GHSL layer with the latitude and longitude obtained for the study locations from each dataset.Because we could not obtain the exact geographic range for study locations in most papers, we generated circular buffer zones around each set of coordinates to associate mangroves forests with population density.To determine the appropriate size for the buffer zone, we conducted a sensitivity analysis on different sizes of circles with 20, 30, 40, 50, and 60 km radius.We found that population density values only slightly increased or decreased with increasing buffer zone sizes, and different buffer zone sizes yielded very similar analysis results.As such, we selected a buffer zone size of 30 km around the center of each set of coordinates to most appropriately represent association with population density.This circle was intended to represent a reasonable buffer zone across all study sites (Luo 2004, Rocha et al 2007, Valenti et al 2023), from the center of each mangrove location.
The GHSL data breaks pixels down into eight different levels of human presence, and each level has its own minimum population density threshold, as shown in the supplemental table 1 (see Pesaresi et al 2022).To calculate population density around each study site, we multiplied the area of each coded density level inside the buffer by its corresponding minimum population density, and then summed the results from each level.We then divided this value by the area of the 30 km-radius circle (around 2827.43 km 2 ) to obtain the average estimated minimum population density for each study location.

Statistical analysis
To explore how human population density relates to mangrove SOC, we performed several statistical tests before running analyses.Because we found high spatial autocorrelation among datasets (Moran's I value higher than 0.2; p-value lower than 0.05) (Chen 2021), we used linear mixed effects models.This approach allowed us to capture the spatial variation among those clusters and control spatial autocorrelation by using study location and dominant tree family as random effects (Dormann et al 2007, Dey et al 2023) through Kenward-Roger approximation (Kuznetsova et al 2017).Based on the ANOVA, AIC scores and R 2 values, we selected the models which best fit our data.Moreover, because we found collinearity between temperature and precipitation, we did not include both climatic factors in the same mixed effects models with population density.
To assess population density with temperature and precipitation altogether, we reported standardized effect sizes for each factor by representing their Hedges' g, which are the values modified from Cohen's d (Lin and Aloe 2021).These effect sizes can show the relative strength of each coefficient, accounting for different units.After calculating the effect sizes, we generated forest plots based on random effects models with a conservative estimation (r = 0.7) (Rosenthal 1986).Moreover, we utilized the total observed variation (I 2 ) and a test of heterogeneity (Q) to verify the heterogeneity of the collected data.A Higgins' I 2 value higher than 75% demonstrates substantial heterogeneity (Higgins et al 2003).Additionally, a significant p-value in Q statistics indicates that effect sizes are not evenly distributed across studies/datasets or that the direction of the effect sizes is different among studies/datasets, further suggesting high heterogeneity (Meisner et al 2014).Finally, to capture the potential non-linear nature of population density on mangrove SOC, we used generalized additive model (GAM).This is a flexible statistical model used to detect the influence of nonlinear effects (Wu and Zhang 2019) to show the relationship over the full distribution.Again, because of the collinearity between temperature and precipitation, we did not include both climatic factors in the GAM with population density.Based on the R 2 value, we included population density and temperature, rather than precipitation, in the GAM.

Distribution of the study locations and descriptive statistics
We first map the study locations with estimated minimum population density (figure 1).The datasets represent a wide global distribution, and most study locations are distributed in tropical and subtropical regions, with approximate latitudinal boundary of 30 • N and 30 • S. The mangrove site with the highest latitude is in southeastern Australia (around 38.3 • S).Study sites with higher human population density in our datasets are mainly located in southeastern China, southern India, Singapore, and southwestern Indonesia (figure 1, see red dots).
With respect to our dataset, the average values of temperature, precipitation, population density, and soil carbon stocks across all study locations are 26.20 • C, 1753 mm, 15.54 (people km −2 ), and 245.50 Mg ha −1 , respectively (table 2).Around 75% of the study sites are located in the northern hemisphere, and the continent with the highest number of study locations is Asia (table 3).Note that only 10% of the study locations are considered relatively isolated mangroves with a minimum population density of 0 people km −2 (table 3).For the dominant tree family of the study sites, Acanthaceae and Rhizophoraceae are the most two common families, while Combretaceae, Malvaceae, Euphorbiaceae, and Fabaceae are relatively rare (table 3).

