Global mangrove root production, its controls and roles in the blue carbon budget of mangroves

Mangroves are among the most carbon‐dense ecosystems worldwide. Most of the carbon in mangroves is found belowground, and root production might be an important control of carbon accumulation, but has been rarely quantified and understood at the global scale. Here, we determined the global mangrove root production rate and its controls using a systematic review and a recently formalised, spatially explicit mangrove typology framework based on geomorphological settings. We found that global mangrove root production averaged ~770 ± 202 g of dry biomass m−2 year−1 globally, which is much higher than previously reported and close to the root production of the most productive tropical forests. Geomorphological settings exerted marked control over root production together with air temperature and precipitation (r2 ≈ 30%, p < .001). Our review shows that individual global changes (e.g. warming, eutrophication, drought) have antagonist effects on root production, but they have rarely been studied in combination. Based on this newly established root production rate, root‐derived carbon might account for most of the total carbon buried in mangroves, and 19 Tg C lost in mangroves each year (e.g. as CO2). Inclusion of root production measurements in understudied geomorphological settings (i.e. deltas), regions (Indonesia, South America and Africa) and soil depth (>40 cm), as well as the creation of a mangrove root trait database will push forward our understanding of the global mangrove carbon cycle for now and the future. Overall, this review presents a comprehensive analysis of root production in mangroves, and highlights the central role of root production in the global mangrove carbon budget.


| INTRODUC TI ON
Mangroves represent some of the most carbon-dense and productive ecosystems in the world. They provide ecosystem services to >200 million people across 123 countries (Costanza et al., 2014;Hutchison et al., 2014;Spalding, 2010). Mangrove functions and services include carbon burial in mangrove soils and its storage for centuries to millennia (Donato et al., 2011;McLeod et al., 2011).
Whether mangroves will continue to act as a carbon store and sink in the future is still uncertain under multiple global changes Lovelock, 2020;Lovelock et al., 2015;Rogers et al., 2019;Saintilan et al., 2020). An accurate forecast of the carbon dynamics and resistance of mangroves to global changes would require a mechanistic understanding of soil organic matter accumulation, notably root dynamics and production (Arnaud, 2021;Cormier, 2021;Kida & Fujitake, 2020).
Root dynamics are important for mangrove resistance to sea level rise (Krauss et al., 2014). Mangroves might withstand sea level rise either through net vertical accretion of soil or by retreating landward within their available accommodation space (Krauss et al., 2014;Middleton & McKee, 2001;Rogers et al., 2019;Saintilan et al., 2020).
Carbonate mangrove soils (i.e. build on karstic environments and Holocene reef tops) include a large portion of dead root materials; therefore, an alteration of root production and decomposition will disproportionally modify their soil surface elevations and resilience to sea level rise (McKee et al., 2021). In mangrove geomorphological settings with more mineral soils (e.g. terrigenous delta, estuaries), the elevation change is dominated by sediment inputs, but a change in soil surface elevation might occur if root production and decay are altered (Lang'at et al., 2014;Rogers, 2021), notably under global changes.
Root processes are also a key component of carbon flow and dynamics in mangroves. The production of roots might represent a third of the net primary production in mangroves (Alongi, 2020;Bouillon et al., 2008). Compared to leaves, roots are not washed away by tides and are likely to be chemically and physically protected from degradation (e.g. via compounds inhibiting microbial decomposition and organo-mineral association-the binding of organic matter with minerals) (Kida & Fujitake, 2020;Middleton & McKee, 2001).
Therefore, roots are believed by some to form the main autochthonous input for mangrove carbon burial (Bouillon et al., 2003;Kida & Fujitake, 2020;Kristensen et al., 2008). Mangrove roots may also accelerate the mineralisation of SOM through the release of (1) oxygen (Pi et al., 2009), which changes the redox condition of the soil (Inoue et al., 2011;Kristensen et al., 2008), and (2) exudates that may remobilise old protected carbon as observed in forests (Keiluweit et al., 2015;Phillips et al., 2011). Upscaling of those root processes with their subsequent effects on carbon dynamics is currently limited due to the lack of an updated estimate of global root production, as well as a lack of understanding of the controls and response of root production to global changes Coldren et al., 2019).
Compared to the aboveground part of mangroves, root production has been little studied (Alongi, 2020). The controls of root production have never been reviewed at a global scale, and the root contribution to carbon burial and losses has been rarely estimated (Cormier, 2021;Ouyang et al., 2017). There is an emerging body of literature on root production and its controls at the site level (Cormier, 2021). Assessing this emerging literature represents an opportunity to critically assess the role of mangrove root processes and forecast the potential impacts of global changes upon them.
Such effort is constrained by the absence of a comprehensive and up-to-date dataset based on a systematic review of recent mangrove root production measurements. Systematic reviews are based on a clearly formulated question and use systematic and explicit method to identify, select, critically appraise and analyse relevant research (Wright et al., 2007, see also Pullin & Stewart, 2006) in contrast to general reviews that aims at presenting the state of scientific knowledge not exhaustively (Pullin & Stewart, 2006). Recent estimates of mangrove root production (Alongi, 2020;Twilley et al., 2017) have contributed to increasing our general understanding of root production, but did not follow the methodology of a systematic review (Pullin & Stewart, 2006). The closest to a systematic review to date is the review of Bouillon et al. (2008), but it included only the five studies available at the time (n = 16 data points) .
Here, we use an alternative approach based on geomorphological settings to critically assess the role of mangrove root production that has largely been overlooked in the literature compared to other aspects of mangrove research ( Figure S1). Through the lens of geomorphological settings Worthington et al., 2020), we quantify the global root production (<20 mm diameter) using a systematic review of root production. We identify environmental controls of root production through a qualitative and quantitative review of the literature. We build on this synthesis to hypothesise the potential effect of global changes on global mangrove root production, and the role of root dynamics in the global mangrove carbon budget. We conclude by outlining a vision of future research directions and actions to increase our understanding of root processes, and ultimately our understanding of the full mangrove carbon cycle now and into the future.

| BU ILD ING THE ROOT PRODUC TI ON DATA BA S E
We collected published studies from the Web of Science and Scopus bibliographic databases using search terms, such as 'mangrove' and 'root producti*' (see Supplementary Material 1.1 and Table S1 for exhaustive description). The search string resulted in 145 articles from Scopus and 131 from Web of Science (as of 23 March 2022). For inclusion in the systematic review, the root production measurement had to be conducted (1) in situ with direct measurement of mangrove belowground root production (e.g. not through allometric equations) that could be converted into mass, and at a depth >15 cm; (2) in mangroves having trees being at least 4 years old; (3) in mangroves that did not have experimental treatments or extreme events (e.g. hurricanes) having led to high tree losses. In total, we had 90 valid root production measurements extracted from 24 articles. We also collected a comprehensive list of 18 factors associated with the root production measurements including methodological, geographical, meteorological, ecological and edaphic factors ( Table S2) that have been shown to control mangrove root production, or that were likely to influence it (Adame et al., 2014;Castañeda-Moya et al., 2011;Coldren et al., 2019;Cormier et al., 2015;Eugenia & Sánchez, 2005;Gleason & Ewel, 2002;Kihara et al., 2022;Lang'at et al., 2013;McKee & Faulkner, 2000;Muhammad-Nor et al., 2019;Rivera-Monroy et al., 2017). We retrieved the geographical coordinates of each root production measurement using Google Maps when they were not reported in the study. All geographical coordinates provided in articles were confirmed using Google Maps (www.google. fr/maps). We used these geographical coordinates to extract for each root production measurement their bioregions from the map of Spalding et al. (2007), their meteorological variables from the WorldClim high spatial resolution climatic map (30 s) (Fick & Hijmans, 2017) and their tidal amplitude data from the map of Rovai et al. (2018). We defined the geomorphological settings for each literature-derived root production rate using the map of Worthington et al. (2020). The geomorphological classification of Worthington et al. (2020) was made by combining remote sensing data, machine learning and reviews of >500 geomorphological settings by mangrove experts. The reported accuracy was   We manually extracted all the other factors from the studies themselves or associated PhD theses and articles. We used Plotdigitizer (https://apps.autom eris.io/wpd) to extract data from the figures. A quality check was done for each environmental data and root production measurement (e.g. Table S3). The root production was measured with ingrowth core (83%) and sequential coring (17%) at a depth often limited to 45 cm (90%) (Figure 1). When coring or using ingrowth bags, large roots (>20 mm) are not well sampled (Adame et al., 2017). Therefore, our analysis does not include roots >20 mm diameter. The reported root production across studies is accessible in Arnaud et al. (2023), which is the database of this article published in the Zenodo repository.