Climatic factors, human population density and mangrove soil carbon stocks
Our analysis shows that both climatic factors are positively associated with mangrove soil carbon stocks (p-values = 0.0026 and 0.0485 for temperature and precipitation, respectively) (table 4, Model 1 and 2).However, population density has a negative relationship with mangrove soil carbon stocks (p-value = 0.0165) (table 4, Model 3).On average, when human population density increases 1 unit (people km −2 ) in mangrove areas, soil carbon decreases by 0.22 Mg ha −1 (table 4, Model 3).This relationship is reduced when it is analyzed with climatic factors (table 4, Model 4, population density p-value > 0.05), suggesting population density is not significantly related to mangrove soil carbon after we account for temperature.
To compare the two climatic factors with population density, we calculate Hedges' g values, create random effects models, and construct forest plots, identifying their partial relationship with mangrove soil carbon.Results reveal high heterogeneity (I 2 = 81.72%;Q statistic p-value = 0.0042).In the forest plot, both climatic factors have positive relationships with mangrove soil carbon, while the population density has a negative, but not significant, association (figure 2).Overall, temperature has the strongest relationship with soil carbon among the three factors and is likely a critical driver of global mangrove soil carbon stocks before considering potential spatial autocorrelation.
In the GAM, we again observe a general negative association between population density and mangrove soil carbon stocks (table 5; figure 3).Mangrove soil carbon substantially decreases between 0-250 people km −2 and starts to fluctuate over high levels of population density.When population density reaches around 300 people km −2 (which is defined as urban domain in the GHSL database), the average  in more detail) in the areas where population density is higher than 800 people km −2 , but again this prediction is accompanied with higher uncertainty.However, it is important to note the GAM includes both population density and temperature's relationship with mangrove soil carbon (table 5).Moreover, as temperature and population density are found to be positively correlated (p-value < 0.001), the line shown in the figure 3 depicting a general decrease in mangrove soil carbon also captures its relationships with temperature (see supplemental figure 2 for the GAM plot of temperature and mangrove soil carbon).1).In addition, with higher annual precipitation (or less drought events), lower decomposition rates due to suboxic or anoxic conditions support soil carbon sequestration (Kida and Fujitake 2020).These ecological interactions help support our finding that both temperature and precipitation positively contribute to soil carbon stocks in mangrove forests (See table 4, Model 1 and 2), which are co-moderated by other environmental factors.
In addition to climatic conditions, mangrove soil and ecosystem carbon stocks are controlled by nutrient availability (e.g.nitrogen and phosphorus) (Weiss et al 2016, Spivak et al 2019) and soil properties (e.g.pH and salinity) (Sanderman et al 2018, Xia et al 2021).For example, recalcitrance of plant biomass (Kim et al 2021) and the presence or absence of allochthonous inputs (MacKenzie et al 2021) from upstream ecosystems play a crucial role in mangrove soil carbon stocks.Furthermore, soil and ecosystem carbon stocks may vary among different tree species.In this work, we found that the forests dominated by Combretaceae family (mostly the white mangrove species Laguncularia racemosa) on average store a higher amount of soil carbon than the mangroves dominated by other tree families (see supplemental table 4).Indeed, white mangroves possess higher soil carbon stocks and bulk density than red (Rhizophora mangle) and black mangroves (Avicennia germinans) (Hernández and Junca-Gómez 2020).We suspect that this may be the result of the preferred habitats of different tree species over environmental gradient (e.g.surface elevation) (Leong et al 2018), varied physiological properties in photosynthesis rates (Mangora et al 2014), and the subsequent allocation of fixed carbon to the aboveground or belowground (Perera and Amarasinghe 2013).However, in this study we do not further explore the relationship between soil carbon and mangrove tree species mainly because the sample sizes among different families are not evenly distributed (see table 3).Moreover, based on the information provided by the collected papers, we were not able to identify the dominant tree species in more than 25% of the total datasets (these were categorized as 'Mixed') (see table 3).As a result, the information about mangrove tree species and their soil carbon may not be well represented in this work, Figure 3. Visualization of the GAM showing the relationship between population density and mangrove soil carbon stocks when accounting for potential confounding effect of temperature.The black line indicates the line of best fit, and the shaded area is the 95% of confidence interval.The icons below the GAM plot represent typical human settlements and anthropogenic impacts, explained in the Discussion, over different levels of population density.For the mangrove forests in highly populated zones (minimum population density is higher than 824 people km −2 , the highest value covered in our dataset), we present two possible scenarios: gradually decreasing (depicted by blue lines and text) predicted by the GAM or entire mangrove area loss (the lower rightmost icon).Please see the results and discussion sections for details, supplemental table 2 for the GAM intercept and coefficients, and supplemental table 3 for values of mangrove soil carbon over population density in more detail.Icon made by Freepik from www.flaticon.comhttps://www.flaticon.com/free-icon/tribe-house_509805?term=tribe+house&page=1& and we hope our preliminary analysis (see supplemental table 4) and data can serve as references for future studies.We believe that the soil carbon stocks of mangrove ecosystems dominated by different tree families may respond differently under climate change or human impacts, but further investigation is warranted.