| MANG ROVE ROOT PRODUC TI ON AT A G LOBAL SC ALE
The mangrove root production measurements spanned all continents having mangroves and were found across all geomorphological settings. Geomorphological settings have allowed a better understanding and upscaling of mangrove structural and functional patterns than latitude or bioregions alone (Rovai et al., 2016Twilley et al., 2018). Geomorphological settings determine mangrove environmental factors, such as the hydrology (e.g. inundation duration and frequency), the nutrient load and salinity of the mangrove soils Twilley et al., 2018;Woodroffe et al., 2016) that are regulating root production (Adame et al., 2014;Ball, 1988;Ball & Pidsley, 1995;Burchett et al., 1989;Castañeda-Moya et al., 2013;Hayes et al., 2017;McKee et al., 2007;Naidoo, 1987Naidoo, , 1990Nguyen et al., 2015;Ola et al., 2018;Twilley et al., 2018). Therefore, we upscaled the local mangrove root production measurements using the framework of the geomorphological setting proposed by Twilley et al. (2018), in addition to the root production per depth and the global mangrove area of 137,600 km 2 (Bunting et al., 2018).
The full methodology for the computation of root production and possible limitation of that methodology is described in detail in Supplementary Material 1.2, and the data of root production reported in each study are provided in Arnaud et al. (2023).
We found that the mangrove root production averaged ~770 ± 202 g of dry biomass m −2 year −1 or 41 ± 11 Tg C year −1 (n = 90) globally ( Table 2,  This study presents an important step forward to quantify root production in mangroves. Yet, our root production estimate is still uncertain due to measurement gaps in very productive mangrove regions (e.g. in Indonesia) ( Figure 1) and at depths below 45 cm, as well as two different methods of measurement used (Figure 1: ingrowth core and sequential coring) (see Section 6). Our root production estimate is two and a half times higher than the previous estimate in Twilley et al., 2017, and one and a half times higher than the one in Alongi (2020) (extracted from Table 2), but ~25% lower than the one in Bouillon et al. (2008) after all value being normalised for mangrove area. Those differences are mostly due to the inclusion of new observations having higher root production measurements (e.g. in the Central Indo-Pacific) and upscaling methodology. Twilley et al. (2017) estimate was based on three and a half times fewer observations with only limited measurements from the Central Indo-Pacific, which is the most productive bioregion in our dataset ( Figure 3). In addition, Twilley et al. (2017) upscaled local root production measurements by doing a mean, not accounting for the impacts of geomorphological settings on mangrove growth and soil properties and hence root production, and therefore likely reducing global estimate of root production . Bouillon et al. (2008) estimate is based on very few data points (n = 16), and was upscaled using a root: aboveground litter production ratio. Yet, biomass allocation between aboveground and belowground responds to environmental stress (e.g. hydroperiod, soil nutrients) and is therefore not constant across mangroves ( In per cent (% of the total) 65 35 100 TA B L E 2 Average (g of dry biomass m −2 year −1 ) of root production in mangroves, as well as its error range and proportion (%) across soil depths.
F I G U R E 2 Root production across geomorphological and sedimentary settings in g of dry biomass m −2 year −1 . Geomorphological settings are defined by Worthington et al. (2020). Carbonate settings are karstic environments and Holocene reef tops. Terrigenous settings are sediment-rich depositional environments as defined in Worthington et al. (2020). Red-filled dots show the sample mean. Bold, horizontal lines show sample medians. The lower and upper hinges correspond to the first and third quartiles of the sample. The upper and lower whisker extends from the hinges to the largest and smallest values, respectively, which is no further from the hinges than 1.5 times the sample interquartile range. Open circles indicate individual measurements, with vertical jitter to reduce overwriting. (Aragão et al., 2009;Cordeiro et al., 2020;Malhi et al., 2011). This suggests that our estimate might be still conservative because mangroves have been shown to allocate disproportionately more carbon to roots than other forests (Twilley et al., 2017).

| ENVIRONMENTAL CONTROL S ON MANG ROVE ROOT PRODUC TION
We investigated potential macroecological controls of root production that have previously been shown to regulate structural and func-

| Global factors explaining the root production
The root production was best explained with a linear model including the geomorphological settings crossed with the maximum air temperature of the warmest month and the minimum precipitation of the driest month (Table S4). This combination explained ~30% of the variability of root production from our dataset (r 2 = .28; p < .001, n = 90). The geomorphological settings crossed with the maximum air temperature explained most of the root production variability (19%; p < .001), followed by the precipitation of the driest month (4%; Combined with geomorphological settings, the air temperature was an important predictor of mangrove root production, as has been observed across terrestrial ecosystems for temperature, if soil moisture and nutrient availability are not limiting tree growth (Pregitzer et al., 2000). The increase in root production with temperature results from an overall increase in net primary production as observed by Coldren et al., 2019. Therefore, it is also likely that root production decreases when aboveground primary production decreases after reaching the thermal photosynthetic optimum of mangroves (i.e. between 25 and 32°C: Alongi, 2009).
Previous local-scale studies have shown that precipitation was related to mangrove root production as in our findings (Hayes et al., 2019;, but not all . Two mechanisms explain the increase in mangrove root production with precipitation at the local scale. Precipitation increases the freshwater availability in soil, and mangrove roots have been shown to preferentially uptake freshwater rather than saline water to support growth for physiological and osmotic pressure reasons (Hayes et al., 2019). Therefore, the input of freshwater might increase overall mangrove tree growth, including roots Simard et al., 2019). The root: shoot ratio (i.e. the root production vs. the aboveground production) is, however, unlikely to decrease with higher precipitation, because low precipitation might rather lead to more allocation of carbon belowground than aboveground (Adame et al., 2017). The tidal amplitude did not improve the model to explain the variability of mangrove root production in our dataset. The resolution of the tidal amplitude data might be spatially too coarse, or might not reflect well enough the inundation conditions of mangroves, because several studies suggest that mangrove root production and tidal inundation might be closely related (Adame et al., 2014;Ezcurra et al., 2016;Rogers et al., 2019;Saintilan et al., 2020). Therefore, it will be critical to report the frequency and duration of daily inundation for future mangrove root production studies.

| Bioregions, sedimentary settings, ecological and soil factors
Mangroves from the Central Indo-Pacific bioregion had the highest root production (604 ± 167 g of dry biomass m −2 year −1 , n = 22,  Table S5) likely because hot-spots of mangrove net primary production have been little studied for root production (e.g. in Indonesia).
Mangroves can also be classified following their sedimentary settings (Worthington et al., 2020). Mangroves in terrigenous settings had a higher root production (469 ± 66 g of dry biomass m −2 year −1 , n = 59, Figure 2) than in carbonated settings (304 ± 38 g of dry biomass m −2 year −1 , n = 31, Figure 2) as reported for the aboveground net primary productivity in the neo-tropics (de Albuquerque Ribeiro et al., 2019). In contrast to the aboveground productivity, the difference in root production between both sedimentary settings was not statistically significant (p = .12, H (5) = 8.7, n = 90, productivity might be less striking for root production than aboveground production, because the root production responds to overall tree growth, but also to carbon allocation that might be higher towards roots in carbonate settings in response to their limited soil phosphorous content . Tree density is often reported in local-scale studies as a potential factor influencing root production in mangroves (Adame et al., 2014;Arnaud et al., 2021), but we did not find a significant relationship between tree density and basal area with root production (tree density: p = .45, r 2 = .12, n = 17; basal area: p = .80, r 2 = .07, n = 15; Table S5). Many studies have reported tree density along with root production (almost 40%), but the threshold of diameter at breast height (DBH) for the inclusion of trees in census was very heterogeneous (from all DBH to DBH > 10 cm) limiting comparisons to 15 measurements.
In contrast, the dominant genus of mangrove sites was often reported (n = 87). The root production was the highest in mangroves dominated by Ceriops (1397 ± 850 g of dry biomass m −2 year −1 , n = 3, Figure 3). The root production of mangroves dominated by Rhizophora and Avicennia was more than three times less important than for Ceriops (Rhizophora: 410 ± 48 g of dry biomass m −2 year −1 , n = 40; Avicennia: 339 ± 46 g of dry biomass m −2 year −1 , n = 30, Figure 3, Table S5). Yet, only a few measurements have been conducted in mangroves dominated by Ceriops.