Population density and mangrove soil carbon
Our analysis reveals a negative association between population density and mangrove soil carbon (figure 3).At low levels of population density, the magnitude of mangrove soil carbon can be greater than the carbon stocks in highly populated areas, likely due in part to the degrees of human activities and the perception of local communities towards nearby mangrove forests.For example, in Kenya, local tribes modestly harvest the timber, non-timber forest products and associated ecosystem goods provided by mangroves and surrounding systems to support their basic needs (Okello et al 2019).Similarly, in Indonesian communities, when people are aware and consider the sustainability of mangrove forests, they initiate mangrove rehabilitation efforts to ensure that the forests are a renewable natural resource (Rosaliza 2018, Sadono et al 2020).Mangrove protection efforts can be more impactful when national government and international organizations are involved.(Romañach et al 2018).Under these contexts, it is expected that mangrove ecosystems are not likely to experience huge carbon loss.
However, lower population density may not always indicate low anthropogenic impacts.A recent study estimates that 62% of global mangrove losses between 2000 and 2016 are attributed to land-use change, primarily through conversion to agriculture and aquaculture (Goldberg et al 2020).The population density around agriculture and aquaculture might not be high, but the negative impacts brought by those land conversions can be substantial.For instance, the release of ecosystem carbon into the atmosphere as a consequence of mangrove loss and degradation has been estimated to be 0.02-0.12Pg annually (Donato et al 2011, Thomas et al 2017).Moreover, the ecosystem health of remaining mangroves in close proximity to aquaculture and agriculture has diminished due to changes in hydrology, salination, and contamination, all of which can influence the ability of mangrove soils to store carbon (Bhavsar et al 2016).Likewise, in the GAM analysis (figure 3), we observe a substantial decrease in soil carbon stocks across lower portions of population density ranges (e.g.0-250 people km −2 ).An 18% decrease at 100 people km −2 is notable, and the level of uncertainty for this estimation is markedly lower than the values estimated for high population density areas (see the confidence interval in figure 3).In the range of low population density, many mangrove forests are well protected under conservation policy and measures, potentially resulting in reduced carbon loss (see Bukoski et al 2017, Murdiyarso et al 2021).That said, this fact can be confounded in the cases of agriculture and aquaculture that are logically found in low population density areas but with high probability of lost soil carbon (Ahmed andGlaser 2016, Arifanti et al 2019).
When population density reaches 300 people km −2 , which is defined as an 'urban domain' by the GHSL database, mangrove soil carbon is estimated to decrease by 37% when compared to relatively isolated mangrove forests (see figure 3).Urban or peri-urban mangroves are often vulnerable to deforestation (Lagomasino et al 2019), overexploitation for timber production and other ecosystem goods, and urban expansion stemming from the growing need to accommodate increasing coastal populations (Richards andFriess 2016, Thomas et al 2017).Moreover, remaining mangroves are subject to other ecological problems, such as forest fragmentation and edge effects, that are likely to impact carbon storage and turn mangrove forests into net carbon sources (Qie et al 2017).Other environmental pollution problems associated with highly populated areas, like microplastic fragments (Kannankai et al 2022), e-waste (Zhou et al 2019), and sewage contamination (Kamau et al 2015), can have devastating impacts on the soil bacterial communities, nutrient availability, metal content (Torres et al 2019), and water quality (Alemu et al 2021) which may further affect soil carbon storage in urban and peri-urban mangroves (Wang et al 2022).Moreover, nutrient enrichment from poorly treated discharges can enhance heterotrophic remineralization of blue carbon stocks from mangrove soils, resulting in CO 2 and methane emissions (Barroso et al 2022).These environmental impacts associated with urban areas might not be the same in areas with relatively low levels of population activity.Therefore, together with those ecological problems, our study highlights the need to raise awareness of low soil carbon in urban mangrove ecosystems with respect to degraded ecosystem health services and the global carbon budget.
At very high levels of population density, the relationship between human activities and mangrove soil carbon can become more interactable, even imbedding with natural factors (Friess et al 2020, Hagger et al 2022) we have discussed above.However, even though the highest population density value we have cited is approximately 825 people km −2 , we do not have sufficient literature support to draw conclusions regarding urban mangrove forests from population density higher than 300 people km −2 (see table 3).Beyond 825 people km −2 , the GAM predicts a gradual decrease of mangrove soil carbon (figure 3, blue dashed line; supplemental table 3), but this is with higher uncertainty.Given the known ecological value of mangrove forests, we investigate if those located in highly populated areas still retain carbon.Here, we reveal the possibility that total erasure of mangrove ecosystems in areas with extremely high populations due to land conversion can lead to near complete soil carbon loss (figure 3, the lower rightmost icon).It is important to note that the negative association between population density and mangrove soil carbon is reduced when we account for typical climatic factors like temperature (see tables 4 and 5).Therefore, population density may not be the best predictor for the direct relationship between human impacts and mangrove soil carbon.We call for further investigations into the effects of anthropogenic activities and the global carbon budget in mangrove ecosystems before conclusions can confidently be made with carbon stocks.