Nutrients, salinity and bulk density have all been shown to exert local control over mangrove root production or primary production in mesocosm and field studies (Adame et al., 2014;Castañeda-Moya et al., 2013;Hayes et al., 2017;Naidoo, 1987Naidoo, , 1990Ola et al., 2018). There was a positive significant relationship between total soil nitrogen and root production (p < .01, r 2 = .73, n = 15, Table S5), but not with total soil phosphorus (p = .33, r 2 = .26, n = 12, Table S5), which is consistent with mangrove field studies that have reported an increase, a decrease or no change of root production (absolute and relative) with an increase in soil nutrients. An increase in soil nutrients is generally expected to decrease the root: shoot ratio in forests (Nadelhoffer, 2000), meaning that the absolute root production can increase with an increase in nutrients, but proportionally less than the aboveground production (Sullivan et al., 2007). This relative decrease in root production in re- The relative root production response to nutrient availability is likely regulated by other environmental factors (e.g. inundation frequency, salinity, anoxia). For instance, the increase in root growth caused by nutrient enrichment appears only in frequently inundated mangrove soils with high salinity (Adame et al., 2014;Hayes et al., 2017;McKee et al., 2007). Mangrove roots might take advantage of an increase in nutrients to increase their growth to access more freshwater and alleviate environmental stress (Adame et al., 2014;Hayes et al., 2017). We did not find any pattern between bulk density and root production (p = .61, r 2 = .07, n = 44, Table S5), despite local-scale manipulative studies showing that bulk density controlled the root production of Avicennia marina and Rhizophora stylosa (higher root production with artificially increased bulk density), but not of Ceriops australis (Ola et al., 2018). Globally, there was no effect of pore water salinity on root production across mangrove studies (p = .46, r 2 = .10, n = 53, Table S5). The response of mangrove root growth to salinity is species dependent, non-linear and not monotonic (Ball, 1988;Ball & Pidsley, 1995;Burchett et al., 1989;Downton, 1982;Naidoo, 1987Naidoo, , 1990Nguyen et al., 2015). More data and consistent methodology of salinity measurements may be necessary to fully examine the effect of salinity. Many other soil factors are likely to influence mangrove root production, such as soil anoxic-oxic conditions, base cations or the age of the mangrove forest Cusack et al., 2018), but they have been rarely reported alongside with root production studies.

| LIK ELY TRENDS IN MANG ROVE ROOT PRODUC TI ON UNDER FUTURE CLIMATE CONDITIONS
Mangrove root production is likely to be affected by global changes that include warming, changes in precipitation regimes, sea level rise, atmospheric CO 2 rise and coastal nutrient enrichment.
Unfortunately, there are limited data available on the mangrove root production responses to global changes. Thus, below we provide perspectives and focus on critical questions that should be addressed about root responses to global changes, guided by our data, previous studies and information from other ecosystems. An increase in mangrove root accumulation with warming has been shown in an outdoor mesocosm (Coldren et al., 2019). We also found a statistically significant effect of air temperature crossed with geomorphological settings on root production, which likely indicates that mangrove root production will be sensitive to temperature variations and extremes in the future. An increase in mean annual air temperature might increase the root: shoot ratio and a rise in root production by increasing the mangrove photosynthetic rate as observed in a mangrove (Coldren et al., 2019) and forests (Norby & Jackson, 2000). Mangroves might grow until a maximum air temperature threshold for photosynthesis (and respiration) as in forests (Norby & Jackson, 2000) assuming that nutrient supply and water are not limited. If resources are limited, the root: shoot ratio might increase under higher temperatures, because mangrove trees might allocate more carbon to roots to explore the soil for nutrients and water uptake. In contrast, if resources are not limited, the increase in temperature might not increase the root: shoot ratio. This hypothesis is important to be tested in mangroves, especially for mangroves close to mega-cities that are prone to eutrophication (Mao et al., 2021).
A lack or a reduction of precipitation may reduce the growth of mangrove trees (Alongi, 2009;Simard et al., 2019) and thus might lead to a reduction of mangrove root production at least temporarily as observed in other ecosystems (Slette et al., 2022). Our data supported this hypothesis with a significant positive effect of precipitation on mangrove root production. For instance, the root production was relatively low (56-79 g of dry biomass m −2 year −1 ) in very arid mangroves of Mexico (i.e. the site with the lowest precipitation in our dataset: Ochoa-Gómez et al., 2019), while the root production was four times higher in very wet mangroves (Kihara et al., 2022). Low precipitation might also increase the amount of carbon allocated to mangrove roots to maintain water uptake as observed in terrestrial forests (Brunner et al., 2015). There were no aboveground production data collected, but the ratio of abo- suggesting that mangrove trees under low precipitation allocate relatively more carbon to root production as observed in terrestrial forests (Brunner et al., 2015).
The effect of sea level rise on mangrove root production is likely not linear and species dependent (Krauss et al., 2014). This might explain why we did not find any pattern of root production and tidal amplitude (a proxy for inundation: Rovai et al., 2018). Palaeorecords have shown that mangrove peat accretion mostly composed of roots at some sites (Ezcurra et al., 2016;Middleton & McKee, 2001) was stimulated by sea level rise up to 6.1 millimetres per year (Saintilan et al., 2020). Other studies have shown that an increase in inundation duration does not ultimately lead to a reduction in mangrove root decay . Therefore, it is likely that the accumulation of roots with sea level rise results from an increase in root production  or a change of root traits (e.g. root turnover, root carbon-to-nitrogen ratio, specific root length, root tissue density). There is a strong need for manipulative studies that modify the inundation duration in mature mangroves to better understand and forecast the response of root production to sea level rise. Reporting the duration or frequency of inundation with mangrove root production will help, because as shown above tidal amplitude data only exist in coarse spatial resolution.
The effect of CO 2 enrichment has been tested only in experiments with mangrove seedlings, which showed that CO 2 fertilisation does increase mangrove root production, but decreased root: shoot ratio (except under low nutrient concentration) (Jacotot et al., 2019;Reef et al., 2016). Similarly, free CO 2 air enrichment has also increased root production in terrestrial mature forests and salt marshes in outdoor experiments (Iversen, 2010;Norby et al., 2004;Norby & Jackson, 2000;Noyce et al., 2019). Finally, coastal nutrient enrichment alone sometimes results in a decrease in mangrove root production with an increase in carbon allocated to aboveground organs (Hayes et al., 2017;McKee, 1996;Naidoo, 1987). However, this was shown to be dependent on local conditions. In our dataset, we found that root production was more important in sites with higher nitrogen, likely reflecting an increase in overall plant growth.
Ongoing global environmental changes include the interactive effects of multiple drivers that might enhance or offset the effect of individual factors (Jacotot et al., 2019;Reef et al., 2016;Twilley et al., 2017). Yet, too few studies have investigated multiple factors simultaneously preventing accurate forecasting of mangrove root production under future environmental conditions.

| IMP ORTAN CE OF MANG ROVE ROOT PRODUC TION AND DYNAMIC S FOR C ARBON BURIAL AND LOSS E S IN MANG ROVE S
The carbon burial rate can be estimated using the global root decay rate and the root litter production per year (i.e. root necromass production), which equals the mangrove root production multiplied by the root mortality rate (Figure 4). Yet, the root mortality rate has been rarely studied in mangroves. A common approach to overcome the lack of root mortality data is to assume that the root system is in equilibrium and that root production equals root mortality (limitations are given below). If we assume that mangrove root production and mortality are in equilibrium, the global dead root production per year equals ~770 ± 202 g of dry biomass m −2 year −1 in mangroves (41 ± 11 Tg C year −1 for the whole mangrove area; Full methodology is given in Supplementary Material 1.4). Ouyang et al. (2017) estimated that the global root decomposition rate was 0.135% day − 1 (~49% year −1 ) in mangroves based on a systematic literature review and gave a range of litter decomposition per genus (Ouyang et al., 2017).