Conclusion
We conducted a global analysis to explore how population density correlates with soil carbon stocks in existing degraded, urban mangrove forests.We observe a notable negative association between population density and mangrove soil carbon stocks.Specifically, when population density reaches 300 people km −2 (urban domains), mangrove soil carbon is estimated to decrease by 37% when compared to the soils in relatively isolated and intact mangrove ecosystems.At very high levels of population density, the relationship is hard to resolve due to insufficient literature on urban mangrove biomes.When we do account for typical climatic factors (like temperature) in the analyses, we see the negative association between population density and mangrove soil carbon is weak possibly because of the various anthropogenic activities at low or high population density ranges that can complicate that relationship.This finding also indicates that the influences of human population density on mangrove soil carbon cannot be isolated from the greater contexts of typical ecological drivers, such as temperature and precipitation.Even still, we further reinforce the notion that mangrove ecosystems in highly populated areas may possess valuable carbon stocks.In the absence of experimental work, particularly for mangroves in highly populated areas, more analyses are required to account for the impact of human activities on mangrove forests before we can confidently define the associations between population density and mangrove soil carbon stocks.
The total amount of carbon that is fixed by mangrove trees and other plants via photosynthesis subtracts the respiratory carbon loss from the plants Aboveground organic carbon 13.2 MgC ha −1

Figure 1 .
Figure 1.Distribution of study locations around the world.The areas in parts of (a) America, (b) Africa and (c) Asia and Oceania are enlarged.Reproduced from [Société OPENDATASOFT, "World Administrative Boundaries -Countries and Territories" (2022).World Food Programme (UN agency)].CC BY 3.0.

Figure 2 .
Figure 2. The effect sizes and confidence intervals of the factors in the mangrove soil carbon stocks.Each effect size is represented by a solid square, and the horizonal line extending from each square represents confidence interval of the effect sizes.The square size in each factor varies inversely with the range of confidence interval.The color of the squares and confidence interval indicates positive (blue) or negative (red) relationships or effects.The weight values and confidence intervals of each effect size for each factor are shown in the right of the figure.The vertical dotted line is the null effect sizes.If the confidence intervals do not cover the null effect size, this indicates the partial effect or relationship of that factor on mangrove soil carbon is significant.(CI: confidence interval).

Table 1 .
Global estimates of carbon flows and budgets in mangrove ecosystems.

Table 2 .
Values of median, mean, standard deviation, minimum, and maximum in each variable.
a Mangrove forests with nearby minimum population density equal to 0 people km −2.b Mangrove forests with nearby minimum population density higher than 0 people km −2.−2 .The GAM predicts a slow decrease of soil carbon (figure3, blue dashed line; see supplemental table 3 for predicted soil carbon values

Table 3 .
Number of samples in sub-groups from each variable.
a If there are more than 2 tree species (from two different families) in a study location and we could not identify the dominant one, we used 'Mixed' .bTheOthers group includes Malvaceae, Euphorbiaceae, and Fabaceae.

Table 4 .
The linear mixed effects models for identifying the effects of factors on mangrove soil carbon stocks.In the Model 4, we do not include precipitation because there is strong collinearity between temperature and precipitation.Based on AIC score and R 2 value, we included temperature in the model instead of precipitation.
(Simard et al 2019)level, * means p-value is lower than 0.05, * * means p-value is lower 0.01, and * * * means p-value is lower than 0.001.aArnaudetal2023).This is echoed in our analysis results, which show a positive relationship between the two climatic factors and mangrove soil carbon.A recent study reveals that temperature, precipitation, and cyclone frequency together can explain 74% of the global trends in maximum mangrove canopy height(Simard et al 2019).Another study suggests more productive(Hutchison et al 2014), allowing soils to sequestrate more carbon via root production, root exudates, and litterfall turnover.Even though higher temperature often leads to greater soil decomposition rates, the deep standing water in mangrove ecosystems can limit soil CO 2 efflux by frequent inundations and shifting aeration (Chimner 2004),

Table 5 .
Approximate significance of smooth terms in the GAM.
(Pregitzer et al 2000evel, * means p-value is lower than 0.05 and * * means p-value is lower 0.01.so the response of soil respiration to rising temperature may not be a linear function(Lovelock 2008).Moreover, a global review describes that mangrove root carbon production exceeds mineralization rate(Ouyang et al 2017), and this can be explained by the fact that elevated temperature, together with other factors like hydrology and geomorphology, leads to greater net primary production and carbon flux to mangrove soils(Pregitzer et al 2000, Rovai et al  2018)(e.g. the data organized in table