We applied the decomposition rate of the dominant mangrove tree genus of each geomorphological setting to estimate that around ~352 ± 94 g of dry biomass m −2 year −1 of the global dead root production (21 ± 2 Tg C year −1 ) can be assumed to be lost through decomposition, either as CO 2 and CH 4 emissions or through lateral loss of DOC and DIC after being consumed by microbes (Full detailed methodology is given in Supplementary Material 1.4). The remaining ~417 ± 108 g of dry biomass m −2 year −1 or 23 ± 5 Tg C year −1 of rootderived carbon is likely buried in mangrove soils, which corresponds to >90% of carbon buried in mangroves based on the 24 Tg C year −1 burial rate of Breithaupt and Steinmuller (2022) (Figure 5). Our root carbon burial rate is around four times higher than the only one previous estimate (i.e. ~50 g C m −2 year −1 equivalent to 100 g dry biomass m −2 year −1 in mangroves: Ouyang et al., 2017). However, in that study, the annual necromass production of mangroves was estimated by multiplying root production (g of dry biomass m −2 year −1 ) by root turnover rate (defined as root production divided by biomass) rather than by root mortality (Ouyang et al., 2017:  tality and by root orders as well as root production at a depth greater than 45 cm. We also considered that our calculation assumes root mortality equals root production for all mangrove roots irrespective of root orders and functional types, but for instance, fine absorptive roots will likely have a higher rate of mortality than fine transportive roots and coarse roots (Sun et al., 2016). There is no measurement for mangroves, but transportive fine roots die every 5-10 year, while fine absorptive roots die likely every 0.5-2.0 year depending on species and ecosystems (Clark et al., 2001;McCormack et al., 2015).
Nevertheless, our estimate strongly strengthens the previous estimate of carbon burial from root-derived carbon (Ouyang et al., 2017) with the inclusion of more data and by using a robust methodology based on geomorphological settings.
Live roots also contribute to mangrove carbon burial and losses through root respiration (i.e. CO 2 emission through autotrophic respiration) and rhizodepositions (i.e. release of labile carbon by roots). Autotrophic root respiration has rarely been quantified for mangrove trees and is highly variable (Lovelock, 2008;Lovelock et al., 2006;Ouyang et al., 2018). In Caribbean mangroves, root respiration accounted for a large portion of the CO 2 efflux from the soil in fringe mangroves, and less than 20% of the CO 2 efflux from scrub mangroves (Lovelock, 2008;Lovelock et al., 2006). Similarly, little is known about mangrove rhizodepositions  and their role in building and stabilising soil carbon stocks (Kida & F I G U R E 4 Root organic matter burial computation in mangroves. Annual root litter production is the root production multiplied by the root mortality rate (a). Annual root litter loss is the annual root litter production multiplied by the root decay rate (b). The root organic matter burial is annual root litter production subtracted from the annual loss of root litter (b).

F I G U R E 5
The importance of roots for the budget of the major carbon fluxes in the world's mangroves. All values are in Tg C year −1 for a mangrove area of ~137,600 km 2 . Red question marks are components that have not been quantified. Abbreviations: DIC, dissolved inorganic carbon loss; DOC, dissolved organic carbon loss; POC, particulate organic carbon loss from soil; CH 4 loss from soil and water. Data on root carbon production and root carbon burial are from our estimates; carbon burial is from Breithaupt and Steinmuller (2022); wood production, litter, soil respiration, CH 4 , DIC, DOC and POC are from Alongi (2020: tables 2, 3 & 5); net primary production is the total of wood, litter and root production.
Fujitake, 2020). Rhizodepositions include the transfer of carbon from roots into the rhizosphere and the soil through root exudations (including mucilage), root cell sloughing and root-associated symbionts living in the soil (e.g. mycorrhizas) (Jones et al., 2009). No in-situ quantifications of root exudations for mangrove trees exist, but root exudations have been shown to represent between 1% and 20% of the net primary production (NPP) in other temperate and tropical forests (Aoki et al., 2012;Jones et al., 2004;Phillips et al., 2008;Yin et al., 2014). The higher end of root exudation values was reported for tropical forest soils (Aoki et al., 2012) that are nutrient-deficient like mangroves (Reef et al., 2010

| UN CERTAINTIE S IN THE G LOBAL E S TIMATE OF MANG ROVE ROOT PRODUC TION
The estimation of global root production in mangroves could be im-  (Jardine & Siikamäki, 2014;Sanderman et al., 2018). Some geomorphological settings have also been understudied, like the Delta (n = 2), while they are likely the most productive in terms of net primary production and root production ( Figure 2).
In addition, most root production studies have limited their investigation to the top 0 to 45 cm depths (90% of the compiled data), but root production has been shown to occur well below this soil layer Castañeda-Moya et al., 2013;Cormier, 2021;Xiong et al., 2013). Shallow measurements are often justified by the assumption that root production is negligible at depth. This assumption requires justification and nuance, because root production below 45 cm depth accounted for up to 45% of the total root production in some mangroves in the United States (Castañeda-Moya et al., 2011) and up to 40% in some mangroves in China (Xiong et al., 2013), Vietnam  and Malaysia (Muhammad-Nor et al., 2019). The controls of mangrove root distribution at depth are not clear , but are disproportionally important for carbon burial. This is because root decay is likely to be slower at depth due to reduced microbial biomass, stabilisation of SOM by minerals and reduced supply of root oxygen and exudates (i.e. labile carbon released by roots) (Rasse et al., 2005;Spivak et al., 2019).
Finally, methods to measure root production can lead to strong differences in root production measurements (Hendricks et al., 2006;Kihara et al., 2022) and could be improved. Sequential coring might reflect spatial and temporal variability of root biomass rather than root production (Hendricks et al., 2006;Singh et al., 1984). Ingrowth cores have several limitations, such as between sampling intervals the roots might die and be unaccounted for production, roots can be lost during core washing (up to 30%) (Cahoon et al., 2003;Sierra et al., 2003) or prune after being damaged during the ingrowth core installation. Those limitations should be acknowledged, but can also be overcome in several ways . Minirhizotrons have proved to be suitable in other ecosystems and recently in mangroves (Arnaud, 2021;Arnaud et al., 2021), but this method needs further developments to convert their result into biomass increment .

| MOVING FORWARD: A ROADMAP TOWARDS CLOS ING E XIS TING K NOWLEDG E G APS ON MANG ROVE ROOT PRODUC TION
(1) We advocate for the creation of a scientific network on mangrove root traits, including root production. This MangRoot Network could facilitate collaboration between scientists, increase capacity building for the measurement of root production (e.g. through workshops) and facilitate the creation and the maintaining of a root trait database for mangroves (such as FRED: Iversen et al., 2017 or TropiRoot: https:// tropi roott rait.github.io/Tropi RootT rait/). Such a database, could aim to assess the growing body of literature on mangrove root production and more broadly on root traits (e.g. mortality, exudations) to understand how mangrove root traits vary over time, space and in response to global changes. The network could also trigger large and coordinated in-situ root trait measurements paired with surface elevation tables (i.e. measuring soil accretion rate) to provide a mechanistic understanding of the role of root production in soil accretion across mangroves.
(2) A significant increase in global efforts to observe mangrove root production is required, prioritising currently under-represented geomorphic settings (i.e. delta) and geographic areas, such as Indonesia, South America or Africa. Specific focus is required on carbon-rich mangroves, but also towards quantifying root production in deeper soil layers (~1 m deep) than current practice. Root production studies will also be more valuable to increase our mechanistic understanding if they include multiple treatment effects (e.g. sea level rise crossed with warming) and report several environmental factors (i.e. inundation frequency and duration).
(3) Improved forecasting of mangrove root production is critical.
More in-situ research needs to be carried out on the belowground carbon dynamics of mangroves under global change scenarios.
Ecosystem-plot experiments using Free Air CO 2 Enrichment (FACE) crossed with warming have indicated unexpected belowground carbon dynamics, notably in salt marshes (Noyce et al., 2019). The next step is to develop a multifactorial experiment involving manipulated sea levels, warming and FACE that will not only cover root production, but the overall mangrove ecosystem response to global changes (Arnaud, 2020).

| CON CLUS IONS
• Global mangrove root production is ~770 ± 202 g of dry biomass m −2 year −1 or 41 ± 11 Tg C year −1 over the entire area The main controls of root production are the geomorphological settings of mangroves in combination with air temperature and precipitation (i.e. ~30% of variance explained; p < .001).
• Burial of root-derived carbon is 23 ± 5 Tg C year −1 representing >90% of the total carbon buried in world mangroves. Vertical and lateral losses of carbon derived from mangrove roots were 19 ± 5 Tg C year −1 revealing the importance of studying roots to better understand global mangrove losses (e.g. DIC, DOC, CO 2 ).
• Warming, changes in precipitation and eutrophication are likely to alter mangrove root production. Yet, there are significant knowledge gaps to predict the interactive effects of multiple environ- Theme of the University of Birmingham.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available